J. (Eds.)-Mineral Processing Plant Design, Practice, And Control Proceedings, Volumes 1-2-Society for Mining, Metallurgy, And Exploration (SME) (2002)

Mineral Processing Plant Design, Practice, and Control

PROCEEDINGS VOLUME 1 - VOLUME 2

Edited by Andrew L. Mular, Doug N. Haibe, and Derek J. Barrett

These organizationsprovided generousfinancial support for this publication.

Platinum Level Canadian Process Technologies Inc. Metso Minerals Industries Inc.

Gold Level Newmont Mining Corporation Norcast OSISOft

Silver Level Pocock Industrial, Inc. WesTech

Coverphoto courtesy of Rick Coleman of P.T. Freeport Indonesia (asubsidiary of Freeport-McMoRan Copper & Gold Inc.). The photo shows the P.T.Freeport Indonesia milling complex located in West Papw, Indonesia. Thefacilities currently process 245,000 tonnes per day of copper, gold, and silver ore, producing 2.8 million tonnes of concentrate annually.

Society for Mining, Metallurgy, and Exploration, Inc. (SME) 8307 Shaffer Parkway Littleton, Colorado USA 80127 (303) 973-9550 / (800) 763-3132 www .smenet.org SME advances the worldwide mining and minerals community through information exchange and professional development. Copyright Q 2002 Society for Mining, Metallurgy, and Exploration, Inc

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or othenvise, without the prior written permission of the publisher. All Rights Reserved. Printed in the United States of America

Disclaimer The papers contained in this proceedings are published as supplied by individual authors. Any statement or views presented here are those of individual authors and are not necessarily those of the Society for Mining, Metallurgy, and Exploration, Inc. The mention of trade names for commercial products does not represent or imply the approval or endorsement of SME. ISBN 0-87335-223-8 ISBN 978-0-87335-223-9

Preface

Nearly 25 years ago, SME published its first major symposium volume on plant design practice: Mineral Processing Plant Design. Two more volumes, the Design and Installation of Comminution Circuits (1982) and the Design and Installation of Concentration and Dewatering Circuits (1986), were published creating a positive worldwide affect on mining, mineral processing, and metallurgy. And although out of print, each of these publications is still in use today. In 1998, SME, with strong encouragement from the Canadian mineral processing division of CIM, agreed to update the plant design series to provide a current overview of all facets of mineral processing plant design, control, and practice. An organizing committee and a general committee were fonned to oversee this huge task. The organizing committee, with input from some 40 members of the general committee, chose sections for the symposium as well as section co-editors. The proceedings editors worked with the section co-editors vetting the content and making the final choices for each section. In 1999, a call for papers was issued in Mining Engineering and various other journals. Authors were contacted by co-editors and encouraged to follow a stringent timetable to ensure that the proceedings would be published in time for distribution at the symposium. Despite busy work schedules, our authors and co-authors managed to meet their deadlines with relatively minor delays. The section co-editors deserve significant praise for this accomplishment. They maintained close contact with corresponding authors and provided help when requested. Readers will find that the results of this hard work are most gratifying and that our authors deserve to be recognized and appreciated for years to come. The useful information included in these two volumes will serve as an up-to-date aid for university-level professors and students in plant design courses; reference material for operators who must consider plant expansion, renovations, and new projects; and a quick reference for mineral processing plant design engineers, engineers from other disciplines, consultants in mineral processing, consultants who have minimal knowledge of mineral processing plant design, suppliers, and manufacturers. The symposium is being held at the Fairmont Hotel Vancouver in Vancouver, B.C., Canada, from Oct. 20-24, 2002. It should be stressed that a comprehensive symposium of this type is organized once every 12 to 15 years and is international in scope. During the meeting, competition is avoided between sessions by not programming them concurrently. Everyone attends the same sessions or roundtables, thereby stimulating interaction and discussion throughout the five-day period. The proceedings editors, section co-editors, and authors wish to thank Ms. Joette Cross, meetings manager; Ms. Tara Davis, program manager; and Ms. Jane Olivier, manager of book publishing, for their excellent assistance on this project. Despite our idiosyncrasies, their help and guidance was provided with professional courtesy and efficiency. We appreciate it. Andrew L.Mular Doug Halbe Derek J. Barratt

V

Contents

The Formal Basis of Design ~

~~

Design Criteria: The Formal Basis of Design J.W. Scott .............................................................................................................................................

3

1 Sampllng

Section Co-Editors...............................................................................................................................

23

Sampling a Mineral Deposit for Feasibility Studies

K.J.Ashley ...........................................................................................................................................

25

Sampling in Mineral Processing

J.W. Merks ...........................................................................................................................................

37

Sampling High Throughput Grinding and Flotation Circuits J. Mosher, D. Alexander ......................................................................................................................

63

Practical and Theoretical Difficulties When SamplingGold F.F. Pitard ...........................................................................................................................................

77

Sampling a Mineral Deposit for Metallurgical Testing and the Design of Comminution and Mineral Separation Processes J. Hanks, D. Barratt .............................................................................................................................

99

2 Bench Scale and Pilot Plant Testwork Section Co-Editors ...............................................................................................................................

117

Overview of Metallurgical Testing Procedures and Flowsheet Development T.P. McNulty ........................................................................................................................................

1u)

Bench-Scale and Pilot Plant Tests for Comminution Circuit Design

J. Mosher, T. Bigg ...............................................................................................................................

123

The Selection of Flotation Reagents via Batch Flotation Tests P. Thompson ........................................................................................................................................

136

Bench and Pilot Plant Programs for Flotation Circuit Design S.R. Williams, M.O. Ounpuu, K.W. Sarbutt .........................................................................................

145

Bench-Scale and Pilot Plant Testwork for Gravity Concentration Circuit Design A. R. Laplante, D.E. Spiller ..................................................................................................................

160

Bench Scale and Pilot Plant Tests for Magnetic Concentration Circuit Design

D.A. Norrgran. M.J. Mankosa .............................................................................................................

176

Bench-Scale and Pilot Plant Tests for Thickening and Clarification Circuit Design B.K. Pocock, C.B. Smith, G.D. Welch..................................................................................................

201

xi

Bench-Scale and Pilot Plant Tests for Filtration Circuit Design T. Kram ................................................................................................................................................

207

Gold Roasting, Autoclaving, or Bio-Oxidation Process Selection Based on Bench-Scale and Pilot Plant Test Work and Costs J. McMullen, K.G. Thomas ..................................................................................................................

211

Bench-Scale and Pilot Plant Tests for Cyanide Leach Circuit Design G.E. McClelland, J.S. McPartland ......................................................................................................

251

Bench-Scale and Pilot Plant Work for Gold- and Copper-Recovery Circuit Design D. Thompson........................................................................................................................................

264

Guiding Process Developments by Using Automated Mineralogical Analysis D. Sutherland, Y. Gu ............................................................................................................................

270

3 Financial and Feasibiiltv Studies

Guidelines to Feasibility Studies J. Scott, B. Johnston.............................................................................................................................

281

Major Mineral Processing Equipment Costs and Preliminary Capital Cost Estimations A.L. Mular ...........................................................................................................................................

310

Process Operating Costs with Applications in Mine Planning and Risk Analysis D. Halbe, T.J. Smolik ...........................................................................................................................

326

Financial Analysis and Economic Optimization L.D. Smith ............................................................................................................................................

346

Mining Project Finance Explained R. Halupka ...........................................................................................................................................

371

4 Models and Simulators for Selection, Sizing, and Design Section Co-Editors...............................................................................................................................

381

Mineral Processing Plant/Circuit Simulators: An Overview

J. Herbst, R.K. Rajamani, A. Mular, B. Flint08 ..................................................................................

383

BRUNO: Metso Minerals' Crushing Plant Simulator D.M. Kaja ............................................................................................................................................

404

PlantDesigner'? A Crushing and Screening Modeling Tool P. Hedvall, M. Nordin ..........................................................................................................................

421

JKSimMet: A Simulator for Analysis, Optimisation, and Design of Comminution Circuits R.D. Morrison, J.M. Richardson ..........................................................................................................

442

JKSimFloat as a Practical Tool for Flotation Process Design and Optimization M.C. Harris, K.C. Runge, W.J. Whiten, R.D. Morrison .......................................................................

461

USIM PAC 3: Design and Optimization of Mineral Processing Plants from Crushing to Refining S. Brochot, J. Villeneuve, J.C. Guillaneau, M.V.Durance, F. Bourgeois ............................................

479

xii

Emergence of HFS as a Design Tool in Mineral Processing J.A. Herbst, L.K. Nordell .....................................................................................................................

495

Reducing Maintenance Costs Using Process and Equipment Event Management O.A. Bascur, J.P. Kennedy...................................................................................................................

507

Enterprise Dynamic Simulation Models D. W.Ginsberg .....................................................................................................................................

528

5 Commlnutlon (Crushlng and Grlndlng) Section Co-Editors ...............................................................................................................................

537

Factors Which Influence the Selection of Comminution Circuits D. Barratt. M. Sherman .......................................................................................................................

539

Types and Characteristics of Crushing Equipment and Circuit Howsheets K. Major ..............................................................................................................................................

566

Selection and Sizing of Primary Crushers R. W. Utley............................................................................................................................................

584

In-Pit Crushing Design and Layout Considerations K.Boyd, R. W. Utley .............................................................................................................................

606

Selection and Sizing of Secondary and Tertiary Cone Crushers G.Beerkircher, K. O’Bryan, K. Lim ....................................................................................................

621

Selection, Sizing, and Special Considerations for Pebble Crushers K. O’Bryan, K. Lim ..............................................................................................................................

628

Selection and Sizing of High Pressure Grinding Rolls R. Klymowsky, N. Patzelt, J. Knecht, E. Burchardt .............................................................................

636

Crushing Plant Design and Layout Considerations K. Boyd ................................................................................................................................................

669

Types and Characteristics of Grinding Equipment and Circuit Howsheets M.I. Callow, A.G. Moon ......................................................................................................................

698

Selection of Rod Mills, Ball Mills, and Regrind Mills

C.A. Rowland Jr. ..................................................................................................................................

710

Selection and Sizing of Autogenous and Semi-Autogenous Mills D. Barratt, M.Sherman .......................................................................................................................

755

Selection and Sizing of Ultrafine and Stirred Grinding Mills J.K.H. Lichter, G. Davey .....................................................................................................................

783

Grinding Plant Design and Layout Considerations M.I. Callow, D.G. Meadows ................................................................................................................

801

Selection and Evaluation of Grinding Mill Drives G.A. Grandy, C.D. Danecki, P.F. Thomas ...........................................................................................

819

The Design of Grinding Mills V. Svalbonas ........................................................................................................................................

840

xiii

6 Size Separation Section Co-Editors ...............................................................................................................................

865

Sizing and Application of Gravity Classifiers W.M. Reed ...........................................................................................................................................

867

Hydrocyclone Selection for Plant Design T.J. Olson, P.A. Turner........................................................................................................................

880

Coarse Screening M.A. Bothwell, A.L. Mular ...................................................................................................................

894

Fine Screening in Mineral Processing Operations S.B. Valine, J.E. Wennen .....................................................................................................................

917

The Use of Hindered Settlers to Improve Iron Ore Gravity Concentration Circuits S. Hearn ...............................................................................................................................................

929

7 Solid-Solid separation Section Co-Editors ...............................................................................................................................

945

Types and Characteristics of Gravity Separation and Flowsheets R.O. Burt ..............................................................................................................................................

947

Types and Characteristics of Heavy-Media Separators and Flowsheets R.A. Reeves ..........................................................................................................................................

962

Types and Characteristics of Non-Heavy Medium Separators and Flowsheets J.K. Alderman ......................................................................................................................................

978

The Selection and Sizing of Centrifugal Concentration Equipment: Plant Design and Layout A.R. Laplante .......................................................................................................................................

995

Sizing and Selection of Heavy Media Equipment: Design and Layout D.F. Symonds, S. Malbon ....................................................................................................................

1011

Photometric Ore Sorting B. Arvidson ..........................................................................................................................................

1033

Electrical Methods of Separation A.L. Mular ...........................................................................................................................................

1049

Selection and Sizing of Magnetic Concentrating Equipment: Plant DesignlLayout D.A. Norrgran, M.J. Mankosa .............................................................................................................

8 Flotation Section Co-Editors ...............................................................................................................................

1095

Overview of Flotation Technology and Plant Practice for Complex Sulphide Ores N. W. Johnson, P.D. Munro ..................................................................................................................

1097

Overview of Recent Developments in Flotation Technology and Plant Practice for Copper Gold Ores A. Winckers ..........................................................................................................................................

1124

xiv

An Overview of Recent Developments in Flotation Technology and Plant Practice for

Nickel Ores A. Kerr .................................................................................................................................................

ll42

Nonsulfide Flotation Technology and Plant Practice J. Miller, B. Tippin, R. Pruett ..............................................................................................................

1159

Design of Mechanical Flotation Machines M.G. Nelson, F.P. Traczyk, D. Lefinski...............................................................................................

ll79

Flotation Equipment Selection and Plant Layout K.R. Wood............................................................................................................................................

1204

Column Flotation G. Dobby .............................................................................................................................................

1239

9 Soltd-Llauld Separatlon Section Co-Editors ...............................................................................................................................

12253

Characterizationof Process Objectives and (General) Approach to Equipment Selection C.E. Silverblatt, J.H.Easton ................................................................................................................

1285

Centrifugal Sedimentation and Filtration for Mineral Processing W. b u n g ..............................................................................................................................................

1262

Characterizationof Equipment Based on Filtration Principals and Theory G.D. Welch ..........................................................................................................................................

1289

Testing, Sizing, and Specifying SedimentationEquipment T. Laros, S. Slottee, F. Baczek .............................................................................................................

1295

Testing, Sizing, and Specifying of Filtration Equipment C.B.Smith, I.G. Townsend...................................................................................................................

2313

Design Features and Types of Sedimentation Equipment F. Schoenbrunn, T. Laros ....................................................................................................................

1331

Design Features and Types of Filtration Equipment C. Cox, F. Traczyk...............................................................................................................................

1342

Plant Design, Layout, and Economic Considerations M . Erickson, M . Blois ..........................................................................................................................

1358

10 Pumping, Material transport, Drying, and Storage Section Co-Editors ...............................................................................................................................

1371

Selection and Sizing of Slurry Pumps M.J. Bootle ...........................................................................................................................................

1373

Selection and Sizing of Slurry Lines, Pumpboxes, and Launders B. Abulnaga, K. Major, P. Wells..........................................................................................................

1403

Slurry Pipeline Transportation B.L. Ricks .............................................................................................................................................

1422

The Selection and Sizing of Conveyors, Stackers, and Reclaimers G. B a ~ o o tD. , Bennett, M. Col ............................................................................................................

1448

Selection and Sizing of Concentrate Drying, Handling, and Storage Equipment M.E. Prokesch, G. Graber ...................................................................................................................

1463

The Selection and Sizing of Bins, Hopper Outlets, and Feeders J. Carson, T. Holmes ...........................................................................................................................

1478

11 Pre-Oxidation ~~~

Section Co-Editors ...............................................................................................................................

1491

Design of Barrick Goldstrike’s Two-Stage Roaster D. Warnica, A. Cole, S. Bunk ...............................................................................................................

1493

Selection of Materials and Mechanical Design of Pressure Leaching Equipment K. Lamb, J. Gulyas ..............................................................................................................................

1510

Barrick Gold-Autoclaving and Roasting of Refractory Ores K.G. Thomas, A. Cole, R.A. Williums ..................................................................................................

l530

Selection and Sizing of Biooxidation Equipment and Circuits C.L. Brierley, A. P. Briggs ....................................................................................................................

1540

l 2 Leaching and Adsorption Circults Section Co-Editors ...............................................................................................................................

1569

Copper Heap Leach Design and Practice R.E. Scheffel .........................................................................................................................................

1571

Precious Metal Heap Leach Design and Practice D. W. Kappes ........................................................................................................................................

1606

Agitated Tank Leaching Selection and Design K.A. Altman, M. Schafier, S.McTavish ..............................................................................................

1631

CIP/CIUCIC Adsorption Circuit Process Selection C.A. Fleming ........................................................................................................................................

1644

CIP/CIUCIC Adsorption Circuit Equipment Selection and Design K.A. Altman, S.McTavish ....................................................................................................................

1652

13 Extraction Section Co-Editors ...............................................................................................................................

1661

Zinc Cementation-The Merrill Crowe Process A.P. Hampton ......................................................................................................................................

1663

Selection and Design of Carbon Reactivation Circuits J. von Beckmann, P.G. Semple ............................................................................................................

1680

Selection and Sizing of Elution and Electrowinning Circuits P. Hosford, J. Wells .............................................................................................................................

1694

Selection and Sizing of Copper Solvent Extraction and Electrowinning Equipment and Circuits C.G. Anderson, M.A. Giralico, T.A. Post, T.G. Robinson, O.S. Tinkler ...............................................

1709

xvi

14 Bullion Production and Refining Section Co-Editors ...............................................................................................................................

1745

Bullion Production and Refining C.O. Gale, T.A. Weldon .......................................................................................................................

1747

Platinum Group Metal Bullion Production and Refining C.G. Anderson, L. C. Newman, G.K.Roset...........................................................................................

1760

Fundamentals of the Analysis of Gold, Silver, and Platinum Group Metals C.G. Anderson .....................................................................................................................................

1778

15 Tailings Disposal, Wastewater Dlsposal, and the Environment Section Co-Editors ...............................................................................................................................

1807

Management of Tailings Disposal on Land B.S. Brown ...........................................................................................................................................

1809

Design of Tailings Dams and Impoundments P. C.Lighthall, M.P. Davies, S.Rice, T.E. Martin ...............................................................................

1828

Hazardous Constituent Removal from Waste and Process Water L. Twidwell, J. McCloskey, M. Gale-Lee .............................................................................................

1847

Treatment of Solutions and Slurries for Cyanide Removal M.M. Botz, T.I. Mudder .......................................................................................................................

1866

Strategies for Minimization and Management of Acid Rock Drainage and Other Mining-Influenced Waters R.L. Schmiemund ................................................................................................................................

1886

Environmental and Social Considerations in Facility Siting B.A. Filas,R. W.Reisinger, C.C. Pamow ............................................................................................

1902

16 Construction Materials for Equipment and Plants Section Co-Editors ...............................................................................................................................

1909

Selection of Metallic Materials for the MiningNetallurgical Industry G. Coates .............................................................................................................................................

1911

Elastomers in the Mineral Processing Industry P. Schnarr, L.E. Schaeffer, H.J. Weinand............................................................................................

1932

Plastics for Process Plants and Equipment G. W.McCuaig .....................................................................................................................................

1953

Commercial Acceptance and Applications of Masonry and Membrane Systems for the Process Industries R.E. Aliasso Jr., T.E. Crandall, D.M. Malone, R.J. Storms .................................................................

1962

17 Power, Water, and Support Facilities Section Co-Editors ...............................................................................................................................

1971

The Development of an Electric Power Distribution System M.N. Brodie .........................................................................................................................................

1973

xvii

Selection of Motors and Drive Systems for Comminution Circuits P. F. Thomas.........................................................................................................................................

1983

Selection of Metallurgical Laboratory and Assay Equipment: Laboratory Designs and Layouts P.F. Wells ............................................................................................................................................

2011

On-Line Composition Analysis of Mineral Slumes T.F. Braden, M. Kongas, K. Saloheimo ...............................................................................................

2020

18 Process Control and lnstrumentatlon Section Co-Editors ...............................................................................................................................

2049

Introduction to Process Control B. Flintoff.............................................................................................................................................

2051

Well Balanced Control Systems T. Stuffco, K. Sunna .............................................................................................................................

2066

The Selection of Control Hardware for Mineral Processing R.A. Medower, R.E. Cook ....................................................................................................................

2077

Basic Field Instrumentation and Control System Maintenance in Mineral Processing Circuits

J.R. Sienkiewicz ...................................................................................................................................

2104

Strategies for Instrumentation and Control of Crushing Circuits S.D. Parsons, S.J. Parker, J. W. Craven, R.P. Sloan ............................................................................

2114

Strategies for the Instrumentation and Control of Grinding Circuits R. Edwards, A. Vien, R. Perry .............................................................................................................

2130

Strategies for the Instrumentation and Control of Solid-Solid Separation Processes G.H. Luttrell, M.J. Mankosa ................................................................................................................

2w2

Strategies for Instrumentation and Control of Thickeners and Other Solid-Liquid Separation Circuits F. Schoenbrunn, L. Hales, D. Bedell ...................................................................................................

2164

Strategies for Instrumentation and Control of Flotation Circuits H. Laurila, J. Karesvuori, 0. Tiili .......................................................................................................

2174

Pressure Oxidation Control Strategies J. Cole, J. Rust .....................................................................................................................................

2196

19 Engineering, Procurement, Construction, and Management Section Co-Editors ...............................................................................................................................

2209

Development of a Mineral Processing Flowsheet-Case History, Batu Hijau T. de Mull, S. Saich, K. Sobel ..............................................................................................................

2211

Specification and Purchase of Equipment for Mineral Processing Plants C. Hunker, S. Maldonado ....................................................................................................................

2223

The Management and Control of Costs of Capital Mineral Processing Plants D.W. Stewart ........................................................................................................................................

2230

xviii

Schedule Development and Schedule of Control of Mineral Processing Plants P. Kumar .............................................................................................................................................

2238

The Risks and Rewards Associated with Different Contractual Approaches P.J. Card .............................................................................................................................................

2245

Success Strategies for Building New Mining Projects R.J. Hickson .........................................................................................................................................

2250

20 Start-up, Commissloning, and Training Section Co-Editors ...............................................................................................................................

2275

Pre-Commissioning,Commissioning, and Training T. Watson .............................................................................................................................................

2277

Plant Ramp Up and PerformanceTesting R.M. Nendick .......................................................................................................................................

2285

Preparation of Effective Operating Manuals to Support Operator Training for Metallurgical Plant start-ups S.R. Brown ...........................................................................................................................................

2290

Planning and Staffing for a Successful Project Start-up K.A. Brunk, L J. Buter, K.M. Levier .....................................................................................................

2299

Maintenance Scheduling,Management, and Training at Start-up: A Case Study P. Vujic ................................................................................................................................................

2315

Operator Training A. Ken .................................................................................................................................................

2328

Safety and Health Considerations and Procedures During Plant Start-up

L.A. Schack ..........................................................................................................................................

2337

2 1 Case Studies Section Co-Editors...............................................................................................................................

2343

Sunrise Dam Gold Mine-Concept to Production W.R. Lethlean, P.J. Banovich ..............................................................................................................

2345

A Case Study in SAG Concentrator Design and Operations at P.T. Freeport Indonesia R. Coleman, A. Neale, P. Staples .........................................................................................................

2367

High Pressure Grinding Roll Utilization at the Empire Mine D.J. Rose, P.A. Korpi, E.C. Dowling, R.E. Mclvor ..............................................................................

2380

The Raglan Concentrator-Technology Development in the Arctic J. Holmes, D. Hyma,P.Langlois .........................................................................................................

2394

Author index ....................................................................................................................................

1-1

Subject Index...................................................................................................................................

13

xix

Sampling a Mineral Deposit for Feasibility Studies K e v i n 1 Ashley'

ABSTRACT Mineral deposits are sampled for several reasons including resource evaluation, determination of the physical and chemical characteristics of materials, and process amenability. This paper discusses the types of sampling generally included in the feasibility phase and how, with coordination and communication, these samples can be used effectively to fulfill the needs of all disciplines involved in the study process. Also discussed are the amounts of sampling required for each type and purpose, methods for making this determination, and presentation of the results for the intended audience. INTRODUCTION Sampling for plant design during the feasibility study phase of a project can be critical to the overall success of the project but this is all too often minimized. Most of the sampling performed in support of feasibility tends to be oriented at identification and quantification of the ore reserve, which is performed usually by geologists and mining engineers, while sampling for comminution and amenability of the ore to processing is left for later bulk sampling programs (as described in the next chapter). While precise quantification of the processing characteristics of the plant feed material does require the additional sampling, there is much information that can be obtained fiom the initial feasibility sampling programs that can guide and even determine the testwork and final plant design. Feasibility studies are performed for the purpose of identifying which potential projects warrant further expenditure. In practice, only 20 to 40 percent of projects that attain feasibility level ever proceed to implementation. As such, it is not surprising that mining companies wish to limit the amount of money they spend during this phase. Also, due to the emphasis placed on ore reserves during the earliest parts of a project, the personnel involved during the early feasibility stage tend to be geologists, mining engineers and some geotechnical and environmental engineers. This budget awareness and compartmentalization of expertise tends to marginalize the collection of plant design data during this earliest phase. This can lead to incorrect characterization of the ore reserves and/or costly delays while work is reperformed to reconcile the in situ content of the target commodity with the recoverable content upon which the revenue of the project is based. It is therefore important for mining companies to involve plant design engineers early in the feasibility process and for plant design engineers to learn how they can effectively gather the sample data they require without substantial additional cost. This paper is oriented to plant design engineers in order to acquaint them with the types of sampling available and the methods by which they can coordinate with geologist, mining engineer, geotechnologist, hydrologist, and environmental colleagues in order to efficiently design sampling programs that support all disciplines. Although this paper is slanted toward the requirements for the final feasibility study phase of a project, the understanding and application of the concepts to earlier stages of conceptual and pre-feasibility studies is encouraged. Several people have developed the methodologies expounded in this paper over the past 40 years working on a cumulative of over 70 feasibility studies, of which the author has been involved with more than 30. In some cases the lessons learned have come after the fact but in

' Bechtel Corporation, San Francisco, California 25

many cases early intervention has allowed mining companies to save large amounts of time and money, and in some cases whole projects, by utilizing these techniques. The plant design engineer is encouraged to add these techniques to their toolbox and to take an active role during the early phases of feasibility studies in order to implement them. TYPES OF SAMPLING AND ANALYSES The purpose of sampling is to determine the physical and chemical attributes of materials associated with a potentially exploitable ore body. Some of these samples involve collection of material for analysis and testing remote from the site and others involve in situ measurement. Materials tested include ore and waste materials associated with the ore body and some materials that are remote from the ore body in areas where plant facilities and in~astructuremay be located. The analyses performed on the test samples vary with the information sought and the types of processes that might be involved in the extraction of the desired commodity. As plant design engineers are often not involved in the collection of this wide variety of samples, the following is a brief review of the types of sampling that are often used in the early stages of a project and the types of analyses that are performed. Types of Sampling Core Drilling. The preferred method of recovering material from within an ore body, core drilling is also the most expensive. Cores normally are taken in the range from 27 mm to 100 mm in diameter but can be as small as 11 mm diameter and as large as 200 mm diameter or greater for special applications. (USACE 2001) Core intervals are usually taken in one to three meter lengths. Cores are generated using a circular drill bit that allows a cylinder of rock to rise within the interior of the drill steel as the drill bit progresses. The drill bit is usually impregnated with diamonds in order to cut the rock cleanly. The cylinder of rock is captured within a core barrel in the interior of the drill steel that is either lifted separately (via a “wireline”) or is extracted with the drill steel as the drill progresses each drilling interval. The resulting drill core generally has a smooth surface and is consistent in diameter along its length. The contiguity of the drill core along its length is dependent on the integrity of the rock being drilled and the skill of the drill operator. The recovered pieces of core are placed in core boxes that hold several intervals side by side so that a drill hole can be stored and observed without laying the cores out to their total length. Core drilling can be performed at any angle including upward in the case of drilling from underground workings. Reverse Circulation Drilling. Due to the high cost of core drilling, some mining companies augment their sampling programs using the much less expensive reverse circulation (RC) drilling method. RC operates in the same manner as methods used when holes are drilled for purposes other than sampling, e.g., for developing a water well, where the chips excavated by the drill bit are removed by a drilling fluid (air, water or mud) that is circulated down the interior of the drill steel and back out the annulus between the wall of the hole and the drill steel except that in RC drilling the path of the drilling fluid is reversed such that it travels up the inside of the drill steel instead of down. This is accomplished by using a specially designed double walled drill steel that allows the drilling fluid to travel down an annulus between the walls to the drill bit where it picks up the chips and transports them back up the middle of the pipe. The product of RC drilling is chips that are usually no greater than 30 mm. In order to keep the samples from each drilling interval separate, it is necessary to allow the drilling fluid to circulate for an interval of time at the end of each interval. Also, in order not to lose the fine material as it comes out of the hole (especially when drilling with air) the material must be passed through a device such as a cyclone. The samples from each drilling interval are bagged separately as they come from the hole. RC drilling is not usable when knowledge of the exact transition from one rock type to another is required, such as in coal. Auger Drilling. Some materials are so unconsolidated that they cannot be sampled by either core drilling or reverse circulation drilling. This is the case with soils, placer deposits of river gravels, and previously placed materials such as mine waste piles and tailings impoundments. In

26

such cases auger drilling is used. Auger drilling is accomplished using a cutting bit to which are attached flights of spiral augers for transport of the material to the surface. Depending on the material being drilled, augering may be performed within a casing in order to support the walls of the hole. The sample may consist of the entire amount of material extracted from the hole or a hollow tube in the middle of the drill stem can be used to collect materials from only desired depths. Other Drilling Methods. Other specialized drilling methods are also used to recover samples of material and to make holes within which in situ measurements can be made. Chip and mud samples are sometimes collected from ordinary rotary drilling which is less expensive than core drilling or reverse circulation drilling, but these are less reliable as to sample location and dilution and are usually used for indicative purposes only. Specialized drills for collection of large pieces of rock, such as in placer sampling, are also used. Channel Sampling. Where the material to be sampled is exposed at the surface or in underground workings, channel sampling is used. In this procedure a channel, of dimensions similar to the diameter of the core or RC hole, is excavated. In solid ground this is accomplished by making two parallel cuts with a rock saw to the desired depth and then breaking the material between the cuts with a pick or rock hammer. The resulting samples are chips similar to those recovered through RC drilling. Generally samples are taken at the same interval as used in the drilling on the project. This is for geostatistical reasons. Channel samples have the advantage over RC samples in that the materials can be geologically logged prior to excavation. Trench Sampling. Where the material to be sampled is close to the surface, trenches are excavated to gain access to the material and samples are taken either as channel samples within the trenches or as selected samples of the excavated material. The result is broken material. Grab Sampling. Where the material to be sampled is exposed in existing surface or underground workings, it is possible to gather material on a random basis as it is mined or transported. Such grab samples can be useful for obtaining overall averages for large amounts of material, however they are not generally useful for identifying material characteristics at specific locations as selection of the samples may be biased by reasons such as ease of access or size of rock pieces. Geochemical and Environmental Samples. In addition to the samples that are collected for direct information about soils and rocks associated with the feasibility study, there are samples that are taken for other purposes that may be useful to the plant design engineer. These include vegetation, animal and sediment samples whose characteristics may indicate underlying conditions that are applicable to the plant design. For instance, the occurrence of certain plants may indicate the presence of levels of selenium that can influence the location of the tailings disposal site for a project. Water Samples. During feasibility investigations samples of surface and ground water are taken for environmental, mine design, and process design purposes. Types of Analyses Most of the sampling methods listed above result in a hole in the ground and a sample of rock or soil that, if good practice has been followed, have been located with an accurate survey and catalogued. Following is a list of the types of analyses that are typically performed. In Situ Measurements. Before addressing analyses of the sample material collected, it is important to note that some analyses are performed using the excavated hole. Primary among these in situ measurements is a downhole survey in order to locate the position of the samples that have been taken since drill holes are seldom straight over long distances. Other types of probes that are inserted into open drill holes include calipers to measure the diameter of the hole along its length, resistivity sensors to detect the presence of water and carbonaceous materials, gamma sources and detectors to measure density and natural radiation levels, expandable cylinders to measure in situ rock strengths and stresses, and gauges to measure groundwater levels. Sophisticated tomographic techniques use sensors placed in multiple drill holes to provide detailed

27

information on the material between the drill holes. Work is also underway to develop sensors that will be able to detect the presence and content of metals. Geologic Logging. Drilling samples, both core and chips, are logged by geologists in order to record the variations in material characteristics along the drill hole. Samples are logged for rock types, mineral types, alteration types, associations of various minerals, and estimates of content of the desired commodity. Core samples are also logged for fracture spacing and intensity, rock quality designation (RQD), and fracture orientation. In many cases the core is photographed in order to record its original condition and color. Chemical Assays. Representative portions of the samples are taken for chemical analysis. In the case of core, this is usually attained by sawing the core lengthwise and using !4 or 'h of the core. Chip samples are split according to appropriate statistical methodologies to reduce the quantity sent for analysis. The chemical analyses can include moisture content, total metal content, soluble metal content, content of non-metal commodities such as potash and lime, heat content for coal and oil shale, and ash content after combustion. In order to save money, many mining companies assay all samples for the main commodities of interest but assay only a portion of the samples for all metals and commodities that may affect the project. In some cases multiple samples are combined to give composite metal values for drill hole intervals greater than the original sampling interval. Mineralogical Assessment. Many metallurgical processes act to separate minerals contained in ores. Determination of the amounts of the various minerals is often made using either the visual estimates recorded during geologic logging, assumptions about combinations of the elements measured during chemical assaying, or visual measurements of thin sections of core using light microscopy, however each of these methods is subjective. Direct measurement of mineral types is possible using a technique called Qualitative Evaluation of Materials by Scanning Electron Microscopy (QEM*SEM). QEM*SEM utilizes a computer controlled scanning electron microscope which can distinguish minerals and their attributes in individual ore particles. These images can also be processed off-line to provide information on modal abundance, grain size, mode of occurrence, liberation characteristics and how much of a particular mineral phase may be recoverable. (MINTEK 2001) Physical Characteristics Direct measurement of physical parameters of the materials sampled, such as in situ density, porosity, permeability, compressive strength, compaction, and grinding index, is also performed. In the case of some commodities it is the physical characteristics that determine value, e.g. kaolin where the value is determined by brightness and slurry viscosity. In some instances these tests require whole core, meaning that the entire sample is destroyed in the testing process. In such cases, it is usual to select samples for these tests at large intervals along the drill holes. Metallurgical Testing. These tests are performed on potential ore materials in order to determine the expected recovery of the desired commodity from the ore for given processes and to determine the characteristics of the tails produced by the process. COMMUNICATION/GEOMETALLURGY Most of the sampling and selection of samples for analysis is generally performed by geologists during early phases of a project. In some cases, even the samples for metallurgical testing are selected by the geologists who collect them. This can lead to test results that are of little value to the plant design engineer and, in the worst case, to incorrect design. One anecdote from the author's experience will illustrate. During development of a coal mine it was determined that the project value could be increased if additional thin seams of coal could be added to the reserve. In order to mine these seams it was determined that it would be necessary to take some dilution of sandstone and shale from above and below the seams but it was also determined that this diluting material could be removed in a coal wash plant. The plant design engineers requested a sample upon which to base their design of the coal wash plant and the geologist was given the task of collecting the sample from the drill core. As the geologist had no experience with coal washing and the scheme of

28

removing the diluting material had not been explained, the geologist selected the coal sample that was requested by taking coal only from the interior of the thin seams with no diluting material added. This material was delivered to the plant design engineers who tested the material, designed a coal washing plant based on the results, and actually built the plant. Needless to say, the plant did not operate correctly and the project did not attain the desired throughput. The preceding example illustrates the need for communication between the various disciplines that work on a project during the earliest phases of the project, especially during the feasibility study phase. This communication needs to be more than a casual acquaintance and for this purpose a methodology called geometallurgy has been formalized. The following describes the steps for implementing geometallurgy within a project. It does not matter which discipline initiates the geometallurgical discussions, however due to the potential negative impact of not conducting this process the plant design engineer is encouraged to learn this process and to take the lead in its implementation. Geometallurgy Process. Geometallurgy is based on mutual respect for other disciplines and a willingness to learn from others and is easy to implement. The first step is to bring the various participants from the geological, mining, metallurgy, environmental, geotechnical, and plant design disciplines together in an initial meeting for a general discussion on the characteristics of the ore body and associated sites. In facilitating such meetings in the past, it has been noted that due to the use of different jargon by the different disciplines (and sometimes similar terms for different things) participants in these meetings may believe they are communicating when they are not. For this reason it is suggested that a facilitator conduct the meeting who is not directly involved in the project. At the initial meeting each discipline should make a general presentation concerning their view of the deposit and the parameters that are important. In these presentations the geologist will talk about the various rock types, mineral types, alteration zones, fracture zones, etc. The mining engineer will talk about access to the deposit, the type of equipment to be used, the drilling and blasting methods, etc. The metallurgist will talk about the importance of work index, size distribution, liberation size, etc. The environmental engineer will talk about limitations on access to specific areas and quality limits for discharge water, etc. And the plant design engineer will talk about the types of equipment that will be used in the process, the foundation requirements for the equipment and buildings, the handling of water recycling from the tailings impoundment, etc. Everyone will learn something new from this meeting. Having established a basis for discussion, it is now important to break down the terms into their simplest forms so that each discipline can understand the goals of the others. In describing work index it is important to ask about the relative hardness of the materials and allow the geologist to discuss and describe the various rock types. In discussing feed size distribution it is important to listen to the description by the mining engineer of how changes in the blasting techniques can make a more acceptable feed. Discussion of the liberation size may lead to understanding of the importance of fracture intensity. And discourse on the need for aggregate during construction may lead to identification of materials close to the project site that have the requisite material characteristics. At this point it is important to formalize the understanding of terms in writing so that all participants have a reference for hture discussions. The goal of these discussions is to identify all the possible classifications of ore, waste, foundation, and building materials in terms of spatial location, geologic interpretation, economic value, and order of exploitation. These classifications can then be evaluated from a plant design standpoint to determine which are potentially important in terms of grinding, separation, recovery, and tailings disposal and a testing program developed that will determine which of these classifications are truly important. The plant design engineer may also discover how to obtain necessary information using sampling programs that have already been planned by asking for additional analysis of those samples. The initial meeting should be followed by additional regular meetings where new or updated information and interpretations of data are shared with all participants in the project. This methodology can save considerable time because it focuses the attention of the entire team on

29

parameters that are essential to the process design and time is not be wasted developing detailed information on unimportant items. Geometallurgical Role of the Plant Design Engineer. Since it is usual for the plant design engineer to enter the feasibility study phase of a project relatively late, it is incumbent upon this discipline to inject itself into the feasibility process as soon as possible and to take a leading role in geometallurgy. No other discipline understands as well the implications of improper or incomplete data on the project design and implementation. The age-old qualification that “the design is guaranteed given that the feed to the plant is similar to the sample tested” aids neither the project owner nor the plant designer. The plant design must be able to respond to variations in the ore, not just the average grade or other characteristic of the ore deposit. In order to understand that variation it is necessary to communicate with those whose responsibility it is to sample the ore body and associated materials. It is also important for the plant design engineer to recognize that the design will be affected not just by the ore characteristics but by the quality of the water, the characteristics of the tailings, the susceptibility of the mine waste to leaching, the quality of the soils in the area of plant construction, and other seemingly minor attributes of the site. In this modem age of mining projects, it is often incomplete knowledge of the minor attributes that can delay or prohibit a project: and collection of this data is often simply a matter of performing additional analyses on samples that have already been collected or are planned to be collected. For instance, during the 1990’s three large copper projects were affected by minor elements that occur in the concentrates. In one instance this was arsenic, in another it was bismuth and in the third it was fluorine. Large expenditures were made in preparation for each of these projects and recognition of the problem was not made until relatively late in the feasibility study process. The geologists responsible for sampling the ore bodies were intent upon characterizing the amounts of copper and other revenue generating metals and had not arranged for the additional cost of the analyses for these minor elements. Had the plant design engineers involved themselves in the sampling for the feasibility studies from a geometallurgical approach early and explained to the geologists the elements which could be important to the beneficiation process and subsequent downstream processing, the problems could have been identified and the work toward finding a solution could have been initiated before considerable hnds had been expended. Incomplete knowledge of geology can have similar consequences. During an evaluation of a gold deposit in Indonesia, the consulting metallurgists who performed the testwork and prepared the flowsheet for the project were asked whether or not they were satisfied with the samples that they had used for these purposes. They replied that they had asked for separate holes to be drilled for samples for metallurgical testwork but had been told that all of the drills had to be kept working defining the ore reserve and that they would have to obtain the sample material they wanted from the rejects that remained from the core sampling program. As was shown later, the integrity of the samples was compromised before the samples were delivered to the metallurgists and as a consequence the metallurgists duly reported that their testwork had shown that more than 90 percent of the gold in this primary, hard rock deposit could be recovered in a gravity concentrate that represented about one percent of the feed material. (Farquharson et al. 1997) This unlikely outcome might have been questioned by the metallurgists had they consulted more closely with geologists familiar with the type of deposit and the manner in which gold usually occurs in such a deposit. The problems of a lack of analysis data that could easily have been obtained during the sampling process are not limited to ore materials. As environmental regulations have become more strict it has become important to be able to characterize the qualities of mine wastes as well. It is common practice in the mining industry for geologists not to have analyzed samples that are obviously waste, in order to save money. However, when it comes to characterizing the acid generating capacity of the mine waste, this lack of data can put the mining company at a disadvantage. In one important project at the end of the 1990’s this lack of data forced the redesign of the mine waste dumps, which resulted in relocation of the plant facilities. In the 2 1St century it has become necessary for all engineering disciplines to become crosstrained. It is no longer acceptable to work only in one’s area of responsibility without

30

understanding the data and processes of the entire project and indeed the entire chain of extraction, processing, use, disposal, and recycling of the commodities that are involved in the project covered by the feasibility study. In a world headed toward sustainable development all disciplines involved in the project feasibility, but especially the plant design engineer, need to be cognizant not only of the aspects of the main commodity being produced as they apply to ore process characteristics but also about the impact of associations of the components and impurities in the commodity which may affect salability and use of the product in the future. For instance, in iron ore mining there are many sources in the world which have very high iron grades so it is not the content of iron which controls the value but the amounts of the impurities, such as silica, alumina and phosphorous. Without some understanding of steel making, it would be impossible to determine exactly which impurities in the samples should be analyzed so that an optimum exploitation strategy could be followed. It is the process and plant design engineers who must bring this knowledge to the feasibility study. AMOUNTS OF SAMPLING The amount of sampling required to support a feasibility study varies with the commodity, the deposit, and the purpose of the samples. Commodities whose characteristics are consistent and are well known from other mines in the area, such as coal in the Powder River Basin, require less frequent sampling than commodities whose presence is more variable such as precious metals. While there are standards requiring that some sampling be performed in order that the presence of an ore body may be claimed in reports of publicly owned companies, the amount of sampling is generally left to the mining company to determine as part of their internal assessment of the risks associated with development of the project and the risks that they perceive will be acceptable to potential investors and lending institutions. Mining companies rely on the advice of their internal and external consultants, including plant design engineers, in determining the amount of sampling required for a particular deposit. While standards vary from company to company, the following outlines a general methodology for determining how much sampling is required for each purpose. Mineral Resources and Ore Reserves. Most of the sampling performed in support of feasibility studies is for the purpose of defining the presence, extent, grade, configuration and continuity of the ore bearing zones of a deposit and is performed under the guidance and responsibility of an exploration geologist. Areas of sampling are categorized by the amount of sampling performed in each area. Most, if not all, mining companies have come to use the categories promulgated by the Australasian Joint Ore Reserves Committee as part of what is known as the JORC Code (See Figure 1). (A complete copy of the JORC Code can be obtained at hm,://www.iorc.org.) In these categories mineral resources are categorized with Inferred Resources having little or no sampling, Indicated Resources having sufficient sampling to reasonably assume continuity of geology and grade, and Measured Resources having sufficient sampling to confirm continuity of geology and grade. Similar codes, standards and terminology have been or are being adopted in Canada, South Africa, the United Kingdom, and the United States of America. (Pincock, Allen h Holt 2000) As shown in Figure 1 , the resources that are defined as Measured may be used to define Proved and Probable Ore Reserves and the resources that are defined as Indicated may be used to define Probable Ore Reserves once the appropriate mining, metallurgical, economic, marketing, legal, environmental, social and governmental factors are taken into account, however Inferred Resources may not be used to define any type of ore reserves at all. During the past ten years, the old category of Possible Ore Reserves has been eliminated entirely.

31

II

1

Increasing level of geological knowledge and confidence

Exploration Results Mineral Resources

Ore Reserves

Inferred r"""""""""'"""

,

Indicated

1

4

Probable

I

/ -

1 1 I

4 # -

Measured

f .

-

8

-. .Proved

I I

I I I I

Considerationof mining, metallurgical, economic, marketing, legal, environmental, social, and governmental factors (the "modifying factors")

Figure 1. General Relationship between Exploration Results, Mineral Resources and Ore Reserves (Reprinted with permission from The 1999 JORC Code, Australasian Joint Ore Reserves Committee, Victoria, Australia).

Although the concept of categorizing resources and reserves by the geologic continuity and the amount of sampling has been generally accepted, the amount of sampling required for each of these categories has been left to the discretion of the individual mining companies and ore reserves professionals. Many, but not all, of those with this responsibility have chosen to use some sort of geostatistical approach to determining the sample spacing required for each category. Geostatistics is used in the mining industry for calculation of mineral resources based on available sampling and it also results in a calculation of the accuracy of the estimates. This accuracy calculation takes into account the natural variation of the commodity characteristic being studied and the volume of the area for which estimates of grade and tonnage are being made. By specifying a required accuracy, this method can be used in reverse to determine the required spacing of samples. Due to a number of variables in this type of calculation it is not possible to perform this calculation on an absolute basis, however it does give relative results that, with judgment, can be used to define the sample spacing for a given project. The required accuracy must be specified by the mining company. The individuality of each deposit having been noted, it is possible to give approximate ranges for different commodities. For instance, the spacing between drill holes for the purpose of feasibility study for sedimentary deposits such as coal and potash are usually in the range of 1000 meters whereas the spacing of drill holes in disseminated precious metals deposits is generally in the range of 50 meters. The spacing for base metals is generally in the range of 100 meters. In making the determination of sample spacing for ore reserves, the geostatistical approach must be tempered by geological knowledge of the nature of the ore deposit since the calculations may not adequately account for discontinuities in the physical shape of the deposit, such as fault blocks and washouts. In these cases additional sampling may be required to define the presence of these occurrences. In most cases, the least expensive sample spacing configuration will be on a regular square pattern. However this may not be the case when access to drilling locations is expensive such as in underground workings or in rough terrain or jungle conditions. In such cases it is usual to drill several holes in a fan from the same location. This increases the amount of redundant data near the collar of the hole but decreases site preparation costs and program duration.

32

The determination of the interval of sampling down-the-hole is also a trade-off between geology, accuracy, and budget. In sampling sedimentary or vein deposits, samples are generally divided at each change in rock type or, when the layers sampled are thick, at a set length of between one and three meters. In sampling disseminated or massive deposits, a set length is used which may be as short as one meter and as long as the nominal bench height planned for the mine. When using core drilling this determination can be made after the fact and different intervals can be used for different purposes. When using a method that creates chip samples, the decision must be made beforehand since the samples cannot be divided into shorter intervals once they have been taken. Whatever method is used for determining the sampling interval, it is important that a consistent sampling or compositing interval be used as this affects the geostatistical calculation of the ore reserve. The sampling program developed for ore reserve calculation generally defines the sample locations that are available for the other purposes described below which deal with the ore body. It is important for the plant design engineer to be involved in the development of the sampling plan early in the feasibility process to ensure that access to specific areas of the deposit are available. In Situ Density. While great care is usually given to determining the grades of the revenue commodity and the volumes of the ore zones so as to make an accurate ore reserves statement, many times the in situ bulk density is not adequately analyzed. This can lead to greater misstatement of ore reserve quantity than errors in grade. The amount of sampling required for in situ density depends on the consistency of the deposit over distance and the relationship between ore grade and density. Where the commodity grade of the materials does not affect the in situ density, it is usually adequate only to take samples from each rock type, which can be applied to the lithologic model of the deposit. For statistical reasons it is useful to have at least ten widely spaced samples from each rock type for this purpose. This is usually the case in low-grade, disseminated deposits. However, where the commodity grade of the material affects the in situ density, such as in massive sulfide deposits, heavily altered deposits, and especially oil shale, it is necessary to measure the in situ bulk density for each sample analyzed for ore reserve calculation. In these cases the in situ density may need to be treated in the same manner as the commodity grade variables in the ore reserve calculation. Occasionally the in situ density cannot be characterized adequately from drill hole materials. An example of this is the in situ bulk density of weathered materials such as bauxite. In such cases a separate grid of special samples must be taken which may involve trench andor channel samples or even in place measurements. In sampling for in situ density, values for waste materials are also important and special drill holes or additional sampling may be required. This is especially true in open pit mines where the stripping ratio can determine the viability of the project. In all cases the in situ density should be calculated on the same basis that the grades are reported, i.e., wet or dry basis. For most metal deposits the assay grades are reported on a dry basis and the in situ density should also be reported as such. Mining Characteristics. The mine design developed during the feasibility study will use mining characteristics that also need to be sampled, including the strength of the rock and the orientation and frequency of fractures. The first source for this information is the drill core, however these cores are usually concentrated on the ore zones. In order to determine the rock mechanics characteristics of the waste materials, especially in the case of open pit mining slopes, special sampling may be required. In addition this may require use of “oriented” core drilling that can show not only the presence of fractures but also their absolute orientation within the drill hole. The amount of sampling is determined by the changes in rock types and orientation of the fractures in the pit area. Such sampling should be performed at least for each rock type that will be encountered in the pit and for each of the principal directions along and perpendicular to the major fracture zone(s) in the area. Since these samples may represent the only cores taken from some waste rock materials, this program should be coordinated with the environmental engineers so that the materials can be tested for important waste characteristics such as acid generating capacity.

33

Metallurgical Testwork The main parameters of the process plant equipment, such as grinding mill size, number of flotation stages, leach residence time, screen opening sizes, and tailings storage requirements will be finalized based on samples specially collected for metallurgical testing, as described in the next chapter. However, in order to investigate process alternatives and variability of ore characteristics early in the feasibility phase of a project the plant design engineer generally has only the samples taken during the geologic drilling program. Since metallurgical tests often use complete core or RC samples and the tests themselves can be expensive, there is sometimes reluctance to provide large amounts of the collected materials for this purpose. This is the point where the geometallurgical approach is most important. When the geologist understands the goals of the metallurgical testwork they can work together with the plant design engineer to develop a sampling program which covers the widest range of deposit characteristics while using a minimum of the collected sample material. The sampling program must provide sufficient materials covering a range of head grades, hardness, and other essential parameters from the different ore types andor rock types in all parts of the deposit in order to quantify the variation in these characteristicswithin each class of material. Only in this way can recovery, ore hardness, and other parameters be modeled to predict the behavior of the ore. To be statistically significant it is desirable to obtain and test several samples from each ore type. Although in some instances it may appear less expensive to make a “representative average sample” for metallurgical testwork which includes a mix of materials in the proportion in which they are encountered in the deposit or in which they might be encountered during a particular period of the mine life, this is discouraged because the mix of material may never actually materialize. A true anecdote will illustrate. “The primary grinding mills for an iron ore operation in northern Minnesota, now closed, were selected on the basis of samples which were composites intended to represent the “average” ore. Soon after startup the mills showed that they could not grind the required tonnage and the mill designers were called in to determine the reason. Finding no flaws in the design or installation, attention shifted to the original basis for design, which was, of course, the test sample. Working backwards to determine how the “composite” sample was assembled, it was found that the sample had been mixed from five separate ore types based on the proportion of each ore in the overall deposit. It turned out that due to the placement of materials in the deposit, the sample represented a mix of ore that would never occur. Further grinding testwork was authorized and it was discovered that not all the ore types needed to be present at all times. By separate testing of the individual ore types it was shown first mathematically, and later in practice, that the desired throughput could be achieved if one of the ore types was always present as at least twenty percent of the feed. Although finally a good outcome, the failure of the engineers to communicate in the early stages of the project resulted in significant cost to the owner both from reduced throughput during the initial operation and in the need to alter the mining sequence to meet the needs of the mill.” (Ashley and Callow 2000) As this example illustrates, metallurgical testwork samples should be taken and tested separately with compositing performed mathematically, if required. One way to examine variability across a deposit at minimal cost is to demonstrate an association between an expensive test which uses much material and an inexpensive test which can be performed on small test samples and then use the expensive test results as a basis for the absolute design parameters and the inexpensive test results as a manner for evaluating the variability of the ore characteristic being determined. For instance, in one gold leach project, final gold recoveries were determined using column tests over many days on a small number of bulk samples that represented variations in ore grade and rock type. A representative portion of this material was used for a bottle roll test in the laboratory using a standardized procedure. The results of the bottle roll tests and those from the column tests matched well. Over one hundred samples spaced on a grid across all parts of the deposit were then tested using the bottle roll test. These samples were taken from the RC samples collected during the drilling program. The results were plotted on the grid and showed that while there was variation in recovery, it was not related to position in the deposit horizontally or

34

vertically or to rock type. Therefore this supported the application of the average values obtained from the column tests to the entire deposit even though those samples had not been collected from all parts of the deposit. Another example is the use of the SAG Power Index (SPI) test provided through MinnovEX Technologies of Toronto, Ontario, Canada. Sizing of SAG mills has historically required tests that use large quantities of material with pieces of rock 150 mm and larger that can only be obtained by tunneling or use of large size core drilling. However the SPI test can be performed using core of the size normally used in geologic exploration. This allows for determination of the variability of SAG grinding characteristics across the spatial extent, ore types and rock types of a deposit at a much lower cost. By correlating the results of the SPI tests with those from larger scale tests, which are still considered essential for optimal design, it is possible to develop a plant design which can handle the variation in feed material hardness without costly overdesign and with confidence that the plant is not underdesigned. (Custer et al. 2001) Foundation Design. While the emphasis during the exploration drilling program is on the area of the ore deposit, it is well to remember that similar types of drilling will be required for foundation design. If possible, the plant design engineer should arrange with the exploration geologist and/or drilling contractor to take preliminary samples from potential plant and tailings disposal sites. Such samples may already be budgeted by the geologists as part of a condemnation drilling program however the handling and testing of the samples may not be included. Final plant design will include more detailed foundation testing; however early sampling using available resources might identify potential problems that could otherwise delay the project schedule. Water Treatment. Water treatment is usually a minor component of plant design, however it is the point in the process that brings all of the various areas together. Water from the mine, the process facilities, the waste dumps, the tailings disposal facility, and the workshops and infrastructure areas all must be evaluated for treatment. In many recent projects the restrictions on quality of discharge water have required considerable effort and alteration of plant design. During the feasibility study phase of the project the quantity and quality of water coming from these various sources is developed by the geologists, environmentalist scientists, hydrologists, meteorologists, and metallurgical testing engineers. It is the responsibility of the plant design engineer to gather all of this data and determine the overall site water balance for quantity and quality in order to form the basis for determining which streams must be treated before discharge. The quantity of required treatment may result in changes in the plant design and these should be incorporated as early in the project as possible. The plant design engineer should evaluate the sampling programs planned by each of the disciplines and augment these programs where necessary to support the information required for the overall site water balance. In particular, determining the quality of the water associated with the tailings should be part of the metallurgical testwork as the levels of minor elements may be the determining factor on process design. PRESENTATION OF RESULTS The sampling programs undertaken in support of feasibility studies develop tens of thousands of data records. Geologists are responsible for reducing the information on mineral resources to an understandable form and the mining engineers are responsible for presenting the basis of the ore reserves. The plant design engineer has the responsibility of reducing the large number of sample data collected to support the plant design parameters into a form which can be easily understood yet retains sufficient detail for qualified analysis. In arranging the information on plant design parameters it is recommended to view the audience as being of three levels. One level represented by bankers and investors will want to know the results of the analysis and the amount of sampling upon which the conclusions are based. Another level represented by engineering managers and other disciplines will want to know methods used for testing the samples and the distribution of the results by ore type and location. And the third level represented by specialists in plant design will want to view the original sample and test run data.

35

For the feasibility study report, therefore, it is recommended that three sections be written. The most complete which contains lists of the tests performed and the original sample data with full explanation should be placed in the appendix of the feasibility study report. A summary of the methods and results of the sampling and testwork, with reference to ?he appendix, should be written as the section of the feasibility study on metallurgy/process and should include final graphs such as gradehecovery relationships. And a synopsis of this section containing the pertinent summary tables, with reference to the metallurgy/process section and the appendix, should be placed in the executive summary of the feasibility study report. All summary tables should include the number of samples upon which the values are based, the average result, and the standard deviation of the results (if the number of values is large enough to make the standard deviation meaningful). Maps and cross sections of the ore deposit showing the testwork sample locations are also essential in order to communicate the coverage of the testwork. REFERENCES Ashley, K.J. and Callow, M.I., 2000. Ore Variability: Exercises in Geometallurgy, Engineering & Mining Journal, 20 1:2,24-28 Australasian Joint Ore Reserves Committee, 1999, JORC Code and Guidelines Custer, S., Garretson, P., McMullen, J., and Bennett, C., 2001, Application of CEET at Barrick’s Goldstrike operation, CMP 2001 Farquharson, G., Thalenhorst, H., and Von Guttenberg, R., 1997, Busang Project - Technical Auditfor Bre-X Minerals Ltd., Interim Report. Strathcona Mining Services Limited. MINTEK, cited 2002, Microbeam Applications, http://www.mintek.co.zd Pincock, Allen & Holt, 2000, Resource or Reserve - the Diference, Pincock Perspectives, Issue No. 2 - January 2000. US Army Corps of Engineers, 200 1, Engineering and Design Geotechnical Investigations, Manual NO.EM 11 10-1-1804.

36

Sampling in Mineral Processing J W Merh'

ABSTRACT The objective of sampling in mineral processing is to estimate grades and contents of sampling units in an unbiased manner and with an acceptable and affordable degree of precision. Sampling units are classified as dynamic and static stochastic systems. The paper examines the most relevant topics of sampling practice and applied statistics such as how to test for bias, how to estimate precision, how to quantify associative dependence between measured values in ordered sets, how to select suitable sampling procedures and how to optimize sampling protocols. Examples are given to illustrate the application of sampling theory in practice. Extensive references to publications describing sampling procedures and guidelines for mineral processing applications are provided.

INTRODUCTION Sampling theory and practice play an important role in mining and metallurgy. The sampling of materials in bulk is well-documented in the literature (Gy 1979; Merks 1985; Visman 1962) but concise definitions, uniform symbols and common rules remain elusive targets. Various Technical Committees (TCs) of the International Organizationfor Standardization (ISO) have developed guidelines on the sampling of coal (TC27), iron ore (TClOZ), and copper, lead and zinc concentrates (TC183). Detailed information can be found in several I S 0 Draft International Standards for copper, lead and zinc concentrates (see References, ISO/DIS), and in standard methods developed by ISO/TC69-Applications of Statistical Methoak. Generally, sampling is the process of selecting a part of a whole such that a measured value for the part is an unbiased estimate for the whole. In mineral processing, a whole is referred to as a sampling unit such as a mass of mill feed, dewatered concentrate or bullion, or a volume of cyclone overflow or tailings slurry. A sampling unit is classified as a dynamic stochastic system when sampled during transfer, and as a static stochastic system when sampled while stationary. The wet mass of mill feed can be estimated in an unbiased manner and with an acceptable degree of precision (Merks and Merks 1992) but SAG mills, and gravity and flash concentrates, have made it difficult to obtain precision estimates for metal grades and contents of mill feed. The variances of metals contained in tailings, concentrates and thickener inventories can be used to obtain reliable precision estimates for monthly mill feed grades. It is beyond the scope of this paper to explain how Monte Carlo simulations can be applied to estimate confidence limits for the metal grade of mill feed on the basis of its wet mass, and of the metal contents and variances of tailings and concentrates (Merks 1991; 1999). The validity of this method depends critically on how slurry flows in mineral processing plants are interrogated, and how the variances of stochastic variables are estimated. On-stream data give valuable statistics that can be plotted in sampling variograms to show where orderliness in the sample space of time dissipates into randomness (Merks 1999). Applied statistics provides scores of powerful techniques to test for bias, to estimate precision, to optimize sampling protocols, and to determine the degree of associative dependencebetween measured values in ordered sets (ASTM 1985; Mandel 1964; Merks 2000; Volk 1980). The statistical analyses applied in this paper are based on comparing F- and tstatistics with values tabulated as a function of degrees of freedom in the F-distributions at 5% and 1R probability, and in the t-distribution (Handbook 1968; Volk 1980). 1

Matrix Consultants Limited, Vancouver, British Columbia

37

A sampling protocol that is based on dividing a set of primary increments into odd- and even-numbered subsets (see Appendix A), and preparing a test sample of each primary sample (see Appendk B), gives an unbiased estimate for the variance of the entire measurement chain. A pair of interleaving (or interpenetrating) primary samples is referred to as A- and B-samples (ISO/DIS 12743). The symbols var(t) and var(spa) refer to the total variance of a measurement procedure (the sum of the variances of the primary sample selection, preparation and analytical stages). Interleaving sampling protocols are equally effective when applied to slurry flows in mineral processing and bulk samples in mineral exploration. Uncertainties in a measurement chain can be partitioned into randomly distributed variations (random variations for simplicity) and biases. The sum of all random variations is statistically identical to zero, and the sum of all biases is statistically different from zero. The variance is the fundamental measure for random variations. Analysis of variance (ANOVA)can be applied to optimize sampling protocols by partitioning the sum of the variance of the primary sample selection stage, the variance of the sample preparation stage and the variance of the analytical stage into its components. This application of Fisher’s F-test, which is the essence of ANOVA, is examined in a separate section.

DEFINITIONS Through the years, probability theory and applied statistics have developed a distinct jargon. Since statistical tests and techniques are applied in all scientific and engineering disciplines, it is unsurprising that vastly different definitions and symbols abound. Elementary concepts such as trueness, accuraq, bias and precision received a great deal of attention and scrutiny from ISO/TC69 -Applications of Statistical Methodr, and from technical committees that deal with the sampling of concentrates, coals and various types of ores. In time, ambiguous terms such as measurement error, margin of error, sampling error, sill value, semi-variogram, nugget eflect, and scores of others, will be replaced with concise definitions, and be assigned the proper statistical symbols.

Accuracy A generic term that implies closeness of agreement between a single measured value or the central value of a set (the arithmetic mean or a weighted average), and the unknown true value of the stochastic variable.

This definition reflects that accuracy is an abstract concept. By contrast, a lack of accuracy can be measured and quantified in terms of a bias or systematic error. Webster defines accuracy as freefrom error. Thus, unbiased measurements are accurate by definition. The term unbiased implies that a properly designed bias test was applied, and that a single measured value or the central value of a set is indeed an unbiased estimate for the unknown true value of the stochastic variable in the sampling unit or sample space under examination.

Bias A statistically signijkant direrence between a single measured value or the central value of a set, and an unbiased estimate of the unknown true value of the stochastic variable.

Testing for the absence or presence of bias is an essential part of sampling in mineral processing. Terms such as random error, or error without adjuncts or adjectives, will not be used to avoid confusion with randomly distributed variations for which the variance is the fundamental and unambiguous measure. Testing for relative bias and estimating analytical precision are key elements of statistical quality control (SQC) and statistical grade control (SGC). Testing for analytical bias demands the use of CertiJed Reference Materials (CRMs). The presence of bias at the analytical stage is easy to detect but sometimes difficult to eliminate at affordable cost. Some sources of bias at the primary sample selection stage and the sample preparation stage are intrinsic to the applied procedure, which makes a bias difficult to detect and impossible to eliminate.

38

Student’s t-test, the bias test par excellence, is described in a separate section. The t-test can be applied to paired test results obtained by employing different analytical methods to replicate test portions taken from each of a set of test samples. The test can also be applied to paired test results obtained by employing different sampling procedures to the same sampling unit, or different sample preparation procedures to the same primary or secondary sample. The presence of analytical bias suggests that at least one of the procedures is suspect, and the absence of analytical bias implies that both procedures are most probably unbiased (ASTM 1985; Davies and Goldsmith 1972; Mandel 1964; Merks 1985; Volk 1980). The t-test should be routinely applied to assays determined at mineral processing plants and control assays reported by commercial laboratories, and to exchange assays between mines and smelters (Merks 1989). One-way ANOVA is applied to test results determined in the same laboratory by employing the same analytical method to replicate test portions taken from each of five up to ten test samples prepared of the same sample mass under carefully controlled conditions. Tests for homogeneity should precede crosscheck programs to ensure that each participating laboratory receives a subset of test samples selected from a homogeneous set. Two-way ANOVA is employed to test for analytical bias when three or more laboratories participate in interlaboratory crosscheck programs by applying the same analytical technique to duplicate test portions taken from each of a set of no fewer than five test samples. Logically, the complete set should pass the test for homogeneity before subsets are submitted to the participating laboratories.

Precision A generic term that refers to the magnitude of randomly distributed variations (random variations) in the measurement procedure applied to estimate the central value of the stochastic variable of interest. Precision, too, is an abstract concept. For example, the precision is low or poor, or the degree of precision is high or excellent, are valid but ambiguous, non-quantitative and vacuous just the same. Quantitative measures for precision such as confidence intervals in absolute values or relative percentages, and symmetric and asymmetric confidence ranges in absolute values, derive from the variance of the central value for the stochastic variable of interest. Sample Apart of a sampling unit or a sample space selected such that a measured value for the part is an unbiased estimate for the sampling unit or the sample space. A sample is often referred to as a representative part of a population or a whole but the concept of representativeness is widely abused and misused in sampling practice (Huff 1954). In reality, the measured value for a sample is an unbiased estimate for the sampling unit if, and only if, each stage of the applied measurement procedure is unbiased. Interleaving Test Samples A pair of test samples obtained by dividing a set of primary increments into odd- and evennumbered subsets (A- and B-primary samples), and preparing a test sample of each primary sample (A- and B-test samples). Selecting a pair of interleaving primary samples (see Appendices A & B) is the most effective procedure to obtain an unbiased estimate for the variance of an entire measurement chain. Since one pair gives only a single degree of freedom (see Degrees of Freedom), the estimate for var(t), the total variance of the measurement procedure, is extremely imprecise (see Table 3). By contrast, 28-31 pairs of interleaving samples give 27-30 degrees of freedom SO that the monthly metallurgical balance is significantly more precise than single daily metallurgical balances.

39

SAMPLING PROTOCOLS Generally, sampling protocols can be divided into random sampling, stratified random sampling and stratified systematic sampling. Random sampling and stratified random sampling are routinely applied to consumer products. Stratified systematic sampling is most effective for all types of materials in bulk and for slurry flows in mineral processing plants.

Stratified Systematic Sampling This sampling protocol is commonly applied to bulk solids such as concentrates, coals and ores, preferably during transfer with a conveying system (dynamic stochastic system) but often while in storage (static stochastic system). Stratified systematic sampling is based on dividing the sampling unit into a set of elementary units (dynamic strata or static cells), and selecting a primary increment from each elementary unit. For example, dewatered or dried concentrate is divided into dynamic strata during transfer with a conveying system by selecting a set of primary increments at intervals of constant mass or time. Dewatered or dried concentrates should not be sampled mechanically during transfer because significant moisture losses will inevitably occur. Concentrate in trucks or wagons is divided into static cells, and a primary increment is selected from the center of each cell with a properly designed probe. Mechanical probe sampling systems make it simple to select pairs of interleaving samples and implement effective risk analysis and loss control at mines and smelters. Interleaving sampling protocols give unbiased precision estimates at the lowest possible cost. It does so at no additional cost if the wet mass of a sampling unit (a lot) is increased by the factor 2. Whenever the set of primary increments is combined into a single primary sample, the variance of the primary sample selection stage and the variance of the sample preparation stage cannot be estimated. Only the analytical variance can be measured and monitored by assaying replicate test portions of a test sample prepared of the primary sample. On-stream analyzers interrogate slurry flows either continuously or intermittently. The large set of on-stream data generated during a shift gives a high degree of precision for the arithmetic mean. However, this central value is an unbiased estimate for the central value only if the analyzer is in a proper state of calibration. Moreover, slurry densities and metal grades in flotation circuits may exhibit associative dependence so that the density weighted average is a more reliable estimate for the metal grade of a slurry than the arithmetic mean. A sampling module, designed to take a pair of interleaving secondary samples from a primary sample flow at constant time intervals during each shift, would make it possible to implement meaningful metallurgical accounting procedures. The degree of associative dependence between on-stream data in the sample space of time (sputiul dependence for simplicity) impacts the variances of ordered sets, and, thus, the precision of central values. Slurry flows are usually interrogated at constant time intervals which simplifies the calculation of variances and central values (see Appendix D). MEASURES OF CENTRAL TENDENCY Measured values tend to cluster around a central value which is often referred to as the central tendency of a set. The arithmetic mean is an unbiased estimate for the central value of a set of measured values with equal weighting factors whereas the weighted average is a more reliable (less bias prone) estimate for a set of measured values with variable weighting factors. Count, density, distance, length, mass and volume weighted averages are important measures of central tendency in mining and metallurgy. Central values are measured with finite precision because each is merely an estimate (an unbiased estimate one would hope) for that most elusive central value, the unknown true value of the stochastic variable of interest in the sampling unit or sample space under examination. The arithmetic mean is the central value of a set of measured values with equal weighting factors. The equation for the arithmetic mean is elementary, and has its own function in spreadsheet software. The weighted average is the central value of a set of measured values with variable weighting factors:

40

where

x =ithweightedaverage measured value

xi =

wli =first weightingfactor for ith measured value Given that wli =mi/Cmi for the mass weighted average grade of =O. 0386*32.1+. ..+ 0.2684*29.8=30.7196 (see Table 2 and Appendix C), and C (l/n)=l for the arithmetic mean grade of 30.90% (see Appendix C), it follows that CwIi is also unity. The second weighting factor of w2i= mi/Cmi=mi/fi is convenient in spreadsheet templates to obtain the variance of a set of measured values (see Measures of Variability), and the variance of its central value (see Variances of Central Values). Weighted averages play a key role in a wide range of applications. For example, the length and density weighted average is the central value of a set of measured values for core samples of variable length and density. Similarly, the distance weighted average is the central value of a set of measured values with variable coordinates in a two- or three-dimensional sample space. In gwstatistics, the distance weighted average transmogrified into the ubiquitous kriged estimate. Table 1 gives, in addition to the set of paired data (dry masses in tonnes and metal grades in percent), W I i , the first weighting factor, which is required to calculate this mass weighted , second weighting factor, which simplifies the equations average grade of 30.71%, and ~ 2 i the for the variance of the set, and for the variance of its mass weighted average grade. Table 1 Weighting factors Unit

1 2 3 4 5

massinmt

12.1 54.5 72.6 90.3 84.2

gradein %

32.1 30.3 30.5 31.8 29.8

WIi

0.0386 0.1737 0.2314 0.2879 0.2684

W2i

0.1929 0.8687 1.1572 1.4393 1.3420

The sum of the first set of weighting factors and the arithmetic mean of the second set are both unity which implies that Cwli=ii2i=l. This relationship makes it simple to check the correctness of both weighting factors in spreadsheet templates. The weighting factors in Table 1 are used in several calculations (see Appendix C).

MEASURES OF VARIABILITY The variance is the fundamental measure of variability. Variances are amenable to mathematical analysis. All measures of variability and precision derive from the variance. For example, the standard deviation is the square root of the variance, and the coefficient of variation is the standard deviation as a relative percentage. The properties of variances are the essence of probability theory and applied statistics. The additive property of the variances of volume, mass and content has scores of powerful applications in mining and metallurgy. Since the occurrence of spatial dependence between measured values in ordered sets is of critical importance in sampling theory and practice, this matter will be examined in a separate section (see Testing for Spatial Dependence). Almost invariably, measured values are ordered either in time (on-stream and production data) or in space (rounds in a drift or a trench; core samples in a borehole; boreholes in a section). The question is then whether a set of measured values displays a statistically significant degree of spatial dependence, or is randomly distributed within its sample space. For example, the set of metal grades in Table 1 is ordered in time but it does not exhibit a significant degree of spatial dependence. Therefore, the fundamental measure of variability is the variance of the

41

randomized set. Thus, the term "randomized" implies that these metal grades constitute a randomly distributed set within this sample space of time. By contrast, the set of on-stream data given in Appendix D exhibits a highly significant degree of spatial dependence. As a result, the variance of the ordered set gives a significantly higher degree of precision for its central value than the variance of the randomized set. The example in Table 1 is more complicated in the sense that each metal grade represents a different mass. In this case, the mass weighted average grade is a measure of central tendency, the variance of the set is a measure of variability, and the variance of its central value is a measure of precision (see Measures of Precision). The sequence in which the various equations for variances are given in the following sections reflects the fact that ordered sets occur more frequently than randomized sets. The terms "ordered" and "randomized are juxtaposed, and combined with "equal weighting factors " or "variable weighting factors " to show how to calculate the corresponding measure of variability.

Ordered; Equal Weighting Factors The variance of an ordered set of measured values with equal weighting factors (equidistant point estimates) is:

C (xi+j-xi)2 varj(x) = 2(n -j) where varj(x) = jth variance term of ordered set xi+j = (i+j)th measured value xi = ith measured value j = jth spacing between measured values n = number of measured valuesfor jth variance term 2(n -j) = degrees offreedom for jth variance term This variance has found many applications in science and engineering. For example, the first variance term of the ordered set of metal grades with equal weighing factors (see Table 1) is varl(x) =[(30.3 -32. 1)2 ... + (29.8-31. 8)2]/[2*(5-I)]= 1.1212 % 2. Given that degrees of freedom for ordered sets of measured values are not universally embraced (Merks 1993, 1997), it is necessary to explore this concept in a separate section (see Degrees of Freedom).

+

Ordered; Variable Weighting Factors The variance terms of an ordered set of measured values with variable weighting factors are computed as follows:

C [ ~ 2 *j (xi+j-xi)2J

B q 31

varj(x) = (2/ C wlj2)-2j where varj(x) = jth variance term of ordered set xi+j = (i+j)th measured value xi = ith measured value j = jth spacing between measured values W I j =first weighting factor for jth variance term w2j = second weighting factor for jth variance term (2/C wlj2)- 2 = degrees offreedom for jth variance term

Variance terms or ordered sets are useful in mineral processing, smelting and refining when variables are measured at different intervals, and in mineral exploration and mining where measured values for core samples of variable length and bulk samples of variable mass are ordered in space.

42

Randomized; Equal Weighting Factors The variance of a set of measured values with equal weighting factors is elementary, and has its own function in spreadsheet software. The correct variance function in spreadsheet software

should be selected to obtain the "sample variance", the variance of a sample of a population rather than the "population variance" itself. These functions are =VAR(xi..xj) in Excel, and @VARS(xi..xj) in Lotus. The denominators in both functions are the degrees of freedom for the set (seeDegrees of Freedom). The variance of the randomized set of metal grades (see Table 1 ) is var(x)=[(32.1 -30.90)2+... +(29.8-30.90)2]/(5-l)=0.9950%2(seeAppendix C).

Randomized; Variable Weighting Factors The variance of a randomized set of measured values with variable weighting factors w2i as defined in Measures of Central Tendency is:

C [ ~ 2 *i (i-~i)2] var(x) = where

WIi

and

IEq 4

(1/ C wIi2)- 1

x = weighted average

xi = ith measured value

wli =first weightingfactor for ith measured value w2i = second weightingfactor for ith measured value ( l / C wIi2)-1 = degrees offredom

x

The weighted average must be calculated before the differences between and xi can be squared and multiplied with ~ 2 i the , corresponding weighting factors. Therefore, several columns in a spreadsheet template are required to obtain not only the central value and the variance of the set but also the variance of the central value (see Appendix C). Given that the order in which squared differences are added does not impact the numeric values of variances, it does make sense to refer to randomized sets. Whenever a set of measured values is ordered, either in space (core samples in a borehole; rounds of crushed ore taken from a drift or trench) or in time (on-stream and production data), testing for spatial dependence becomes an important element of statistical analysis (see Testingfor Spatial Dependence). Table 2 gives, in addition to the mass weighted average grade of 30.71 % for the data set in Table 1, the most common measures of variability (see also Appendix C). Table 2 Basic statistics Statistic

Symbol'

Mass weighted average grade in %abs

-

x Variance of randomized set in ( %abs)2 var(x) Standard deviation in %abs sd(x) Coefficient of variation in %re1 cv 1 text

2

Symbol2

Value

xbar var(x) sd(x)

30.71 0.8143 0.9024 2.9

cv

template

Due to its squared dimension the variance is not a user-friendly measure for variability. By contrast, the coefficient of variation (the standard deviation in relative percent) makes it simple to check and compare different degrees of variability at a glance.

TESTING FOR SPATIAL DEPENDENCE A statistically significant degree of spatial dependence gives a lower variance of the ordered set, and, thus, a higher degree of precision for its central value. Testing for spatial dependence is also an important element of statistical analysis when optimizing sampling protocols. Fisher's F-test is applied to assess whether two variances are statistically identical or differ significantly

43

10

Variance Terms

Is

Figure 1 Sampling Variogram by comparing the ratio between the highest variance and the lowest variance (the F-statistic or calculated F-value) with values tabulated in the F-distributions at 5 % and 1% probability with the applicable degrees of freedom. Applying the F-test to var1(~)=1.1212%2,the first variance term of the ordered set of metal grades in Table 1, and var(x)=0.9950%2, and the variance of the randomized set, gives F= 1.1212/0.9950= 1.13. Since this F-statistic is lower than F0.05;8;4=6.04 at 5% probability, these variances are statistically identical. By implication, the ordered set of metal grades does not exhibit a significant degree of spatial dependence. Thus, this set is classified as "randomized"within its sample space. The tabulated values in the F-distribution at 5 % probability rank from F0.05; 1;1= 161 to F 0 . 0 5 ; 0 1 ; ~ = 1 and , those in the F-distribution at 1% probability rank from F0.01;1;1=4,052 to FO.O1; 01;03 = 1 (Handbook 1986). Therefore, the F-statistic is always the ratio between the highest variance and the lowest variance (Volk 1980).

Sampling Variogram The sampling variogram in Figure 1 is based on a set of 96 on-stream data obtained at 15 min intervals during a 24-hour shift. The variance terms of the ordered set, the variance of the randomized set, and the lower limits of the asymmetric confidence ranges at 95% and 99% probability are given in Appendix D . Since the F-statistic of F=0.0790/0.0038=20.79 for var(x) =O. 0790, the variance of the randomized set, and var2(x)=O. 0038, the second variance term of the ordered set, is higher than the tabulated value of F0.01;95;188=1.33 at 1% probability, the degree of spatial dependence at a spacing of 30 min is statistically significant. By contrast, the F-statistic of F=0.0790/0.0618= 1.28 for the variance of the randomized set, and var20(x)=0.0618, the 20th variance term of the ordered set, is lower than F0.05;95;154= 1.34 at 5 % probability. Evidently, the ordered set of on-stream data no longer exhibits a significant degree of spatial dependence at a spacing of 5 hours. When the variance of the randomized set (var(x)=0.0790%2), the lower limit of its asymmetric 95 % confidence range [95 % ACRL =var(x)/FO.05;dJ 00 =O. 0790/1.24=O. 0581 962], and the lower limit of its 99% confidence range [99% ACRL=var(x)/FO.O1;dfi~=0.0790/ 1.36) =O. 06379621, are also plotted, the sampling variogram illustrates whether the degree of spatial dependence is statistically significant and where orderliness dissipates into randomness. In fact, a sampling variogram is a visual interpretation of Fisher's F-test when applied to check the degree of spatial dependence in the sampling unit or sample space under examination. Generally, the existence of spatial dependence at spacing j is verified by applying the Ftest to the variance of the randomized set and the jth variance term of the ordered set. If the

44

calculated F-value is higher than the tabulated F-value with the applicable degree of freedom, either at 5 % or at 1% probability, then the degree of spatial dependence at the jrh spacing is statistically significant at the corresponding probability level. Although the first variance term determines the intrinsic variability of a dewatered sample for a shift, it does not give a variance estimate. Only a pair of interleaving samples gives 8n unbiased variance estimate that takes into account the second variance term of ordered on-stream data. Whenever the variance of a randomized set and the first variance term of the ordered set are statistically identical, the differences between consecutive on-stream data become random numbers that cannot be used for process control. Higher variance terms have fewer degrees of freedom than lower terms because the last but one datum is not used for the second term, the last but two for the third, and so on. Therefore, F-statistics for small sets should be interpreted with caution. In addition, mathematical analysis ought not to be applied to differences between statistically identical variances (Merks 1993).

DEGREES OF FREEDOM The concept of degrees of freedom in applied statistics is the corollary of the fundamental requirement of functional or mathematical independence in probability theory. The difference between var(x), the variance of a sample, and g2, the population variance, explains why degrees of freedom are finite in applied statistics but deemed infinite in probability theory. The differences between n measured values and the arithmetic mean of the set are XI -X, ..., xi-X, ..., x n-x so that the sum of n differences equals (XI+ ..._ x i + . .. + m)-&. By definition, the arithmetic mean of a set of n measured values is x = ( x l + ... +xi+...+xn)/n. Hence, (XI+...+xi+ ...+xn)-n.f=O. Logically, if n-1 differences are given, the missing one is determined because the sum of n differences is zero. Because a set of n measured values has n-1 independent differences and a single dependent difference, a randomly distributed set of n measured values has n-1 degrees of freedom. By contrast, the first variance term of an ordered set has 2(n-1) or (2/C wlj2)-2 degrees of freedom because all but the first and last datum are used twice which implies that each higher term has two fewer degrees of freedom than the preceding term. One measured value does not give any information on that elusive population variance , is indeterminate as it ought to be. It can be simply because C ( ~ l - . f ) ~ / ( n - l ) = O / Owhich proved by induction that adding any number of functionally (or mathematically) dependent values to a set of measured (or independent) values does not add a single degree of freedom. Degrees of freedom are positive integers for sets of measured values with equal weighting factors but become positive irrational numbers for sets of measured values with variable weighting factors.

+

VARIANCES OF CENTRAL VALUES The variances of central values are pivotal statistics in sampling theory and practice because they play a critical role in bias testing of mechanical sampling systems, manual sampling procedures, sample preparation techniques and analytical methods. The variances of central values also underlie confidence intervals and ranges, bias detection limits as measures for the power or sensitivity of Student’s t-test, and probable ranges as intuitive measures for the probabilistic limits within which an observed bias is expected to fall.

Variance of Arithmetic Mean The variance of the arithmetic mean of a set of n measured values with equal weighting factors is elementary:

where var(.f) = variance of arithmetic mean var(x) = variance of set n = number of measured values in set

45

The intermediate term in Equation 5 reflects an important step in the derivation of this equation from the variance of a general function as defined in probability theory (Volk 1980). This simple relationship between the variance of a set of measured values with equal weighting factors and the variance of its arithmetic mean is often referred to as the Central Limit neorem. Perhaps ironically because the variances of weighted averages are required in many applications in mineral exploration, mining, processing, smelting and refining.

Variance of Weighted Average The variance of a count, density, distance, length, mass or volume weighted average of a set of measured values with variable weighting factors is:

where var(i) =variance of weighted average var(x) = variance of set W I i =first weightingfactor for ith measured value The variance of the weighted average of any set of measured values with variable weighting i l / n . When factors converges on the variance of the arithmetic mean when all w ~ approach formulated as the sum of two or more products of squared weighting factors and variances, the central limit theorem proves that the variance of the central value of two or more sets of primary samples, selected from multinomial, binomial or Poisson distributions, converges on the normal (or Gaussian) distribution. How to determine whether or not a set of measured values exhibits a Gaussian or normal probability distribution is described in a draft standard under development by ISO/TC69. Whenever a set of measured values departs from normality, it can be partitioned into subsets such that each subset approaches a straight line segment in a log-normal plot of a numerically ordered set. The next step is to calculate the central value of the set and its variance from the arithmetic means and the variances of all subsets (Merks 1998).

VARIANCE OF CONTAINED METAL The additive property of the variance of contained metal (the mass of contained metal or metal content) underlies various applications in mineral exploration, mining, processing and smelting (Merks 1985, 1988, 1991, 1999; Merks and Merks 1991). ISO/DIS 13543 describes how to determine the variance of the mass of metal contained in a lot. This method is based on the premise that pairs of interleaving primary samples are routinely taken from all lots. Pairs of interleaving samples also give the variances of metal contained in tailings, concentrates and thickener inventories, which can be used to calculate reliable precision estimates for metal grades of mill feed. The mass of metal contained in a quantity of crushed ore or mineral concentrate is a function of its wet mass, moisture content and metal grade:

m 71

Me=Mw+MF+GF where Me = mass of contained metal in mt Mw = wet mass in mt MF = moisturefactor :1 -[O.Ol * %H20] (dimensionless) %H20 = moisture content in percent GF = grade factor :0.01 * %Me (dimensionless) %Me = metal grade in percent on dry basis

The variance of contained metal is obtained by substituting in the equation for variance of a general function the squared partial derivatives for Equation 7 and the variances of these stochastic variables :

46

where var(Me) = variance of contained metal in mt2 var(Mw) = variance of wet mass in mt2 var(MF) = variance of moisturefactor (dimensionless) var(GF) = variance of grade factor (dimensionless) Multiplying the mass term in Equation 8 with Mw2/Mw2,the moisture term with MF2/MF2,and the grade term with GF2/GF2,dividing each term by Me2, and multiplying the sum of all terms with Me2, gives the following equation for the variance of contained metal:

var(Me) =Me2 [var(Mw)/Mw2+ var(MF)/MF2+ var(GF)/GF2]

Eq 91

Extreme care should be exercised to ensure that variables and variances are correctly entered into Equation 9. Changing from grades in percent (%Me) to grade factors (GF), from precious metal grades in g/mt to contained metal in kg, and from moisture contents (%H20)to moisture factors (MF), demands close attention to derivatives, dimensions and decimal places. Scale calibration data can often be used to obtain the variance of wet mass (Merks and Merks 1992). The additive property of variances also makes it simple to determine the variance of the mass of metal contained in an ore deposit, and to calculate confidence limits for its metal content and grade as a measure for the risk associated with the least precise measurement procedure in mining and metallurgy. The variances of dividing whole core samples into halves, preparing test samples of selected halves, and taking and assaying test portions of test samples, are extraneous to the sample space, and add to the variance of the stochastic variable within its sample space. Extraneous variances may be subtracted from the variances of the randomized and ordered sets before Fisher’s F-test is applied and confidence limits for contents and grades of ore deposits are computed (Merks and Merks 1991, Merks 2000).

VARIANCE OF GY’s SAMPLING CONSTANT Gy’s sampling theory proposes that the primary sample mass required for a specified degree of precision can be estimated a priori. In sampling practice, however, it can only be determined experimentally because the degree of heterogeneity of a stochastic variable within a sampling unit defies a priori estimation (Merks 1985, Visman 1962). The variance of the primary sample selection stage is the sum of the composition variance (the variance between particles within primary increments) and the distribution variance (the variance between primary increments within a sampling unit). It is the latter variance that defies a priori estimation in heterogeneous sampling units, and that causes the ordered set of on-stream data to exhibit spatial dependence (seeFigure 1) and gives a higher degree of precision for the central value (see Appendix D). Gy’s sampling theory suggests that CT~(FE),his fundamental error, is a function of C, his sampling constant, and d3, the cube of the top size of the particulate matter. Gy’s sampling constant C, in turn, is a function of four factors (Gy 1979). The variance of Gy’s sampling constant, too, derives from the variance of a general function (Volk 1980):

var(C) = C2* [var(c)/c2+var(l)/12+varfl/f + var(g)/g2]

[Eq 101

where var(C) = variance of sampling constant var(c) = variance of mineralogical compositionfactor var(l) = variance of liberationfactor v a r n = variance of particle shapefactor var(g) = variance of size rangefactor Logically, C is a constant only if each of these variances is infinitesimally small but, in the real world, variances are finite. In fact, a single pair of interleaving primary samples gives an imprecise estimate for var(spa), the sum of the variances of the primary sample selection, preparation and analytical stages because it has but one degree of freedom. In sampling practice,

47

the question is not so much whether Gy’s sampling constant is indeed a constant but how imprecise a variance estimate with a single degree of freedom really is. Fortuitously, applied statistics gives a relationship between degrees of freedom and confidence limits for variances. Symmetric 95% confidence ranges for variances are computed from values of the X2distribution at different probability levels (Handbook 1968). Table 3 gives the tabulated +values at 2.5% and 97.5% probability, and the lower limits [95% CRL= df * var(spa)/x20.975;dfl and upper limits [95% CRU=df * var(spa)/x20.025;dfl of the symmetric 95% confidence ranges for var(spa)=0.10 when estimated with increasing degrees of freedom (Volk 1980). Table 3 Symmetric 95 % confidence ranges for var(spa)=O. 10 Degrees of freedom 1

5 10 25

~~0.975

~~0.025

5.02 12.8 20.5 40.5

0.001 0.83 1 3.25 13.1

03

95% CRL 0.020 0.039 0.049 0.062 0.10

95% CRU 101.8 0.602 0.308 0.191 0.10

Table 3 underscores the astounding precision of variances when degrees of freedom are infinite. Given that Gy’s sampling constant is a function of four stochastic variables whose variances have finite degrees of freedom, it is implausible that var(C), the variance of Gy’s sampling constant, is infinitesimally small. The more so because the variance of the cube of the topsize is 32=9 times larger than the variance of the topsize itself. Neither is it plausible that U ~ ( F E ) ,Gy’s fundamental error, gives a meaningful a priori estimate for the primary sample mass required for a specified degree of precision for a heterogeneous sampling unit. After all, the distribution component of the variance of the primary sample selection stage, which is a measure for the degree of segregation or heterogeneity in a sampling unit, can only be estimated from a sampling experiment based on taking 20-30 pairs of small and large increments (ASTM D2234, Merks 1985, Visman 1962). When this experiment is applied to a dynamic sampling unit, the distribution variance, the very statistic to be estimated, is reduced. This is the corollary of Heisenberg’s uncertainty principle in sampling practice where the measurement procedure impacts the outcome. The same experiment does give an estimate of the composition variance but does so at high cost.

MEASURES OF PRECISION The fundamental measure of precision is the variance of a central value but derived measures of precision such as confidence intervals and ranges are more intuitive and transparent than variances. For many applications, 95% confidence intervals (95% CI) and 95% confidence ranges (95% CR) are acceptable, but if the risk associated with a wrong decision is high, confidence intervals and ranges at 99 % or 99.9 % probability should be considered. Confidence intervals are given in absolute values and relative percentages whereas confidence ranges are given in absolute values only.

Confidence Interval The calculation of a 95 % confidence interval for the central value of a randomized set requires a tabulated value of the t-distribution at 5% probability with df=n-I or df=(1/CwIi2)-l degrees of freedom. However, if the first variance term of the ordered set is significantly lower than the variance of the randomized set, the t-values at df=2(n-l) or df=(2/C wIi2)-2 degrees of freedom may be used. Since t0.05;60=2.000 for 60 degrees of freedom, and t 0 . 0 5 ~ = z 0 . 0 5 =1.96 for infinite degrees of freedom, the z-value of normal distribution is can be rounded to 2, the following equation applies to all sets:

48

95% CI=sd(i) * to. 05;df where 95% CI = 95% confidence interval (absolute value) sd(x) = standard deviation of central value to. 05;df = tabulated t-value at 5 % probability df = degrees offredom Table 4 gives the mass weighted average grade of 30.71 % for the set of paired dry masses and metal grades in Table 1, its 95% confidence interval of f1.3796abs (absolute percent), and its 95% confidence interval of 95% CI=1.37*1oo/jO. 71 = f4.596rel (relativepercent). Table 4 Confidence interval Statistic

Symbol'

Mass weighted average grade in %abs

x 95% CI 95% CI

95 % Confidence interval in %abs 95 % Confidence interval in % re1 1

text

-

Symbol2

Value

xbar 95% CI 95% CI

30.71 f1.37

f4.5

templates

Appendix C also gives 95 % CIS in absolute values (%abs in this case), and in relative percent (%el) but without f - symbols. Comparing the 95% CI of f4,5%rel in Table 4 with 95% CI= f1.05%rel for the randomized set of on-stream data (see Appendix D ), and 95% CI= f0.24%rel for the ordered set illustrates how a large data set and a significant degree of spatial dependence impact the precision of the central value of 5.22%. Confidence intervals at 99% and 99.9% are obtained by multiplying sd(i) with t0.01;df and t0.001;df respectively.

Symmetric Confidence Range The lower and upper limits of a symmetric 95 % confidence range are obtained as follows:

95% CRL=X-95% CI 95% CRU=x+95% CI where 95% CRL = lower limit of 95% confidence range 95% CRU = upper limit of 95% confidence range x = central value of set 95% CI = 95% confidence interval Table 5 is based on the mass weighted average grade of 30.7 1% for the set of paired wet masses and metal grades in Table 1, and on the derived statistics in Appendix C. Table 5 Confidence range Statistic

Symbol1

X Mass weighted average grade in %abs 95 % Confidence range 95% CR 95% CRL Lower limit in %abs 95% CRU Upper limit in %abs 1

text

Symbol2 xbar 95% CR 95% CRL 95% CRU

Value 30.71 29.3 32.1

template

Confidence ranges at 95% and 99% probability are convenient control and action limits in statistical quality control (SQC) charts.

49

Asymmetric Confidence Range The lower limit of an asymmetric 95% confidence range is the central value of the set minus its 90% confidence interval. Similarly, the upper limit of an asymmetric 95% confidence range is the central value of the set plus its 90 % confidence interval: 95% ACRLG--W% CI 95%ACRU=x+W% CI

where 95% ACRL = lower limit of 95% confidence range 95% ACRU = upper limit of 95% contdence range X = central value of set 90% CI = 90% confidence interval These lower limits and upper limits are mutually exclusive. In other words, either the lower limit or the upper limit is valid. Together, however, the same limits give a symmetric 90% confidence range. For large sets, the tabulated value of to. 10;df converges on &.lo= 1.645.

STUDENT’St-TEST The t-test is applied to examine whether the difference between identifiably different paired test results is due to random variations or caused by the presence of bias. Typical examples are test results for reference increments and system samples, for different laboratories, for the same laboratory but at different times or by different technicians, or for different analytical methods. In every case, the question is whether two central values differ significantly, and, thus, whether their difference indicates the presence of bias (reject null hypothesis). Alternatively, the difference between central values is statistically identical to zero (accept null hypothesis), and its numeric value merely reflects the effect of random variations in measurement procedures. An observed bias is either significantly higher than an accepted value such as a certified value for a Certified Reference Material or the central value of a set of reference increments (a positive bias), or significantly lower than the certified or reference value (a negative bias). When a set of paired test results reported by two laboratories fails the bias test, the t-test does not reveal which laboratory is suspect but only that the difference between their central values is higher than random variations alone could explain. Most textbooks on applied statistics give the t-distribution with tabulated values for probabilities ranging from 90% to 0.1% for one degree of freedom to infinite degrees of freedom. If the t-statistic is much lower than t0.90;df, it may reflect the too-good-to-be-rrue effect, which could be indicative of tampering with test results. Since the t-statistic (the calculated t-value) is the ratio between the difference between two central values and the standard deviation of the difference, the following equations apply:

d[var(b)/n]

sd(hx)

where t = t-statistic ;I = central value offirst set x2 = central value of second set di: = diperence between central values var(Ax) = variance of diperences n = number ofpairs sd(Ai) = standard deviation of diTerence The central limit theorem also underlies the relation between sd(&), the standard deviation of the difference between two central values, and s d ( b ) , the standard deviation of the differences between paired test results. Given that sd(&) = d[var(b)/n]= s d ( b ) / d n , it follows that three variables interact and determine the t-statistic and the power of the t-test.

50

Appendix E gives the t-statistics for a test program designed to test for bias between reference increments and final system samples. This bias test program is based on comparing 30 pairs of test results determined in reference increments removed from a stopped belt with the aid of a sampling frame, and in final system samples obtained with a multistage mechanical sampling system. Table 6 gives the basic t-statistics for the test program. Table 6 Basic t-statistics Statistic Central value in % : reference increments Central value in % : final system samples Difference in %abs Difference in % re1 Calculated t-value Significance f

text

f

template

+++

Symbol'

Symbolz

Value

f(R)

xbar(R) xbar(s) dxbar dxbar t

8.10 7.49 -0.61 -7.5 11.296

x(s) Ax Ai t

***

significant at 0.146 probability

The calculated t-value of 11.296 exceeds the tabulated value of t0.001;29=3.674 at 0.1% probability (Handbook 1968; Volk 1980) so that the difference of -0.6l%abs or -7.5%rel implies the presence of bias. The spreadsheet template in Appendix E gives three asterisks to indicate statistical significance at 0.1 % probability. Two asterisks would have been printed for statistical significance at 1%, and a single one at 5 %. In addition, 11s (not significant) would have been printed in the same cell if the t-statistic were lower than t0.05;df at 5 % probability. The variance of differences is calculated from the differences between paired data. This variance is the sum of the variances of all systems and procedures used to obtain the set. Given that the variance of differences and the number of paired data determine the power of the t-test, it is possible to prove that even a small and commercially insignificant difference is a bias if the number of pairs is large enough. A preliminary bias test may be needed to estimate the number of test results necessary to prove statistical significance at a specified probability level. The t-test can also be applied to pairs of measured values with variable weighting factors such as central values for on-stream data and test results for slurry samples for the same production period, or to the exchange assays for lots of variable mass. It is beyond the scope of this paper to present a numerical example.

Bias Detection Limits Bias detection limits (BDLs) are intuitive measures for the power or sensitivity of the t-test to detect a bias or systematic error between two central values. BDLs are defined for the Type I statistical risk only, and for the combined Type I and Type I1 statistical risks. A simple analogy exists between these statistical risks and the role of a fire alarm. The Type I statistical risk refers to the event that a fire occurs but the fire alarm does not sound. The Type I1 statistical risk refers to the event that the fire alarm sounds but no fire occurs. Finally, the combined Type I and Type I1 statistical risks refer to a fire and the sound of a fire alarm. The effect of the number of pairs on the power of the t-test becomes evident upon realizing that these statistical risks are obtained by multiplying the standard deviation of the difference either with the tabulated t-value at 5 % probability, or with the sum of the tabulated t-values at 5 $6 and 10 % probability. A symmetric two-sided 5 % probability for the Type I risk only, and an asymmetric one-sided 5 % probability for the Type I1 risk, are widely accepted. Several I S 0 Standards on bias testing of mechanical sampling systems and manual sampling procedures specify statistical risks in the same manner and at the same probability levels. Based on this convention, the bias detection limits for the Type I risk only, and for the combined Type I and Type I1 risks, are defined as follows:

BDL(I) =sd(&)

+

to.05;df

IEq 171

51

BDL(I& II) =sd (&)* [to. 05;df+ to.1O;dfl where BDL(I) = BDL for Type I risk only BDL(I&II) = BDL for combined Type I and Type II risks sd(Ai) = standard deviation of direrence tO.05;df = tabulated t-value at 5 % probability tO.lO;df = tabulated t-value at 10% probability df = degrees offredom Table 7 gives the bias detection limits in absolute and relative percent on the basis of the differences between the central values of 8.10% for reference increments and 7.49% for final system samples (see also Table 6 and Appendix E ) . Table 7 Bias detection limits Statistic Difference Bias detection limits Type I risk only Type I and I1 risks

%abs

%re1

-0.61

-7.5

fO.ll f0.20

f2.5

Symbol Ax BDLs BDL(1) BDL(I&II)

f1.4

Bias detection limits are effective control and action limits for SQC charts in which precision and bias of measurement systems and procedures are monitored as a function of time. A strong case can be made that metal grades and contents of concentrate shipments should be measured and monitored to ensure that biases are detected before losses become punitive.

Probable Ranges Probable ranges (PRs) define the limits within which an observed bias is expected to fall. Whenever a difference between two central values turns out to be statistically significant, and exceeds either the bias detection limits for the Type I statistical risk only, or the combined Type I and Type I1 statistical risks, the following relationships give the lower and upper limits of the corresponding probable ranges for the observed bias:

PRL(I) = ~;_-BDL(I) PRU(I) = Ax+BDL(I) PRL(I&II) = hx-BDL(I&II) PRU(I&II) = &+BDL(I&II) where PRL(I) = lower limit of probable rangefor Type I risk only PBU(I) = upper limit of probable rangefor Type I risk only PRL(I&II) = lower limit of probable rangefor combined Type I and II risks PBU(I&II) = upper limit of probable rangefor combined Type I and II risks hx = observed bias BDL(I) = bias detection limitfor Type I risk only BDL(I&II) = bias detection limit for Type I and Type II risks Reporting probable ranges for an observed bias makes sense only if the difference between two central values is indeed indicative of the presence of bias, and the null hypothesis is rejected. If a difference exceeds the bias detection limit for the Type I risk, the lower and upper limits of the corresponding probable range are reported. For example, the difference of -0.61 %abs between 8.10% for reference increments and 7.49% for system samples is lower than BDL(I)= -0.11 % and BDL(I&II)= -0.20% (see Table 7). Therefore, the lower and upper limits of the

52

probable ranges are defined not only for the Type I risk but also for the combined Type I and Type I1 risks. Logically, the difference of -7.5 %re1between reference increments and system samples is also lower than BDL(I)= -1.4%rel and BDL(I&I)= -2.5Arel (see Table 7) which implies that the probable bias ranges are defined for the Type I risk only and for the combined Type I and Type I1 risks. Table 8 gives the probable ranges for the observed bias. Table 8 Probable ranges Statistic

Symbol

% abs

%re1

Difference Probable range Type I risk only Lower limit Upper limit

Ai PRs PR(I) PRL(1) PRU(1)

-0.61

-7.5

-0.72 -0.50

-8.9 -6.1

PR(I&II) PRL(I&II) PRU(I&II)

-0.81 -0.41

-10.0 -5.0

Type I and I1 risks Lower limit Upper limit

If a difference between two central values is statistically identical to zero, and the lower and upper limits of the probable range for the Type I risk are not defined, the abbreviation nu (not applicable) may be printed in the appropriate cells of the spreadsheet template (see Appendix E ) . Whenever an observed bias in moisture content or metal grade impacts the cumulative mass of metal contained in concentrate production, it would make sense to convert probable ranges into monetary units.

FISHER’SF-TEST Fisher’s F-test is applied to determine whether two variances are statistically identical or differ significantly. The F-test is based on comparing the ratio between the highest variance and the lowest variance with tabulated values from the F-distributions at 5 % and 1% probability and with the applicable degrees of freedom for each variance (Handbook 1968; Volk 1980). If the calculated F-value is lower than the tabulated value of F0.05;dfi;dfl at 5 96 probability, then the variances are statistically identical. The probability that this statistical inference is true exceeds 95 %. Conversely, the probability that this inference is false is less than 5 A. Alternatively, if the F-statistic is higher than F0.05;dfi;dfl at 5 % probability, the variances differ significantly, and the probability is less than 5% that this inference is false. Similarly, if the F-statistic is higher than F0.0l;dfi;dfl at 1 % probability, the variances differ significantly but in this case, the probability is less than 1% that the inference is false. Tabulated F-values, too, reflect that dfi and dfl are the degrees of freedom for the numerator and denominator in the F-test. The fact that F0.05; 00;00 =F0.01; 00; 00 = 1 explains why the concept of degrees of freedom is of critical importance when analysis of variance is applied to test for spatial dependence and to optimize sampling protocols.

Optimizing Sampling Protocols Suppose that a sampling experiment gives vur(spu)=0.075 for the sum of the variances of the primary sample selection, preparation and analytical stages, and vur(a) =O. 050 for the variance of taking and assaying a test portion of a test sample. The question of whether these variance estimates are statistically identical or differ significantly can only be solved if the applied sampling protocol and the degrees of freedom for vur(spu) and vur(u) are taken into account. For example, a pair of interleaving primary samples gives a single degree of freedom for vur(spu)=0.0750, and duplicate test portions taken from each of a pair of test samples give two degrees of freedom for vur(u)/Z=O.O50/2 =O.O25, the analytical variance of the arithmetic

53

mean of duplicate test results. This is the reason why the analytical variance (the variance of taking and assaying a single test portion of a test sample) is divided by the factor 2. Since F=0.075/0.025=3.00 is lower th.an F0.05;1;2= 18.51 at 5% probability, the difference of 0.075-0.025=0.050 between vur(spu) and vur(u)/2 is not a valid estimate for vur(sp), the sum of the variances of the primary sample selection and preparation stages. Therefore, no statistical significance should be attached to this difference of 0.050, nor should mathematical analysis be applied to such differences (Merks 1993). By contrast, 20 pairs of interleaving primary samples would give 20 degrees of freedom for var(spu)=0.075 whereas duplicate test portions taken from each of 40 test samples would give 40 degrees of freedom for vur(u)/2=0.050/2=0.025. In this case, the F-statistic of F=0.075/0.025=3.00 exceeds not only F0.05;20;40= 1.54 at 5% probability but also F0.01;20;40=2.37 at 1% probability. Hence, the same difference of 0.075 -(0.050/2)=0.050 is a valid estimate for vur(sp), the sum of the variances of the primary sample selection and preparation stages. The probability that this statistical inference is false is less than 1%. The latter F-test shows that vur(sp), the sum of vur(s), the variance of the primary sample selection stage, and vur(p), the variance of the sample preparation stage, adds most to var(spa), the variance of the entire measurement chain. The most effective method to reduce var(s) the variance of the primary sample selection stage, is to increase the number of primary increments (Gy 1979, Merks 1985, Visman 1962). The variance of the sample preparation stage can be estimated by preparing duplicate test samples of each of a pair of interleaving samples, and assaying duplicate test portions of each test sample. For example, 10 pairs of interleaving primary samples would generate 20 pairs of test samples and 40 pairs of test portions, and give 10 degrees of freedom for vur(spu), the sum of the variances of the primary sample selection, preparation and analytical stages, 20 for vur(pu), the sum of the variances of the sample preparation and analytical stages, and 40 for vur(u), the analytical variance (see AppendicesA & B). The variance of the sample preparation stage can be reduced to a minimum by comminuting and homogenizing dried sample masses prior to division. The key is always to find a compromise between acceptability and expediency, a task that requires some understanding of experiment design and statistical analysis of test results. Sample preparation procedures are prone to bias due to cross contamination and loss of dust, moisture or native metal while comminuting, homogenizing and dividing sample masses (Merks 1985, 1988, 1993).

SUMMARY Sampling in mineral processing is based on scientifically sound elements of probability theory and applied statistics. The properties of variances are the quintessence of sampling theory and practice. The additive property of the variances of volume, mass and contained metal play a key role in metallurgical accounting procedures. Combining a set of primary increments into a single primary sample does not give a variance estimate. Dividing a set of primary increments into a pair of interleaving primary samples is the most effective procedure to estimate the variance of the entire measurement chain. Interleaving sampling protocols are equally effective for slurry flows in mineral processing and bulk samples in mineral explorations. Sampling protocols can be optimized by applying analysis of variance to partition the sum of the variances of the primary sample selection, preparation and analytical stages into its components, and by examining which variance component should be reduced to improve the precision of the measurement procedure. On-stream data almost invariably exhibit a significant degree of spatial dependence. Metal grades of contiguous sets of core samples within a borehole, or a set of adjoining rounds in a drift or trench, may also display a significant degree of spatial dependence. When plotted in a graph the variance terms of an ordered set display a sampling variogram. When the variance of the randomized set and the lower limits of its asymmetric 95% and 99% confidence ranges are also plotted, the sampling variogram shows whether the degree of spatial dependence is statistically significant and where orderliness in the sampling unit or sample space under examination has dissipated into randomness.

54

The computations discussed in this paper are carried out with spreadsheet software. Setting up effective spreadsheet templates is an important element of sampling practice in mining and metallurgy. A strong case can be made that sound elements of probability theory and applied statistics be implemented in all the measurement procedures commonly applied in mineral exploration, mining, processing, smelting and refining.

ACKNOWLEDGMENT The author dedicates this paper to the memory of Len Green, formerly with Falconbridge Limited, and for many years the driving force behind the Canadian Advisory Committee to I S 0 Technical Committee 183 on copper, lead and zinc concentrates.

REFERENCES ASTM Standards on Precision and Bias for Various Applications, 1985, Second Edition ASTM D2234, 1997, Standard Practice for the Collection of a Gross Sample of Coal, Annual Book of ASTM Standards, Volume 05.05, Gaseous Fuels; Coal and Coke Davies 0 L and Goldsmith P L, 1972, Statistical Methods in Research and Production, Longman Group Limited, London Gy, P M, 1979, Sampling of Particulate Matter; Theory and Practice, Elsevier Scientific Publishing, Amsterdam Handbook of Tables for Probability and Statistics, 1968, The Chemical Rubber Company, Cleveland, Ohio Huff, D, 1954, How to Lie with Statistics, Penguin Books, Middlesex, England ISO/DIS 10251, Determination of Mass Loss of Bulk Material on Drying ISO/DIS 12743, Sampling Procedures for Determination of Metal and Moisture Content ISO/DIS 12744, Experimental Methods for Checking the Precision of Sampling ISO/DIS 12745, Precision and Bias for Mass Measurement Techniques ISO/DIS 13292, Experimental Methods for Checking the Bias of Sampling ISO/DIS 13543, Determination of Mass of Contained Metal in the Lot Mandel, J, 1964, The Statistical Analysis of Experimental Data, Dove Publications, New York Merks, J W, 1985, Sampling and Weighing of Bulk Solids, TransTech Publications, ClausthalZellerfeld Merks, J W, 1988, Sampling and Weighing of Mineral Concentrates, Bulk Solids Handling, Volume 8, Number 2 Merks, J W, 1989, A New Approach to the Exchange of Assays, Sampling in the Non-Ferrous Metals Industry, TransTech Publications, Clausthal-Zellerfeld Merks, J W and Merks E A T, 1991, Precision Estimates for Ore Reserves, Erzmetall, October Merks, J W, 1991, Simulation Models for Mineral Processing Plants, CIM Bulletin, September Merks, J W and Merks E A T, 1992, Precision and Bias for Mass Measurement Techniques, Matrix Consultants Limited, Vancouver Merks, J W, 1993, Abuse of Statistics, CIM Bulletin, January Merks, J W, 1994, Precision and Bias-The Keys to Total Quality Management, Proceedings of the Eighteen International Precious Metals Conference, International Precious Metal Institute, Allentown, Pennsylvania Merks, J W, 1997, Applied Statistics in Mineral Exploration, Mining Engineering, February Merks, J W, 1998, Private communication prepared for litigation Merks, J W, 1999, Process Simulation with Spreadsheet Software, Minerals & Metallurgical Processing, May Merks, J W, 2000, Borehole Statistics with Spreadsheet Software, SME Transactions, Volume 308

Visman, J, 1962, Towards a Common Basis for the Sampling of Materials, CANMET Research Report R93, July Volk, W, 1980, Applied Statistics for Engineers, Robert R Krieger Publishing Company, Huntingdon

55

.-m ~

L

c W

0 c

v)

56

APPENDIX B Interleaving Sampling Protocol

d Sample preparation stage

F

w B-sample

A-sample

I

[

Homogenize

1

Incremental division

Select 1 kg test portion

Dry at 105

Mass loss on

I

+ I

Symbol : var(MF) I

Measurement variance of moisture factor

D r y a t 105 centigrade

1

+ Pulverize

Symbol : var(GF1 I

for grade

Measurement variance of grade factor

57

Y Test samples

-1

Test results for grade

I

APPENDIX C Measures of Variability and Precision Value

Statistic Arithmetic mean grade in % Variance in %^2 Standard deviation in O h Coefficient of variation in %re1

30.90 0.9950 0.9975 3.2

Number of measured values Variance of mean in %^2 Standard deviation in O h

n vadxbar) sd(xbar)

95% Confdence interval in %abs # 95% Confdence interval in %re1 95% Confidence range Lower limit in %abs Upper limit in %abs

95% CI 95% CI 95% CR 95% CRL 95% CRU

1

5 0.1995 0.4467 1.24 4.0 29.7 32.1

Degrees of freedom Tabulated t-value at 5% probability

Statistic

Symbol

Mass weighted average grade in % Variance in %^2 Standard deviation in O h Coefficient of variation in %re1 Sum of squared weighting factors Variance of mean in %^2 Standard deviation in O h 95% Confdence interval in %abs # 95% Confdence interval i n %re1 95% Confidence range Lower limit in %abs Upper limit in %abs

I

xbar var(x) sd(x)

cv

30.71 0.8143 0.9024 2.9

sum(wliA2) var(xbar1 sd(xbar)

0.2401 0.1955 0.4422

95% CI 95% CI 95% CR 95% CRL 95% CRU

Degrees of freedom Tabulated t-value at 5% probability # #

Value

df t0.05;df

1.37 4.5 29.3 32.1

I

O 3 :.l:

# based on 95% CI=sd(xbar)*tO.O5;df

Sampling Unit 1 2 3 4 5

Mass in m t 12.1 54.5 72.6 90.3 84.2

Grade in % 32.1 30.3 30.5 31.8 29.8

wli 0.0386 0.1 737 0.2314 0.2879 0.2684

50

wlP2 0.0015 0.0302 0.0536 0.0829 0.0720

w 2i 0.1929 0.8687 1.1572 1.4393 1.3420

w2i (xbar-xbar)"2 0.3709 0.1 484 0.0526 1.6997 1.1194

APPENDIX D Measures of Variability and Precision for On-stream Data Symboi

‘Statistic Arithmetic mean grade in O h Variance of randomized set Standard deviation Coefficient of variation in %re1

xbar var(x) sd(x) CV(d

First variance term of ordered set Standard deviation Coefficient of variation in %re1

varl (x) sd 1(x) CV(0) 95% CI 95% CI 95% CR 95% CR 95% CR

95% Confidence interval in %abs # 95% Confidence interval in %re1 95% Confidence range Lower limit in %abs Upper limit in %abs

0.057 1.a9

0.01 2 0.24

5.1 6 5.28

5.21 5.23

L

# based on 95% CI =sd(xbar)*t0.05;df

Sampling Variogram for On-stream Data ‘ariance terms of ordered set

Symbol

1

varl (x) var2(x) var3(x) var4(x) var5(x) var6(x) var7(x) var8(x) var9(x) varl 0(x) varl 1(x) varl2(x) varl3(x) varl4(x) varl5x) varl6x) varl7 (x) varl8x) varl9(x) var20(x)

5

ia

15

20

1st 2nd 3 rd 4th 5th 6th 7th 8th 9th 10th 11th 12th 13th 14th 15th 16th 17th 18th 19th 20th

15% ACRL

0.0038 0.0038 0.0070 0.0106 0.0140 0.01 61 0.0205 0.0253 0.0313 0.0355 0.0385 0.041 7 0.0440 0.0483 0.051 1 0.0535 0.0548 0.0550 0.0587 0.0618

59

0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790 0.0790

0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637 0.0637

0.0581 0.0581 0.0581 0.058 1 0.0581 0.058 1 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581 0.0581

APPENDIX D - cont'd Set of 96 On-stream Data Measurad at :

Test result

in I

00:15 00:30 00:45 01:00 01:15 01:30 01:45 02:OO 02:15 02:30 02:45 03:00 0315 03: 30 03:45 04:OO 04: 1 5 04: 30 04:45 05:OO 05: 15 05:30 0545 06:OO 06:15 06:30 06:45 07:OO 07: 15 07:30 07:45 08:00

5.67 5.64 5.54 5.49 5.59 5.71 5.71 5.78 5.45 5.47 5.45 5.45 5.47 5.37 5.41 5.33 5.32 5.49 5.41 5.41 5.33 5.35 5.52 5.59 5.66 5.44 5.45 5.43 5.43 5.54 5.37 5.40

Measured

a: 08:15 0 0 30 0045 09:OO 09: 15 09:30 09:45 1000 10:15 10:30 10:45 11:oo 11:15 11:30 1 1 :45 12:oo 12:15 12:30 12:45 13:OO 13:15 13:30 13:45 14:OO 14:15 14:30 14:45 15:OO 15:15 15:30 15:45 16:OO

Test result in 56

5.42 5.32 5.33 5.32 5.32 5.29 5.26 5.31 5.30 5.33 5.22 5.30 5.29 5.28 5.29 5.20 5.27 5.27 5.28 5.38 5.26 5.28 5.28 5.18 5.00 4.89 4.89 4.72 4.72 4.59 4.59 4.49

Degrees of Freedom and Tabulated F-values Statistic Degrees of freedom for : Variance of randomized set First variance term of ordered set Tabulated F-value at : 5% Probability 1 % Probability

F0.05;95;00

60

1.24 1.36

Measured at :

1615 16:30 16:45 17:OO 17:15 17:30 17:45 18:OO 18:15 18:30 18:45 19:OO 19:15 19:30 19:45 20:oo 20: 15 20: 30 20:45 21:oo 21:15 21:30 21:45 22:oo 22:15 22:30 22:45 23:OO 23: 15 23:30 23:45 24:OO

rest

in 96 4.49 4.68 4.68 4.78 4.70 4.78 4.78 5.07 5.07 4.97 4.97 4.95 4.95 5.02 5.02 5.06 5.06 5.16 5.16 5.25 5.25 5.17 5.17 5.17 5.17 5.24 5.24 5.15 5.15 5.08 5.06 5.06

APPENDIX E Student's t-test for Paired Data Statistic

Symbol

Arithmetic mean in % : reference increments Arithmetic mean in % : system samples Difference in %abs Difference in %re1

xbar(R) xbar(S1 dxbar dxbar

8.10 7.49 -0.61 -7.5

Variance of differences in %^2 Standard deviation in % Coefficient of variation in %re1

var(dx) sd(dx1

0.0865 0.2942 3.6

n vaddxbar) sd(dxbar1

30 0.0029 0.0537 11.296

cv

Number of paired data Variance of difference Standard deviation Calculated t-value Significance

t

Value

***

I

Bias detection limits in %abs Type I statistical risk only Type I & Type II statistical risks

BDLs BDL(I) BDL(I&II)

0.1 1 0.20

Probable bias range in %abs Type I risk only : lower limit Type I risk only : upper limit Type I & II risks : lower limit Type I & II risks : upper limit

PBRs PRL(I1 PRU(I1 PRL(I&II) PRU(I&II)

-0.72 -0.50 -0.81 -0.41

Bias detection limits in %re1 Type I statistical risk only Type I & II statistical risks

BDLs BDL(I1 BDL(I&II)

1.4 2.5

Probable bias range in %re1 Type I risk only : lower limit Type I risk only : upper limit Type I & II risks : lower limit Type I & II risks : upper limit

PBRs PRL(I) PRU(I) PRL(I&II) PRU(I&II)

-8.9 -6.1 -10.0 -5.0

'Degrees of freedom Tabulated t-values a t : 10% Probability 5% Probability 1% Probability 0.1 % Probability

df

to. 1O;df t0.05;df tO.O1 ;df to.001;df

* * * significant at 0.1 % probability

61

29 1.701 2.048 2.763 3.674

APPENDIX E - cont'd Test results for bias test program reference increments and system samples

Pair Number

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Reference increments

6.97 6.85 6.69 6.75 7.39 7.34 7.98 7.70 7.95 7.91 7.51 9.22 8.82 8.77 8.80 8.65 8.74 7.40 8.39 7.94 8.23 8.42 8.28 8.96 7.71 8.26 8.93 8.52 8.14 9.63

62

6.54 6.38 6.59 6.34 7.13 6.73 7.58 7.23 7.32 7.23 6.73 8.44 8.33 8.41 8.77 7.76 7.68 6.63 7.39 7.04 7.61 7.85 7.21 7.59 7.19 7.82 8.57 7.83 7.76 8.97

Sampling High Throughput Grinding and Flotation Circuits John Moshei DanAIwndcr’

ABSTRACT Consistently operating a plant at peak efficiency requires knowledge and understanding of circuit performance. Conducting surveys allows analysis of plant p d m a n c e , and presents an opportunity to identify areas for improvement in circuit operation, maintenance, and conuol. A d d i t i d y , 8s the use of modeling and MmpUtM ~ h u h t of h & d t s becomes more prevalent, the need for sound survey data on which to base simulation models becomes even more important. With larger and larger comminution and flotationcircuits, sound survey procedures are essential in orda to collect good data. All surveys have three stages: planniag and preparation, survey sampling. a d data analyses. This p discusses thc process and technique of conducting circuit surveys for large comminution circuits and minaalprocessing plants.

INTRODUCTION Circuit surveys are essential in gaining an undmwdm . g of circuit performance. Also, with advances in circuit simulation techniques, modern computer models based on survey data can often be used to reliably extrapolate performance to other ore types and operating conditions. Development of robust simulation models raquireS sound survey data and ore characterization. Surveys define circuit respoase under a set of Laown operatingconditions for a known ore type. This paper describes circuit survey methodology for large plants in operation today. There is little doubt that circuits with capacities in excess of 50,OOO mtlday and producing more than 1oO.OOO m’lday of flotation feed present special sampling challtnges. On the other hand, with application of sound sampling techniques and some degree of ingenuity, these plants can generally be sampled quite successfully. Whik every plant is different, and survey tecbniqucs must be developed for psrticular circumstaaces, the methodology for conducting SOuILd circuit surveys is the same. This paper discusses survey techniques that an?applicable to my size comminution or mineral processing to larger circuits. Mast of thc principles discussed are general in plant. but with specific refnature, and apply to both comminution and minaal pocesSing circuits. All circuit surveys have three general phases: p h m h g and Preparation, sufvcy sampling, and sample and data analyses. Each phase, along with special considedons for comminution d miaaal processing circuits, is discussed in the following sections.

’ A.R. M.cPbason ConsultantsLtd.. Golden, CO ’JKTcch Ry.. Lndoomopilly. QLD.

63

PLANNING AND PREPARATION While the benefits of m b t a n d'ng ~ plant operations and identifying areas for improvement by conducting circuit surveys am large,surveying a plant has a cost. Surveys require significant expenditures of effort in planning and execution, in sample analyses and characterization (either by internal or external laboratories), and in disrupted produdon. Therefore, thorough planning and Preparation are critical to ensure that sound data is coUected at a minimUm cost. Survey planning and preparation are closely related, and di!'fer mainly with regard to focus. Survey planning refers broadly to scheduling the survey, enswing the survey is suited to meet defined objectives, and that appropriate resources are available to conduct the survey successfully. Survey prepafation is more focused on the details of sampling and executing the survey successfully.

-

The first step in planning circuit surveys, whether for a comminution or metallurgical circuit, is ideditsation of the survey objective. Once the survey objective is defined, the survey planning s h d proceed accordingly. Additionally, since conducting surveys and making process modifications are often iterative, past survey data should be examined when planning future surveys. If data from one set of samples were suspect during past surveys, or if one aspect of conducting the survey was particularly problematic, survey practice should be modified to account for the deficiency. In many cases, circuit surveys are conducted in order to provide baseline data used to plan for the processing of future ore types, or to solve problems related to a specific portion of the ore body. To provide the maximum benefit, the plant conditions surveyed should be relevant to the objective. For example, if the survey objective is to provide baseline data for evaluation of a pebble crusher installation, the survey should be collchtctcd while milliag ore that is anticipated to benefit from pebble crushing. Similarly, if the survey is meant to address general plant operations in &r to identify areas for improvement. the ore selected for the survey should not be the hardest on in the pit or the most recalcitrant from a metallurgkal standpoint; ore selected should be representative of typical plant operations. Survey planning should be done in conjunction with other dcpamnents. Joint planning with various departments will minimk conflicts with the survey. 'on with the mine will ensure that appropriate ore is being mined, that blasting and mucking techniques employed are typical,and that the survey does not conflict with si@cmt maintenance events, shovel movements, etc. Mill Maimtumnce. Coordination with maintenance staff is essential to avoid conducting the survey at an inopportune time in the maintenance cycle. There may be less benefit in conducting a survey in the days immedratt * ly before or afta lifter or grate change outs, since these conditions may not be indicative of typical circuit operation. Finally, if a necessary crash stop can be i n e m with scheduled maintenance downtime, production losses associated with the survey will be mitigated. MilrqpeRboas The survey should not conflict with training periods, safety meetings, shift changes. OT any other sisnificant mill event. Calibration of mill instrumentation (flowmeters. weightometers, on-line analyzers, etc.)should be conducted as close as possible to the survey date. For comminution circuits, this is padcularly important for mill feed and recycle belt load cells and hydrocyclone feed mass flow meters. For flotation circuits, on-stream analyzer calibration is Critid. MiUMcbllprgcrrl Term.Metallurgical technicians must be available for sampling, sample preparation, and analyses. Also, laboratory space and logistical assets must be available to handle survey samples. Plan on using 3 to 6 people for a large circuit survey under the supervision of one engineer. Labor requirements for survey preparation arc often undetestimated. Labor requirements will be most intensive during comminution circuit crash stops; au~fulplanning and sufficient labor will minimize circuit downthe.

---

64

Having specified that joint planning is essential to minimize conflicts, there will never be an id& time to conduct a circwt survey. The objective is to minimiZeCoLLfliCts. During the planning process, survey log sheets should be developed. In well-instrumented plants, much of the relevant survey data will be collected and stored by a data collection or control system. Nonetheless, there will likely be additional data that must be manually recorded from the control system, or collected from meters or insbumentation in the field. Log sheets should include both data drat is automatically recorded,as well as field collected data. m e act of using survey log sheets, even for data that are automatically rccordcd will help ensurc that key circuit indicators are monitored during the survey, and can provide early warning of developing problems. Some items, such as consistency of SAG feed size distribution, stockpile condition, and flotation froth stability are difficult to quantify. but can be very hpoxtant in interpreting survey data. Often, photographs taken during a survey can be useful when reviewing data and circuit Operating conditions. Survey planning should include a thmugb review and analysis of which samples to collect during the survey. The analysis should cllsurt that all relevant streams in the circuit can be defined (either by direct sampling, or by mass balaacing otha streams) with the specified sampling plan. Generally, some degree of data redundancy is d e s i i l e in a sampling plan; redundant data can provide useM crosschecks, and can be invaluable during mass balancing. For example, even if cyclone underflow is diluted with a known water addition, the ball mill feed sample provides a useful crosscheck of both particle size distribution and water balsnce data. The desirability of redundant samples must be weighed against the cost of the additional samples, as well as any sampling uncertainties intmduccd by the redundant samples. If the planned redundant samples have a low &pee of confidence (due to difficult sampling points, for example), collecting the redundant samples is probably senseless. If survey sample points arc identikd that require maintenance or cleaning, construction or modificatioo,such work should be completed well in advance of the planned survey. For example, pump discharge sample points in a flotation circuit survey need to be flushed and sample hoses consbnrcted. In many cases,flotation circuit survey samples can be collected at the same points used by cm-stre8m analyses Units. On-Saam analyzer sample points arc typically well-engineered, but do not assume that this is so. Likewise, sample collection devices should be fabricated or ordered. Generally, cutters where the bottom of the coktion vcssel is slightly longer and wider than the top work much better than cutters of a ‘half-moon’ design; tbe shape of the vessel helps minimize sample splashing. An exccucIlt g e d design is presented by Napier-Mum, et, 4. (1996). Additionally, it is better to use cutters with robust constnmion in high velocity streams. Not only will the cutters be mote durable, but the extra mass of a well constructed sampler will mist in consistent sampler motion through heavy, turbulent flows. For sfreams with high flow and contabhg coarse particles (and therefore requiring large sampler apatures), mechanical sampling devices BZC often requkd. For example, SAG mills lugex than 22 feet (6.7 m) that have pebble ports buxme nearly impossibie to sample using handheld sample devices. For such streams mechanical sampling devices arc the only practical way to sample the stream. In a few cases, a short duration diversion (thereby providing a snapshot sample) of such streams (to a large collection vessel) can be morc practical than attempting to sample them. Alternately, of course, such a stream can be calculated (assuming sufficient data about other streams). In some cases, circuit operation can be simplified for the survey. Far example, if the focus of a survey is comminution circuit perfomancc, and a gravity concentration circuit on cyclone undeflow greatly complicates the circuit v l h g plan, the gravity &uit ~ 8 be n bypassed during the survey. Such a simpliiication has little effect on the comminution performance, but may greatly simplify collection of !Wund semples. Alternatively, the scope of some surveys can be expanded. For example, if the focus of a survey campaign is flotation circuit performance, a grinding survey could be conducted in c o n j d o n with the flotation survey at the s ~ m etime. A complete circuit data set can then be collected with relatively little extra effort in planning, coordination, and execution.

65

Sumey Prepmtion

Survey preparation generally deals with rehearsing and implementing the overall survey plan: practicing sampling technique. practicing donning and using required safety equipment, and acquiring and marking sample containers. particularly with the large volume flows in single-line comminution and flotation circuits, sound sampling technique will gnxtly enhance the quality of data collected. In addition to general sampling procedures (for a succinct review of sampling procedures as applied to mined processing circuits, see Kelly & Spotiswood, 1989). the following guidelines are particularly important: 1. Entire stream width should be sampled. For samples beiig cut from weirs, samples should be cut across the entire stream from left to right, and include the entire depth of

flow. Sample cutters should allow cutting the entire stream once without overflowing; if this cannot be done, the stream should be divided into sectors for individual sample passes. 3. Total volume sampled should be consistent with sample top size.

2.

In genaal, sample points are p f d in the following ordet: sample collected across the entire sa-eam, thief samples collected from a ”perfect-mixing”zone (sufficient agitation to ensure a homogenous mixture), and sampling h m a tee f i h g at a ninety &p angle. Sample collected from a valve perpendicular to the slurry flow, or “dipcup” sampks of streams with coarse particles are the least reliable, and should be avoided except as a last resort. When sampling pipes, the smallest inner dimension of the valve should be at least six times the diameter of the largest particle to be sampled. Particularly for the largest plants, custom designed samplers can facilitate collection of representative samples. Examples include samplers of mechanical cross chute samplers, or custom “dip” samplers using overhead cranes for large comb& sumps. pahaps the most difficult comminution circuit survey sample to collect is SAG mill circuit product (screen or trommel uadersize). Collecting this sample is complicated by the fact that slurry flowsinvolved contain colvse particles in turbulent, high-volume streams. In some circuits, acceptable samples can be collected fromspecific points, but segregation in this sample stream is often issue. Direct s a m p l i of sumps, as opposed to sampling the streams that feed the sumps, almost always has a sampling bias associated with segregation and flow patterns in the sumps. Design of sample points and samplers for such streams should include a careful analysis of potential bias. UnforNnately, while there is no ‘’best p r a c t i c e ” for sampling SAG circuit product, acceptable samples can be collected from most circuits. Samples collected should always be carefully compared to mass balance results. Good c m in sampling the ball mill circuit will allow much more confidence in calculatiag the SAG circuit product. often, with repeated sampling of the same circuits,patterns of sample bias will emerge. and these can be used to judge the validity of a set of samples. Survey preparation should include rethe collection of every sample. This Serves that all survey personnel have the several purposes. First, the sampling rehearsal eappropriate sampling equipment, containers, and safety equipment in position for the survey. Any neceSSBty equipment to clear sample points that fncluently plug should be on hand as well. Secondly, the rehearsal serves to train and practice sound sampling technique. It is often prudent to use retaining &vices for samplers,particufarly for high volume streams. sample “leashes” can prevent loss of cutters and other sampling &vices. If retaining cords are used, it should be obvious that the sampling device should not be tehred directly to the person sampling, and that the retention cord should be m e d in such a fashion that deployment of the cord to retain a dropped sample device will not entangle the operator. Just before the survey (preferably on the day of the survey), a final rehearsal should be conducted. This final rehearsal should include a review of prooadures to be used during the survey, and should include optrationalstaff. survey staff, and my suppwt staff involved with the survey. The rehearsal should include the following: general survey sequence, designated

66

responsibility for specific tasks during the survey, actions upon shutdown, lock-out procedures, checking for appmpriate sampling and safety equipment, and a review of safety procedures and actions to take should a safety incident occur. Any modifications to typical plant Operating procedure should also be reviewed. Such modification may include rUnniag the pebble crusher full time (instead of ordoff), shutting of sump return water (which may not be measured), or diversion of certain parts of the circuit not relevant to the survey. Support personnel that have duties associated with the survey should be included in the final rehearsal. Such personnel will be different at each site, but may include electricians (for mill crash stop lock-outs) and safety staff (for confined space atmosphae checks and pn>ceduns). As a finalcheck prior to the survey, make sure that no forseeable circuit upsets are pending, that the circuit is in steady-state (by both inmumentation and control indications and operator consensus), and that the circuit is operating in a condition collsistcIlt with the survey objectives. For example, in a flotation circuit survey, it is often useful to check the recent history of selected streams that indicate the steady-state nature of the circuit. In most cases, the cleaner tail and scavenger concentrate assays provide a good indication of flotationcircuit stability. Similarly, in a &ding circuit survey, good indicators of steady-state include SAG mill power (the measwed control indicator of SAG mill load) and cyclone fed pressures (a direct result of cyclone feed volume). Finally,give any Operations, maintenance, or mining staff who are not directly involved with the survey a last reminder that the survey is b e i i conducted. After survey staff are in position, enthat there is communicationwith all survey mcmbcrs (either visually, or with radios).

SURVEY SAMPLING A circuit survey must define the response of an ore type at a known set of operating conditions. Most details concerning equipment, such as Size and configuration,are known before the survey. operatiag variables, such as cyclone Operating presswe, mill speeds, power draw, flotation cell airflow, and reagent additions typically operate within a certain range, but actual values must be measured or rtcorded during the survey. During a circuit survey, the engineer in charge must enme that survey samples are collected, that opaating conditions and circuit response are recorded, and they must direct any necessary changes to circuit operation. If crash stops are included in the survey phn, the engineer in charge must orchestme measurement of mill loads, recording of equipment data, and sample collection during the shutdown period.

sprpeys.ppks

A circuit survey is typically most representative Of stcady-state Opaation when conducted over a time interval during which several sample cuts are collected. The time period and number of samples cuts comprising a survey composite is not fixed. At one end of the extreme are shift composite samples. Such a sampk period is often reasonable for Circuits with long residence times, such as leach circuits. For more dynamic circuits such as comminution and flotation depiction of average circuit circuits, however, such a long sample period may rtsult in an ~ccura&e performaace. but it may not repmen&Circuit performance at any given instance. At the other end of the emme are ‘snapsh’ samples that represent instantaneous circuit performance. Though snapshot samples can be useful, particularly for unit operations with essentially zero residence time such as hydrocyclones, there are two drawbacks in using them for circuit surveys. First,the Circuit may not be in equilibrium, so samples collected may not be representative of equilibrium c~aditionsd h g Operating conditions. Secondly, by collecting only one sampie set, the risk of increased sampling ermr is introduced (collecting multipk samples over a defined period can reduce random sampling error, but any systemic sampling bias will be unaf€ected). Depending on the survey objective. however. shift samples, instantaneous samples, and composite s w e y samples all have valid roles. In selecting appmpriate sampling periods, things to c o n s i b are the total circuit residence time, thc time it takes for changes in circuit operation to become evident, and the time for which there is reasonable assumwe of maintaining s W y state

*

67

circuit operation (without such things as chrrnghg are types, mill feed size, shift changes, or periodic events in the mill or concentram that change recycle loads). Generally, comminution circuit periods of one-hour are reasonable. Flourion circuit surveys times are a function of the total circuit residence time.Generally, a survey period that allows two or more turnovers of the circuit are required, with slightly longer survey periods preferred. If the required survey period is still uocemin, it may be desirable to change an operating parameter and watch to see how long the circuit takes to reach steady state. Examples of such changes include changing feed rate to maintain a higher SAG mill power draw,or increasing water flow to the hydrocyclone feed sump. The survey period should be mughly dvee times longer than the period required to establish steady state. A survey period using an even number of interval samples is more useful than a period with an odd number of cuts (i.e., four samples instead of three). Such a method allows equal weighting of samples collected in an dtemate fashion (such as sampling e v q other cyclone in a cluster). It is not mxswy that all sampling points in a circuit be sampled at precisely the same time in a circuit survey, but it makes control of the survey eesier and helps maintain communication between survey p e m ~ l . Resspdless of the circuit survey duration, some samples, either by nature of their top size, mass, or accessibility, arc typically collected only OMX duting the survey, most commonly during a shutdown period at the end of the survey. In a comminution circuit survey, such samples typically include SAG mill feed, pebble crusher feed, and crusher product. Panicularly for SAG mill feed and pebble crusher feed samples, the particle top size and mass of these streams makes sampling by any mechanism other than belt cuts difficult. Tbe SAG mill feed belt sample should be long enough to cover any cyclical belt loading, and equivalent to the distance (visually estimated) between the warsat rocks on the belt. If coarse rocks are present, but infrequent, include the wanest rock on the belt in the sample and make a note ofthe nominal distance between the coacsest rocks. For all belt samples, measure the sampled distance precisely (going from center to center on rollers is expedient and precise), and include fines in the samples (using dustpans and brooms). A modifkd method allows better definition of the coarse end of the panicle size distribution with less sample than a collecting a complete belt cut over tbe entire sampled belt length. First.a reasoMbly sized belt cut (2 to 5 m) is taken. This sample &fines the size distribution of finer material, and the mass split between coarse and fine material. Secondly, larger particles arc collected o v a a longer stretch ofbelt (at least long enough to include a complete cycle, if the belt has variable loading, o f h w.w long enough to include several of the coarsest partick on the belt). 'Ibis sample &hes tbe size distribution of coarse material. A good rule of thumb for the size split between coarse and fine is plus and minus 75 mm (Napier-Mum 1996). Pebble crusher feed can be sampled by either belt cut, or in some cases, by means of a diverter chute. lhe mass of the pebble crusher feed is actually morc important than the size distribution (as the top and bottom size are bounded by the grates and screen size, respectively). Depending 011 plant layout, pebble crusher product can be collected by belt cut, or at the end of the return belt. In some cases, the mane ore reclaim feeders must be shut off prior to the shutdown to facilitate collection of pebble crusher product on a clean belt. The mass of belt cut samples can often be used as a crosscheck against belt scales. A study of relevant sampling tkozy will generally indicate that samples collected from streams with coarse particles must be quite large to be statistically representative. Collecting and processing exmrnely large sample masses can be quite costly. In general, the sample mass required for sound sampling is typically only an issue with SAG Circuit samples; for most other streams in aplant, typical sample sizes will be more than adequate. Appopnate sample sizes as a function of particle top size are addressed in the sample and Data Analysis section. o p m t b l g c33a&*as There are two broad categories of data concerning Oprmting conditions during a survey. One set of data is collected from process instrumentation, and the other from direct measurements. Generally, much of the relevant survey data can be collected from plant insmentation.

68

With advances in instrumentation and control, survey data m often available digitally, at frequent time intervals, and inclusive of tnnds. With such systems, survey data can generally be readily recalled from data archives. When using this data, it is essential to understand the origin of each data point. Some data will be measlned directly. some will be calculated from other data points, and some data may have arbitrary ‘‘correction” factors applied. The use of control system data in detailed data evaluation and mass batanchg can be chaIknging if the underlying data source is not well understood. For data measured directly, knowing when the measuring instnunent was last calibrated, and the likely bias in measurement (if applicable) is essential. A second set of data must be measured. For comminution circuits, these data include items such as the operating dimensions (noting the difference firom new dimensions) of cyclone spigots and vortex fiders, pebble crusher closed side set, mill ball charge volumes, and total mill load volumes. For flotation chuits, these data include pulp levels, bnpller speeds, air hold-up percentage, cell gas velocities, reagent smngths, etc. While ball charge volumes and total mill load volumes are often known within a range, rarely does plant practice and control allow precise estimation of these parameters. Even if such data are reasonably well known, it is useful to occasionally verify models or the indirect measurements that are used to estimate them.Comminution circuit surveys offer an ideal opportunity to accurately measure ball charge volumes and total mill load volumes, evaluate sluny void filling, gme and liner condition, and allow estimation of the composition of the mill load (either size distribution, rock type, or both). In some cases, depcndhg on m e y interval and grate and liner condition, the crash stop may be used to measure key mill in& dimensions. Depending on the nearest scheduled maintenance downtime, checks of mill load parametas during these periods may be sufficient. Mill load data are best evaluated by crash stopping the circuit. Until the crash stop, the operator should continue Operating the mill as it was operated during the survey. SAG mill feed wafer should be shut offjust before the crash stop (long enough fot the water to stop-visuaily test this how long this takes in sampling rehearsals). During the crash stop, the SAG mill total load volume is measured. Photographs of the mill load may also be useful for evaluating the size distribution of the mill load (although it should be realizad that,in such endeavors, there is likely to be a high degree of segregation in the mill load). As neccsmy, test pits can be dug to obtain samples usually at the various points in the mill. In addition to the size analysis of the SAG mill charge samples (which when compared to that of the SAG mill fdcan be indicative of the build-up of critical sizes), examination for lithology and mineral alteration in harder rock coapnents can sometimes point to the source of critical size. After this is accomplished, the SAG mill can be ground out to check the ball charge volume. Although each operation has prefances for conducting grind outs, variable speed drives generally allow complete grind-outs with a low risk of liner damage. Ball mills can continue to grind out during the SAG mill load measurements. After mill grind-outs, ball charge volumes can be checked, and if necessary, mill internal measurements can be vaified. Mill entry after crash stops d grind-outs can be hazardous. Visibility is often limited after crash stops due to watw vapor. Mill roofs should be inspected for the potential of falling balls, mud, and rock. After a grind ouk mill contents are often hot enough to cause balls to spall due to thermal expansion and contraction. Hosing the mill roof and sides with a high-pressure jet can remove recalcitrant lifter and liner packing, and p K h g sheets of plywood over the mill or ball charge can offer some protection from ball spalling. Alternately, after the ball load is checked, ore can be fed to the mill to “bed” the charge before entering the mill to take measurements. The benefit of data collected during crash stops and grind-outs cannot be overemphasized. In addition to the benefits of direct evaluation of mill loads, detaikd power data (in conjunction with mill opaatiag conditions) during the crash stop and grind-out can often be used diagnostically to evaluate slurry pooling. Despite the importaace of crash stops and grindouts, they need not be excessively onerous. For example, c o m p k entry into the mill is typically not required to check the total mill load. For those operations with good maintenance recards and understanding of grate and lifter configurations and w w paaerns, detailed measurements of mill intemals do not need to be conducted frequently. A well rehearsed crash stop to check total mill load volume can be done

69

in l a thaa 10 minutes, including lockout-tagout procedure^. A t h m g h m h stop to check total mill load volume, followed by a grind-out to check ball laad volume and measure mill intemals can be accomplished in less than an hour by a trained crew.

Survey Coor
SAMPLE AND DATA ANALYSES lmntedindy after the survey has been conducted, all samples shoukl be c o v d and moved to a consolidated location. After consolidation, samples can be moved (or shipped) for sample preparation and analyses. For most survey samples, the mass collected will be larger than that required for particle size or chemical analysis. Ohen, collection of larger COmpoGite samples (either as a result of larger or more frequent cuts) can minimize sampling variability. Once collected, however, the samples must be analyzed. There are two philosophies for handling samples collected during surveys: split the sample down to a manageable size prior to particle size and chemical analysis, or process the entire sample. Some take the latter approach, believing drat they are being most thorough. This approach usually backfires. If the entire sample is sized, either the Scrcen decks will be grossly overloaded, or multiple sizings must be done. Conducting multiple sizings takes time,and introduces the potential for confusion and errors, particularly with multiple tare weights. A better approach is to split the samples Using sound sampliag practice. Samples can be split wet, or the entire sample dried, then split. Wet splitting takes less time and requires less sample handling. but requires conderable experience to do well, particularly with samples containing coarse material. Drying the entire sample followed by dry blending and splitting is more cumbersome, but in conjunction with mechanical splitters (either riffle or rotary), is more c e h n to d t in representative subsamples. Based on experience aad balancing sample theory with practical considerations, minimum mass for particle size analysis of survey samples as a function of top size are summarized in Table 1. Because of the difficulty of sampling coarse material, the entire sample should be sized in the case of samples containing a top size coarser than 100 mm. Collection of a 500 kg sample is generally acceptable for most SAG mill feed samples. Samples can be split to smaller mass once the sample has been screened down to finer SQttll sizes. Samples with a top size of less than 12 mm should generally be wet screened (typically at 37 pm) prior to screening at coarser sizes; depending on the ore characteristics, wet scxeening at coarser sizes may be necessary. Comminution circuit samples should be scrttlltd using a complete A series; the same screen series should be used for all samples witbin each circuit. Standard laboratory screens (2OOmm diameter) are convenient for sizings, with punch plates or steel rod grids being convenient for samples with the coarsest top sizes (above 50 mm). Data analysis has several sequential stages. The fmt stage occu~son the plant floor during the survey. If process upsets OCCUT during a survey paid or if samples are collected incomctly, the samples very likely will not be representative, and should be discar&d. It is more timeefficient to discard obviously bad samples than to process the samples and then attempt to make sense of them.

70

Ta > 100mm 100mm-1212 mm - 0.420 0.420 - 0.150 < 0.150 um

N/A Fract. Shoveling

Entire Sample 50 kg

Cone & Quarter Riffle RoraryorRiffle

5-10 kg 1 kl3

RotarvorRilBe

500 e

The second stage of data analysis occurs after the samples have been analyzed. This stage is based on a common sense appraisal of the data; such an evaluation can provide a usefui insight into the validity of the data. For example, in corDminution circuit survey samples, the cyclone feed size didbution must lie between the underflow and overflow size distribution. Where possible, this comparison should be made on a size by size basis (as opposed to cumulative or bulk assays). Similarly, in flotation circuit survey samples, the assays of a flotation feed must lie between the concentrate and tailing values. Mass balance programs and calculations, both commercial and proprietary, are often used for a third stage of data analysis. Mass balancing packages are powerful tools in data analysis, but are dso very dangerous in the hands of an untrained user. Mass balaacingpackages can be extremely otha sound data).Improperly useful for identifying bad data, or for calculating missing data (h used, mass balancing can also be used quite successfullyto convert a data set that is 90% sound into a data set that is almost impogsible to interpret. In mass balancing, it is most important to be able to identify bad data than to attempt to minimhe regression errors across the entire data set. In mass balancing flotation circuit survey data,the following guidelinesare useful: Obtain the average flotation feed flowrate from the comminution circuit product mass flow readings. 'Ibis value is typically quite sccurate, and will be fixed during mass balancing. If the comminution circuit has more than one smam feeding the flotation circuit (for example, flash flotation concentrate or gravity concentrate), then a balance of the comminution circuit is required first to estimate the flotation feed flowrate. Check the overall assay data by ensuring that for each bank tbe feed assay is in between the concentrate and tail assay. If there is a discrepancy in these values (incorrect assays or sampie mislabeling), this must be resolved before continuing. Cooduct a simple water balance over the entire circuit using flotation feed, final concentrate and final tailings sample solidspercent and any water addition flows into the circuit. Using the main element of a site or most reliable statistical element (for example, copper in a copper c(Mccntrator, nickel in a oickel concentrator, iron or sulfur in platinum or gold sulfide cxmccnmuor), mass balance the flows in tbe Circuit. 'his method should p v i d e a good indication of the flows in the circuit; somc plants are difficult to mass balance using one element, however, but they may balance well with multiple elements. A good example of such a situation is a IeacI-ZiaC conccntratoc balancing using lead in the lead circuit, and zinc in the zinc circuit is a much more rational approach than a single element balance. After assessing the overall mass balance with a single elanent, introduce other elements sequentially. With approPriate assessment of the likely confidence level for each element in each sfream, balance the whole flotation circuitFor example, the confidence level for the copper assay in a final concentrate is likely much higher than a gangue element in a recycle stream. In some cases,W i g each bank first often provides a good estimate upon which to base the full flotation c M t mass balance. It can be useful to mass balance minor elements (for example, gold, silver, platinum, arsenic, etc) after the main elements have been balanced.

71

An assay by size mass balance should be conducted using the overall mass balance as the basis. As with difficult overall assay balances, mass balaocing each bank separately on a size and assay basis can lead to an optimal solution.

SPECIFIC CONSIDERATIONS Some additional notes are in order on insmunentatiOn. Many plants do not have flowmeters installad at all water addition points; relative to measuring slurry flows, flowmeters for water are very accurate and inexpensive. An accurate water balance can be quite useful for circuit surveys, and aaaining accuracy in the water balance can genaally be done at little expense. In this regard, all relevant control valves should be checked for correct action. maintenance, or replacement. If a plant is not sufficiently instrumented. this deficiency should be identified and rectified during the survey planning stage. Many circuits are now very well instrumented. Such instrumentstion clearly helps the survey team to understand what is occurring in a ckuit, but do not assume without verification that instrumentation data is accurate. Conducting surveys provides ao easy opportunity to verify that samples coUected by on-stmm units are representative; whik most system are well engineered, occasionally significant bias is u n c o v d during surveys. Also. wbile on-strcam analyzer data are generally good for the main mineral, data for other minerals are oftea of lower quality. On-stream size data are generally quite pod for relative shifts, but may not comspond well with sieve data fromthelaboratory. 'Lbe focal point of a pod comminution circuit survey should be the cyclone-ball mill circuit. With relatively little preparation outside of good sampling tools and training, excellent samples can be collected around the cyclone and ball mill circuits. With solid data for these unit operations, the SAG circuit product can be reliably calculated, regardkss of sampling difficulties with that stream. The ability to have a calculated SAG circuit product with a high degree of confideace is invaluable, at a minimum to cross check the physical sampIe collected, and also (quite possibly) because a Feprcscntative sampk cannot be collected. Along with sound data around the cyclone-ball mill circuit, a good circuit water balance and the mass of the SAG mill recycle load are the u n d c r p i i of a good comminution circuit m y . d similarly, the focal point of a good flotation survey sho~klbe the cleaner circuit. ~ o o data on the cleaner tail stre8ms in the flotation circuit should provide the survey engineer a reliable basis to estimate recircuhbg loads in the mass balance. High variations in these streams are most h l y caused by flucatationS in the re-ckulating loads in the cleaaer circuit. Variations can generally be easily i n f d by monitoring solids density or recycle mass. Alternately, the combined cleaner tailing assay from the on-stream analyzer is generally a good indicator of stability. Of course, a good cleaner circuit survey also requires stability in the upstream processes including the comminution circuit. Because of this linkage, conducting surveys of the comminution circuit conjunction with the flotation surveys allows tbe survey engineerthe ability to assess the overall stabiity of the plant. This also provides valuable data connecting the two -P To check survey variability, or to assess the ability to discern vay small changes in circuit performance, it can be useful to assess sample variability. Kbeping each sample cut separate when colkcted allows each cut to be analyzed scpratcly. Assuming sound sample techniques and no gross errors, the sample variability between cuts is indicative of the variability of the overall survey. Of course, the overall survey cannot have kss variability than that observed between sample cuts. Typically, this procedure would only be done for key survey streams. Conducting replicate sample analyses adds very little incremental cost to a survey, yet can greatly facilitate estimating the confidence interval for critical assays. 'Ihe results of such a study can also be used in conjunction with sensitivity analyses of mass balance data.

72

CONCLUSIONS conducting surveys is almost always an iterative proass. At the completion of each survey, review the survey pn>cadures, and make notes to improve the process the next time. Regardless of the survey objective, a by-product of the Circuit survey process is a better knowledge of plant operation and maintenance status, and a more thorough u m h m d m g of the overall linkage bemen mining, mineral ptocessing, and extractive metallurgicalfunctions in a plant. The importance of regular surveys of both the comminutim and flotation circuits cannot be u n d e w . Building a dabbase of survey knowledge provides Site personnel a fum basis for future decisions or analyses concerning circuit improvements. Repetitive surveys also ensure the skills aaained from planning, conducting and analyzing survey data are maintained at a high level.

Kelly, E.G.,and Spottiswood, DJ. 1989. Inaoduction to Mineral Processing. Mineral Engineering savices. Kalgoorlie.

Napier-Munn. T.J.,et al. 1996. Mineral Comminution Circuits: Their Operation and Optimization. Julius Kniaschnitt M i Research Centre. Indooroopifly, Queensland. Seidel, D.C. 1995. Laboratory Rocaduns for H y ~ t a U u r g i d - R o c e s s i n gand WasteManagement Experiments. Informatioa Circular 9431. United States Department of the Interior, Bureau of Mines.Washington, D.C.

73

APPENDIX 1: SURVEY EQUIPMENT 1.1 Geatrrl Survey Eqpipment

Persorrrl Safety Equipment Gloves Dust Masks / Respirator Eye protection Safety harnesses RccordhgTods Pen, paper, and survey logs

Calculator

Camera

Flashlight

Hearing Protection Hard hat Rubber boots & Raingear Lockout Locks

1.2 Gripding Survey S p t d k 4ui-t

!3amphgdwices Cutters: Hydrocyclone samples, ball mill discharge samples Custom Samplers: Screen UnQcrsiZcsamples Shovels, brooms, dustpans: Belt samples M-tdevirrS

Tape Measure

Tachometer (for belt speed) Long Level (for measuring from mnnion) spmpkcoat.iwrs Barrels for belt samples Buckets (with lids) for slurry samples Barrel markers (paint markers) Forklift sIJetyEq.ipwnt Amospbae sniffer (for confind space entry) Pump (water removal after mill roof cleaning)

Stopwatch Scales M m y Cup

Bags (dried sample splits) Metal marking tags Permanent markers pallets

High p r e ~ ~ uhose r t (mill roof cleaning) Ladder (SAG mill entry)

1.2: Flobaoa * Survey Speciac Equipment

snmphgdtvices Pump Hoses: Selected concentrate and tailings samples Cutters: Cleaner feed / Recleaner samples, On-sueam samples Dip Samplm: Selected concentrate samples Dip Samplers: Selected tailings samples MensrurmclPtdevices Tape Measure Stopwatch Tachometer (for impeller speed) Scales Pulp Level Stick (for measuring from froth depth) Marcy Cup Superficial gas velocity probe optional) Air hold-up Robe (optional) RccordingTods CalCUlator Pen, pape, and survey logs Camera Flashlight spmpkconbiptrs Metal marking tags Bags (dried sample splits) Buckets (with lids) for slurry samples Permanent markers

74

APPENDIX 2: SURVEY PREPARATION After the survey plan has been coonhated with the mine plan, mill operations and maintenance, and labor schedule considerations, final details can be taken care of during final survey preparation. 2 1 Grindirag Surveys

Not Later " I One Day prior to the SIlrrrey 0 Check all sampling ports, access hatches and panels, etc. 0 Practice sampling 0 Check to make sure all supplies are on hand 0 Tare and mark survey sample containers 0 Final preparation of survey logs Day o f h e y 0 Review survey procedure and objectives with mill control aperator 0 Review DCS data to ensure that the circuit is at steady state, and operating under the desired conditions for the survey objective 0 Distribute sample containers 0 Rehearse mill crash and lockout/tagout procedures 0 Check all sample ports 0 Timebelts 0 Mark belts for crusher feed and product sample points (SAG feed sample point will be detcxmined visually after the crash stop) Checkmillspeeds 0 0 Sketch screen panel configuration 0 Slug pebble crusher 0 Final review of survey procedures with all staff 0 Wety briefing with all staff

N o t h m T h n O n t h y Prior tothe Sprpey 0 Check all sampling ports. accesshatches and panels. gratin%s ~ t c . 0 Practicesampling 0 Check to make sure all supplies are on hand 0 Tare and mark survey sample containers 0 Final preparation of survey logs Day d s o r v t y 0 Review survey procedure and objectives with mill control operator 0 Review DCS data to ensure that the circuit is at steady state, and operating under the desired conditions for the survey objective 0 Distribute sample containers Measure pulp levels and compare to DCS data 0 Checkallsampleports Measureimpellerspeeds 0 Measure air rates with superficial gas velocity probe (comparewith DCS) 0 Check mill operating conditions (ata minimum for the regrind circuit) M e a ~ ~ air r e hold-up (optional) 0 Final review of survey procedures with all staff 0 Safety briefing with all staff

75

APPENDIX 3.1: GRINDING CIRCUIT SAMPLING P U N (EXAMPLE)

Flotation Feed

Launderoron Stream Analyser cyclone overflow Combinedslream at each nest Cyclone Underflow At.every other cyclone cyclone Feed Blanlred cyclone Ball Mill Discharge Dischargehousing SAG Screen 4 Samples, UR & Undersize FIR ofScreen CnrsherFeed Feed Belt

Sm. Cutter

Buckets

4 x (15 min. int)

Sm.Cutter

Buckets

4 x (15 min. int)

Lg. Cutter

Buckets

4 x (15 min. int)

Buckets Lg.Cutter Box Cutter

Buckets Buckets

4 x (15 min. int) 4 x (15 min. int) 4 x (15 min. int)

Shovel,

h

1 x (belt cut)

h

1 x (belt cut)

Drums

1 x (belt cut)

Buckets

broom Crusher Discharge

product Belt

SAG Mill Faed

Feed Belt

Shovel, broom Shovel,

broom

APPENDIX 3.2 EXAMPLE FLOTATION CIRCUIT SAMPLING PLAN

/ Pdnt

FlotationFeed

on Stream

Sm. Cutter

Buckets

SampleHose

Buckets

Sm. Cutter

Buckets

I Sm.Cutter 1 Lipsampler

I Dip Sampler

Buckets BucLets Buckets

4 x (30min. int) 4 x (30 min. int) 4 x (30 min. int)

Sample Hose

Buckets

4 x (30 min. int)

Sm. Cutter

Buckets

4 x (30 min. int)

Sm. Cutter

Buckets

4 x (30 min. int)

Dip Sampler

Buckets

4 x (30 min. int)

AnalYsCr

Roughs Concentrate Rougber Tail

Scavenger Concentrate Scavenger Tail

RMP

on Sonam AMlySa cleaner Feed Box

CleanerFeed CleanerConcentrate CellLaunder E M of Cleaner Cleaner Tail Bank Pump CleanerScavenger concentrate Cleaner Scavenger On Stream Tail M p X FinalConcentrate Ons-

I

AnalySer

Recleaner Tail

End of Recleaner

76

Practical and Theoretical Difficulties When Sampling Gold Francis F. Pitard’

ABSTRACT Everyone is aware of problems created by a bias when using non-optimized sampling protocols for precious metals, but poor precision may be as bad. Poor precision is responsible for grade reconciliation problems between the geologist, the mine, and the mill. These difficulties are minimized when sampling protocols are implemented that are based on solid theoretical foundations. Recent literature suggests difficulties calculating the variance of the Fundamental Error using Gy’s Sampling Theory. There is a mineral grinding explanation, which is addressed in the Sampling Theory. Another problem is the poor practice of sampling in gold mills around the world, This paper offers clarifications, solutions, and a reference for managers of gold projects. INTRODUCTION Identification of the Problem When referring to Dr. Pierre Gy’s work, users of the Sampling Theory generally refer to a simplistic formula to calculate the variance of the Fundamental Error in a sampling protocol. This formula has its limitations and is based on two important assumptions. The first assumption states that the content of interest, say gold, varies much more from one density fraction to another than from one size fraction to another. This assumption is almost always true. The second assumption states that the proportion of a given size fraction within a given density fraction does not vary in a significant way from one density fraction to another. This assumption may not be correct when the constituent of interest does not grind as fast as the other constituents. This paper presents Dr. Pierre Gy’s work in greater depth, and suggests new versions of the important Heterogeneity Test. This will improve the reader’s ability to respond to a delayed comminution problem, and to investigate the contribution of other errors that may occur in addition to the Fundamental Error issue, especially the Grouping and Segregation Error, the Delimitation Error, the Extraction Error, the Preparation Errors, and the Analytical Error. Disregarding these other errors will result in poor sampling practices at the mill, and will make it very difficult to obtain accurate metallurgical balances that are required to optimize the process. A Solution in Line with the Sampling Theory This document intends to show that the original Sampling Theory as presented by Dr. Pierre Gy can solve any sampling problem, such as sampling of gold, soft minerals, and very hard minerals that do not comminute at the same rate as their surrounding matrix. In this paper, a carefbl synthesis of Gy’s important and old publications is made (Gy 1956, Gy 1967, Gy 1979, Gy 1983). This paper concentrates on gold, but similar problems can be encountered with other precious metals, molybdenite, galena, chromite, etc.

’ Francis Pitard Sampling Consultants, LLC, Broomfield, Colorado, USA

77

SAMPLING PROTOCOL AND SAMPLING ERRORS The sampling protocol is a set of parameters that defines and quantifies the comminution and subsampling steps that culminate in the assaying process. Because the final assay aliquot is a very small fraction of the original lot it is to represent, it is subject to sampling errors. The assay is never “true”. Sampling errors are the result of either Constitution Heterogeneity (e.g., differences in gold content between individual rock fragments) or Distribution Heterogeneity (e.g., differences in gold content between groups of fragments), as shown by the Sampling Theory (Gy 1979, Gy 1983, Gy 1992, Pitard 1993). A thorough mineralogical study of gold particles and their associations is essential to constructing a sampling protocol. The following characteristics of gold mineralization should be known:

0

How many types of gold mineralization have been identified in the deposit? What is the largest size of the gold particles? Are gold particles clustering? Are some areas showing coarse gold with no fine gold around? Is visible gold associated with quartz veins? Is gold showing enrichment along a geological contact, or a major quartz vein contact? Is gold associated with another major mineral (e.g., sulfides such as pyrite or arsenopyrite, pyrite, spinels such as chromite)? Is gold alloyed with another metal?

One class of sampling errors occurs as the result of the choice of parameters for the sampling protocol, as outlined below: The Fundamental Error. This error is the direct effect of small-scale Constitution Heterogeneity caused by differences between individual fiagments in the immediate neighborhood of where a sample is collected. In addition, a composite sample may reflect differences between fragments collected far apart in time or space. In such a case, the Fundamental Error would also be affected by what is called the large-scale Constitution Heterogeneity. The variance of the Fundamental Error is a function of sample weight, top fragment size, fragment size distribution, fragment shape, fragment density, mineral of interest density, mineral of interest content, and mineral of interest liberation level. Careful tests are necessary to predict and minimize the variance of the Fundamental Error. Failure to minimize the variance of the Fundamental Error would make good sampling at the mill impossible, regardless of how suitable the sampling equipment might be. The Grouping and Segregation Error. This error is the direct effect of small-scale Distribution Heterogeneity caused by differences between groups of fragments in the immediate neighborhood of where the sample is collected. Therefore, the Grouping and Segregation Error is a function of the local Fundamental Error, the amount of local segregation, and the number of increments collected in that neighborhood. Again, a composite sample may reflect differences between groups of fragments collected widely apart in time or space. In such a case, the Grouping and Segregation Error would also be affected by what is called the large-scale Distribution Heterogeneity. Often, samples taken at different places are sub-sampled and analyzed separately to assess local trends over distance, tons, or time. In order to overcome the large-scale Distribution Heterogeneity, it is necessary to select an appropriate sampling interval (called the basic “stratum” of observation), and a sampling mode, either strict random (i.e., taking all the increments in the middle of each stratum), random systematic (i.e., taking the first increment at random within the first stratum, then taking all subsequent increments at the Same place in the subsequent strata), or stratified random (i.e., taking each increment at random within each stratum).

78

The large-scale Distribution Heterogeneity may be periodic in character, which renders a systematic sampling mode hazardous. In such cases, the stratified random mode is by far the safest approach. For example, a 12-minute cycle of the gold content generated by the discharge of a SAG or Ball Mill in phase with a 12-minute sampling interval would have devastating effects on the calculation of a material balance. A second class of sampling errors occurs during the practical implementation of the sampling protocol when sampling devices are not properly designed, built, used, maintained, and cleaned. These errors are: The Delimitation Error. The statistically correct selection of one increment to be collected as a sample requires that all parts of the lot have exactly the same chance of being selected. Therefore, the geometric boundary of an increment must coincide with an isotropic module of observation. Modules are spherical for a three-dimensional object (e.g., a stockpile), cylindrical for a two-dimensional object (e.g., a mining bench), and a complete and uniform slice for a onedimensional object (e.g., a flowing stream). As a result of omnipresent gravity generating transient segregation, any boundary deviation from the ideal module generates a non-constant bias. For example, it is not rare to find that a blasthole sample represents half of an active mine bench and part of the next bench, generating total havoc in ore grade control. This error can become significant at the mill unless correct cross-stream samplers are used. Therefore this problem will be carefully addressed in a later section of this paper. The Extraction Error. As the sampling tool contacts the material to be sampled, all material inside the isotropic module of observation, or every fragment with its center of gravity inside that module, must be perfectly recovered. Any deviation &om that behavior generally results in an incremental recovery loss. Sampling devices that are poorly designed, built, used, maintained, and cleaned result in a deviation from the rule of center of gravity and generate a non-constant bias (e.g., diamond core recovery loss, an auger systematicallyrejecting coarse fragments, too narrow a cutter for a cross-stream sampler). The Preparation Errors. Preparation processes applied to the increment or to the sample, such as crushing, grinding, pulverizing, drying, screening, and packaging can introduce a loss, contamination, or an alteration of the physical or chemical integrity of the sample (e.g., the cyclone of a reverse circulation drilling machine while eliminating most of the fines from the sample can introduce a large bias). The Weighting Error. Increments must be proportional to the stream flow rate for a onedimensional stream, and to the thickness of a two-dimensional object. Any substantial deviation from proportionality can introduce a bias. The prime objective of this paper is to discuss the important sources of gold sampling problems resulting from the Fundamental Error, the Grouping and Segregation Error, the Delimitation Error, the Extraction Error, and the Preparation Errors. A large section of the paper is devoted to the Fundamental Error, which must be clearly understood and its effects minimized before the other errors can be addressed and corrected.

THE FUNDAMENTAL ERROR The Fundamental Error (FE) is the direct effect of the Constitution Heterogeneity of a lot (CHL), which is expressed as follows:

79

where NF is the total number of fragments F in the lot L, a, is the gold content of one fragment,

a, is the unknown true gold content of the lot, MIis the mass of one fragment, and ML is the mass of the lot. All masses are expressed in grams. All contents are dimensionless, expressed as part of one. CHL is a relative, dimensionless variance. This formula is complete and does not contain any approximation or assumptions. It is a very important starting point in the Sampling Theory. To eliminate the task of estimating NF, a new, more pragmatic term ZHL called the Constant Factor of Constitution Heterogeneity is defined:

In order to estimate IHL in a practical way, the lot may be divided into a number of size fractions and density fractions. The estimate of ZHL then becomes:

where V, is the average volume of one fiagment in a given size fraction a,;la is the density of a given density class p, a,, is the average gold content of fragments in the size-density fraction

L@, and ML@is the mass of the size-density fraction L,. This is the reference formula. It is complete, and carries no assumptions. In order to estimate ZHL, it is possible to perform a complete size-density analysis on a representative composite from a given geological unit. Such a test is well documented in Gy’s work, but it is cumbersome and very expensive. Approximations can be made with two assumptions, which, in combination, lead to a more practical approach. The validity of these assumptions for gold will be discussed later in the paper. Assumption 1: Experience shows that the gold content

a,, varies much more from one

density fraction to the next than from one size fraction to the next. Therefore, in equation [ 11 ap is substituted for a,, . This assumption is nearly always true. 0

Assumption 2: The study of a large number of real cases shows that the proportion

M~~plM~p does not vary from one density fraction to the next in a manner that would significantly alter the estimation of ZHL. Therefore, in a simplified test, it is assumed that M L ~ M L s can be replaced by their average MLJML, which gives ML~sM = L F L~ ~M L . Now, ZHL is estimated as follows:

80

Cases where Gold is not Liberated The calculation of V, involves a fragment shape factorf, to correct for the volume d,' of a cube, where d, is the average size of a fragment in a size fraction a . Because of the third power applied to d, , the coarsest size fraction has an overwhelming effect on the final outcome of ZHL. Therefore, a test to estimate the order of magnitude of IHL for gold can concentrate on the coarsest fragments. The determination of the gold content in a series of size fractions in what has come to be known as the Heterogeneity Test will provide data to decide whether this assumption is realistic. The following cases are suggested versions of a Heterogeneity Test, depending upon the fragment size investigated. Case #1: Estimating IHL for Single, Large Fragments. Collect Q fragments (e.g., 100 fragments) selected one by one at random (i.e., there is no Grouping and Segregation Error) from the top size fraction al (e.g., 10 cm). Weigh each fragment q, which gives M4, and assay it for gold, which gives a4. Then, assuming each fragment q has a density 4, equation [2] can be written as follows:

where

Mq M L = x M 4 ,but Vq = and g = 4

A4

c4

M, ML

If the size d of the largest fragments is defined as the size opening of a screen that retains 5% of the collected fragments and, if the referenced material is a calibrated fraction between two consecutive screens, the value of g is around 0.5. In the original lot, where many size fractions are present, the value of g is around 0.25. After substitution and simplification ZHL becomes:

If the selected number of fragments is 100, then M L = 1 0 0 ~: q

Case#2: Estimating IHL for Groups of Small Fragments. Case #1 looked at individual large fragments for the sake of simplicity. But, if ZHL is to be estimated for material from a blasthole, or from a reverse circulation drill hole, or from a jaw crusher, groups of small fragments must be collected to involve enough mass and to have a reasonable chance of including enough gold particles. Prepare 100 groups, each made o f p = 50 fragments collected one by one at random from the top size fraction al (e.g., 1 cm). p is selected in such a way that the entire group can be submitted to a single fire assay, avoiding sub-sampling errors. A prior mineralogical study of the size of the gold particles and the expected average gold grade should give a preliminary idea about the necessary sample mass for the test. This study may determine a value for p which is different from

81

the value 50 used in this example; p = 50 is generally sufficient only for finely disseminated gold, or for cases where gold is included in another mineral like sulfide. For coarse gold in quartz for example, it may be necessary to select p = 5,000 fragments or more. In that case, weigh 100 fragments and constitute p = 5,000 fragments by collecting 50 times the mass of the 100 fragments. Each group of 5,000 fragments is submitted to a bottle roll cyanide leach procedure followed by a 50-gram fire assay of the leached residues. If

M , = 1OO&fp, in the same way as for fonnula [4], ZHL becomes:

If d is defined as the size of a screen retaining 5% of the fragments, and if d4 is the real size of each fiagment, g can be calculated as follows:

Calculation of a Sampling Nomograph The Constant Factor of Constitution Heterogeneity, IHL,can be divided into the following components:

IH, = fgctd’ = Cd’

171

where f is the fragment shape factor (around 0.5 for most minerals and 0.2 for liberated gold particles), g is the fragment size distribution factor (around 0.25 for non-calibrated material and 0.50 for calibrated material), c is the mineralogical composition factor, t is the liberation factor, d is the size opening of a screen that would retain 5% of the material to be sampled at any given sampling stage, and C is the sampling constant for a given stage of comminution. It is obvious that the size of the factorsfand g is approximate, and therefore the solution of equation [7]is as well. An additional, even greater difficulty lies in the estimation of the liberation factor f? . Literature in mineral processing led Gy in early publications to suggest the following formula:

where do is the liberation size of gold. At this stage, it is appropriate to recall the definition of the liberation size do which is defined as the size below which 95% of the material must be ground to liberate at least 85% of the gold (Gy 1967). Therefore do may become extremely small when a substantial amount of gold cannot be detected by microscopic observations. It should be emphasized that the grain size of the coarsest gold may have little to do with the liberation size defined earlier. Indeed, regardless of the size of the coarsest gold particles, the determination of the liberation size do may become a difficult task. It is not rare indeed that a substantial amount of gold is so fine as to be almost “in solution” in the rock and does not occur as discrete particles of pure gold that can be observed under the microscope. This very important

82

property has been described extensively in various articles and in a textbook (Ingamells and Pitard 1986, Pitard 1993).

0.1

2

on1

Figure 1. Testing an existing sampling protocol for blastholes with a nomograph A slightly conservative and pragmatic solution can be used, as follows: 0

0

0

From experience in a given deposit, estimate the proportion of gold that is easy to sample (i.e., invisible under the microscope) as suggested by both Ingamells and Pitard. This gold is the “low background gold content” which is finely disseminated in the ore. Call the proportion of gold, which is visible under the microscope, the coarse gold. This could range in grain size fiom several millimeters to a few microns. This is the gold for which the sampling characteristics need to be identified in order to optimize the sampling protocol while the fine gold can be ignored. The liberation size of the gold, which is difficult to sample, can then be defined.

Knowing If€, and all its other components listed in equations [7] and 181 will allow the calculation of do. The product of the following factors:

is relatively constant fiom one stage of comminution to another (not to be c o n b e d with one size fraction to another for which c can change substantially). Then, the sampling nomograph can be calculated using the following formula:

83

where S ’ , is the variance of the Fundamental Error. Then, for any selected, acceptable value of the variance, the sample mass it43 can be calculated. There is no guarantee that equation [8] provides the ideal model. This model has been s h o w to work well in practice for comminution stages not too far from the value of d = 1 cm as used in the heterogeneity tests. It is therefore often useful to perform simplified heterogeneity tests on fragments one centimeter in size, typical of the value for d in primary sampling stages for blasthole material, reverse circulation chips, and laboratory jaw crusher products. Using equation [lo] makes it easy to calculate a sampling nomograph, similar to the one illustrated in Figure 1, and optimize the sampling protocol in order to achieve a desired level for the variance of the Fundamental Error. Guidance for Reading Figure 1 The X-axis shows the mass MS of the various sub-samples of a particular sampling protocol. The Y-axis shows the variance SZ, of the Fundamental Error. Each oblique line represents a given stage of comminution. The thick, dotted line shows the “safety line”, a pre-selected, maximum acceptable variance of the Fundamental Error (e.g., PFE= 0.01 or a % relative standard deviation SFE= f 10%). The broken line of medium width represents the existing sampling protocol (i.e., collecting an 8-kg sample from an 800-kg pile with top fragment size d about 1.27 cm; crushing the 8-kg sample to a top size d of about 0.17 cm; splitting out a 500-gram sample; pulverizing the 500-gram sample to a top size d of about 106 microns; finally assaying a 30-gram analytical subsample). The broken line should not but does extend above the safety line. The wide broken line represents a better protocol that does not cross the chosen safety line. The use of equation [lo] to predict the variance of the Fundamental Error for a particular sampling protocol should be restricted to cases that lack evidence of gold particle liberation. This is usually the case for crush sizes of about 1 or 2 millimeters. As soon as there is mineralogical evidence that a few coarse gold particles become liberated, it is advisable to use equation [16] which is discussed in the next section. This may, however, initially produce conservative results. When the solutions to equations [lo] and [16] are in reasonable agreement, the nomograph obtained using equation [lo] does not need to be corrected. When the solutions to equations [lo] and [16] are in disagreement, the exponent of d could be adjusted downward from 3 in equation [7] or from 2.5 in equation [lo], to a lower value. However, such adjustment is an artifact compensating for problems (i.e., delayed comminution making the selected value for d completely wrong, Grouping and Segregation Error, Analytical Error) that have nothing to do with the scientific definition of that exponent. Cases where Gold is Liberated In cases where most of the gold reports to size fractions coarser than 80 microns, it is likely that the fine pulverization performed with laboratory mills will have liberated the gold. Alluvial gold would also be included in this category. Assumption 1 made earlier is still valid, while Assumption 2 becomes weak. Therefore, equation [2] can be restated as follows:

By definition the gold is liberated, therefore the shape factor is a function of the density, thus:

84

Developing this relationship for the density class ;lg of the gangue and for the density class ,IAu of the gold, and calling E the infinitesimal gold content of the gangue:

Obviously, the first term of the sum is negligible when compared to the second term. Furthermore, aAu= 1 by definition, and aL is usually very small. Therefore equation [13] simplifies as follows:

By definition:

Thus, the very useful simplified formula is obtained:

wherejL is the shape factor of gold, g A u is the gold partkle size distribution factor, dAuis the largest size of the gold particles defined as the size opening of a screen that would retain 5% of the total gold content. IffAu= 0.2, g A u = 0.25, and AAu = 16 (in practice native gold often alloys with other metals), useful sampling nomographs can be calculated with the following formula:

Figure 2 shows a case where the selected variance of the Fundamental Error is S2, = 0.0225 (i.e., %SFEf 15%). The gold deposit represented in Figure 2 contains gold particles up to 700 microns. The example uses a gold grade of 1.2 g/t, which is critical fiom a grade control point of view. The nomograph indicates that the total sample mass, which should be subjected to a metallics screen assay, is 10 kilograms. Figure 2 also shows that a 100-gram sample is required for fire assaying the minus 150-micron fraction. Contrast this with using only a 30- to 50-gram sample for a fire assay of an eight-hour shift composite sample of the grinding circuit product as mill feed. The resultant gold assay would have a %SFE = f 450%. When such a large value for %SFE is obtained, the assay result is no longer accurate. A sampling error of this magnitude makes it impossible to obtain a representative sample of the mill feed that can be used to accurately assess metallurgical performance.

85

Motnograph 1 .oe5

1oe4

10

1

1

OaoI

Figure 2. Nomograph for 700-micron liberated gold Guidance for Reading Figure 2: The X-axis shows the estimated size of the largest liberated gold particles. The Y-axis shows the minimum sample mass to be assayed to ensure a relative standard deviation % SFEof the Fundamental Error no larger than f 15%. The nomograph applies to this value of % SFE alone. Each diagonal line represents an expected average gold content expressed as part of one (e.g., 1.2 g/t is written 1.2~10"). The thick, dotted line shows the minimum sample mass to be assayed fiom the minus 150-micron size fiaction. The thick line shows the minimum sample mass to be screened if 700-micron gold particles are present. Similar derivations could be performed for other minerals of interest, and would lead to their own version of equation [la]. Representing all Size Fractions in a Sample

In equation [ 131 the particle size d, = dg of the gangue had a negligible effect. However, there are limitations to this simplification. For example, for a small sample containing fine liberated gold mixed with large gangue fiagments, the Sampling Theory (Pitard 1993) demonstrated that the following relationship must be hlfilled:

1 d 3 <-IHLAu

- 25 In other words, regardless of the sampling requirements for gold, the sample must always be representative of all the size fiactions present in the original lot. Attempts to Empirically Compensate for the Delayed Gold Comminution In a case with unrecognized delayed comminution for gold, formulas [7] and [16] may appear to give the wrong results because they result in estimates for the variance of the Fundamental Error that are much smaller than variances obtained fiom careful duplicate sampling. This problem becomes serious when, by pulverizing the material very fine, a substantial amount of gold liberates

86

before the liberation size as defined above is reached. In addition, other errors to the observed variance can contribute to the variance of the Fundamental Error. The corollary to this observation is that results fiom duplicate sampling after fine grinding cannot be used reliably to determine the sampling characteristics that are required for the calculation of the variance of the Fundamental Error. This is because duplicate sampling results in a large over-estimation of the sampling constants, especially those for fine materials. This is clearly demonstrated in results fiom the Heterogeneity Test recommended later in this paper. Therefore, it is of the utmost importance to investigate the nature of the coarsest particles in pulverized material in a Heterogeneity Test. A scanning electron microscope can be used to perform such a study on polished sections that are prepared from the coarsest fraction of a pulverized sample. Various authors in the past have made attempts to change the usual exponent “3” applied to d in the calculation of the variance of the Fundamental Error (Richards 1908). Some authors who have recently revived the idea (Franqois-Bongarwn 1998) propose to use the following formula:

where K and x are two constants to be determined through appropriate experiments. If careful precautions are taken, this approach has its merits and may lead to a different model than the one suggested by equation [8]. The purpose of this discussion is to point out that there is often a problem if the classical formulas [7] and [ 161 are taken for granted. However, the focus should not be on the exponent ‘‘3”) but on the value selected for d which may become completely inappropriate if delayed gold comminution takes place. The exponent “3” may also seem too large for finely ground material when other errors are encountered in addition to the Fundamental Error: Values for d much smaller than 1 cm become extremely sensitive to variances other than the variance of the Fundamental Error (e.g., variance of the Analytical Error and other sampling errors such as Grouping and Segregation, Delimitation, Extraction, and Preparation Errors), and the precision of these variances. Particularly, as some gold begins to liberate, the Grouping and Segregation Error may become overwhelming and nearly impossible to minimize. Changing the value of x to compensate for variances that have little to do with sample mass or 6agment size, is a direct departure fiom Gy’s approach. A complex sampling problem should instead be divided into its basic components. Once the Fundamental Error has been properly quantified, then the other sources of variability can be addressed separately. Important remark: The reader may notice x = 3 in formulas 171 and [16], and x = 2.5 in formula [lo]. This is no contradiction and is due to the fact that the selected model for the liberation factor in [lo] is equation [8].

THE GROUPING AND SEGREGATION ERROR The Grouping and Segregation Error (GE) is the direct effect of the Distribution Heterogeneity of a lot (DHL), which is expressed as follows:

87

where Y is the grouping factor, Z is the segregation factor, and NG is the number of groups of fragments in the lot, NG is set by the size of the increment collected to constitute the sample. Thus, in the Sampling Theory the variance of the Grouping and Segregation Error is expressed as follows:

s;, = ns;, This new variance is impossible to calculate accurately because the segregation factor Z is determined by the amount of segregation taking place in the lot. Segregation is a transient phenomenon and is dependent upon density and other factors. It is usehl however to determine the likely domain of that variance. DHL varies between a maximum that equals the Constitution Heterogeneity CH, and a minimum called [DH~lmin(Gy 1979, Gy 1983, Gy 1992). Because of large differences in mineral densities and in fragment sizes, DHL is never zero. [DH~]minis a random variable with a mean that can be calculated as follows:

rnean(DH, )fin = N G CH, NF where NF is the number of fragments in the lot. will ever be reached for liberated gold as the factor 2 is likely to It is unlikely that [DHLlmin increase during the homogenization of the lot. Unless samples are collected from a very large number of increments using a rotary splitter, pGE is likely to be the dominant sampling variance. The Sampling Theory demonstrates that it is impossible for DHL to become greater than CH,. From this, it does not follow that SZGEcannot be larger than S2,. Indeed, without taking precautions to minimize Y (by taking as many increments as practically possible) and 2 (by homogenizing the material prior to taking the increments), yGE may completely overwhelm 9,. ,when sampling gold or other It is the author’s experience that the difficulty of minimizing heavy minerals, is often greatly underestimated and leads to operational short-cuts such as the misuse of a riffle splitter, homogenizing by rolling or by coning, or by speeding up the vibroconveyor of a rotary splitter.

THE DELIMITATION ERROR AT THE MILL The Delimitation Error (DE) can generate the largest sampling bias encountered at the mill. It can be introduced when the use of inappropriate sampling equipment does not guarantee that all parts of a material stream have exactly the same chance of being selected as a sample. The correct sampling of mill feed, concentrates, and tailings for metallurgical accounting purposes requires an investment in cross-stream samplers. Multi, stationary cutters across the stream, as illustrated in Figure Nos. 3 and 4, may provide acceptable data for process control purposes. They are, however, not suitable in cases where a very small sampling bias can have a serious effect on the accuracy of metallurgical balance calculations, and therefore on the ability of plant supervision to effectively manage the entire operation. Systems shown in Figure Nos. 3 and 4, if well designed, are far better than sampling systems such as pressure pipe samplers, and any kind of sampler using a single, partial diversion of the stream. These latter inherently flawed systems are based on the wrong assumption: “the stream is well mixed and homogeneous, therefore diverting or pumping a small part of it is just fine,” and their shortcomings are primarily caused by a transient sampling bias that cannot be overcome. For further details on Quality Assurance with respect to the Delimitation Error the reader is referred to the appropriate textbooks on the subject (Gy 1979, Gy 1983, Gy 1992, Pitard, 1993). Figure 3 shows a system equipped with five stationary cutters across the stream. The five cutters minimize segregation problems across the stream (i.e., Z dimension). The entire height of the stream is sampled (i.e., Y dimension). The stream is sampled all the time (i.e., X dimension).

88

Figure 4 shows a similar system in a compact, circular configuration. Three sampling stages allow a 0.002% sampling ratio, which is very attractive to feed an on-line analyzer. This sampling system would be ideal for material Balance and plant feed mass determinations if the radial cutters, located under the inverted cones and which continuously sample the entire width of the stream, were able to rotate. Such a sampling system would be perfectly proportional. It would therefore provide a sample mass, which would be completely proportional to the total mass passing through the plant, and thus permit verification of tonnages that are determined independently at the mine and by mill weightometers.

Five Cutters

Sample stream going to a similar, smaller samphg stage

Overflow inducing turbulence

Main stream

Figure 3. Suggested sampling system for process control sampling of slurries (Courtesy of Thermo Gamma-Metrics, Adelaide, South-Australia)

THE EXTRACTION ERROR The properly delimited increment or sample needs to be correctly collected. The rebounding rule and center of gravity rule were established and discussed in the Sampling Theory to ensure that the Extraction Error (EE) remains small. In the case of gold, this error can become quite large and introduce a large bias and a significant additional variance, because this bias is never constant. During an exploration campaign, any problem during drilling (Le., with recovery of core or cuttings) is synonymous with extraction bias. The missing material may be completely different than the recovered material. To give an example, a 95% core recovery may lead to a 50% loss of gold. A case in Canada showed a 20% increase in the gold assay after losing fines (mainly sterile clay) during reverse circulation drilling. Severe consequences can result if this problem is ignored. The variance 9, of the Extraction Error cannot be calculated and it may, very likely, be quite large if strict precautions are not taken. A convenient statement, such as “precautions were taken to render this variance negligible”, should raise suspicion, unless these precautions are listed and described in a comprehensive way. The volumetric flow of mill feed or tailings streams can be very large, 10,000 m3/hour is not rare for a copper mine. Under such conditions, a cross-stream sampler can develop serious problems unless the foliowing precautions are taken by the manufacturer of sampling equipment and by the engineering firm:

89

Velocity of the Flowing Stream to be Sampled. At the sampling point, the stream flow must be laminar. Under no circumstances should the stream velocity be faster than 2 meterdsecond. This requires the introduction of a system, similar to the one illustrated in Figure 5 , to slow down the stream. I

l

l

PROCESS FLOW _____)

FIRST STAGE CUT

---

It

SECOND STAGE CUT

_----

-b

Figure 4. Illustration of a simple sampling system, offering correct delimitation at all times (Courtesy of GR Sprenger Engineering, Inc., Louisville, Colorado, USA)

90

Figure 5. Suggested system for slowing down stream velocity

91

For small flow rates (below 500 m3/hour),a minimum cutter opening of Wo = 3 d + 1 cm is recommended, where d is the size of a screen opening retaining no more than 5% of the solids. Because d is usually very small in plant pulps, this rule simplifies to Wo = 1 cm. However, large flow rates need significantly larger cutter openings. Very little research has been done on this subject. The author’s experience leads to the guideline illustrated in Figure 6, which has been successfully applied in many large copper flotation plants in Chile.

r

Slurry stream flow rate in m3/hour

Recommended cutter opening W, in centimeter

1

Figure 6. Suggested guideline for minimum cutter openings for very large slurry rates The cutter must not travel across the stream faster than 45 cmlsecond, which is a good ASTM guideline. Under such condition, the cutter fills up quickly. Under no circumstances should the

92

cutter overflow during its trajectory. Therefore, its capacity must be generous, and the diameter of the cutter exit must be quite large, as suggested in Figure 7.

Figure 7. Typical cross-section of a properly designed pulp sample cutter THE PREPARATION ERRORS Operations such as crushing, grinding, pulverizing, packaging, transfer of the sample fiom one place to another, screening, drying, and filtering can lead to Preparation Errors. These errors are

93

the result of contamination, losses, alteration of the physical or chemical composition of the sample, or of unintentional mistakes made by a poorly trained operator. For gold, the major recommendations for minimizing these errors are to: 0

0

Prevent the loss of fines from drilling machines, jaw crushers, roll crushers, plate pulverizers, dusty homogenizing procedures, and ventilated hoods. Prevent contamination between successive samples from drilling operations, jaw crushers, roll crushers, drying ovens, plate pulverizers, ring and puck pulverizers, and screens. Prepare samples with low gold contents with different equipment and in different rooms than samples with high gold contents, this is of the utmost importance.

PURPOSE OF HETEROGENEITY TEST For trace constituents, such as gold, it is impossible to accurately determine the sampling parameters required to calculate the sampling constant C in equation [7],for a given state of comminution, without performing carefully designed Heterogeneity Tests. These tests are designed to generate the necessary knowledge for optimizing sampling protocols and minimizing some of the common problems, which can be experienced during the implementation of these protocols. There are two approaches to perform such tests 1. Collect duplicate samples at each stage of comminution, repeating the procedure until at least 30 pairs are obtained. A variance analysis can then be performed to establish the variance affecting each sampling stage. This is often referred to as the “sample tree experiment”. 2. Select a size fraction that is most likely to introduce the largest error into the sampling protocol (e.g., a 1-cm size fraction for sampling of blasthole cuttings) and collect 100 samples made of n fragments. Such fragments should be collected one by one at random if the value selected for n is smaller than 50. If the value selected for n is greater than 50, then the mass of an increment containing the desired number of fiagments can be determined by weighing, and the 99 further such increments be collected by carefully weighing each increment. The number of fragments n should be carefully selected, following a thorough mineralogical investigation.

The second procedure should be combined with replicate sampling of all size fractions for which a sub-sampling stage may be required. Furthermore, this approach should be augmented by a careful investigation of the exact nature of the coarsest particles after the material has been pulverized very tine at the laboratory. This investigation can be performed with a microscope, and conclusions may lead to a correction of the sampling nomograph for the last sampling stages, which involve very fine material and where the coarsest particle of all may be a gold particle. Both approaches have their advantages and disadvantages, and neither is perfect. The sample tree approach has been described (Franqois-Bongarqon 1998) and many others have used it, including the author, in several cases where valuable information was available fiom quality control duplicate sampling at all sampling and sub-sampling stages. However, the results are often strongly influenced by the precision of the large variance that is caused by the Grouping and Segregation Error, and which can become erratic when gold is liberated in finely ground material. It is therefore important to determine the appropriate sample mass at any given sampling stage. In order to solve this matter effectively, it is necessary to identify the variance of the Fundamental Error with minimum interference from other sources of variance.

94

SUGGESTED HETEROGENEITY TESTS FOR GOLD The author has developed a Heterogeneity Test to calculate sampling nomographs and to optimize sampling protocols. Such a test should be performed on every major type of gold mineralization. The following test closely follows a real case. A preliminary mineralogical study has revealed that the test can be performed on a 340 kg composite, using samples from the top size hction (d = 1 cm),made up ofp = 50 fragments each. At this stage, the reader should refer to case #2 for non-liberated gold, so there is no ambiguity about the choice of the composite mass and of the value for p. The following case is a customized, real case, and needs to be adjusted to fit other projects. 1. The composite is generated from material from 50 different locations within a single, coherent type of mineralization. Dry the composite overnight, at 110°C. After drying, the composite weighs about 340 kg. Crush the entire composite to roughly minus 2 cm using a clean jaw crusher with the opening adjusted to 2 cm. 2. Divide the composite into four lots, using a fractional shoveling procedure, and call the lots A, B, C, and R respectively: A = 136 kg, B = 68 kg, C = 68 kg, and R = 68 kg. The R lot is saved for potential tests that may need to be rerun. 3. Screen lot A through 1.25-cm, 0.63-cm, 0.335-cm, 0.17-cm, 0.085-cm, 0.0425-cm, and 0.02 12-cm screens, creating eight size eactions. 4. Weigh each size fraction and record the weights. 5 . Spread the -1.25-cm +0.63-cm hction on a clean surface: The Heterogeneity Test is performed on this fraction, where d = 1.05 cm. 6. From this fraction, collect 100 samples. Each sample is made ofp fragments selected one by one at random, until the sample mass is approximately 50 grams. Number these samples fiom 1 to 100, weigh each of them, and record the values for p. 7. Pulverize each of the 100 samples directly in an enclosed ring and puck pulverizer to about 95% minus 106 microns, (do not use a dusty plate pulverizer which is known to smear gold too much). 8. Assay each of the 100 samples for gold by fire assay and gravimetric finish, using the entire sample. Atomic Absorption (AA) finish is not recommended for the test, since the most relevant assays are the ones showing high gold contents. If AA is used, the dilutions have to be monitored very carefully. 9. Crush each of the seven size fractions remaining fiom lot A, and what is left of the minus 1.25-cm +0.63-cm hction, to 95% minus 0.30 cm, and collect a split of about 1,000 grams from each original size fraction. If the coarsest size fraction weighed less than 1,000 grams, use the entire fraction. Pulverize each 1,000-gram split to 95% passing 106 microns. Perform a screen fire assay using the entire sample and a 150-micron screen. Weigh and assay the entire +150-micron fraction, and perform two 50-gram fire assays on the minus 150-micron fraction. Record all weights and assays, which are very reIevant for the interpretation of the test. Important remark: The size opening of the screen to perform metallic screen fire assays should be selected for each project using a nomograph like the one illustrated in Figure 2. There is no such thing as one size fitting all cases. 10. Take lot B and split it into 16 equal splits, using a rotary divider equipped with 16 segments. A riffle splitter may be used but it is not recommended. Record all weights. 11. Crush each split to minus 0.30 cm, then pulverize the entire split to about 95% passing 106 microns. 12. Spread each split in a large pan, and collect one 50-gram sample made of about 12 random increments (assay it using the entire sample), and one 100-gram sample made of about 24 random increments (assay it entirely in two 50-gram aliquots). Important remark: Results from these three assays for the 16 splits often lead to the conclusion that the observed variance has very little to do with the variance of the Fundamental Error.

95

13. Take lot C and crush it to about 4 . 3 0 cm. Then, split C into two equal lots C1 and C2 of about 34-kg each using a rotary splitter. 14. From lot C1 take a 16-kg split, divide it into 16 equal splits, using a rotary divider. Record all weights. 15. Take one split and perform a size distribution analysis using 0.63-cm, 0.335-cm, 0.17-cm, and 0.085-cm screens. Record all weights. Recombine the size fractions, so that this split can participate in the following steps. 16. Pulverize each split to about 95% passing 106 microns. 17. Spread each split in a large pan, and collect one 50-gram sample made of about 12 random increments (assay it using the entire sample), and one 100-gram sample made of about 24 random increments (assay it entirely in two 50-gram aliquots). 18. Crush lot C2 to about 95% passing 0.085 cm. 19. From the crushed lot C2 constitute a 10,000-gram fraction by collecting fifty 200-gram random increments (a square scoop can be used), and divide it into 16 equal splits using a rotary splitter. Record all weights. 20. Take one split and perform a size distribution analysis using 0.17-cm, 0.085-cm, and 0.0425-cm screens. Record all weights. Recombine the size fractions, so that this split can participate in the following steps. 21. Pulverize each split to about 95% passing 106 microns. 22. Spread each split in a large pan, and collect one 50-gram sample made of about 12 random increments (assay it using the entire sample), and one 100-gram sample made of about 24 increments (assay it entirely in two 50-gram aliquots). 23. Recombine all rejects from the 16 C2 splits, so that a lot D is obtained that weighs approximately 7.5 kg and is pulverized to about 95% passing 106 microns. 24. Divide lot D into 16 equal splits, using a rotary splitter. Record all weights. 25. Take one split and perform a size distribution analysis using 0.0425-cm, 0.0212-cm, and 0.0106-cm screens. Record all weights. Recombine the size fractions, so that this split can participate in the following steps. 26. Spread each split in a large pan, and collect one 50-gram sample made of about 12 random increments (assay it using the entire sample), and one 100-gram sample made of about 24 random increments (assay it entirely in two 50-gram aliquots). 27. From any rejects from these tests, prepare a 10,000-gram sample (e.g., 10,000 grams from lot C2). Screen this sample on a 212-micron screen. There is always a substantial amount of material that does not grind well. Wash this coarse material, discarding the fines. Separate the heavy minerals by panning. Prepare polished sections from the heavy concentrate and investigate the nature of the material that does not comminute well. Interpretation of the Tests The results of the recommended Heterogeneity Tests provide data for the calculation of the variance of the Fundamental Error and for the estimate of the other errors encountered when sampling gold ores, as follows:

1. Steps 5 through 8 give all the necessary information to arrive at a good estimate of the sampling constant K in equation [lo]. 2. Step 9 provides the knowledge of the gold grade variability as a fbnction of fragment size and will point to the potential presence of significant Extraction and Preparation Errors. 3. The results of steps 10 through 26 provide all the information necessary to estimate the contribution of variances other than the variance of the Fundamental Error. 4. Step 27 detects any problems due to gold delayed comminution, which would necessitate the use of equation [16].

96

ADVANTAGES OF THE RECOMMENDED APPROACH

1. The tests accumulate an enormous amount of information at a reasonable cost. 2. The Fundamental Error for the coarsest fragments (i.e., about 1 cm for blasthole chips, reverse circulation chips, or the product of a jaw crusher) becomes well known. It is indeed the primary sampling stage in ore grade control and exploration that is likely to introduce a large error. 3. The sampling constant C can be calculated using one-centimeter fiagments, therefore the sample mass and K remain the only contributors to the variance of the Fundamental Error, d being one. This gives a good estimate of K. 4. Differences in gold content between various size fractions become well documented. For example, ifthe minus 212-micron material contains five times more gold than the coarser fragments, then any loss of fines during the implementation of the sampling protocol would introduce large Extraction Errors and Preparation Errors, all of which can generate a significant bias. 5 . Screen fire assays performed on 1,000-gram pulps give information about the behavior of gold particles after the material has been pulverized. Delayed comminution of gold may dictate the use of equation [16]. 6. Polished sections can provide the necessary information to correct the calculated nomograph using equations [8] and [14]. 7. The variance between 16 replicate splits fiom different comminution stages provides information about variances due to sampling errors other than the Fundamental Error. 8. The Heterogeneity Tests document the variance of the Analytical Error, which becomes negligible between the 100 handpicked samples made o f p fragments. 9. The histogram of the 100 assays performed on p one-centimeter fragments can provide a good estimate of the proportion of gold that is easy to sample (i.e., the Ingamells’ low background gold content). 10. Since thep one-centimeter fiagments are collected one by one at random, at least in many cases, there is no Grouping and Segregation Error included in the estimation of the sampling constant K. Perceived disadvantages of the recommended approach

1. The one hundred samples made of p one-centimeter fragments often generate a few outliers. This is not a problem but represents quite well the behavior of the gold, and could even lead to a good modeling of the Poisson distribution using the histogram of the 100 assays. While these outliers lead to a relatively large variance, this actually minimizes the negative effect of the Analytical Error as experienced in the duplicate sampling approach. 2. The Heterogeneity Tests are performed on a calibrated size fraction, which is not representative of the grade in the total sample. This leads to a particle size distribution factor of 0.5 instead of 0.25, but this is easy to correct for in the calculations. 3. If the gold content of the finer size fractions is higher than that of the tested fraction, then the mineralogical factor applied to the data fiom the tested fraction would be slightly conservative, and therefore the sampling constant K would also be conservative. Again, this is easy to correct for in the calculations. 4. The entire test is performed on a single composite. If the material to be sampled at the mine and especially at the mill comes from sources with sufficiently different sampling characteristics, then the Heterogeneity Tests will need to be repeated for the other major ore types. The common sampling protocol will be dictated by the characteristics of the most heterogeneous ore.

97

CONCLUSIONS There is a degenerative process involved in proceeding fiom equation [l] to equation [7]. This is due to the use of an increasing number of approximations. Therefore, the classical formula [lo] is not the universal solution for all sampling questions and should be used with these approximations in mind. Equation [16] provides a reliable backup when gold begins to liberate. Heterogeneity Tests must be conducted in such a way that the Analytical Error and sampling errors such as the Grouping and Segregation Error, Delimitation Error, Extraction Error and Preparation Errors do not interfere with the calculation of sampling constants. Once the sampling constants are known, the variance of the Fundamental Error can be determined, and once this variance is known, the effects of the other errors on observed variances can be estimated. Frequently, the focus is entirely on the Fundamental Error, and the seven other sources of sampling errors are ignored. Indeed, it is not rare to see attempts at solving problems associated with the Fundamental Error while at the same time creating far more severe Grouping and Segregation, Delimitation, Extraction, and Preparation Errors. Improvements in the reliability of the results of sampling programs in metallurgical plants thus require the application of all aspects of the Sampling Theory. All sampling errors must be addressed to guarantee that accurate metallurgical balances can be calculated so that problems in the milling process can be identified and corrected. Stringent and uncompromising sampling practices at the mill are required to achieve this goal. In the author’s experience, this will more than pay for itself by allowing the identification of hidden costs associated with a poor mill performance and concealed by flawed sampling systems. REFERENCES

Gy, P.M. 1955, 1956. Poids a Donner a un Echantillon - Abaques. Revue de I’Industrie Minerale 38 (1956), and Erforderliche Probemenge - Kurventafeln. Internationaler Kongress fur Erzaufbereitung - Goslar (8-1 1 Mai 1955) et Erzmetall8 (1955), p. B 199-220. Gy, P.M. 1967. L’echantillonnage des Minerais en Vrac. Tome 1: Theorie Generale. Revue de 1’IndustrieMinerale - 15 Janvier 1967. Page 75. Gy, P.M. 1979, 1983. Sampling of Particulate Materials - Theory and Practice. Elsevier Scientific Publishing Company. Developments in Geomathematics 4. Gy, P.M. 1992. Sampling of Heterogeneous and Dynamic Material Systems: Theories of Heterogeneity, Sampling and Homogenizing. Amsterdam, Elsevier, Pitard, F.F. 1993. Exploration of the “Nugget Effect”, Geostatistics for the Next Century, An International Forum in Honor of Michel David’s Contribution to Geostatistics, Montreal, 1993. Edited by Roussos Dimitrakopoulos, McGill University. Kluwer Academic Publishers, Boston. Ingamells, C.O., and Pitard, F.F. 1986. “Applied Geochemical Analysis”, Volume 88. Chemical Analysis: A Series of Monographs on Analytical Chemistry and its Applications. A Wiley-Interscience Publication. John Wiley & Sons. Pitard, F.F. 1993. Pierre Gy’s Sampling Theory and Sampling Practice. Textbook published by CRC Press, Inc., 2000 Corporate Blvd., N.W. Boca Raton, Florida 33431. Second edition, July. Richards, R.H. 1908. Ore Dressing. McGraw Hill. New York. Franqois-Bongarqon, D. 1998. Extensions to the Demonstration of Gy’s Formula. January.

98

Sampling a Mineral Deposit for Metallurgical Testing and the Design of Comminution and Mineral Separation Processes Jerry Hanks' and Derek Barrat+

ABSTRACT Proper samples are essential for metallurgical testwork for the successful design of a mineral processing plant. To select an economic process, several alternatives require investigation to predict metallurgical performance, develop design criteria and flowsheets, estimate operating costs, and size process equipment. This paper discusses sampling requirements for different phases of process development, co-ordination with the exploration team, required sample mass for various tests, the advantageddisadvantages of drilling methods, core size, and bulk samples. Bench-scale testing and profiling a deposit is discussed in relation to process simulators to optimize a sampling campaign and the economics of pilot testing bulk samples. INTRODUCTION Sampling a mineral property in order to test comminution and other metallurgical properties is an essential step in project development (Hanks 1997). In addition to ore samples, it may also be necessary to locate and test samples of other raw materials that might be essential to the metallurgical process; e.g., limestone, magnesium oxide or similar bulk commodities. The presence or absence of such material can have a strong influence on process selection, facilities siting, and project economics. As an example, a nickel-cobalt laterite project using the pressure acid leach (PAL) process will typically require a tonnage of limestone in the range of 60% to 80% of the tonnage of ore processed. Water from the site should be used in the testwork whenever possible to establish any potential problems that may arise with its eventual quality. Metallurgical characterization should begin early enough in the project to identify potential fatal flaws in processing. Some familiar examples of metallurgical problems include: highly refractory gold deposits which make recovery by heap leaching impossible; a copper sulfide flotation concentrate with high arsenic or fluorine contents making it difficult to market the concentrates; or very fine-grained massive sulfide mineralization which does not permit effective liberation of chalcopyrite from sphalerite at any practical grind size. Early identification of such problems can prevent a project from becoming a black hole for exploration and development budgets. On the other hand, most exploration projects never become mines, so there is no point in spending process development dollars too fast. The best way to develop a schedule for exploration, sampling, and testing involves scoping the work to a realistic schedule of milestones for performing engineering studies in order to progressively evaluate the project and to decide whether to continue or terminate the work. This paper begins with a description of some of the more important sampling methods used for comminution and metallurgical testing. The second section deals briefly with the types and approximate unit costs of testwork that are required for the range of engineering studies typically used in evaluating a new project. The last section summarizes the testwork required for

' Mineral Processing Consultant, Phoenix, AZ, USA DJB Consultants, Inc., North Vancouver, B.C., Canada

99

comminution, mineral separation, and metal extraction processes, with emphasis on sample sizes and types, the purpose(s) of the tests and time required for performing the tests.

SAMPLE TYPES AND CHARACTERISTICS General Guidelines The sample types that are most often taken for comminution and metallurgical testing include:

0 0 0

0

Grab samples Reverse circulation drill cuttings Small and large diameter diamond drill core Auger drill samples, as used for shallow deposits such as placers Channel samples, as another useful tool, particularly for sampling underground drifts or surface outcrops Various types of bulk samples.

In the early stages of a project, the same samples that are used for resource definition are normally used for metallurgical testing. The cost of extensive large diameter drilling andfor bulk sampling for metallurgical tests can only be justified after a project has been shown to be feasible. Given the high cost of obtaining samples, it makes good sense to obtain the maximum amount of realistic information possible from any given sample, however small it might be. Even though resource definition and understanding of the deposit geology will always be the driving force in planning a drilling program, the need for metallurgical samples must also be considered. The cost of obtaining such samples can have a major impact on the project budget. In general, the less costly the sample, the less suitable it is for metallurgical testing. There can be a fundamental conflict between geologist and metallurgist over the proper use for exploration samples, particularly core. Geologists may want to keep the material largely intact forever, while the metallurgist wants to destroy the core and immediately perform testwork on it. A compromise is always needed. The cardinal characteristic required of any sample is that it represents some definite portion of a mineral deposit. Representation should not solely be confined to the grade of the sample. Geological parameters that should also be representative are: 0 0 0

0

Lithology Alteration Degree of oxidation Mineralization Hardness Geotechnical competence.

Other important sample characteristics that should be considered include: 0

0 0

Mass of the sample Particle size distribution Maximum particle size The cost of obtaining the sample.

100

Table I shows a comparison of characteristics by type of sample.

Table 1 Characteristics of samples for comminution and metallurgical testing Type of Sample Reverse Circulation Large Sample Drill Small Diameter Characteristics Grab Cuttings Diameter Core Core Bulk Varies Good Varies Good Coverage Poor Low Better Good Best Mass of sample Low-good Good Good Poor Fair-good Particle size Poor-good Moderate-high Moderate-high High Moderate Relative cost Low No matter how extensively and meticulously the comminution and metallurgical tests are carried out, the results are no better than the quality of the samples that are used. Often too much significance is attached to testwork that is performed using inadequate or inappropriate samples.

Sample Types Grab Samples. These are often taken very early in a project for preliminary assays and metallurgical evaluation. Grab samples can be very useful, but test results must be used with care. The acquisition and selection of grab samples should be decided with due consideration and documentation of their significance vis I? vis other methods of taking samples. A common mistake is to take a “specimen” rather than a “sample.” The geologist should always ask himself: “Why am I submitting this particular sample and what do I expect to learn from the test results?” Grab samples, with proper preparation, are suitable for agitation and small diameter column leach tests, flotation tests, crushing, rod mill, and ball mill work indices, and abrasion tests. Preparation consists of crushing to the proper size and splitting out the required mass of material. Reverse Circulation Drilling (RVC). Cuttings from these holes are often used in mineral exploration to obtain assay data and limited geologic interpretation. RVC is usually selected based on cost and, with respect to samples for metallurgical testing, “you get what you pay for.” Only preliminary data can be obtained from RVC cuttings. Preparation of RVC cuttings consists of drying and mixing the material. Portions can then be split out for metallurgical testing or pulverizing for assay. RVC cuttings can be used for agitation leaching including bottle roll tests, small diameter column leach tests, flotation tests, and Bond ball mill work index determinations. However, the results of all these tests should be considered preliminary on account of the possible risk of losses incurred during sample acquisition, and also the inherent bias away from a natural size distribution in the sample. Diamond Core Drilling. Drill core is the workhorse for producing samples for early-stage and mid-stage process development testing. Diamond drill holes (DDH) can be cored in the following diameters: AQ BQ NQ HQ PQ 6-inch 6.5-inch 8-inch

27mm 36mm 48mm 64mm 85mm 150mm 165 mm 200 mm.

Except for very short holes, the minimum size used in exploration is often NQ, 48 mm (1.88-in.). For longer holes, it is common to begin the hole at a larger size; e.g., PQ, and reduce

101

the diameter in stages as the hole gets deeper (and the drilling gets tougher). Larger diameter holes are common in soft deposits such as laterites, but are not normally drilled in hard rock until the project has been shown to be feasible. After logging, the core is prepared by splitting, or preferably, sawing the core lengthwise, then crushing one half to about 13 mm (0.50-in.) A portion of the crushed material is split out and pulverized into assay pulps, while the remainder is available for comminution and metallurgical testing. This material, termed “coarse reject,” is suitable for Bond rod and ball mill work indices, Bond abrasion indices, flotation testing, agitation and small diameter column leach tests, and similar work. The results of flotation and grinding tests, as well as agitation leach tests, can often be used with a good degree of confidence as long as a sufficient number of tests are performed, and the appropriate scale-up factors are applied for equipment sizing and design criteria. Some of the tools for comminution evaluation require particle sizes larger than split core, for instance, Bond crushing work indices and the JK Tech drop weight test. Fairly short intervals of whole core can be used. Often a large number of samples must be tested so it may be necessary to retain and assay the test products. This is to avoid having a large number of drill intervals without assay data, a situation that would impact adversely on the geologic block model and resource estimate. It is becoming common practice to twin holes, say NQ core for resource definition, preliminary testwork, and assessment of ore variability, followed by twinning of selected holes with larger diameter core; e.g., HQ or PQ, for detailed testing of the full range of comminution, geotechnical, and metallurgical parameters using sample intervals that can be related to a preliminary or more detailed mine plan. All of this takes a lot of co-ordination, but it maximizes the amount of data that can be obtained from a given sample, and is well worth the effort. Large Diameter Core. Diameters of 6-inch (150 mm), 6.5-inch (165 mm), or &inch (200 mm) can be used for most comminution and metallurgical testing except for column leach tests at run-of-mine (ROM) size. Included are pilot plant tests for autogenous and semiautogenous grinding characterization. For this purpose the core is not split, but rather stagecrushed to represent a primary crushed ore. The proper size distribution can be estimated from manufacturers’ tables or simulated from basic strength of materials data, but it must take into account the mining method; e.g., coarse blasting in open pit or finer blasting underground. Depending upon the type of drilling equipment used and rock characteristics, large diameter core drilling is limited to depths of 150 m (500 ft) or less. Large diameter core drilling may be used to obtain large amounts of material for testing, over 100 tonnes in some cases. A tendency to drill down dip sometimes prevails in order to maximize the amount of material intersected from each drill hole and this must be done with care, or avoided, to guard against taking nonrepresentative samples. Channel Samples. These are similar to core samples with respect to sample preparation and the tests for which they can be used. Depending upon selection of the channel cutting equipment, it may be possible to obtain samples with a larger maximum dimension compared to that from small diameter core. Bulk Samples. These can be obtained by large diameter core drilling, pitting, ripping, trenching, blasting an exposed face (ROM), or from a stockpile; e.g., primary crushed ore, sinking shafts, or driving adits. Obviously the cost to obtain such samples can be very high, and can only be justified for projects that are very likely to proceed. Bulk samples, properly taken and prepared, can be used for any type of metallurgical testing. However some problems are often encountered in obtaining and preparing such samples. One very common problem arises from material that has been blasted or crushed too fine. This is especially common when contractors are used to produce and prepare the sample. A tight contract and good supervision are required to ensure that a good representative sample is obtained. The authors know of more than one bulk sample that was over-crushed for heap leach tests, or over-blasted as feed to pilot plant testwork for autogenous and semi-autogenous grinding, resulting in questionable test results. Technical and economic reasons alone do not always determine which type(s) of bulk sample(s) should be taken. For a recent fast-track copper project, initial plans called for obtaining

102

bulk samples by driving adits into the deposit. However, this would have required a timeconsuming effort for permitting that might have delayed the project. Since drilling was already permitted, large diameter core drilling was selected instead. A common fallacy is to assume that a bulk sample is somehow better than other types of samples just because of its size. However, in the context of a large deposit even a large bulk sample may not represent much of the mineralization. For example, a 50,000 tonne bulk sample is only about 0.01% of a 400 million tonne resource. When this approach is compared to the acquisition of a large number of core samples from several drill holes, the core samples will give better coverage of the extent and depth of the resource in terms of representing ore variability within the mine plan.

Additional Sample Considerations Tests on an adequate number of individual samples are recommended and will give more information, provided that sufficient material is available, than tests on a composite. For largescale testing, compositing of samples is almost always required. Tests on composites should be interpreted with an understanding that individual components of the composite may behave very differently than the composite as a whole. Frequently, samples covering a very long interval of a single drill hole (often the highest grade hole) are used to make up a test composite. However, deposits are usually mined (at least in open pits) for extended periods over a limited vertical distance. It is better practice to make up composites using equal elevation intervals from several nearby drill holes. It is pointless to test a composite that is unlike anything that will be encountered at a particular time interval in a production schedule. Sample handling must be done with care. One common problem is inadequate removal of contaminants such as oil, grease, drilling mud, and surfactants. Such material can seriously impact the response of a sample to flotation testing or cause preg-robbing in cyanidation tests. Repeat tests should be done on splits from the same samples at intervals during the process development period to determine if samples are changing in storage. Massive sulfide samples in particular oxidize quickly. This can impact both the assays and the response to testing. A nitrogen purge and double sealing in plastic can be used to protect samples, or metallurgical samples can be, and are often stored in freezers. When sampling an exposed face, the weathered surface should be slabbed away first, then the sample taken from a fresh surface. Whenever possible, the actual source of water that is to be used for production should also be used for testwork, or at least shown to be similar to water available at the test site. Since surface water may show seasonal variations, samples should be taken and tested throughout a full year. Well water should be pumped long enough to remove any contaminants before sampling. Water samples should be tested promptly to avoid changes that may occur with ageing. Refrigeration is often required to reduce ageing effects. In a few cases, processing will be done in seawater. Preliminary tests can be done using “Instant Ocean” or similar products available at aquarium supply houses. However, seawater varies greatly from one location to another, and the actual source to be used for production should be sampled and tested. Where the water source is likely to be very cold or very warm, or both, testwork should be done at the appropriate temperatures. One aspect that affects testing for comminution in particular, is the effect of northern hemisphere continental climate and permafrost (Antarctica is excluded for the time being). Frozen ore in open pits can elevate inherent rock strengths (UCS) by a factor of two or three. This effect is very important, for example, in the Minnesota Iron Range where it adversely affects autogenous grinding performance during winter. In northern Canada, permafrost has been known to occur at mine depths of 1,250 feet in an underground primary crushing station.

103

TESTWORK REQUIRED FOR ENGINEERING STUDIES A number of engineering studies are usually required to evaluate the resources, economics, process selection, and other aspects of a project as it advances from discovery to production. Both the exploration and the process development program should be designed to provide enough data at the right time to adequately evaluate a project before committing to the next, more expensive phase. Often there are project milestones such as option payments or other contractual requirements that dictate the timing of the studies. For example, it is common for a joint venture agreement to require completion of a feasibility study within some specified time frame. The nomenclature and requirements of engineering studies vary from company to company, but generally follow the sequence of 0 0 0

0

Scoping Studies Pre-feasibility Studies Feasibility Studies Basic Engineering.

These studies require an increasing degree of process definition as the schedule progresses towards Basic Engineering. Also, more sample material will be necessary to test for ore variability and different processing conditions as studies become more detailed. A summary of the testing required for each type of study is shown in Table 2, and is discussed in the following section of the paper. Sample weight requirements, top size in the feed, and approximate unit cost per test are shown in Table 3.

104

Table 2 Testwork required for engineering studies

Type of Test Comminution Bond work indices (rod and ball) Bond abrasion indices SAG power index (SPI variability) Bond work indices (crushing) MacPherson autogenous index Autogenous media competency (advanced)* UCSPLI Fracture frequency Bond work indices (rod & ball) 2ndfacility JK Tech drop weight tests AG/SAG pilot plant Final process design criteria Flotation Preliminary reagents/pH Rougher grind-grade-recovery Regrind and cleaner flotation Locked-cycle Ultra fine grinding Optimization of major ore type (variability) Pilot plant (compleddifficult ores) Final process design criteria Dewatering Concentrate thickening Concentrate filtration Tailing thickening Tailing filtration Final process design criteria Heavy Minerals Gravity gold recovery(GRG) Heavy liquid separation Gravity/magnetic/electrostatic Pilot plant Final process design criteria Leaching Small diameter columns (variability) Intermediate diameter columns Large diameter columns Bottle roll (variability) Batch agitation leaching (ClL/CIP)** Semi-continuous (CIL/CIP)*** Agitation design tests (rheology, suspension) Final process design criteria

Scoping

TvDe of Studv Prefeasibility Feasibility

Basic Engineering

X

X

X

X

X

X

X

X

(XI

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

(XI (x>

X

X

X

X

X

X

X

X

X

(XI (x>

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X X

X

X

X

X

X

X

X

X

(4

X

X

(x>

X

X

(XI

X

X

(XI

(4

(x)

X

X

(4 (x)

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

(4

X

X

X

X X

X

Note: (x) as required * Includes Bond work indices, Bond abrasion index, UCS, and PLI. ** As required for special situations; e.g., flotation concentrates, high clay content ores. *** As required to determine carbon loading adsorption kinetics, carbon fouling, etc.

105

X

Table 3 Approximate unit sample weight, top size, and cost of comminution and metallurgical tests Type of Test Suggested Sample Weight Top Particle Approximate Cost (US$) per ROM/Core dia. Requirement kg Sue for Test, mm Test (dependent Type/mm per Test upon laboratory location and local currency) Comminution Bond work indices (rod) PQMQMQ 12-15 12.7 750 per mesh 85/64/48 Bond work indices PQMQNQ 12-15 3.36 750 per mesh (ball) 85/64/48 Bond abrasion indices PQmQMQ 1.6 19 x 13 350 85/64/48 Bond work indices PQMQ 15-30 76x51 550-1,000 (crushing) 85/64 SAG power index (SPI PQMQMQ 2 12.7 500 for variability) 85/64/48 MacPherson autogenous PQmQ 225-250 32 6,800 index 85/64 Autogenous media gl'/6"/PQ 200 152 x 140 3,000 competency (advanced) 200/150/85 (four size classes) UCStPLIlFracture PQMQMQ Field selected 510 x 152 Field testing WCS) frequency 85/64/48 JK Tech drop weight 6"/PQ 200-250 65 x 65 (five 5,500 150185 size classes) AGISAG pilot plant r/6" 1OO,Oo&3 00,000 200 100,00&300,000 2001150 (program) HPGR PQMQ 350 25 Vendor testing 85/64 Flotation 1-2 3.36-1.65 400-600 Rougher PQMQMQ 85/64/48 1,000 15 3.36-1.65 Cleaner (grind-gradePQMQMQ recovery) 85/64/48 15-25 3.36-1.65 2,000 + Locked - cycle PQMQMQ 85/64/48 100-500 3.36-1.65 14,OOO-70,000 Circuit design PQmQMQ 85/64/48 Ultrafine grinding Concentrate From pilot plant Dependent on Vendor testing test N/A 8"/6" Pilot plant 50,000 - 100,000 3.36 2,500Iday 2001150 Dewatering Concentrate or tailing From laboratory 1 Dependent on Vendor testing thickening tests test Concentrate or tailing From laboratory 1 Dependent on Vendor testing filtration tests test

106

Type of Test

Heavy Minerals Gravity gold recovery (GRG) Heavy liquid separation

Gravity/magnetic/ electrostatic Pilot plant

Leaching Bottle roll Batch agitation (CIL/CP) Semi-continuous (CIL/CIP) Small dia. columns (variability) Intermediate dia. columns Large dia. columns

Note:

Suggested ROWCore dia. Typelmm

Sample Weight Requirement kg per Test

Top Particle Size for Test, mm

Approximate Cost (US$) per Test (dependent upon laboratory location and local currency)

PQMQINQ 85/64/48 Beach sand or PQMQNQ 85/64/48

40-70

12.7

1,000-1,200

0.100-0.200

Size classes from 850 microns and finer Dependent on mineralogy

25

Beach sand or PQn-IQNQ 85/64/48 8"/6"IPQ 200/150/85

50-100

400

5,000-20,000

Dependent on mineralogy

2,500Iday

HQNQ 64148 HQNQ 64/48 PQMQmQ 85/64/48 HQNQ 64/48 HQNQ 64/48

0.05-0.100

3.36-1.65

500

2-5

3.36-1.65

500

30-50

3.36-1.65

5,000-6,OOO

9

6 - 19

3,000

80

51

7,000

8"/6"PQ 2001 150185

60,000

200

75,000- 100,000

Costs shown in this table do not include any allowances for the acquisition and delivery of samples, or for any interpretation of test results, but do allow for the disposal of pilot plant products.

METALLURGICAL PROCESSES AND TESTING Introduction While a large number of processes are available for the treatment of ores, only a few are commercially important for treating the more common types of mineralization encountered in present day exploration projects. Most processes fall into the broad categories of physical and hydrometallurgical treatment. The most important factor in process selection is the mineralization of both ore and gangue. However, knowledge of the mineralization is only the starting point; testing is required to determine how any particular sample will respond to a specific treatment scheme. The mineralogists employed by exploration geologists are usually not sufficiently familiar with mineral processing to provide all the data that a metallurgist requires. This can best be done by specialists in process mineralogy. Since many deposits may contain several principal minerals, more than one process may be required to treat the entire deposit. In most cases, mineral separation and metal extraction processes are preceded by comminution in order to physically and economically liberate or expose the minerals of economic interest.

107

The sample requirements and the procedures that are used for comminution, mineral separation, and metal extraction testwork are discussed in this section. It is intended to give the reader an overview of the relationship between sample size and the test purpose(s) for which each type of sample is taken. Detailed test procedures can be found in the relevant chapters elsewhere in this book.

Comminution Comminution tests are necessary to quantify the amount of energy required to grind the ore to the required size, as well as to determine the configuration of the comminution and classification circuits. Bond Work Indices. Initial tests are usually done using the Bond ball mill work index (WiBM) procedure. The Bond rod mill procedure (Wi,) is used to establish the extent of variability in work index for coarser incremental stages in size reduction. This becomes important in assessing the potential for accumulation of critical sizes in AG/SAG mills, and in the use of certain power-based models for the determination of specific power consumptions in AG/SAG/Ball mill circuits. These tests require at least 12 kg to 15 kg of feed for each sample tested, depending upon the number of cycles that are run to establish equilibrium, the mesh sizes tested (usually two or three), and mineralogy and liberation analysis in each procedure. These samples should be selected to permit stage-crushing to pass 12.7 mm (0.50-inch) for the rod mill test, and 6 mesh (3.36 mm) for the ball mill test. It is considered advisable to dispatch 25 kg to 30 kg in each sample to allow for repeat or additional tests. For scoping studies, results from a single laboratory are sufficient. If a project goes to the pre-feasibility stage, a second laboratory should be used to confirm the results of the first. For detailed feasibility studies, the Bond low energy impact crushing work index (Wi,) and abrasion characteristics should also be determined. Each crushing test requires at least ten pieces between 51 mm and 76 mm (2-in. to 3-in.) in size, although twenty pieces are preferred to establish the frequency of variability. Short sections of whole core are suitable. The abrasion test requires at least 1,600 g of 19 mm x 13 mm (3/4-in. x 112-in.) particles of the same material that is used for the crushing test. The purpose of this abrasion test is to provide an index (Ai) from which wear rates of grinding media and mill liners can be estimated. This test can also be included in scoping studies if the probability of high wear rates exist and autogenous grinding might be a candidate. Test results for Wic, WiRM,and WiBMcan be used as input to power-based grinding circuit design models that can specify equipment sizes, motor powers, operating conditions within the mills, mill speeds, power distribution between primary mills and secondary mills, and the necessity for pebble crushing or pre-crushing, for a range of ore types and mill throughputs. Such mill sizing exercises can be very economical and effective in the absence of pilot plant testwork. For grinding circuit design (circulating loads in cyclones, confirmation of transfer size as new feed to secondary grinding, pebble recycle loads and the effect of pebble crushing, etc.), these powerbased methods can be supplemented by JK SimMet simulations and, when combined, can replace pilot plant testwork with a high degree of confidence. Case History: Bond Work Index Tests This approach has been applied successfully to a 120,000 tpd greenfields copper-gold project in Indonesia for which sampling criteria was dictated by minimal disturbance to the environment and the project schedule prior to completion and approval of the feasibility study (MacLaren 2001). Pilot plant testwork using bulk samples was not practically possible within the study schedule. This meant that only drill core could be utilized, with insufficient quantities being available to permit pilot plant testwork for grinding. Consequently, 5.1 1 km of core was obtained in stages using PQ, supplemented by HQ core at depth, from fourteen holes that were within the five-year and ten-year pit

108

envelopes. Many of these holes were twinned with NQ resource definition holes. A total of 105 intervals were selected according to lithology and alteration types. Each interval was sampled for Bond’s low energy impact crushing, rod milling, and ball milling work indices, the Bond abrasion index, and point load indices. These work indices were composited according to mining depth and bench level for each mine year up to year 10 for mill sizing and mill throughput exercises, and prediction of a potential expansion, using a power-based model. Rejects from sampling core and test remnants were composited for each lithology in preparation for flotation pilot plant tests in sea water (Barratt 1996).

Batch Amenability Tests for Autogenous and Semi-Autogenous Grinding. There are several techniques used for evaluating autogenous or semi-autogenous (AG/SAG) grinding characteristics. These include the MacPherson Autogenous Work Index test, the JK Tech Drop Weight and Abrasion tests and in Australia, the Advanced Autogenous Media Competency test (AMCT). The first of these is mostly used for general amenability, while the next is becoming a useful tool in the design of large-scale grinding circuits and predictions for existing mill performance when new mineralization is encountered, and the last, the AMCT, is used to distinguish the amenability of an ore for autogenous grinding as opposed to semi-autogenous grinding, and to size grinding mills using power-based methods. The Minnovex SAG Power Index (SPI) is used to predict the throughput of an ore type in an existing or new grinding circuit. The JK Tech simulator is useful in evaluating proposed circuit changes; e.g., addition of a crusher to a SAG-ball mill circuit, whereas the SPI test can be used as an indicator for grinding circuit optimization. The MacPherson Autogenous Work Index. This index is obtained by testing 225 kg to 250 kg of sample, either drill core, ROM, or primary crushed ore, that has been stage-crushed to pass 32 mm (1.25-in.), in a 460 mm (18-in.) dia. dry mill that is closed-circuited with a 14-mesh screen and air classifiers, and which contains a ball charge. Power draw is measured and converted to a power consumption per tonne (kWNt) of net product, from which an autogenous work index (AWi) is calculated. MacPherson classifies ores into those that are amenable to autogenous grinding and those that would benefit from a ball charge (SAG) if significant variability in Bond work indices exists. The MacPherson test results require interpretation by A.R. MacPherson Consultants Ltd. and should be supplemented by Bond work index tests for low energy impact crushing, rod milling, ball milling, and the Bond abrasion test. The MacPherson test does not directly measure the competence of an ore at coarser sizes or its ability to generate and maintain autogenous grinding media in the mill charge. For these reasons, the authors consider that this test is most appropriate for pre-feasibility studies. The SPI Test. This test is performed in a small standard mill operated dry for a specified period of time, after which the mill is emptied and the charge is dry screened. The feed is stagecrushed to pass 12.7 mm (1/2-inch) and oversize is returned to the mill for another grinding period, etc. The total time in minutes that is required to grind the charge to a specified product size is used to relate the grinding characteristics of a particular ore to a large database of other materials, from which the performance of an existing or new mill can be estimated. On some projects, several hundred samples are tested, and the results entered into a block model from which predictions can be made regarding the throughput capability of a given mill, either existing or proposed. The SPI test requires 2 kg of coarse rejects from drill core or RVC cuttings. The JK Tech Drop Weight Test. As the name implies, this test utilizes a standard weight dropped from a standard distance onto a test sample. Previously a pendulum test was used, and can still be used if there is a lower limitation on top size, which reduces the number of size classes for which energy levels are analyzed. An abrasion test is also conducted using a small, low speed test mill. This is a different test with a different objective from the Bond abrasion test referred to previously. The drop weight and abrasion tests generate breakage function numbers that are used in the JK SimMet program to simulate different process conditions in a grinding circuit, and to

109

predict plant capacity and power draw. The best accuracy is obtained by “calibrating” the model against a similar circuit for which actual operating data are available. About 200 kg to 250 kg of sample is required. It is necessary to ensure that sufficient material is available in the coarsest size classes, in particular a minimum of 3 kg (10 pieces) of 65 mm x 65 mm (2.5-in. x 2.5-in.), and also that overall, material is available for duplicate tests. Five size classes are tested ranging down to 14.3mm x 12.7 mm (9/16-in. x 112-in.). In addition, a minimum of 3 kg of 65 mm x 38 mm (2.5-in. x 1.5-in.) is required for the abrasion test. Although 6-in. diameter drill core is preferred by JK Tech, PQ core can be accepted as a minimum diameter, or ROM and primary crushed ore can be tested. Core is broken up, usually in one-pass crushing, to minimize the bias of cylindrical surfaces and to maximize the beneficial effect of sharp edges on test results. With smaller diameter core, the number of effective size classes that are tested may be reduced. For the less expensive abbreviated test procedure, the 22 mm x 19 mm (0.875-in. x 0.75-in.) size class is subjected to breakage using five energy levels as a means of testing ore variability economically. Each of the two methods has its particular application and has to be supplemented with proprietary interpretation, but it is generally agreed that the JK method is superior (at least at present) for grinding circuit design, while the SPI approach is an economical method for assessing ore variability in terms of mill throughput. Case History: SPI and JK Tests For one recent grass roots project, the decision was made to utilize both the SPI and JK Tech test procedures, using the same samples, for a bankable feasibility study. Sections of whole core were taken at regular intervals and shipped first to the JK Tech test center for approximately 150 tests. After drop weight and abrasion testing, samples from the same core sections were sent to the Minnovex laboratory for tests using SPI procedure. The material was then returned to the exploration sample preparation facility where it was carefully split. One split was pulverized for assay, while the remainder was cornposited and sent to still another facility for Bond work index determinations. In this way the samples provided a maximum amount of data. This data in turn was used to help plan the AG/SAG pilot plant campaign, as well as to provide a check against pilot plant results.

The Advanced Autogenous Media Competency Test. The AMCT has its origins in the former Allis Chalmers organization and, in its present form, is offered by Amdel in South Australia. Orway Mineral Consultants had input to later developments jointly with Amdel. The procedure involves the following individual tests, the results of which are subject to interpretation by grinding circuit designers. Standard Bond media competency test in which about 200 kg of sample is sized into the following size classes: 152 mm x 140 mm (6-in. x 5.5-in.), 140 mm x 127 mm (5.5-in. x 5-in.), 127 mm x 114 mm (5-in. x 4.5-in.), and 114 mm x 102 mm (4.5-in. x 4-in.). The percentage weight retained and the number of rocks in each size class are recorded. If drill core is used, minimum PQ size, it is broken up into 150 mm maximum length pieces for sizing. The sample is loaded into a drum, 1.83 m (6 ft) dia. x 0.3 m (1 ft) length that is rotated for 500 revolutions at 26 rpm (83% C.S.). Net power draw is recorded and a specific power consumption is calculated. At the end of the test, the mill charge is sized from 102 mm (4-in.) to 75 microns. Product size analysis and the number of survivors in each size class between 152 mm (6-in.) and 19 mm (0.75-in.) are referenced to ores of known amenability to autogenous grinding or semi-autogenous grinding. Twenty surviving rocks in each size class between 102 mm (4-in.) and 19 mm (0.75-in.) are subjected to the Bond low energy impact crushing work index (Wid test for comparison with the results on equivalent size classes on fresh ore. Differences in the value of this index between size classes for survivors could be indicative of competence

110

0

0 0

at a coarser size or at a “critical size” relative to results on fresh ore and to standard Bond work indices for grinding (rod milling and ball milling). Standard Bond grindability tests for rod milling (WiRM) and ball milling (WiBM) to a specified mesh(es) of grind on composited survivors and fresh ore. The ratio WiRM: WiBMis important in assessing the expected power efficiency of autogenous or semi-autogenous grinding, pursuant to the implementation of pebble crushing. Standard Bond abrasion test. Standard tests for unconfined compressive strength (UCS) and point load index (PLI) for reference to PLI results that have been determined at regular intervals on drill core in the field. These tests can be conducted on survivors and fresh ore for comparison.

In comparison with MacPherson, SPI, and JK Tech procedures, more value is obtained per sample with the AMCT across a range of size classes and types of tests in terms of a spectrum of comminution criteria for less cost. The AMCT test, in its complete format, has been used for a considerable number of projects, particularly Australian low-grade gold ores of high competence. These test results have been used to create power-based designs that have correlated well with actual operating data and to design criteria that have been scaled-up from pilot plant testwork (Siddall 1996, Barratt 1999).

Pilot Plant Testwork While some small-scale test methods give adequate AG/SAG information for scoping and prefeasibility studies, others generate data for use in power-based and simulation methods for sizing grinding equipment and circuit design for detailed feasibility studies. Pilot plant testing is still regarded as being necessary for inclusion in a detailed feasibility study program by most grinding circuit design experts in certain circumstances; e.g., autogenous grinding, pebble milling, precrushing, and pebble crushing. Pilot plant testing methods have improved greatly in the last ten years, and the number of different circuit configurations and mill variables to be investigated has been reduced. This is the result of much more experience in operating comminution circuits as well as numerous changes to circuits that were designed earlier. For these reasons, the amount of sample required and the cost of pilot plant testwork have both been reduced. For porphyry copper and copper-gold deposits, the most common circuit at this time is the SAG mill-ball mill-pebble crusher, or SABC circuit, in which the SAG mill discharge is screened, with the oversize going to one or more stages of crushing and the crusher discharge returned to the SAG mill. One approach that is receiving a lot of attention recently is to pre-crush a portion of the SAG mill feed to a size that is finer than the “critical size” (this is usually in the range 65 mm to 89 mm, or (2.5-in. to 3.5-in.). One operation segregates 152 mm x 51 mm (6-in. x 2-in.) material from primary crushed ore, crushes it to pass 38 mm (1.5-in.), and blends the crushed product into the SAG mill feed. Pre-crushing may increase the mill feed rate considerably, by up to 50% in one case on hard ore, compared to normal SABC operation. For pilot plant testwork for such a case, particular attention should be paid in ensuring that the size distribution in the bulk sample is representative for testing. There are relatively few facilities available for pilot plant testing of AG/SAG milling. It is not generally economical for an operating company to build and operate a pilot plant. Additionally, a lot of experience is required to operate a pilot plant successfully. 0

Case History: Drill Core Bulk samples are required to obtain the mass of material required for AG/SAG pilot plant testing. One recent project utilized 165 mm (6.5-in.) core drilling to produce 120 tonnes of sample for this purpose. The test program consisted of two ore types and four main circuit configurations and cost about $100,000, not including the cost of drilling and transporting the samples to the test site. The amount of material required for testing depended on the mill feed rate, which varied from about 180 kg/h to 1,200 kglh (400 lb/h

111

to 2,700 lb/h) or more. Feed rate was influenced by the hardness of the ore, type of circuit tested (e.g., autogenous or semi-autogenous), the ball charge characteristics, and the option of including pebble crushing. From the net power consumed and the feed rate the specific power consumption was calculated, usually expressed in kWh/t. This is one of two main parameters used for scaling up to an industrial-scale mill size, the other being the transfer size of material from the AG/SAG mill to the ball mill. Transfer size is obtained from screen analyses of the AG/SAG mill circuit product. Sample preparation required two weeks, and the actual testwork took three weeks. An additional two weeks were required for data analysis and reporting. The cost of pilot plant testing is dependent upon the number of different ore types and circuit configurations to be investigated, and the pilot plant location; e.g., Chile vs Canada, and local currency equivalent. The cost of sample preparation and disposal of products is significant.

High Pressure Grinding Rolls (HPGR) Testing for the potential application of high pressure grinding rolls (HPGR) is usually performed gratis on a laboratory-scale by the equipment manufacturer. Machine sizes are 0.10 m (4.0-in.) dia. x 0.03 m (1.2-in.) width, or more usually, 0.30 m (1.0 ft.) dia. x 0.07 m (2.75-in.) width. Pilotscale tests using a 0.71 m (2.33 ft) dia. x 0.21 m (8.25-in.) width are usually performed at an operating site or in a commercial pilot plant, with the HPGR machine on loan from the manufacturer. In ferrous and non-ferrous mining, HPGR has found application in crushing diamondiferous kimberlite prior to physical separation, gold ores prior to heap leaching, and iron ores prior to ball milling, with potential for copper and gold ores. Potential benefits arise from the creation of microcracking in crystal lattices and associated savings in power consumption in downstream comminution stages. Typical sample size is 350 kg for the larger laboratory-scale machine (0.30 m dia.) for a series of tests in 25 kg lots, each prepared to pass 25 mm (1-in.). Tests are usually in open-circuit with closed-circuit operation as an option. The effects of variables such as feed top size, grinding force, and machine speed are assessed.

Gravity Gold Recovery In many grinding circuits for both small-scale and large-scale plants, gold or other precious metals occur in sufficient quantities that permit their recovery by gravity concentration according to size class and specific gravity. Centrifugal gravity concentrators are currently used extensively for this purpose; e.g., Knelson or Falcon. A typical batch test procedure requires 30 kg to 70 kg of ore that has been stage-crushed for rod mill feed (minus 12.7 mm), the actual tested weight being dependent upon the actual head grades and size classes of gold in the sample. The procedure (Laplante 2000) for a gravity recoverable gold (GRG) test involves grinding the sample in a rod mill to 100% minus 850 microns for gold recovery in Stage 1 using a laboratory Knelson centrifugal gravity concentrator. In Stage 2, 20 kg to 27 kg of the tailing from Stage 1 is ground from 45% to 60% passing 75 microns as feed to the second pass, and 20 kg to 24 kg of the tailing from that stage is ground to 80% passing 75 microns as feed to the third pass in Stage 3. All concentrates and samples of tailing are assayed. A metallurgical balance is produced to show stage-by-stage recovery and cumulative recovery by size class down to 20 microns. Such cumulative recovery will most probably exceed that which is obtainable on an industrial scale because, in practice, only part of the grinding circuit stream (usually the circulating load) is processed.

112

Flotation Tests Flotation is the most common physical separation method for minerals in use today. Flotation is applied to recover copper, lead, zinc, molybdenum, nickel, tungsten, precious metals, iron ore, coal, and numerous other commodities. Bench-Scale Rougher and Cleaner Tests. Laboratory-scale grinding and flotation tests are almost always done together because flotation response is so strongly dependent on particle size and the testing of freshly ground pulps. The first stage testwork is usually performed on the coarse rejects from core drilling or channel sampling. Grab samples, reverse circulation chips, and even assay pulps are also used, but with less confidence in the results. The amenability of an individual ore type to a simple rougher flotation procedure can be determined with as little as 3 kg of sample. A fairly good grind-grade-recovery relationship, suitable for scoping or pre-feasibility studies can often be established for a simple ore with about 15 kg of sample. The basic flowsheet for the grinding and flotation processes can be defined well enough for such an ore and for use in feasibility studies, and even plant design, with 100 kg of samples, while not accounting for ore variability tests. Complex or poly-metallic ores require three to five times as much sample to design a circuit, size equipment and support a feasibility study. In cases for which ore variability is paramount, usually as a result of differences in mineralogy and head assays, it is necessary to perform a more extensive series of tests using a standard rougher flotation procedure to characterize the variability of the deposit. In one recent project, approximately fifty such tests were performed prior to a scoping study, with fifty more prior to pre-feasibility, and a like number prior to the detailed feasibility study. As the mine is developed, standard flotation tests will continue to be performed in order to test the variability of the ore. Ideally, a sufficient number of tests should be run to provide statistically significant metallurgical input to the geologic block model. Open-circuit tests give reliable grade and recovery data for rougher flotation (most rougher circuits are open circuits). Open-circuit cleaner tests will usually give a good indication of the ultimate final concentrate grade that can be achieved, but will give low recovery figures, because of the amount of valuable material tied up in middlings and circulating products. Locked Cycle Tests. Locked-cycle tests are required to give reliable overall recovery figures. These tests should be run until the circuit is in equilibrium, usually with five or more cycles. Even extended locked-cycle tests tend to give unreliable figures on the grade and tonnage of intermediate products because of difficulty, which is often experienced, in controlling the regrind size on a small-scale. Mini-scale and full-scale pilot plants, properly operated, give excellent data for flotation plant design. Pilot Plant Tests. For copper and precious metals deposits, flotation pilot plants are rarely technically necessary, but may be required for project financing, concentrate marketing, or hydrometallurgical testwork. Process development for flotation circuits can be fast tracked. If a sufficient number of samples is available, testwork for a scoping study can easily be completed in two or three months. Pre-feasibility studies require four to six months, and a detailed feasibility study less than one year. However, the rate at which test materials become available from exploration usually extends these periods by a considerable margin. Heavy Minerals Testing Tests are usually conducted on samples of beach or river sands. The most common sampling techniques are: grab samples, auger drilling, and bulk sampling to obtain test materials. Crushing and grinding are usually not required so no tests are performed for these unit operations. Processing generally follows the sequence: washing and desliming, gravity upgrading, drying, electrostatic and magnetic separation. Testing follows the same sequence. Initial gravity tests usually consist of screening and heavy liquid tests to determine the specific gravity analyses and grade of a number of samples of a few hundred grams each. Assays and mineralogical examinations are performed on the test products. Larger scale gravity tests are performed using

113

laboratory-scale or pilot-scale equipment such as tables, spirals, and cones. One advantage of gravity testing, as compared to flotation or hydrometallurgical testing, is that the test products can be recombined for additional testing. The gravity concentrates (the heavies) are usually dried and further concentrated by electrostatic methods. These tests are often performed while the material is still hot from drying, because some materials, such as rutile, respond differently at higher temperatures. Several stages of magnetic separation with or without additional electrostatic separation usually complete the testwork. A good material balance can be obtained using only laboratory equipment. However, the scale of testing may be dictated by the need to produce a fairly large quantity of final concentrates for marketing samples. A simple pilot plant campaign can generate the required samples and sufficient data for equipment sizing and plant design. Samples for larger scale tests are obtained in the same manner as for the initial tests. Depending upon the ratio of concentration and the amount of concentrate required, the amount of sample required varies from about 5 tonnes to 20 tonnes. Pilot plant and laboratory equipment is readily available, as are laboratories with the equipment and expertise to conduct a test program.

Hydrometallurgical Process Tests Various types of leaching processes are used extensively to extract precious metals, copper, nickel, and cobalt and numerous other elements from their ores. Leaching conditions and design parameters are largely determined by the type of mineralogy and the grade of the ore, particularly in the case of gold. This paper discusses only atmospheric leaching. Whereas this process is used very extensively for copper and precious metals extraction, and while the process chemistry is totally different between the two, they are nearly identical in terms of sampling and testing procedures. Agitation Leach Tests. Assay type procedures such as hot and cold cyanide leaching, and a number of similar procedures for acid or ferric sulfate leaching of copper minerals are used to assess the relative response of samples to standard procedures. These tests can be run on virtually any material from assay pulps to bulk samples. Typically, these tests require only an hour or two to perform. The next stage is often bottle roll testing using RVC cuttings, coarse rejects from core, or similar material. Bottle roll, or other types of small-scale agitation leach tests, require a minimum of 50 g to 100 g of sample for scoping tests on copper ores and are run for periods from 48 hours to 7 days or longer. For gold ores, 1 kg to 1.5 kg of crushed ore is tested. For non-refractory ores, the variability of metal extraction across the resource is most economically assessed with simple, small-scale test procedures. In the case of refractory ores, more complicated test procedures no doubt prevail. Batch-scale agitation leach tests are performed sometimes on ores, but more generally on flotation concentrates, to generate carbon adsorption hysteresis curves for gold in CILICIP. These tests are run from 24 hours to 48 hours and require 2 kg to 5 kg per test. Semi-continuous tests on a larger scale, each test with up to twelve stages, can also be run in situations that require resolution; e.g., the risk of carbon fouling and the necessity for a higher degree of confidence in carbon adsorption kinetics and carbon loading. From 30 kg to 50 kg may be required for these tests. Column Leach Tests. For potential heap leach projects, small-scale column tests are the next step. Fine samples, typically 6 mm to 19 mm (0.25-in. to 0.75-in.) are leached in columns, often 76 mm (3-in.) diameter columns 1 m to 1.5 m (3 ft to 5 ft) high. These require only about 9 kg (20 Ib) of sample, equivalent to about 4 m (12 ft) of split HQ core. In addition to determining the leach extraction and rate of extraction, these tests have the objective of optimizing the test variables (acid and ferric ion or cyanide concentrations, need for a cure step, etc.).

114

As work progresses two factors come into play which control the column size and the amount of sample required: 0

0

Larger particle sizes are tested requiring larger column diameters. The column diameter should be at least 4, preferably 5 times the maximum particle size. Taller columns are used to get more realistic data of pregnant leach solution (PLS) grade, pH, and acid or cyanide consumption. (The ideal column height is the same as the lift height to be used in actual operation. This is rarely possible at the early stages of a project.)

As a result of these two factors, the sample requirements increase rapidly. For example, a 152 mm (6-in.) column that is 3 m (10 ft) high requires nine times as much material as the smaller column mentioned previously. Both small and intermediate size columns are usually run for periods of 45 days to 120 days, but may extend for a year or more. An additional 30 days is required to set up and take down a column test, assay the residue, and report the results. The cost is dependent upon the column size, frequency of monitoring, and duration of the tests. Screening and assay by size fraction is recommended for the heads and residues. For advanced ROM heap leach projects, large diameter columns are the preferred test method. One mining company uses 1.8 m x 9 m (6 ft x 30 ft) columns for this work. These take about 60 tonnes of sample. Test duration is usually 69 days to 120 days, but can be over a year. Tests are expensive, typically $75,000 to $l00,OOO per test, not including the cost of obtaining the sample. The columns can be easily controlled and generate dependable data that is suitable for designing full-scale heap leach operations. Test heaps are even more expensive, $400,000 and up, and are so difficult to construct and control that the data generated is often suspect. It is far better and less expensive to run several large diameter column tests than one or two test heaps. For copper projects, solutions may be prepared from laboratory or commercial chemicals, or raffinate from an existing SXRW operation may be used. The solution from column tests may or may not be recycled. If it is recycled, assay samples are taken first, the copper or precious metals are removed by SX or carbon adsorption respectively, and the acid and ferric ion concentrations, or cyanide and alkalinity are adjusted before the solution is recycled. Test columns may be “stacked” to simulate multiple lifts. In this procedure a column is leached for the usual period, then a second column with fresh ore is started. The PLS from the new column is measured and sampled, then applied to the top of the first column. After an additional time period a third column may be added. Obviously, it takes a long time to complete tests of this type. Test columns are also “rested’ at times to simulate intermittent leach cycles. This may increase recovery, but the main benefit is usually a higher PLS grade. Column tests generate enough solution for bench testing of downstream process steps, SX shake-out tests, or carbon adsorption tests for sample. Electrowinning can usually be designed without testwork. For scoping studies, roughly 10 to 20 small diameter columns are required. Pre-feasibility studies usually require about 20 additional small diameter tests and 5 to 10 intermediate column tests. Detailed feasibility studies require up to 100 small and intermediate size columns. If leaching is to be done at ROM or primary crushed ore size, two to five large diameter columns are also required. The factors that influence these numbers include the variability of the deposit, whether or not the study is required to be “bankable,” and the company’s experience with similar projects.

115

CONCLUSIONS Exploration, sampling, testing, and evaluation of mineral deposits are closely tied together. Good planning and co-ordination are required to provide the right information at the right time for proper evaluation of the project. Metallurgical tests should start early enough to identify fatal flaws, but should not progress too fast since the project may never become a mine. Process selection is largely a function of ore and gangue mineralogy, but testing is required to confirm the amenability of a particular mineralogy to a given process. Generally, the lower the cost of obtaining a sample, the less suitable it is for metallurgical testing. The importance of scheduling properly conducted testwork on representative samples is vital for the complete and thorough evaluation of a project. New tools such as the SPI test as well as established batch test procedures, used in combination, are available for evaluation of ore grindabilities and the design of grinding circuits. Pilot plants are still required for AG/SAG circuit design in certain circumstances. The links between sourcing the samples and the purposes for which the samples are taken have been emphasized. The relative cost of metallurgical testwork is often low in comparison to the cost of drilling and obtaining samples. Therefore, it is the fervent hope of the authors that the detail in this paper will encourage the wise and cost-effective use of the mineral exploratiordproject development dollar in order to make samples available, and to maximize the quality and the amount of test information obtained at any stage of project study. ACKNOWLEDGEMENTS The authors wish to acknowledge material that has originated from their work on projects associated with the following companies: Phelps Dodge Corporation, AMAX, Inc., Newmont Mining Corporation, Cia. Minera Los Pelambres, Minera Alumbrera, Boddington Gold Mine, Cia. Minera Dona Ines de Collahuasi SCM, CVRD, Newcrest Mining Limited, KCGM, TeckCominco, Echo Bay Mines Ltd., J.K. Tech, Krupp Polysius, Amdel, Lakefield Research of Canada, A.R. MacPherson Consultants Ltd., The Knelson Group, Minnovex, and Fluor Daniel Wright Ltd. REFERENCES Hanks, J.T. 1997. Process Development for Exploration Projects. SME Annual Meeting. MacLaren, D., Mitchell, J., Seidel, J., and Lansdown, G. 2001. The Design, Start-up and Operation of the Batu Hijau Concentrator. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, IV : 316. Barratt, D.J., Matthews, B.D., and deMull, T. 1996. Projection of AG/SAG Mill Sizes, Mill Speeds, Ball Charges, and Throughput Variation from Bond Work Indices. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, I1 : 541. Siddall, B., Henderson, G., and Putland, B. 1996. Factors Influencing Sizing of SAG Mills from Drill Core Samples. Proceedings lnternational Autogenous and Semi-Autogenous Grinding Technology 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, I1 : 463. Barratt, D.J., Basic, J., Dunlop, G.A., and Phillips, R. 1999. Autogenous and Semi-Autogenous Grinding: Laboratory and Pilot Plant Studies. Mineral Processing and Hydrometallurgy Plant Design, World’s Best Practice. Australian Mineral Foundation. July. Laplante, A.R., Woodcock, F., and Huang, L. 2000. Laboratory Procedure to Characterize Gravity - Recoverable Gold. Transactions, Vol. 308. SME-AIME.

116

Overview of Metallurgical Testing Procedures and Flowsheet Development Terence P.McNulty’

ABSTRACT Bench scale laboratory ttSting reveals fxtual information about a sample’s characteristics and its behavior under well-defined and carefully conmlled conditions. Oace applicable chemical and physical data have been obtained and a prclhumy economic analysis has justified a more detailed evaluation, a pilot plant program can be developed. Pilot-scale confirmation of bench test results enables well-informed completion of material and energy balances, equipment selection, and plant design. Both bench and pilot programs are intended to minimize risks, although a pilot plant may be unnecessary if sufficient engineering experience and Operacing history already exist.

INTRODUCTION Metallurgical testing is done to satisfy questions concerning: The behavior of a sample under a well-defined set of chemical and physical conditions, The technically and economically optimum conditions for concentration or separation relating to specific project requirements, and The ultimate plant design incorporating well-informed selections of processing unit operations, equipment types and sizes, materials of construction,and physical arrangement.

Such testing can be conducted at a variety of treatment rates, ranging from bench-scale, through pilot-scale. and demonstration-plantscale. mere are no hard and fast rules distinguishing the size at which experiments are bench-scale or pilot-scale. Generally, progression from bench-scale to pilot-scale to demonstration-scale increases the complexity and expcnsc of a program. In a welldesigned progmn, each stage will also reduce the risk associated with the commercial development of a process.

BENCHSCALE TESTING Testing for nearly all plant design pwposes should be conducted on samples of solid material or solutions that represent as faithfully as possible the anticipated feed for the process. This requirement cannot be overemphasized, as failurt to use a representative sample essentially negates the applicability of the test results. The principal exceptions are tests meant to examine extremes in composition or chemical d o r physical properties. Tests carried out in beakers, flasks, bottles, and small process simulators like batch flotation cells and autoclaves are referred to as “bench-scale” and are intended to provide reliable information about underlying principles. Once the chemistry and physics have become clearly understaod and a preliminary economic analysis has justified a more comprehensive evaluation. a p i l o t - d e program may be designed. Ideally, the bench work should be suficiently comprehensive to allow preparation of a material balance on the basis of which a pilot circuit can be designed. PILOT PLANT TESTING Pilot plants typically use larger samples than bench-scale testing, and are most often a collection of fully continuous units. In some cases, however, it is often difficult to match the scale for all ~

1

_

_

T. P. McNulty and Associates. Inc., Tucson. Arizona

119

unit operations (due to unit capacities, or consideration of physical dimensions that are not easily scaled). If so. some unit operations may be conducted semi-continuously, while others may be evaluated separately from the pilot plant. Reasons for conducting pilot plant programs vary. The pilot plant may, for instance, reveal previously unrecognized information about the influence of recycled intermediate products or slurry streams on yield or quality of the final product However, the primary objective of a pilot program is not testing, per sP. but confirmation of batch laboratory conclusions and the development of information that can be USBd confidently by the designers of a demonsmation plant or a commercial plant. If the desired flowsheet is based on mature technology such as carbon-inleach cyanidanon of an oxidized gold ore or solvent extraction and electrowinning of copper from the leaching of oxide ore heaps with dilute sulfuric acid, a pilot plant probably will not be necessary. If, however, the technology is novel, or the feud to the commercial plant will be highly variable in some respect, or prototype equipment is envisioned, or if any aspect of the proposed flowsheet has not yet been successfully engineerad, designed, and operated, a pilot plant evaluation is mandatory. Furthermore, novel technologies, especially those employing aggressive chemical and/or physical processing conditions, add considerable risk to engineering and design and to the prediction of equipment capacity and reliability. Almost invariably, a failure to correctly account for the risks posed by new tachnology will retard the rate of achievement of design production rate. This last consideration poses the greatest threat to the financial performaace of the plant and the entire project of which it is a partThe outcome of flowsheet development and eventual plant design is always subject to a degree of risk. Beyond the usual risks associated with ore reserve definition, permit applications, political instability, commodity price projections, and estimateS of capital, operating, and maintenance costs, there are technical uncertainties. Bench tests provide us with a way of minimizing intellectual risks such as incomplete appreciation of process chemistry. Pilot plant programs should reduce the risks related to improper equipment selection, insufficient retention times, inadequate solids suspension, and other engineering/design issues. In addition to the rather traditional reasons given above for justifying a pilot program, there are others that deserve brief mention: 0

0

0 0

0

0

Large samples of product can be made for purposes of further testing for subsequent merit such as smelting, or for the recovery of byproducts; Product can be made for evaluation by potential customers; he owners of some commercial plants have derived great value from the guidance provided by pilot plants that operate intermittently or continuously on samples of plant feed; A pilot plant can be used to assess the technical and economic effects of changes in flowsheet. process conditions, or reagent choice that can be made in an existing operation; Pilot plants can facilitate the training of operators and supervisors. Tbe generation of some types of information needed to satisfy questions of a regulatory nature may only be possible in a pilot plant; If debt financing of a project is necessary, lending institutions and their technical advisors may require the results of a pilot plant program; Equipment designs that cannot be evaluated at a smaller scale can be developed and tested.

Sometimes, a hybrid between batch laboratory testing and a pilot plant can provide useful information without the expense of the latter. In some cases, madrematical modeling of laboratory data will suffice. At the time of this publication, modeling of gadliquid and vaporAiquid behavior for some unit operations such as distillation has become an accepted practice. However, modeling of processing steps for suspensions of solid particles is usually considerad to be the responsibility of equipment manufacturers. Suppliers of mixing equipment, for instance, usually have pilot-scale batch apparatus in which different impeller types and sizes can be rotated by variable speed dnves equipped with power measurement devices. Transparent

120

vessel walls allow visual estimation of degree of suspension of drc coarser pamcles. Experience gained by the manufacmr’s laboratory personnel in using such information to estimate the scale of commercial mixers then provides valuable guidance for future scale-up recommendations. Effortsto reduce costs will inevitably lead to increasing use of mathematical models based on anticipated particle size distribution, liquid density and viscosity, and desired level of homogeneity of the particle suspension. However, the very depeodence of models on underlying assumptions requires great diligence in ensuring drat the model faithfully represents the range of values that is likely to be experienced by each variable.

DEMONSTRATION PLANTS Another hybrid testing and evaluation technique is the “mini-” or “demonstration” plant. As the term implies, mini-plants are small relative to commercial capacity. but are continuous collections of apparatus that can be used to examine the performance of s e v d successive unit operations at low feed rates. Feed rates range from a few kilograms per hour to many tons per hour. The advantages of such p h t s include reduction in the risk of project development through identification of problems not observed in smaller plants or short-ducation campaigns. While larger demonstration plants can bc quite expensive, some mini-plants have relatively low equipment and labor costs. Some highly automated and instrumented smaller scale plants can be operated with only one or two technicians per shift and can be opcrakd for extended periods with little attention. However, mini-plants should only be designed and built by clever professionals with considerable experience. Successful smaller-scale examples include countercurrent decantation circuits and CIP cyanidation circuits. Some demonstrafion plants are conducted in highly specialized research and development facilities, while others are conducted on-site as a precursor to the development of a deposit.

THERESEARCHANDDEVELOPMENTENVIRONMENT Prior to the domestic mineral industry’s deep recession in the mid-l98os,many mining and metal producing companies had established corporate RdkD laboratories that were well-equipped and staf€ed with experienced technicians, chemists, engineers, scientists, and managers. Also, the United States Bureau of Mines (USBM) had extensive facilities and hundreds of professionals in a number of locarions across the country. All of these corporate and government entities were able to conduct many types of bench tests, to perform chemical analyses and mineralogical characterizations, and to d u c t pilot plant campaigns with endless combinations of physical and

chemical processing unit operations. There were at least 15 privately owned contract laboratkes with a total of about 60@700 employees and with broad capabilities for bench and pilot testing. Some of these groups specialized in fields such as iron ore or phosphate, but most conducted testwork with many physical separation techniques, flowion and hydrometallurgy. A few also had continuous grinding or slurry pumping circuits. Furthermore, most of the suppliers of equipment for crushing, grinding, classification, flotation. and mixing had pilot facilities supported by specialized laboraton‘es. The same was trite to a lesser extent of the developers and suppliers of flotation reagents and solvent extractants. This all changed as a result of fundamental shifts in coporate and government priorities. By the mid-199os. only a few mining companies owned laboratories and those had become narrowly focused. The USBM ceased all minerals-related research and process development and terminated nearly all of the technologists who had been involved in those activities. Most of the suppliers of equipment and reagents either curtailed or eliminated their product testing and technical services staffs a d laboratories. M a n y domestic conlract laboratories either closed or became much smaller and more aligned toward niche markets. However, a few wellestablishad and respected laboratories remain in the United States, Canada, and Australia The survival of these few private conlractors is vindication of the principle that testwork is best left to those who practice it daily in a very competitive b u s h e s environment wherein clients return only to the contractors who provide the highest quality services. The testwork and the subsequent interpretations and reports are judged by harsh standards, and careless performance is

121

not quickly forgotten. However, the quality and timeliness of work done by these laboratories probably is generally superior to that done earlier by the various mining company and government entities. Moreover, slulled and experienced people who have no financial interest in the outcome of a client’s project and who can convey bad news as well as encouragement conduct that work. Their product is information unfettered by prejudice or vested interest Unfortunately, the demise of so many technical organizations and the resultant loss of several thousand research personnel and process developers have irreversibly deprived the world’s minerals industry of a reservoir of expertise and judgment whost value is incalculable. Many mining companies no longer have high-level staff who can design or interpret laboratory and pilot plant programs so they must rely almost entirely on contractors’ laboratories and consultants for minimization of technical risk. Of greatest concern is the fact that some companies have already reduced technical staffing so drastically that they no longer are capable of directing and coordinating contractors engaged to solve problems or to develop projects.

CONCLUSIONS Due to the decreasing in-house technical capabilities of many project developers and operating companies, it is more important than ever before to exercise great diligence in conducting project research and development programs. Careful collection of a representative sample must be followed by competent and thorough bench testing, and, if necessary, by pilot plant programs designed and conducted by professioaals. The approPriate b a l m between levels of appropriate process development, cost, and risk reduction will have to be carefully considered for every project. In each case, however, a communicarive and constructive relationship between the client and the service contractor will ensure that time and money carefully invested at the outset will avert exptnsive blunders and will maximize the profitability of nuly worthwhile projects.

122

Bench-Scale and Pilot Plant Tests for Comminution Circuit Design John Mosher' and Tony Bigg'

ABSTRACT Ore characterhiion for comminution circuit design is a complex compromise between sample representativeness, sample top size, sample quantity. cost, and utility. Ore characterization for SAG milling is particularly complex due to the broad size range of SAG f e d , and by the varying mechanisms of breakage. 'zhis paper discusses bench- and pilot-scale testing for the m g e of comminution circuits commonly used in mineral procesSing plants, focuSing on design of AG/SAG circuits. The sample requirements, data generated, and the relevance of various test procedures are discussed. In addition to tests commonly used for conventional and SAG circuit design, procedures for some specialized types of grinding equipment are also discussed.

INTRODUCTION Ore h c ' 'on tests, both bench- and pilot-scale are required for optimum design and operation of comminution circuits. The role of ore characterization in circuit design is readily apparent. The circuit design must allow achieving the throughput and grind that is suited to the balance of the plant's equipment and capabilities,and must do so economically. operating at peak efficiency implies satisfying two conditions: minimum operating cost to attain the desired final grind (typically in terms of coI1sum8blesof steel and power), and the efficient use of installed capital. Overall, in terms of capital and operating costs,the most efficient circuit is the one that allows the greatest rate of return for a project. In addition to providing the gr#ltest rate of return for a project, the comminution circuit design should offer mbust operation with the range of ore types that the circuit was designed for. Notably.constructing a circuit with a gross excess of capcity is an inefficient use of capital, and will decrease the rate of return on the pht. ore characterhion tests facilitate benchmarking plant operations, evaluating grinding efficiency, enabling conDinuous improvements in circuit operation, and planning for future operations. Efficient use of installed capital does not imply that grinding circuits should be designed without appropriate safety factors; on the contrary, inclusion of approPriate safety factorsis likely to maximize the risk-adjusted rate of return. Selection of an a p p q u h safety factor is beyond the scope of this paper, but should collsidec the variability of the ore body, mtdwgkd sensitivity to grind size, circuit flexibility, maintenance planning. and expansion plans. Tbe sensitivity of the first three aspects can be investigated with bench- and pilot-scale testing. Design and operation of an efficient circuit (in terms of both capital and operating costs) requires knowledge of ore breakage and grinding characteristics. There arc a variety of techniques applied for ore charactexiation. Techniques used depend on the design philosophy and approach, the type of circuit being designed, and to some extent, tbe personal preference of the designer or opaator. In addition to the direct . 'on datacan be compared results of a test program, significant value can be added if the ore with a databsse of other results, in terms of both labotatory measurements and operating plant data.

' A.R. Macpherson Consukants Ltd.

123

Certain circuit configurations requite mote ore chamtemm * ‘onthan others. A conventional crusherrod mill-ball mill circuit typically requires less charactenzatl ‘on effort than an autogenous or SAG circuit. This is due in a large part to the fact that dre conventional Circuit perfonnanoe is less sensitive to changes in feed size distribution and ore hardness, and because the conventional circuit does not rely on w to act as media. In this regard, conventional circuits are atguably more robust. and are therefore less sensitive to feed changes. UdiLe rod mills or ball miUs which an?essentially constant power draw devices, SAG mill power draw is dynamic, and affected by changes in Opaating conditions. As the grinding performance of a SAG mill is strongly affectedby the O T ~characteristics (in terms of feed size distribution, hardness, and density). changes in feed ChSraCteristicscan rapidly change circuit operating conditions. This papa discusses a number of experimental proceduns now in use for bench- and pilot-scale ore hardness characterization for conventid and AG/SAG (throughout this paper. whenever a reference to SAG milling is made, it refers to both AG and SAG milling) mineral processing milling circuits, dong with other coasiderations in the ore characterizationprocess. Additionally, test procedures for some s p e c i d i d unit operations in mineral processing are briefly summarizbd.

TESTOBJEcfIvEs . One of the &reat challenges in SAG mill ore ‘OIL is that SAG milling accomplishes breakage by both impact and abrasion; typically a SAG mill product has a much gnatex quantity of fines than a rod mill product with an equivalent Pm. Additionally, autogcnous griadins media must survive in the mill environment (in terms of both competency and wear by amition). Ideally, ore characterization tests for SAG milling would accomplish the following: 0

Test particles over the en&

size range of SAG mill feed for both impact and abrasion breakage,

at energy levels expected in commer~ialmills,

Determinemediacompetency, Allow examination of steady-state mill load charactaisth (particularly for the buildup of a certain rock fraction or “critidy sized“ material), 0 Gemate a breakage versus energy level “map’’ fause in Circuit simulations, 0 Be conducted on a sample mass sufficientin relation to particle top size to allow repeatable, statistically robust test results, 0 Determine total grindingpowerrequired. 0

0

In addition to collecting the above data. charactenzatl * ‘on tests wwld ideally use a small sample mass. allow the use of core. and employ simple and expedient test procedures. The depth and breadth of ore characterization data desired is not always cmsktent With the prrctical oonsiderations. Therefore, d l ore testing and characterization for SAG Circuit design involves compfomistsover the ideal. Any test that measures somc aspect of rock herdness will have somc correlation to mill throughput. An acceptable test,however, must measure hardness with relevaace to commercial grinding, as well as being a robust pcedure capable of generating accumte and repeatable results. This paper will focus on test procedures,and not on design procedunsbased 011 test d t s . A numberof mill designers use the same or overlapping data sets in markedly Werent ways Nonetheless. some refaences to design procedures are relevant, and BIC unavoidable.

SAMPLE SELECTION The first consideration in design involves development of a sampling plan. Samples selected for testing should bear some logical relationship to a mine plan, pmctkalconsuch as sample access and cost will undoubtedly be factors in the decision, but samples should not be selected solely on the basis of avaiIabiIity. In sample selecti~n,the variability of the orebody and the Popomons of the ore types within the orebody should be considaed. when evaluating the results of variabiility studies, designing the comminution circuit based on the hardest sampk tested is g e d y folly. If the hardest sampk repsen6 5.10, or 209b of the ore body, this means that 95.90, or 80% of the time the capital assets in the circuit will be underutilized. It is equally fallacious to completely disregard the hardcr fractions of an orebody. W h e n determining ore blends for design, simpk mathematid averaging of hard aod soft ote breakage and grindability characteristics is not an unreasonable first approximation. As a rcsult of accumulation of harda ore types in mill loads and recycle loops, however, the effect of hard ore on milling rates may be snore pronounced than the average would indicate.

124

Ore sampling for operating plants is more Straightforward. Ore sampled should be representative of the ore that the circuit is milling, with the best method of sample collection being the samplingof the mill f d . Of course, the degree to which ore characten'stics can be extrapolated to other ore types is a function of the variability of OR within the deposit When drill con is selected as a feed for testwork it should be cartfully prepared by d i n g to give a natural size distribution.

TEST PROCEDURES FOR COMMINUTIONCIRCWTS There are two general classifications for ore characterizahioa: bench-scale and pilot-scale testing. Benchscale implies a smaller sample requirement Some bench tests a~ conducted batch, some are locked-cycle, and a few are operated fullycontinuously. Generally, the sample requirement for bench-scale charactcrizationis 250 kg or less. Some of the larger batch and locked cycle tests can require up to 1 mt of sample; while these tests are relatively large-scale, they are generally included with laboratory and benchscale charanerization procedures because they are not fuuy continuous. A pilot-plant program is charactenzed ' by two primary factors: fuuy continuous operation, and circuit operation with generally the same (or similar) unit operations and configumion as the envisioned full-scale process. The sample mass required is a function of the sample top size tested and the sizeNvoughput of the pilot plant equipment. A reasonable planning figure for AGSAG pilot plants is 5-10 mt per test condition. Pilot circuits with a finer feed size. such as a r t g d mill pilot plant, can often use much smaller samples.

. .

Labo~d&nchsakchllcteclpfioD In gtneral, comminution tests can be broken into two broad categories those that measure total mill power input, feed rate, and the resuiting o v d d l produdsize distribution (the work index approach), or applying a known energy input to a known particle mass and maswing the resulting breakage (singleparticle bre&age tests). Typically. work index tests start with a feed of a horn size distribution, apply a known amount of energy, and then mc8sun the product she distribution. Single particle breakage tests typically endeavor to quantify breakage as a function of energy input. Design methodologies vary, but typically rtly on scaling bascd on power input. computer models generally quire the input of fundamental bnalcage vs. energy input, but may use power based ore hardness work indices to scale the magnitude of grinding rates. A few tests present a hybrid approaEh (such as the Bond impact test), using a single-particlebnalcage test to calculate a work index. The work index approach has been historically used with excellent results, as a basis for designing conventional (several stages of crushing. followed by rod and ball milling) and A W A G mill grinding circuits. Simulation offas the expanded potential to more closely evaluate the effect of operational variables, and allows examination of the interaction between unit operations. Thae are numerous test protocols that have been used for comminution circuit design. Some tests have had WiGeSpread use, and been employed for the design Of a variety of different circuit configurations. Table 1 smmarhs salient points concerning bench- and pilot-scak programs with the most widespread use for SAG and conventional Circuit design, with more detailed treatment of test typedcategories in the S griading tests. and each has touched on a following sections. mere have been a numbex of past I E ~ ~ C Won variety of topics. This review endeavors to present each test in light of sample top size, sample mass required, type of test, and how well the test addresses breakage as it occuls in commercial SAG mills. Additionally, the deliverable of each test, whether qualitative or quantitative (in terms of work indices or simulation parameters) is discussed. In one way or anotha, all of the tests described are a compromise to the traditional pilot test program, and each test bas inherent strengths and limitations. Media Compdmq Test. At kast thne media competency tests have been widely used in AGISAG s i chalmers drum media comptency test, and Amdelcircuit design: Kilborn pebble competency test, M Orway advanced media competency test The Auischalmers test batch mills ore in a 6 by I-foot h, with the pmduct of milling qualitatively e v a l d The Amdel-Onvay test is conducted similarly, but adds impact testing on various size ranges of the mill products. Tbe Kilborn test is conducted in a smaller mill, and is a Semi-coatinuous test (ground ore is removed and replaced with fresh ore until an equilibrium condition is met). In genaal. these tests arc e v a l d qualitatively to suggest competency and circuit configuration. S i l ( 2 0 0 1 ) pvides more details concaniag intapreraton of the Amdel-Orway media competency test Essentially, these tests seek to evaluate the competaq of c o ~ ~ s eparticles r with a minimum of ore. ~n this respect they a~ a compromise of sample top size and sample mass required. Any of the three tests can

125

Table 1. Summary of Grindability Test Sample Requirements, Type,and Energy Level

Top Size Tert Pilot Testing

100-

Sample Requested

Closing Slze

I

Varies

I50

I

-1O,OOO/

Sample

Tvpe

Peak

UA'

I

Varies

Energy

1

Continuous

1

WID ~

Peak E ~

l%-TP

50

I

1.83

test

Media Competency

165

NIA

750

400

Batch

100

18

I .83

Bond Impact

75

NIA

2050t075

7.5

Single Particle

200

500

NIA

NIA

mrn Rocks 75

24

Single Particle

450

1400

NIA

Drop Weight

64

& Batch

MacPherson Autogenous SAG Power Index (SPU Bond Rod Mill Bond Ball Mill

32

1.2

135

100

Continuous

3.3

70

0.450

25

to

2

Batch

0.2

8

0.305

13

NIA Pm=1.7 1.2

u3

10

Locked Cycle

1.6

500

0.305

3.3

0.149

tO

4

Locked Cycle

0.6

--

0.305

F,=13

'Sample Consumed in licates sample actually used during a typical test, for an ore of moderate hardness vviith an Sa of -2.8, without repeats 'Per unit mass,based on particles in the largest size fraction

*

Ore Competeacy without the provide useful information about sample m a s q u i r e d for pilot testing. The tests, however, provide neither a direct indication of power requirements for milling, nor generate dam suitable for simulation. TWOof the three tests are batch tests; batch tests provide little insight to steady-state mill load composition.

Impact, Rod Mill, and Bail M U Tests. Fred Bond started development of the Bond grindability tests in the 1920s after he observed sh~rtcsmiogSin Using batch tests as a reliable indicator of energy requirements for closed circuit grinding. The Bond impact test was developed as a measure of crusher energy requirements. The tests predated large autogenous mills and were never intended to be used for SAG design, but many designers use the Bond indices in various capacities in analyses of overall power required for SAG circuits. The low energy impact work index (W,) is determined by a single particle test in which rocks subjected to progressively higher energy levels until hcture. Thc energy input at fracture is used to and ball mill (BWJ tests are similar semicontinuous (locked cycle) tests. calculate the IW,.The rod (RW,) Based on the rate and size distribution of product generation at steady state, in conjunction with known power input in a standard mill size at standard ~oditions,the work index is calculated. The work index can then be used to calculate lpindins circuit powa m p k w n t s (Bond, 1960). 'Ihe c o k t i v e use of the three indicts provide an indication of brealrsge over a large range of M c l e sizes. The BW,and RW, are well established robust, aad repeatable. 'Ibe scatter of the IW,makes this test most useful far evaluating the range of hardness as opposed to a design number. The work indices can be used directly for power calculations, and for qualitatively suggesting circuit configurationsthat will grind at the optimum power consumption. While the IW,test requires hale sample (20 particles), it can be difficult to get a representative sample of coarse OR with a low particle count., W i g to results with a high &gree of scatter. Also, sample d IUC~S),@cUlarty when sampling core. There is no selection can be discriminatory ( t o ~ harder measurement of the size dishibution of breaLage pmgeny, OT any hdkation of change in progeny size distribution with hacasing energy above first fracture enagy. h past development of SAG mill circuits, the ball mill work index (BWJ was often used as the basis for design, with predictable flaws. As previously described, a SAG mill effects bfealrage in three ways: impact, amition, and abrasion (wearing away of the surface for semi-autogenous grinding madia). The BW, does not address impact breakage of coarse particles, which may not correlate to en= requirements for fina particle breakage by abrasion. There are many examples of plants where the batl mill index is not a reliable indicator of SAG mill perfOnnaaCe.

Also worthy of mention are "comparitive Bond ball d work index" tests or "onecycle Bond ball mill work index tests." These tests are essentially batch grinds conductad in a SEandard Bond ball mill. Because they do not 8ccoullt for the effects of a s d y - s t a t e d bad, they are generally unacceptable For use as a design bask In certain circumstances, however, they can be quite useful for indicative variability testing. Yap etal. (1982) provides a review of the many comparative tests that have been used. All tests involve comparison of the batch grind d t s to the results from a reference sample, or the use of a correlation based on several samples. When the batch grind is conducted in a Bond ball mill. it is sometimes referred to as a "one-cycle Bond test." Rowland has published numerous d c k s 011 the use Of the Bond work indices for design of c o n v e n t i d circuits, most recently elsewhere in drese volume^. B o d work indices also play some tole in most design medrodologies. The design methodologies of most engineering companies. consulting companies, a d simulation programsrely 011 Bond data in some capacity or another. Drop Weight Test. The drop weight test was developad by the Julius Kruttschnitt Mineral Research Centre (JKMRC) and has three components: single j t d c l e breakage tests, a batch abrasion test on a unimodal set of particles (53 by 375 m),and a specific m~ty twt 011 co~fseparticles (22.4 by 19 mm). The single particle breakage tests (on sets of particles in five size ranges between 64 and 12 mm)are used to develop two items: SAG milling parameters aad the ceushet appmoce function.Rior to calculating SAG milling parameters, single @ C k tests data a reduced to produce a plot oft,,, (defined as a particle with a &meter 109b the size of the original particle size)vs. energy input Parameters ("A" and "b") are fittedto experimental data The two parameters arc not independent, and the product of the two (A x b) is the best meawe of o v e d hardaess;lower values iadicate a highex resistence to breakage. Implicit in fitting parametas to expaimental data is the assumption that particles of all sizes break in a like fashion. This is

127

typically a reasonable assumption, but is undoubtedly a simplification. A typical plot, and the form of equation for the parameters. is pccscnted in F i i 1. The abrasion test is conducted autogemously in a batch mill; the product size distribution of the test is to calculate the parameter "t,". s m a k viduts of "t," B1C indicative of a higher resistance to breakage by abrasion (Napier-Munn, et all%). In summary. the drop weight test evaluates an ore's response to both impact and abrasion breakage. Impact bmakage is evaluated for rocks as large as in any other test except for piloting. The test has the highest energy input of any of he tests described, and appmxhates the peak nominal energy levels observed in large commercial mius. The abrasion test has some shortcomings in that the test is strongly affected by sample surface roughness, and suffers from the drawback of defining steady-state breakage in a batch environment. The assumptions that all particle sizes and rock types have s b h breakage patterns (implicit in determining A and b values) can lead to e m for some ore types. As dte test r e k on single particle breakage and a batch test, there is no opportunity to evaluate the steisdy-state mill load composition.

0.00

0.50

1.a0

200

1.50

250

3.00

wFigrup 1 T y p i a l B m a k a g e v s . ~ M.pfor.n(hMcCke

While drop weight test results provide data suitable for simulation.they do not provide a direct measure of the power requifed for grinding h m one size to another. Additionally, the parameters generated are suitable only for the JKSimMet ~ ~ m p u t&hg/simulation ef paclage. (parameters do have value in relation to the overall database of tests, and raw brralcage vs. enagy data may well be suitable for other simulation work.) Required grindhg power for a circuit,however, can be estimated by predicting the power draw of the mills that meet design thmugbput and grind in circuit simulations. This approach is somewhat circuitous, and does not provide a benchmarlr for power efficient grinding. Aatogamas Mill Work Iada Test.The autogenous mill work iadex (AWJ test (the MacPherson test) was developed by Art MacphaMM while evaluating AGlSAG mill design at Amfall Mills Ltd. This test uses a 450 nun dry air-swept SAG mill with an 8% ball charge. The test is run fully continuously in closed circuit Mill load is sound mnuolled to B 2%% total charge Using a comput~ controller. in this test, the mill is ~ 1ContinuousIy 1 uatil sfeady-state conditions are met. Once steady state is attained, the bench-scale mill circuit is surveyed over a oab-bout period. At the completion of the test, the for comparison to the mill feed. Based on the mill load mass, density, and size distribution arc ~CSURYJ size disaibution of the feed and product, and a known power input, tbe autogenous mill worlc index is calculated. Duc to the shortfall of high energy impact events occuning in the bench-scale mill (relative to a full scale mill), the bench-scale work iadex for harder ote tvpes is coirected based on a correlation developed for efficiently operated full-de p h t s (Macphasoa1989).

128

In addition to providing an autogenous work index (which can be used analogously to the Bond grindability work indices), a number of semiquantitative data are produced by the test. As this test is nomnily conducted in a series including the Bond rod and ball mill work indices, differences between energy requirements for breakage at different particle sizes can be discerned. Along with the throughput and size distribution of the test product, the relationship between the work indices suggests the circuit configuration and power split for grinding at peak power efficiency. Additionally, the test allows evaluation of the steady- state mill load in twms of size distribution, bulk density, and m i n e d r w k type composition. These considerations are important for pradiction of the potential for critically sized material. A drawback to the MacPhersoa test is drat the test is limitad to parkles finer than 32 mm. This limitation is a direct result of the test using a reasonable sample size, both in terns of attainable impact energy levels and of the top size that can be fed to the mill. Use of a larger top size would require use of a larger mill, which would in turn require use of a larger sample mass to ensure that steady-state milling conditions are met. Also, the test does not produce 811 energy-breakage map suitable for direct computer simulation of MacPherson test results. The AW, is used to determine the power down to the ball mill feed size. Expexience in selection of the transfer size for dculabing the power is required. This selection is based on the observed ore characteristics in the bench-scale test, and after considering other aspects of the proposed full-scale design. The Primary consideratioa in this case is the substantial additional grinding power that a SAG mill requires to perform by tines generation (in compathn to a rod mill product with an equivalent 80% passing size); this issue is discussed in detail by McKM (2001). Designing to an actual T, size i n d of a "theoretical" T, size (accoUnting for the fines generation) runs the risk of designing too small a SAG mill.

SAG Power Ida Test. "he SAG Power Index Test (SPITest)is a batch test conducted in a 305 mm mill. The test is conducted with a 15% ball charge of 25 mm balls and a total millcharge of 24%. Feed for the test is prepred to have an F , of 12.5 m m (In inch), with the test run until the batch ore charge is milled to a P, of 1.7 m m (10 mesh). The time required to reach a Pmof 10 mesh is then converted to a power index via the use of a proprietary transformation (Starkey 1996, starkey & Dobby 1996). The SPI test has the lowest nominal peak enagy event of any of the characterization tests described, and also the lowest energy level in terms of jouledkg for the largest particle size in the mill. As such, the test is essentially an indicator of an om's breakage rtsponst to abrasion events. As with other batch tests, the test is limited by the fact that a steady-state mill load is never reactred. Minnovex' CEET program has been developed to use the results of the SPI test. At present, use of test results (reported as either a griading time in minutes or as a unit power hasis) is essentially in the CEET program, although results can be compand to others in the test database. Rock Mcchrdes Tests. Rock mechanics tests endeavor to define the mataial properties of the rock itself. This approach includes measurements such as the unconfined co-ve strength (UCS), point load strength, as well as a n u m k of direct stress tests. Many of these tests have been adapted from rock mechanics test programs, civil or mechanical engimhg tcsts, and mataial property tests. These tests appear to be quite attractive since they diractly measwe rock strengdr, instead of indirectly infemng rock strength based on grinding or breakage energy required. Direct meaammmts of rock breakage properties should meawe the energy required for -e, whik g h h g tests includes the inefficiency of energy application. Beannan (1989) provides a review of various rock mechlllllcs-based methods of strength testing. Taking a fundameatal approach to measuring these Properties is sound, but the approach has difficulty in accounting for rock impafections. In other wofds, rock sfrength typically has some variation with particle size. This variation may be a result of the distribution of macre or micro-imperfections in the nonhomogenous material, a result of weadrering. or other factors. Such imperfections may significantly reduce the energy requirements down to a Certain particle size. This change in arergy requirements for breakage (or alternately stated, a change in breakmge as a function of energy input) at different particle size is well documented. It is a weakness of not only rock mechanics tests. but otha test protocols as well. At present, rock mechanics tests are not widely used as a direct basis for design. A number of designers, howeve. do use test results to Categorize ores into g d classes for which certain circuit configurations are more apptopriate. Wot Testing .. Pilot testing has long been the benchmark for detemmmg the power requimments for a commercial circuit.

129

The scale of pilot testing has often been debated. For AG-SAG circuits, the scale of pilot testing ranges from appmximately 10-30 kg/hr to 500-5000 kg/hr. At the lower throughput range is the 450 MacPherson test, sometimes referred to as a "minkplant" (Siddall and White, 1989) due to the fully continuous, steady-state nature of the test. Note that the MacPherson test is conducted under one standardid set of conditions. At the higher end of the scale are systems based on 1.83 rn mills; this size mill has become almost industry standard for pilot testing. whie installations vary, the best of these installations are highly instrumented and connected with data collection system. Among the drivers for the development of the 450 mm mill test was the requirement to conduct testing that eccouated for steady-state load effects (and particulady critical-size buildup) while using reasonably sized core samples. While the 1.83 m pilot mill has the advantage of milling the full commercial SAG feed size range (other than those circuits without primary crushing, and the caveat that, for hard ores, the top feed size should be reduced). it follows that larger sample masses are required for testing. Larger ssmple ~llllsscsIVC required to reach steady state operations with the larger miU volume, and to a lesser extent. larger sample masses are required to maintain sample representativeness. A number of intermediates between these two continuous tests have been examined over the years (SiddalI, et. d.); periodically, the use of smaller pilot mills is revisited. Clearly, there is nothing inherently wrong with intamediate sizes, and intermadiate sizes offadifferent tmk-offs between sample top size, sample requirements, and similarity to full-scale operations. An advantage. however, of using either the standard full-size 1.83 m pilot plant or the 450 mm $ot plant test IVC the extremely large database (including correlation to full-scale installations)that exists for these two mill sizes. The 1.83 m pilot plants are operated in many laboratories and research centers arouod the w d d , including Australia, Canada, Chile, the United States, and others. The 450 m m test is currently offered by A.R. MacPherson Consultants. Pilot testing is useful for evaluating a range of Circuit Operating variables. For SAG mill circuits, the most common variables examined are ball charge. mill speed, pebble crushing requirements, SAG circuit closing size, and the power split between mill sactions (psuming that more than one stage of grinding is being conducted). The effect of feed size can also be evaluated (With some caveats). Typically. pilot testing focuses on determining Operating conditioa~that result in meeting the desired grind with the least power. Despite the demonstrated utility of pilot testing, not all variables can be tested at the pilot scale. Such items as grate open area. pebble port size and open area, liner and lifter design, and pulp lifter configuration can be tested to some degree but are very difficult to evaluate and scale-up from pilot data, and iue generally fixed. Assuming a representative sample (which has long troubled mill designers). the pilot program provides the best opportunity to develop a circuit design that minhbxs capital and operating cost. A pilot p r o m offers the least d e - u p and the greatest degree of certainty in pndicting commercial milling response. Therefore, the most robust design methodology likely includes pilot testing. Pilot testing. however, is only one component in an o v d program that should include variabiity testing, detailed bench-scale ore characterization, and simulationswork prior to final c h i t design. The ideal test scenario includes variability testing thrwghout the ore body. identification of samples of in(based on piqmtion of the onbody, mining sequence, and range of hardness), and the conducting thorough bench-scale charactenzatr . 'onof selectbd samples. Following the bench-scale program, preliminary circuit simulations would be conducfed to focus pilot work. Pilot work would use the sample of greatest relevance for the bulk of pilot plant testing, With smaller samples used to evaluate circuit response for otha ore types. Additionally. pilot plant d t s clill be used for computer simulations. only predictive models that account for the changes in mill diameter, spead, and other operating conditions are useful for designing from pilot plant data. Historically, pilot testing was almost always c o n d d for large grinding circuit designs. For smaller projects, the cost of sample collection. shipping, and pilot testing may be beset directed to more variability testing of the orcbody and to specifyins larger mills. In this case, larger mills am essentially an insurance policy against the uncertainty asso~iatedwith less detailed testing. For largecircuit designs developed with an objective of Opaating at peak efficiency in tenns of capital productivityand lowest operating costs, pilot testing is often essential. Recently, some notable high-throughput SAG circuits have been designed and commissioned without pilot tcsting. The Batu Hijau SAG circuits, for example, were sized and e v a l u a with a combination of powa-based bench-scale tests (iacluding the MacPherson 450 nun mill and the Bond series). some rock mechanics tests (such BS ucs d g ) , and mop weight testing. The final design was developed and a n a l y d by Fluor, MacPhason. and JI(SimMet Circuit simulation techniques. Power-

130

based methods and simulation-based methods were both successful in predicting circuit throughput (McClaren, D. et. al. 2001). selection of psld Eqaipwnt. Designing an integrated pilot-plant presents more of a design challenge

than many people appreciate. Typically. the pilot plant design will face incompatibility of equipment types selected for the flowsheet. One example is incompatibility between the SAG mill and ball mills selected for the pilot circuit (in terms of capacity for either throughput or griad). Another example of the problem relates to using cyclones for closing grinding circuits; attaining a coarse closing size with a small (75 mm or smaller) cyclone is difficult. This difficulty is not insurmountabk; there are a number of valid approaches to solving this problem, some of which rely OIL maintaining large recycle loads around the cyclone. Additionally, material handling issues are often formidable. Pumping of coarse screen undersize streams with pilot plant-sized pumps can be a challenge. Most l a b o d e s familiar with pilot testing will have solved problems of this nature in the past, and are invaluable in proper set-up and design of the pilot plant. Often,grinding pilot testing can be conducted in concert with otha metallurgical testing. To provide some level of decoupling between the milling and metauwcal tests, some degree of surge capacity is recommended between the circuits. Dtcoupling the two circuits can minimize interruptions to down stream processes from grinding circuit upsets, and having to d c t griading circuit operations due to short duration process upsets in the metallurgical circuit Such decoupling must be well thought out to consider slurry suspension, materials handling. and any effects of slurry aging a oxidation. The latter consideration will typically dictate that any slurry hold-up be of short duration.

Elements of a !hccedd Pilot Test pr0g.m. A successful pilot test program entails far more than simply assembling pieces of equipment in the saquen~edepicted on a flowsheet. Once pilot plant equipment has been selected, the pilot plant s h o u l d be commissioned with an eye toward simple and accurate sampling. ease of equipment cleanoutMushing between samples, ease of clearing plugged lines and pumps, and simple and accurate measurement of Operating data. Of course,the ability to safely operate the plant should be of paramount importance. In actuality. assembling the key pieces of equipment for a typical pilot plant is far easier than assembling a traiwd.welldirectad work force that understands how to operate the pilot plant, how to w m t l y sample the p h t , how to d y z e pilot plant samples, and how to interpret pilot plant data. In orda to collect good pilot plant data, the pilot plant engineer must have a thorough knowledge of the desired Operating conditions for the circuit W e not absolutely essential, a good pilot plant will have automatic data recording of such variables as mill speed mill load, beariag pressure, mill noise. power draw, etc. In some cases. pariicdady when a comminution pilot plant is integrated with a metallurgical pilot plant, on-line analyzers may also be useful. Pilot plants can be SUCCCSSfully operated with a minimum of instnrmentaton and control. but doing so requires progressively h i e skill levels of the operating engineer and crew. In short, assembling the approPriate equipment for a pilot program represents only a small &on of conducting a successfulprogram. Pilot testing offers the best o v d data package, but still falls short in terms of generating a breakage otba than those tested.The pilot program versus energy kvel map, and in direct extrapolation to c o d t i o l l ~ requires the largest mass of sample, and requires a large Commitment of test equipment and skilled operators to conduct. Resuming a good experimental design, howeva. pilot results are well suited for computer modeling, and the d t s of the pilot test can often be used for simulation of the full-scale plant. Caveats of Wot Test Rog.ms When pconsider.

a pilot program. then are several important items to

sample selection. Since a large bulk sample is often collected for pilot programs, it is cxucial that ~ ~ ~ how the sample used for pilot testing Elates to the project designers and future o p a t understaad c more for convenience than as the best the balance of the ore body. Often, pilot samples ~ r collected repxscntation of the orebody. 0 Scak effects. Some variables do not scale particularly well from the pilot mill to the large-scale mill, or cannot be effectively tested at the pilot-scale. One example is top feed size. While pilot-scale testing can generally evaluate the entire size range of a commercial plant, feed sizes are generally minus 150 mm, this size is may be too coarse for extremely hard ores. In a similar vein, evaluating 0

131

whether or not to go Without a primary crusher and d l run-of-mhe ore cannot typically be tested at the pilot scale. 0 Feed Preparation. It is quite difficult to describe mill feed size distributions with a single number such as the F,. 'Ihe feed size distribution to the mill should match the design (or the best approximarion thereof) throughout the size distribution, not just at the Fa. 0 Feed Characterization. ?he pilot test sample should be fully characterized at the bench-=&. This allows better selection of pilot test conditions, as well as providing a benchmark for pitot operation. Such characterization data can be useful when troubleshooting the pilot plant. 0 Classifier Performance. Often, classifiers used in pilot plants are ill matched to the flowrates that they handle. As classifier performance has a significant impact on overall circuit performance, this parameter should be closely monitored. Evahaation of Hot Data. When evaluating pilot data, it is important to understand how the program was conducted. In many cases, circuit configurations may have bnited the operating window for the test. For example. for soft ores, feed rates may have been mificially restrictad by feed belt capacity. Alternately, the same mill canditions of total volumetric loading may not have been maintained between ore types, even if other test variables are relatively constant. Generally. even internal recycle streams should be analyzed for Size distribution and solids density. The screen sizts used for each strcam should be the same, and should typically represent the full root of 2 Screen series. Coarse st~eamsshould be sized completely down to the finest size. Any minor incremental savings gained by restricting analyses and Sizing of pilot plant samples e l 1 be lost by extra work and interpretation during mass balancing. Detaild sizing around classification devices is particularly important. For SAG mills, the mill load should be thotoughly charactaized in terms of size distribution, void spacc, density. and bulk mincral composition. Any differencesbetween the feed and mill load mined composition should be noted.

Mill Power Mcsmxnent. Most pilot plants include a measunmcnt of power draw. It is important to note just exactly what power draw is being measured. The most convenient power draw measurement is at the motor taminals. Because pilot mills often have ovefsiztd motors relative to their power draws, the power may be drawn at a very inefficient point on the motor efficiency curve. The most thorough pilot piants meastux net power input to the charge (generally by measuring torque), or consider motor inefficiency aad transmission losse~by conducting Prony brake tests.

SPECIALIZED TEST PROCEDURES There a number of unit operations that are ikqueutly employed in specialized mineral processing applications. Some operations. such as regrind Circuits, are almost universally employed in specific applications (most complex sulfide flotation c e t s have a regrind c h i t ) . Other applications, such as High Pressure Grinding Rolls or Hicom high-g d h g , use specific equipment types for which other test procedures are not necessarily applicable. For most of these spec~alizedoperations, specific machine- or application-specific test protocols are required.

Regriad/Fiacciiading Fine @ding and regrind circuits comprise a broad subset Of m h a d processing milling applications. of a Concentrate after at least oltc stage of concentration (typically Regrind g e d y implies re-g flotation) in order to improve liberation for subsequent qwations. Regrind applications typically have a final gnnd size with a P,of less than 75 pm. Fine as a broad category.includes milling to Pa's of less than 75 pm, and often down to sizes of less than 10 pm. There are several universal considerations when conducting tests for fine g f k h g applid01E: 0 Use media of the same top size. steady state composition, and material as anticipated for the find design. Ball size distributionhas a large effect on the grinding efficiency, and on the shape of the mill product size distribution. If closed circuit milhg is being consideraQit should be conducted in the pilot plant, even if it is on a semi-continuousbasis. The method of size analyses must be defined. It is beyond the scope of this paper to discuss and c o q a r e the results and pros aod cons of sub-sieve analyses, lasa diffraction analyses, sedimentation techniques, or Cyclosizcr sizing. Suffice to say that Cyclosiza data is generally the most consistent

132

with conventional screening, but is best conducted only for ores having minerals of uniform density. Most practical fine sizing techniques only indirectly measure size, so it is important to compare results on the same basis. Actual feed to the full-scale regrind circuit may vary considerably as plant feed, primary grinding, and flotation conditions vary. Testing for several fine-grinding techniques is discussed in the following sections. In general, scoping tests (typically grind curves plotting energy vs. P a can be conducted with 10-100 kg of material, while pilot programs would require on the order of 50-2000 kg of sample. Conventiod Ball Milling. Although the Bond ball mill work index (including work index tests conducted at a very fine closing size) can be usad as the basis for unit power requirements, such an approach is useful only for a gross approximation of grinding energy. Efficiency factors used with Bond's third thexry equation are only approximations, and are often not accurate predictors of grinding power. Pllot testing can be conducted in relatively small, fully-continuous mills. The major difficulty for these types of tests is accurately measuring the power draw of small mills, followed by the difficulty of matching

cyclone Optrating conditions and throughput to the mill throughput and desired final grind. S t i m d M i l k While each of the specific units in this broad category have key differences and distinctions, they are similar in that they all employ media stirring with a mill with a vertical shaft. This category t o r (SMD) mills, as well as includes devices such as Metso's Vertimills and Stirred Media M Metprotech's stirred mills. Each mill type. has different mechanisms for stirring media, and in some applications, certain mills have gained more market share. Test protocols are also generally similar. At the coarsec end of regrinding, initial power estimates ace sometimes based on empirical savings over conventional mills. Occasionally. some mill vendors and designers will use batch-grinding tests fo~dss?,on. Several of the mills of this type. have small pilot units, and continuous pilot testing provides the most refined cstimatt of power requirements for a specific application. While small-scale pilot tests for these mills are sometimes not directly scaleable, in conjunction with a dambase of existing operations, good estimates of full-scale performance can be made. Ispmills The MIM Isamill is a continuous, pressurizd, horizontal stirred mill, with media stirring accomplished by discs instead of pins. The mill was developed by MIM. and is based on Netszch milling technology. As such, small pilot-scale Netzsch mills (with proper instnunentation) can be used for pilot testing. Typically, an open circuit grindcurve is developed with progressively higher energy inputs. Based on thcse data. pilot tests can be conducted. In conjunction with specific power measurements, the results of these tests can be used for design.

HighPm~mutGriadingR&(APGR) KruppPolysius developed the HPGR technology, and markets the unit in several mineral processing applications. Both laboratory and pilot units erist The laboratory unit is the called the LABWAL laboratory high pressure grinding roll. It is operatsd in h h test mode. The data logging associated with the unit allows the calculation of specific throughput rate. required grinding force. and required specific energy input for the generation of a cutain product fineness. The data can be used for IKSimMet modeling. Largex units are available for pilot testing. HicomNUt8tiEgMill The Hicorn mill is a centrifugal mill that employs a high speed rotation about a nutating axis to grind in acceleration environments of a 3 0times gravity. The unit has found commercial applications in milling of d i a m d ores,and is being investigated for a variety of other applicptions (including both wet and dry grinding). Two stages of testing are geaaally employed. The first stage is generally conducted in a batch laboratory Hicorn mill. ?his stage of testing is conducted to scope the effect of various operating variables, and to dcvelop a grind curve. Morc refined estimates of required grinding power can be made by fully continuous closed-circuit pilot testing.

OthtrRehttdRacdues A variety of other s p e c i a l i d pieces of equipment are used for various comminution functions. Of these, of greatest interest are those unit opaarions that focus on selective grinding. Of these, ateition cells and

133

scrubbers are most common. While both units are used in disaggregation and selective grinding applications, tbeu mode of operation is very different. Attrition cells an typically used to impart a high intensity scrubbing action, and are often used to clean surfaces or remove clayey or less competent minerals in subsequent size separations. Scrubbing is less intense, and is a combination of slurrying and light autogenous grinding. In both cases,initial test work is often conducted in batch mode. with tbe results expressed in terms of residence time instead of specific power requirements. Batch tests should be regarded as indicative only. The next step for either unit is continuous pilot testing. Pilot testing allows variable residence time as a function of particle size and short circuiting.

CONCLUSIONS This paper has discussed a range of bench- and pilot-scale testing for mineral processing comminution circuits. The focus has been on conventional and AWAG grindii circuit design, with emphasis on AGISAG circuits. Ore characterization for AWAG circuits is considerably complicated by the larger number of operating variables, and the significant effect that feed vahbility has on circuit performance. There are a variety of bench-scale -c . 'on tests availabk, and most tests are ultimately a compromise between sample top size, sample mess, test enviromnt (in temns of batch, semicontinuous, or continuous), breakage mcchama . and energy kvel. Roperiy developed pilot-scale test programs can increase the certainty of circuit p a f m predictions. pilot testing is the capstone of comprehensive circuit design programs. In pilot testing, availability of appropriate equipment, sound test design (to include selection of test variables. plant sampling, and data analyses), and skill in conducting the program and the results ~ a r r yapproximately equal weight in conducting a successful pilot program. When conducted in conjunction with thorough bench-scale ore characterization,variability testing of the onbody. and complemented by circuit simulation, pilot testing offers the most robust basis for final plant design. In comparison to pilot testing, bench-scale testing allows evaluation of more ore types than typically possible in the pilot plant, and with mnsiderably less sample mass. Again, there are trade-offs betwen sample top size, sample mass, and test relevance, particularly for SAG milling. For comminution circuits with a smaller topsize feed, pilot programs c ~ be n conducted with significantly less sample mass than required for a SAG circuit. Ultimately, the most robust test program includes variability testing of an orebody, thorough bench. .on of samples of interest (either nprcsenting large parts of the orebody, initid scale -c . . production, or problem ores), pilot testing (to include bench-scale charactenzatlon of the pilot sample), and simulation of both pilot-plant and bench-scale test d t s . It should be evident from the discussion in this paper that no single approach adequately addresses all aspects of ore characterization,that all tests have inherent compromises, and that more comprehensive test programs an an exercise in risk reduction.

REFERENCES Anon., 2002. Glossary of Comminution and Ore Hardness Terms. Lalrefield, Ontario: A.R. MacPherson consulmts Ltd., 2002. Bearman. RA., Pine, R.J., and Wills. B.A., 1989. "Use of Fracture Toughness Testing in Characterising the Comminution Potential of Rock". Today's Techwlogy in the Mining & Metallurgical Industries, Kyoto, Japan 2-4 October. Published Loadon IMM, MMJ, pp. 161-179. ," ChemicalEngineering, Vol. 6,(Rev. 1961 Bond, F.C.. 1960. "Crushingand Griading c a l c ~ ~ a sBritish by AK Pub. 07R9235B). MacPherson. AR., 1989. "Autogenous Gnndmg. 1987-Update," CIM Bulletin, Vol. 82, No. 921, Jan., p ~75-82. . Man, Y.T. & S. M m l l . 1997. "Using Modelling and Simulation for the Design of Full Scale Ball Mill Circuits" Minerals Engineering, Vol. 10, No. 12, pp. 1311-1327. McClaren. D.,Mitchell. J., S e i l . J.. and Lansdown, G.. 2001. "Ihe Design, Startup. and Operation of the Batu Hijau Concentrator," Fkceediings Autogenous and Semi-Autogenous Grinding 2001, UBC, Vancouver, BC, Canada. McKen, A., Raabe, H.& Mosher, J. 2001. "Application of Ojbxatkg work Indices to Evaluate Individual Sections in Autogenous-SemiautogenoudBd Mill Circuits." PhKXedings Autogenous and SemiAutogenous Grinding 2001. UBC, Vanmuver, BC, Canada

134

Masher, J.B. & Bigg, A.C.. 2001. “SAG Mill Test Methodology for Design

and Optimization,”

Roceedings Autogenous and Semi-AutogenousGrinding 2001. UBC, Vancouver, BC, Canada. Napier-Munn, TJ et al, 1996. Mineral Comminution circuits, Their operation and Optirnisation. UQ/JKMRC, Qld, AWL, p ~8.1-85 Siddall, B., G.Henderson, and Brian ht.land, 1996. %ctors Influencing Sizing of SAG Mills from

Drillcore Samples.” Proceedings Autogenous and Semiautogenous Grinding 1996, UBC,Vancouver, BC, Canada, pp. 463-473. Siddall, G.B., and M. White, 1989. ‘The Growth of SAG Milling in Australia.” Proceedings Autogenous r , Canada. pp. 169-179. and SemiautogenousGrinding 1989, UBC, V a n ~ ~ u v eBC, Starkey, J.C.. 1997. “Getting Mom from Drill Core Rehhmy SAG Design,” Proceedings Randol Gold F o 1997. ~ pp. 67-72. Starkey, J.C. & G. Dobby, 19%. “Application of the M i v e x SAG Power index at Five Canadian SAG Plants,” Roceedings Autogenous and Semiautogenous Grinding 1996, UBC, Vancouver, BC, Canada.pp. 345-360. Yap,R.F., J.L. Sepulveda, and R Jauregui, 1976. Deteamination of the Bond Work Index Using an ordinary Laboratory Batch Mill.” In Design and Installation of Comminution Circuits, SME. pp. 176-203.

135

The Selection of Flotation Reagents via Batch Flotation Tests Philip Thompson'

ABSTRACT Batch bench scale flotation tests are the most common method for the selection of flotation reagents during flowsheet development. However, selection of the best reagent for the ore type and flowsheet configuration is often overlooked once a workable reagent suite has been developed. This paper describes methods for reagent selection using both an experience and a statistical approach. A brief description of reagent functions and the types of bench scale tests available for reagent selection will also be presented, along with the limitations of these tests. A case study example of a commercial application will be discussed. INTRODUCTION The selection of a suitable flotation reagent scheme can be a daunting task due to the abundance of reagents available to the test engineer. This is particularly true when a new orebody is being developed. Statistical methods that have been developed in the last 50 years can often aid in reagent optimization once a reagent scheme has been delineated, however, the initial screening process for reagent choice is usually based upon the experience of the test engineer. A successful testing program begins with a clear statement of the study objective. The selection of a reagent scheme for a new orebody will require a much more extensive test program than will the optimization or modification of a reagent scheme for an existing operation. Statistical analysis methods are much more suited to modification of existing schemes due to the limited number of variables and the well defined parameters. This paper focuses on the selection of a reagent scheme for a new sulfide orebody, although the techniques discussed may be applied to other flotation systems such as coal and other industrial minerals. TEST PREPARATION A successful flotation testing campaign on a new ore begins with a search of the literature. A good starting point for the selection of a reagent scheme is often the evaluation of reagent schemes that have been previously developed for similar ores. Although caution should be used when applying existing reagent schemes to a new orebody, this approach can often provide a good starting point. Reagent vendors usually posses a wide range of experience and should be consulted early in the test program. The selection and documentation of ore samples is an extremely important part of the testing process. Many flotation projects have been compromised due to poor sample documentation, careless preparation, and improper storage. Diamond drill core samples are preferred when testing a new orebody. These types of samples are easier to store and preserve than reverse circulation or bulk samples. Core samples usually resist oxidation better than other types of samples, but care should be taken to ensure that drilling or cutting fluids do not contaminate the core. Several drilling companies offer special procedures for drilling core for metallurgical testing purposes. Most sulfide ore samples should be stored in a freezer during the testing campaign to prevent oxidation. This is particularly important for massive sulfide samples. 1 Dawson Metallurgical Laboratories, Inc., Salt Lake City, Utah.

136

Sample preparation techniques should minimize the amount of fines that are generated. Although it is often impossible to accurately simulate full scale crushing and grinding operations (particularly SAG or AG milling), stage crushing of core sample to either 10 or 20 mesh using a rolls crusher in closed circuit with a screen will minimize the amount of fines generated during coarse sample preparation. Test charges weighing one (1) or two (2) kilograms are usually used in batch tests, although 10 or 15 kilogram charges may be used if the ore is low grade, or if a significant amount of rougher concentrate cleaning is anticipated. Grinding is performed in sealed batch rod or ball mills. Rod milling is usually preferred unless very fine grinding is required. Laboratory rod mills do not generate as many fines as ball mills. Occasionally stage grinding will be required to minimize the generation of slimes from ores that contain soft, easily slimed minerals such as chalcocite and argillite. Fresh reagents should be used. Most reagent vendors supply small samples with specific lot numbers and expiration dates. Reagents with expired dates should never be used. This is most critical for liquid reagents that may oxidize or evaporate upon prolonged storage. Batch laboratory flotation machines are usually sized for 500, 1000, or 2000 gram solid sample charges, although larger batch test units capable of testing 15 kilograms or more are available. These are conventional mechanical machines. Column cells have gained wide acceptance in cleaning applications, but their use in preliminary batch flotation testing is not common, due to their specialized application and relatively large sample requirements.

FLOTATION REAGENT TYPES AND FUNCTIONS Flotation reagents may be classified into the following categories: 0

0

0

0

0

Collectors or promoters. These reagents react with the valuable minerals, rendering them hydrophobic for collection into the froth phase. The term “promotor” has historically been used for reagents that act as auxiliary collectors, but in current usage the terms collector and promoter are interchangeable. Depressants. These reagents react with undesirable gangue minerals that collect with the valuable minerals. They render the gangue minerals hydrophilic and prevent their collection into the froth phase. Activators. These reagents react with valuable minerals that cannot be rendered hydrophobic by collectors without prior surface modification. Pulp dispersants. These reagents reduce slime coatings on valuable minerals via their dispersion back into the pulp. Frothers. These reagents reduce the surface tension of the pulp and create a stable froth into which the collected minerals float.

Some reagents may perform more than one function depending upon the dosage and the ore type. For instance, sodium hydrosulfide (NaHS)is an activator for metallic copper, but it can also be used as a depressant for arsenopyrite and other sulfides under certain conditions. Many collectors have frothing characteristics, and some frothers have mild collecting power. The following sections briefly describe the more common reagents used in sulfide flotation. More detailed information on flotation reagents is readily available in the literature (Avotins, Wang and Nagaraj 1994, Nagaraj 1994, Suttill 199, Bulatovic and Wyslouzil 1985). Collectors Xanthates. These collectors have been used for over 75 years and are still in use today. Amy1 xanthate is a powerful, relatively non-selective reagent that will collect most sulfides. It is commonly used for bulk sulfide flotation, where all the sulfides are recovered, such as gold bearing pyrite ores. Shorter chain xanthates such as isopropyl, butyl, and ethyl xanthate are more selective and are often used in flotation of copper, lead, and zinc ores where selectivity against pyrite is important.

137

Dithiophosphates. These reagents, also known as promoters, are usually more selective against pyrite than the shorter chain xanthates. They are characteristically weaker collectors than xanthates, and some ores require the addition of both dithiophosphate and selective xanthate for acceptable recovery. Several dithiophosphates have been developed specifically for flotation of gold that is present as either liberated grains or associated with sulfides such as chalcopyrite or pyrite. They are also used in the flotation of lead, zinc, and silver sulfides. Thionocarbamates. This class of collector is very stable and has a long shelf life. It was developed specifically for copper sulfide flotation and is selective against pyrite in alkaline pH pulps. Recent modifications have showed that modified thionocarbamates allow for selectivity against pyrite at pulp pH values as low as 7. These reagents are also excellent collectors for easily slimed copper sulfides such as chalcocite. Dithiophosphinates. These relatively new reagents are extremely selective against pyrite and other iron sulfides. They are widely used in lead flotation and have recently found acceptance in copper and zinc flotation, particularly complex polymetallic ores and massive sulfide ores. These reagents are quite weak and are often used in combination with selective xanthates or dithiophosphates to achieve acceptable metal recovery. Trithiocarbonates. This class of reagents was developed in the last 15 years. They exhibit very high selectivity against pyrite in the flotation of copper ores. Mercaptobenzothiazole. These reagents were developed to improve recovery from tarnished or semi oxidized copper and zinc ores. They are often used with a selective xanthate to improve recovery. Mercaptans. The use of mercaptans has generally been limited to specialized applications due to their strong objectionable odor. They are used in copper circuits with selective xanthates. Xanthogen formates. These reagents have been used to float copper sulfides in both acidic and basic pulps. They have been used to recover cement copper and precipitated copper sulfide from acidic pulps. Monothiophosphates. This class of collector exhibits strong collecting power for sulfides at low pulp pH values down to 2. It is also used with dithiophosphate to float copper at neutral or alkaline pulp pH. Flotation oils. Reagents such as diesel or fuel oil are used in combination with a variety of collectors for molybdenite flotation. Activators Sulfuric acid. This reagent is widely used to enhance the flotation of tarnished sulfides, particularly gold bearing pyrite in bulk sulfide flotation applications. Acid is usually added to provide a pulp pH of no lower than 5 . Small amounts of sulfuric acid will often destroy all selectivity in complex selective sulfide flotation circuits. In fact, tarnishing or mild oxidation of complex sulfide samples will often release minute amounts of sulfuric acid and render the sample useless for selective flotation testing. Copper sulfate. Tarnished sulfides are often activated with copper sulfate in alkaline or neutral pH pulps during bulk sulfide flotation. This reagent activates sphalerite in basic pulps, and has been used to activate orpiment. Sodium sulfide. Sulfidizing reagents such as sodium sulfide (Na2S) and sodium hydrosulfide (NaHS) have been used to activate metallic copper from ores and slags. It is also used to activate cerrusite. It has been used occasionally to enhance the flotation of tarnished sulfides during bulk sulfide flotation, and also for activation of lead carbonate. Lead nitrate. Lead salts such as lead nitrate or lead acetate have been used historically to activate stibnite. However, the use of lead salts in flotation pulps has been essentially eliminated in recent years due to environmental considerations. Depressants Lime. This reagent is the most common depressant used in the mineral industry. It is effective for depression of pyrite, arsenopyrite, and pyrrhotite. Some valuable minerals such as

138

galena and liberated free milling gold may be depressed by excessive amounts of lime. Pyrite is the most common sulfide gangue and this mineral is sensitive to both the hydroxide and calcium ion. In some cases a pulp pH of approximately 8 or 9 is sufficient for pyrite depression. Sulfur dioxide based reagents. Sulfite reagents such as sulfur dioxide, sulfurous acid, bisulfite, and metabisulfite are commonly used to depress pyrite, particularly in copper circuits. There is published evidence that these reagents actually improve copper recovery while acting as pyrite depressants (Broman, Hultqvist and Marklund 1985). These reagents will also depress pyrrhotite. Ammonium bisulfite is used to depress copper activated pyrite in copper circuits. Cyanide compounds. Sodium cyanide is an extremely powerful depressant for pyrite. When used in excess this reagent will also depress valuable minerals such as copper, lead, and zinc sulfides. It will also severely depress the flotation of liberated gold. Mixtures of zinc sulfate and cyanide, usually in a three to one (3:l) ratio, have been used successfully to depress pyrite from silver sulfide bearing ores. Zinc sulfate. This reagent is used to depress sphalerite during galena flotation from lead and zinc bearing ores. Organics. Reagents such as dextrin, starch and lignin sulphonates are used in combination with cyanide, sodium sulfide and zinc sulfate to depress pyrite, pyrrhotite and arsenopyrite from massive sulfide ores. Carboxymethylcellulose (CMC) is often used to depress talc during the flotation of copper ores. Sodium sulfide. This reagent, alone or in combination with sodium sulfite, has found application as a depressant for zinc in ores containing both copper and zinc. Sodium sulfide is also used to depress copper during the separation of molybdenite from copper concentrates. Oxidizing agents. Pulp oxidizers such as dichromate, permanganate and peroxide have been used historically in specialized applications, such as the depression of lead sulfide during copper flotation from ores containing both lead and copper. These reagents will also depress pyrite to some extent, however, they are not widely used currently due to environmental concerns.

Pulp Dispersants These reagents are used to disperse slime coatings from mineral surfaces. Their action is based upon the surface area of minerals rather than mineral type. The most common dispersants include soda ash and sodium silicate. Tripolyphosphate is used occasionally to control particularly troublesome slimes such as montmorillonite clays, however, the dispersing action of this reagent is powerful and often irreversible. Frothers Although there are many frothers available to the test engineer, currently marketed frother types are usually alcohols, polypropylene glycol, or polyglycol ethers. The most common frother is a methyl amyl alcohol known commercially as MIBC (methylisobutylcarbinol). This frother is relatively weak, but it possesses the least collecting power. Frothers are often custom designed for a particular orebody and are proprietary mixtures of the frother types mentioned above. Other types of frothers such as pine oil and cresylic acid have some specialized applications in industrial mineral flotation, but are not generally used in sulfide flotation circuits. EVALUATION OF BATCH FLOTATION TESTS Two important criteria for evaluating a reagent scheme are collecting power and selectivity. The collecting power of a scheme is determined by recovery and the selectivity is determined by the grade of concentrate produced. The two types of tests that are most commonly used to evaluate these criteria are rougher kinetic tests and multiple stage cleaning tests. Rougher kinetic tests These types of tests provide a significant amount of information in a single test. Some preliminary grade and recovery relationships can be established and flotation times can be delineated in a single test. Multiple concentrates, taken over a measured time period, also provide better

139

metallurgical balances than tests that generate only one or two concentrates and a tailing. Grind fineness optimization, reagent type and dosage, and reagent addition points (grind, single point conditioning, multiple stage addition, etc.) can all be established with this type of test. The classical evaluation of rougher kinetic data has been the generation of a rate curve that fits the first order kinetic equation (Arbiter and Harris 1962) indicated on the following page:

In (CJC) = kt Where C, is the concentration of valuable mineral in the pulp at the beginning of the test (zero minutes), C is the concentration of valuable mineral in the pulp at time t, and k is the rate constant. Plotting In (CJC) against time produces a straight line with a slope equal to the first order rate constant. This evaluation is used extensively in fundamental flotation research to determine the collecting power of a reagent scheme because the greater the rate constant, the faster the flotation kinetics. For more practical evaluation of reagent schemes, the author prefers the incremental grade approach that was originally developed to determine rougher flotation time. The grade of the individual timed concentrate samples and the cumulative grade of the overall concentrate are plotted against time. The cumulative recovery is also plotted against time on a second axis. Flotation is considered complete when the incremental concentrate grade curve crosses the head grade of the ore feed to the test. At this point no additional concentration from the ore is occurring. The time at which this intersection occurs determines the cumulative concentrate grade and cumulative recovery for the test. This evaluation is illustrated in figure no. 1. Staged reagent additions and scavenging can also be incorporated into this test if rougher recoveries are low.

25 f i 100 90 80

20 a 0

70

8 15

60

8

=-

$ 0

50

2

U

c 10

40

0

30

; Q

0

4

0

20

5

10

0

0 0

1

2

3

4

5

6

7

8

9 1 0 1 1 1 2 1 3 1 4 1 5 1 6

Rougher Time, minutes

Figure 1 Rougher flotation kinetics for a copper ore using incremental grade evaluation Multiple stage cleaning tests This type of test is used to determine the potential for upgrading concentrates. Usually rougher concentrates from a test are combined and subjected to multiple cleaning stages. This type of test is usually performed when a few promising rougher reagent schemes have been identified and

140

refined. Factors such as concentrate regrind, pulp dispersion, and additional collector or depressant addition during cleaning must be addressed in these tests. Relatively large scale rougher tests using up to 15 kilograms of sample are occasionally required for low grade ores to ensure enough concentrate for cleaning. Results are evaluated by comparing the recovery at a fixed concentrate grade for various reagent schemes. Flotation time in each cleaning stage is usually not optimized due to the fact that cleaner flotation kinetics are characteristically rapid and most commercial cleaning circuits that use mechanical cells are oversized. An example of grade recovery evaluation for a silver ore is illustrated in figure no. 2.

-ReagentSchemeB

17,500ppm con grade 0

loo00

20000

30000

40000

50000

Concentrate Grade, ppm Ag F

i

2 Grade recovery nkbionshrp * for a silver ore psing muhipk deaning stages

DELINEATION OF AN APPROPRIATE REAGENT SCHEME Experience approach The selection of a reagent scheme for a new orebody begins with a review of existing operations that are treating ore with similar characteristics. This requires a detailed background study of the ore including mineralogy, ore variability, liberation characteristics, and potential byproduct values. Assistance from reagent vendors should be solicited prior to commencing actual testing once the mineralogy and liberation characteristics are understood. Preliminary screening tests rely heavily on the experience of the test engineer. Most of this screening work is performed via rougher kinetic tests. Some cleaning test work is performed to determine reagent requirements during cleaning, but these requirements are usually minor. In precious metal bulk sulfide flotation systems collectors, activators, and frothers are usually the only reagents used. A powerful xanthate such as amyl in combination with a dithiophosphate is usually all that is required, although occasionally an activator such as sulfuric acid or copper sulfate is used to enhance the flotation of tarnished pyrite. Individual collector dosages for this type of system usually vary from 10 to 100 grams per tonne. Frother (usually MIBC or polypropylene glycol) is used in batch testing as required to produce a stable froth, Frother addition is not optimized because frother usage in small batch tests is usually much greater than in actual plant operation.

141

Selective sulfide flotation systems generally rely on combinations of depressants and collectors to achieve differential flotation of valuable mineral from sulfide gangue such as pyrite. Lime is the most common pyrite depressant and the use of this reagent is a good starting point in most testing programs. Activators are usually avoided, except in special cases such as sphalerite flotation. Depressant dosages usually vary from 50 to 1500 grams per tonne, while collector dosages can vary from 5 to 30 grams per tonne. Often more than one collector and more than one addition point is used. Collector dosages are often critical, for too much collector can reduce selectivity by floating too much pyrite and not enough collector will result in poor recovery. Reagent dosages are often optimized by statistical methods.

Statistical Approach The use of statistical analysis in the selection of a reagent scheme is usually limited to optimizing reagent dosages after a preliminary scheme has been selected. Statistics are also extensively used to evaluate the relative importance of various reagents in existing reagent schemes at operating plants. The two most common types of statistics that are used in reagent scheme evaluation are screening tests and response surface design. The most common type of screening statistic used is the Plackett Burman analysis (Plackett and Burman 1946). This is a special two level factorial design that is used as a preliminary screening tool to determine which reagents have a significant effect on recovery, grade, or both. Relatively large numbers of variables can be screened at two levels. The variables evaluated can be either discrete, such as pulp dispersant addition or not, or continuous, such as two addition levels of dithiophosphate. This type of analysis cannot optimize a reagent scheme. More sophisticated statistics such as response surface design are used to optimize reagent additions. These methods are based upon factorial design of experiments and may be used to evaluate several variables at several levels. This type of analysis is quite complex and software packages that are available from companies such as ECHIPB are used to simplify the experimental design and interpret the results. LIMITATIONS OF BATCH TESTING Small scale batch test results should be evaluated with caution due to several limitations. The most important limitations are associated with recycle streams, site water, and test size. Recycle streams Single batch tests cannot evaluate the effect of recycle streams on rougher or cleaner flotation performance. Recycle streams usually have a bigger impact on flotation circuit design than on reagent selection, however, this is not always the case. For example, a build up of lime and pyrite in cleaner tailings from flotation circuits treating gold bearing copper ore can adversely affect gold recovery due to the depressing effect of calcium and hydroxide ion on liberated gold. This may require modification of the flotation pH or the addition of an auxiliary collector for gold. The recovery of silver from pyrite bearing silver sulfide ores is often affected by recycle of cleaner tailings middlings, particularly if lime is used during cleaning. Site and process water Most batch flotation tests are performed in tap water that is available at the laboratory. Confirming tests using site water should be performed when the preliminary reagent scheme has been developed. This will allow the effects of site water pH and dissolved solids to be evaluated. Process water should be used if it is available to determine the effect of residual reagents, particularly depressants such as lime. Obviously test work on a new orebody cannot be performed in process water and locked cycle testing with adequate water recycle is essential.

142

Batch test size The size of the test cell can influence the recovery of valuable mineral. Larger batch tests of 10 or 15 kilograms have indicated higher recovery compared to small one (1) or two (2) kilogram batch tests. This is probably due to the different hydrodynamics in the larger 40 liter cell compared with the smaller three (3) and six (6) liter cells. An example of this phenomenon for two silver ores is presented in table 1. Table 1 Effect of test size on silver rougher flotation Test Size Silver Tailings Assay Ore (kg) (PPm) A 2 46.3 A 10 34.0 B 2 55.9 B 10 47.3

Silver Recovered

(%I 82.8 87.5 82.6 86.4

CASE STUDY OF A SOUTH AMERICAN SILVER ORE This section illustrates an experience approach to the selection of a reagent scheme for a South American silver ore containing approximately 350 ppm silver. The ore contained several silver sulfosalts such as freibergite, proustite, argentite, and pyrargyrite, along with abundant pyrite, some sphalerite and silicate gangue. The objective of the project was to produce a selective silver concentrate that assayed 17,500 ppm silver at a minimum of 80 percent recovery by flotation. Discussions with reagent vendors indicated that a mining operation in North America processed similar ore by flotation. This plant used a mixture of dithiophosphate and ethyl xanthate collectors for the silver minerals and a sodium sulfite depressant for pyrite control. This reagent scheme was the starting point for the new ore. Preliminary scoping tests indicated that silver recovery approaching 80 percent was achieved with 5 grams per tonne dithiophosphate and 10 grams per tonne ethyl xanthate and bisulfite Unfortunately this choice of depressant produced a low grade silver depressant for pyrite. concentrate assaying only 3,000 ppm silver. Other depressants such as lime and a zinc sulfatekyanide complex were tested. These tests were performed at an arbitrary grind size of approximately 80 percent passing 74 microns. Results indicated that the zinc cyanide complex was extremely effective in depressing pyrite and produced a highly selective silver concentrate from this ore. These results are presented in table 2. Table 2 Effect of several pyrite depressants on silver rougher flotation Depressant Silver Concentrate Assay Silver Recovered (1000 gramshonne) (PPm) (%I 71.3 CaO 3,300 76.4 NaHS03 2,750 ZnS04/NaCN (3/1) 12,005 85.9 The grind size and pulp pH were then optimized in two series of tests using the zinc cyanide depressant reagent scheme. A natural pulp pH of approximately 6.0 and a primary grind size of 80 percent passing 120 microns were established. Additional collectors were evaluated once the depressant type, primary grind size and pulp pH were established. Results presented in figure 3 indicate that the original choice of dithiophosphate and ethyl xanthate produced the most selective silver flotation at the highest recovery. Each collector in the combination was added at a dosage of 10 grams per tonne.

143

Table 3 Effect of various collector combinations on silver rougher flotation Silver Concentrate Assay Silver Recovered Collector Combination (PPm) (a) Dithiophosphate/ethyl xanthate 12,000 85.9 Thionocarbamate/ethyl xanthate 9,600 82.0 Dithiophosphinate/ethyl xanthate 11,600 78.0 Multiple stage cleaning tests were performed on rougher concentrate produced with the dithiophosphate and ethyl xanthate collector combination. Grade recovery relationships indicated that a silver recovery of approximately 82 percent was achieved at the target concentrate grade of 17,500 ppm silver. Rougher concentrate regrinding was not required. These results have been illustrated previously as reagent scheme A in figure 2. Subsequent locked cycle testing and continuous pilot plant testing confirmed the results of the batch tests and produced concentrates assaying up to 20,000 ppm at recoveries of up to 83 percent.

CONCLUSIONS The batch bench scale flotation test is a valuable tool in the selection of an appropriate reagent scheme. A combination of experience and background information is usually adequate for the preliminary reagent selection. Once a reagent scheme is chosen the addition points and dosages may be optimized by statistical methods. However, the test engineer should be aware of the limitations of batch testing, particularly the effect of recycle middling streams and recycle process water. Reagent schemes that are selected by batch bench scale testing should always be confirmed in locked cycle testing. Pilot plant testing may also be required for delicate reagent schemes that may be adversely affected by recycle streams. REFERENCES Avotins, P.V., S.S. Wang, and D.R. Nagaraj. 1994. Recent advances in sulfide collector development. In Reagents for Better Metallurgy, ed. P.S.Mulukutla, Chapter 6. Littleton, Colorado: Society for Mining, Metallurgy, and Exploration, Inc. Nagaraj, D.R. 1994. A critical assessment of flotation agents. In Reagents for Better Metallurgy, ed. P.S. Mulukutla, Chapter 10. Littleton, Colorado: Society for Mining, Metallurgy, and Exploration, Inc. Suttill, K.R. 1991. A technical buyer’s guide to mining chemicals. E&MJ. Aug 1991. 192:8. Bulatovic, S.M., and D.M. Wyslouzil. 1985. Selection of reagent scheme to treat massive sulfides. In Complex Suljides Processing of Ores, Concentrates and By-products, ed. by A.D. Zunkel. 101 - 137. The Metallurgical Society, Inc. Broman, P.G., J. Hultqvist, and U. Marklund. 1985. Experience from the use of SO2 to increase the selectivity in complex sulphide ore flotation. In Flotation of Sulphide Minerals, ed E. Forssberg, Elsevier, Amsterdam. 277 - 291. Arbiter, N., and C.C. Harris. 1962. Flotation kinetics. In Froth Flotation Anniversary Volume, ed. D.W. Fuerstenau, Chapter 8. New York:The American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc. Plackett, R.L., and J.P. Burman (1946). The design of optimum multifactorial experiments. Biometrika. 33. 305 - 325.

144

Bench and Pilot Plant Programs for Flotation Circuit Design S.R. Williams, M.O. Ounpuu, and K. W. Sarbutt

ABSTRACT Aspects of flotation test programs, ranging from preliminary evaluations through to full feasibility study development programs, are discussed. These aspects include: the objectives; level of data obtained, including evaluation of this data; limitations of the test programs; and practical requirements, including sample quantities, time frame and methodology. Discussion of necessity and reasons for pilot plant evaluation and types of metallurgical deposit mapping programs are included as well as pitfalls encountered in bench and pilot plant flotation programs. INTRODUCTION This paper discusses flotation flowsheet development for new mining projects and expansions from conceptual (or preliminary) flowsheet development through to pilot plant campaigns. Each phase is discussed with a strong focus on the objective of that phase, and a conceptual analysis of the methodology that one would apply in each phase. The single most important issue to be addressed before any phase of metallurgical testwork is that of sample selection. Testwork results reflect the sample tested! The corollary of this is that poor sample selection can lead to poor or misleading metallurgical results. To get a truly representative sample, metallurgists must work together with the project geologist(s) and mine planner(s) to carefully select material for metallurgical testwork and establish logical compositing criteria. The investigator should incorporate good qualitative and quantitative mineralogy in the flotation flowsheet development process. Understanding the nature of the sample mineralogy should drive testwork development and ultimately lead to the optimal flowsheet. The phases of the flotation process development that will be discussed are: 0 0 0 0

Preliminary scoping studies, Pre-feasibility studies, and laboratory-based feasibility studies, Pilot plant testing, and Metallurgical mapping programs.

This paper focuses on sulphide mineral flotation but many of the comments can be applied to other mineral types. PRELIMINARY SCOPING STUDIES Objectives The conceptual, or preliminary, stage of metallurgical flowsheet development has a few very specific objectives. These are to:

' Lakefield Research Limited, Lakefield, Ontario, Canada KOL 2HO; [email protected] 145

Provide some broad understanding of the metallurgical deportment of the desired metal(s), or alternately, that there are no serious metallurgical concerns, (for example, ‘is this a refractory gold ore?’) b) Confirm that the metal@) desired can be recovered from that sample(s) using classical flowsheets and technology and identify which specific flowsheet route is indicated, c) Establish an order-of-magnitude level of recovery for the desired metal(s) and some indication of the type of concentrate quality, and d) Produce indications for future testwork. What flowsheet parameters need to be studied/optimized and are most economically sensitive? a)

Data Quality Given that the objective of this phase is to understand the metallurgical deportment of the sample(s), it is not expected that statistically rigid methodology can and will be applied. It is important to ‘scatter-gun’ different approaches (i.e. flowsheetsheagent regimes) to look for what appears to be working and to establish preliminary understanding of how the sample@) responds to the various tests. What is its ‘sensitivity’ to those tests? Data Evaluation The fundamental tool for data analysis is the ‘grade-recovery’ curve. It is preferable to prepare this curve with grade on the x-axis because frequently the flotation concentrate grade is set or controlled by smelter contracts. If gold were the desired metal, grade is typically recorded on the y-axis, as gold recovery by flotation often depends on the recovery of another mineral. For complex polymetallic flotation, another key tool is a “selectivity criteria” or the index of the desired metal versus an undesired metal, Gaudin’s Selectivity Index is such an index. (Taggart 1945) Test series are limited in the conceptual phase of flowsheet design and frequently tests are not duplicated. Test series focus more on various reagent combinations than optimization of reagents. Therefore, data evaluation in these cases is limited to identifying the tests which gave better metallurgical performance. Where test series are used (e.g. primary grind-size suite), these are best evaluated using a simple grade-recovery curve, that shows all tests. Limitation Testwork at this stage is preliminary. The most significant limitations arise because (a) sample size is small and can be biased because this testwork occurs at an early stage of project development, geological sample selection is limited, or (b) the number of tests performed is limited. Often for primary grind and regrind, a finer-than-optimum size is selected in this conceptual phase testwork to “minimize” poor liberation effects. These parameters are, of course, optimized in the pre-feasibility stage. Practical Requirements Samples. Sample size is usually small and too often is less than 40 to 50 kg per sample. Sample selection is critical, but is often limited by sample availability. Composite samples are preferred to small, individual (one-meter) drill-core intersections. Compositing must be carefully considered and should be done after consultation with project geologists. It is best to composite drill core material. Samples such as assay rejects, reverse circulation drilling chips or old samples are usually very poor samples for metallurgical testwork and should be avoided if at all possible. Timing. Conceptual flowsheet development should be undertaken early in a project development history. It is important to understand the metallurgical deportment of the metals early in a project, as this can influence the economic decisions pertaining to the project. For example, “if we have a refractory gold ore, do we have sufficient precious metal value to sustain mine development?’ If the answer to this questions is “no”, then can anything metallurgical be done to significantly improve the economics?

146

PRE-FEASIBILITY AND LABORATORY SCALE FEASIBILITY TESTING Purpose and Objectives The key objectives of the pre-feasibility and laboratory scale feasibility phase are to: Identify the probable flowsheet and reagent regime required for the economic recovery of the desired metal(s) or, establish if there are any fundamental problems in establishing economic flowsheets, 2) Establish, with a higher degree of confidence the expected recovery(ies) and concentrate quality(ies) of the proposed flowsheet, 3) Study the variability of the metallurgical performance throughout the (known) orebody. This metallurgical mapping or (as it sometimes referred to) geo-metallurgical mapping program will be discussed in further detail later in this article. Effective laboratory-based pre-feasibility and feasibility studies routinely incorporate such mapping programs, and 4) Establish preliminary concentrator design parameters (i.e. grinding information [not discussed here], flotation retention time, reagent requirements etc.). I)

Data Quality Pre-feasibility and feasibility studies require larger sample quantities and large sample sets. This permits more statistically based testwork with the better inherent data analysis and metallurgical conclusions. Some statistical based methodologies that can be applied are given in Griffith (1962). The application of statistical methodology is still limited and there are a number of possible reasons for this. There is a general discomfort in dealing with statistics, but a larger reason is that experience indicates that if we understand the nature of a certain type of ore we will apply a certain (known) flowsheet and reagent regime to that ore. This methodology is often the most expedient. However, in some instances, an ore does not respond well to conventional technology. In this case, the more rigorous methodology of statistical based testwork design and data analysis should be used to direct the testwork and resolve the complexity of flotation pulp chemistry. Data Evaluation The recommended sample size for this phase of testing is large bulk composites from which multiple charges can be composed. Normally, these charges are one or two kilograms each, but can be up to 10 kg. The composite must be well homogenized so that there is minimal variation in head assay and flotation testwork calculated heads. For non-nugget situations (i.e. not Au, Ag, PGE or Mo), the overall‘ variation should be f5% (absolute), whereby expected analytical variation is about k3% (absolute). For nugget samples, the variation could be much greater. Reliable analytical methods must support the metallurgical testwork and evaluation. Both the analytical method used and the statistical QA/QC used are important. It is also important to state the analytical method used and stress the need to always compare assays based on the method used. The authors have seen numerous examples of misleading information based around differences in analytical methods (for example, Mo by acid digest, AAS versus Mo by XRF (briquette or pyrosulphate fusion preparation)). Data evaluation then follows simple analysis based around what set of conditions gave better metallurgical response (usually presented in tables). It is important to maintain ‘bridges’ or linking tests between ‘series’. This is usually achieved by repeating a ‘standard’ test. It is also important to track the results and consistency of the ‘standard’ test(s) throughout the program. (For clarification, it is possible that a ‘standard’ test will change throughout a large testwork program). In analysis of metallurgical results, it is important to assess “to what extent is my less-thandesired metallurgical response (recovery and concentrate quality) a result of less-than-desired liberation and/or less than optimum chemical environment”. Quantitative information on liberation can be obtained by mineralogy. Even qualitative mineralogy can effectively guide the flotation investigation, (for example, “the sphalerite in the copper concentrate is mostly liberated.”) Therefore, use of mineralogy is one of the most important tools available to flotation investigators. The two critical questions that mineralogy helps answer are “what is the nature of my losses to the tailings,” and “what is the nature of my concentrate contamination?’

147

Knowing quantitatively the influence of the liberation on a metallurgical response means the effect of the given chemical processing environment can be inferred. This type of analysis is continuous and interactive during flowsheet development. Generally, it is more expedient to focus on liberation early in a flotation flowsheet development program (i.e. primary grind and regrind(s)), to the point at which an economically acceptable trade-off point is established for these parameters. This is particularly important for the fine-grained polymetallic ores as lack of liberation can mask effects of reagent regime change (for example: “was my poor selectivity against pyrite due to poor liberation or poor chemical environment?”) Limitations As previously discussed, all metallurgical testwork is limited to the validity and furthermore, representivity of the sample(s) tested. Testwork is also limited to the breadth and completeness of the reagent regimes tested: The intent of the paper is not to discuss the enormous number of permutations of collector, frother, pH and type of modifier used, Eh, water chemistry, depressant, activator, dispersant conditions which can be tested to achieve the desired flotation control selectivity. A number of selected references on reagent regime selections are available are given for this purpose (Bulatovic and Wyslouzil 1985; Bulatovic and Wyslouzil 1988; Bulatovic and Salter 1990; Bulatovic and Wyslouzil 1999; Bulatovic, Wyslouzil and Kant 1998; Bulatovic, Wyslouzil and Kant 1999; Agar et al. 1996). Practical Requirements Samples. As previously mentioned the best sample(s) for testwork are composite samples. The compositing for a porphyry copper deposit metallurgical testwork program will be distinct to that of a massive sulphide ore or that of a vein hosted Au/Ag ore. Sample selection will take place for both grinding and flotation testwork at the same time and will follow similar logic. Relevant criteria include: a) Rocktype b) Alteration type c) Mineralogy and/or head grade to assess variation in desirable metal(s) content, or major gangue mineral content (i.e. pyrite/pyrrhotite host). d) Oxidation states (for example, oxide zone versus a supergene zone versus a primary sulphide zone). e) Mining plan (such as year of mine production criteria). f, Unusual occurrences (e.g. highly faulted/fractured or folded zones, different mineralogy, etc.) These should be studied only if they are deemed to be geologically and economically significant to the ore distribution. These same criteria can be used for identification of samples for a metallurgical mapping program. Metallurgical mapping programs are usually incorporated in a complete pre-feasibility or feasibility program. Excessive compositing (i.e. production of large, overall composites) can mask valuable metallurgical response information and can give misleading conclusions about the actual plant performance. Therefore, it is generally recommended that the team create four to six composites. The amount of sample required for pre-feasibility and laboratory scale feasibility testwork can vary from as low as 100-200 kilograms to as great as one to two tonnes of sample (per composite to be tested). An example of the latter extreme is testwork that studies Cu/Mo separation after production of a bulk Cu/Mo concentrate. Sample preservation is important as testwork can span many months. Surface oxidation will occur on exposed sulphide mineral surfaces with time. This has been found to compromise test results and give incorrect metallurgical information (Table 1).

148

Table 1 Selection of samples Wt Test Sample 1 New Drill Core

Samples 3'"CI Conc RoConc Sample 1 Old RC Chip Sample 3'" CI Conc Ro Conc Sample 2 New Drill Core 3 " CI Conc Ro Conc Sample 2 Old RC Chip Sample 3Id CI Conc Ro Conc Sample 3 New Drill Core 3'" CI Conc Ro Conc Sample 3 Old RC Chip Sample 3rdCI Conc Ro Conc

%

1.85 9.47 1.49 13 2.08 11.7 1.4 14.8 2.14 10.7 1.61 12.3

Grade %Cu 38.6 7.9 37 5.68 30.5 5.86 30.9 4.4 36.1 8.38 37.3 6.47

Recovery %Cu 92.3 96.5 68.2 91.4 87.2 94.3 55.9 84.5 83.5 96.9 68.1 90

Head Calc %CU

KgO pm

0.78

21 1

0.8 1

197

0.73

202

0.77

194

0.92

184

0.88

191

It is clear from this, that sulphide mineral samples must be preserved. To achieve this, the samples must be kept as coarse as possible until testwork begins. Reverse circulation drilling and laboratory-reject products make very poor sample(s) for metallurgical testwork. When the sample must be crushed (often to -10 mesh for flotation testwork) it should be preserved in sealed bags, ideally at sub-zero degrees Celsius (in a freezer), in an inert atmosphere (N2or Ar). Timing Embarking on a pre-feasibility or feasibility study is a corporate decision (outside of the domain of metallurgical investigation). Given that sample requirement for these program(s) and the associated metallurgical mapping program(s) are large, it is important that the sample availability and/or acquisition be considered before undertaking these programs. Methodology A typical example of a pre-feasibility program for a porphyry copper ore will have the following components : Sample characterization:

- chemical

- mineralogical identification - mineralogical liberation with respect to different size fractions

- petrography Grindhime relationship: Rougher flotation optimization:

- primary grind - reagents - flotation time

Cleaner flotation optimization:

Locked cycle tests:

- number of cleaning stages - cleaner scavenger - regrind (together with liberation analysis) - reagents - flowsheet 'balance' - recirculation load - optimization of collector and frother - recycle water

- final grade/recovery curve

This methodology places importance on the use of both locked cycle tests and mineralogy. These aspects are further discussed here.

149

LOCKED CYCLE TESTING A locked cycle test is a repetitive batch test used to simulate a continuous circuit. The basic procedure consists of a complete batch test performed in the first cycle which is then followed by similar batch tests which have “intermediate” material from the previous cycle added to the appropriate location in the current cycle. These batch tests, or cycles, are continued in this iterative manner for a number of cycles until, ideally, steady state is reached. The final products from each cycle, i.e. final concentrate and final tailings, are filtered and thus removed from further processing. At the end of the test, all the products, final and intermediate, are dried, weighed and subjected to chemical analysis. The test is balanced and a metallurgical projection is made. Typically we think of flotation for locked cycle testing, but any procedure can be locked cycle tested. A Bond grindability test is an example of a locked cycle test. While the above description can be found in many classical textbooks (Taggart, 1945; Coleman, 1978; MacDonald and Brison, 1962; MacDonald, Hellyer and Harper, 1983, no discussion beyond the basic procedure is provided. In fact, comments such as “It is questionable whether in any case it approximates mill results any more closely than the standard butch test. (Taggart 1945) arise. It is truly surprising that our classic textbooks indicate that locked cycle tests are more art than science, and suggest that they can be of dubious value. None of the textbooks provide meaningful insight or discussion in: I’

Preparation €or a locked cycle test, The number of cycles to perform, How to assess if the test has achieved steady state, How to produce a valid metallurgical projection, Assess if the metallurgical projection is valid. Objectives of Locked Cycle Testwork The purpose of the locked cycle test is to simulate continuous circuit behaviour from batch testwork. There are at least three objectives in a locked cycle test:

Metallurgical projection of continuous circuit behaviour, Assessment of circuit stability or “robustness,” and Flowsheet or “network” development. Locked cycle testing is the preferred method for arriving at a metallurgical projection from laboratory testing. The reason for this is simple: the final cycles of the test mimic a continuous circuit. In a batch test, the deportment of the intermediate streams to concentrate or tailings is unknown. In locked cycle testing these streams are recycled and, at the end of the test, the material in these streams should report to either concentrate or tailings. Thus it will be clear how the intermediate streams divide between concentrate and tailing. Cycle tests are also used to assess the suitability of a flowsheet and reagent suite. If the cycle test does not come to steady state, then this indicates there are problems. Typical flowsheet problems stem from recovery intensive flowsheet (countercurrent) for ores with challenging mineral selectivity, or aggressive flotation in the recovery stages and too selective in the latter cleaner stages which forces a circulating load. Typical reagent problems stem from either too much or too little added, or a build-up of reagent in the circuit. Batch Tests Leading Up to Locked Cycle Tests The batch testwork prior to a cycle test must be adequate to insure a reasonable chance of success. A failed locked cycle test is far more memorable in people’s mind than a successful test. Ideally, the batch tests leading up to a locked cycle test, have each separation stage optimized for reagents and flotation time. A general trend in most batch testing is trying to achieve the highest possible recovery fiom the test. This emphasis provides an early estimate of the likely metallurgy for a sample, but is not optimum for cycle testing or a continuous circuit operation. “Ultimate” batch test recovery is achieved by targeting very high stage recoveries (beyond optimum, i.e. incremental flotation rate

150

of non-valuables greater than valuables), and often leads to more cleaner stages than may be required to produce the targeted concentrate grade. When this procedure is locked cycle tested, it invariably fails because the individual stage recoveries are too high and low grade concentrate and poor mass conservation usually result. In the batch test procedure, the intermediate streams report to tailings while in the cycle test there is no effective means of escape for this material as it will remain in the circuit. It has been observed that stage recoveries, of 85% to 90%, work out well for locked cycle and pilot plant testing. Steady State, Stability and Mass Conservation Although these terms are used interchangeably when locked cycle tests are discussed, they have different meanings. Mular and Richardson (1 986) provide an excellent description of steady state. “At steady-state, the mass input rate equals the mass output rate, whether it is entire process that is being considered, or individual unit operations. For a system at steady-state, no material accumulates internally; each unit operation is functioning with an unchanging volume of material already in the circuit. This description of steady state highlights the need for stability and mass conservation. Stability implies constancy. For example, the concentrate weight and grade remain the same for the last three cycles of the locked cycle test. Mass conservation implies “what goes in .. . must come out.” In the context of a locked cycle test, this means if 1000 grams of sample goes in, then 1000 grams must come out as final concentrate and tailing. However, mass conservation must also apply to the metal units. Thus, if 100 grams of chalcopyrite goes in, 100 grams of chalcopyrite must come out. Invariably, most people look for stability when studying locked cycle test results, as it is easy to see by looking at the data. Most people ignore mass conservation because it is not easily determined by quickly glancing at locked cycle test data. Steady state implies both stability and mass conservation. A good locked cycle test achieves steady state. ”

Metallurgical Projections Producing a valid metallurgical projection is one of the most important components of the test. It is the final numerical summary of the test’s metallurgical performance. There are at least three different procedures used to generate the metallurgical projection. n-product formula (balance on assays of final products), SME procedure (balance on final product weights and assays), Concentrate production balance (balance on final concentrate weights and assays). All three procedures will produce the same metallurgical projection for a test at steady state. None of the procedures are ideal for a test, which is not at steady state. The following provides a brief description of the procedures. N-Product Formula. The n-product formula is a simple material balance technique that utilizes the assays from the final products to determine the mass balance. Taggart (1945) provides an excellent description. In the case of a simple ore with only a concentrate and tailing, the procedure uses the assay of the feed, concentrate and tailing; C = F * (f-t)/(C-t)

The remainder of the balance is calculated once C (the concentrate mass) is determined. In application to locked cycle test balancing, the weighted average assay for the final two to four cycles is used. One of the important requirements for using the n-product formula is that the circuit must have mass conservation, i.e. input material = output material. If the circuit does not have mass conservation, then the n-product formula will provide an erroneous result. Computer mass balance programs such as MATBAL, BILMAT or JKSimmet use essentially the same approach as the n-product formula when applied to locked cycle tests. SME Procedure. The SME procedure is described in the SME handbook (Weiss, 1985). The procedure is more direct and should be easier to apply than the n-product formula. In the case of a simple ore with only a concentrate and a tailing, the concentrate is projected as the average mass and assay of the concentrate produced in the last few cycles of the test, and the tailing is projected

151

in a similar basis. The feed for the test is then calculated as the sum of the products. This procedure is acceptable as long as the test has come to steady state. If the test has not achieved mass conservation, then it will be erroneous because it completely ignores the material that does not report to the final products. Concentrate Production Balance. This procedure is a derivative of Weiss (1988) in which the concentrate is projected in the same way. However, the tailings are then calculated as the difference between the feed and the concentrate. This procedure does not overstate the metallurgy when the test does not have mass conservation. An overall premise is that the concentrate produced is the only concentrate produced. All other material must therefore be tailings. In many respects, this procedure resembles a month end production balance at an operating plant, because the smelters only pay for the concentrate received. Study of Balance Procedures. Ounpuu (200 1) compared the effect of these balance procedures on a test that achieved steady state and another test that did not. The steady state example had all three balance procedures yield the same metallurgical projection. The second example, which did not come to steady state, had the projected lead recovery vary from 75% to 85%. The difference arises from the assumptions inherent in each of the three balance procedures. Most balance procedures assume that the process is in steady state, or more importantly, that there is mass Conservation. If this is not so, then most balance techniques overestimate the recovery. The concentrate production balance does not overestimate the recovery, but may yield a conservative estimate. This technique has the benefit of at least predicting a laboratory recovery than can be achievable recovery. A reasonable metallurgical balance can be made for a test, which is unstable but has mass conservation. A test which oscillates around 100% mass conservation can have a valid metallurgical projection by using the number of cycles which comes closest to 100% mass conservation. Hence, a cycle test can use any number of cycles (greater than 2) for the projection, and the guideline for how many cycles to use for the prediction should be dictated by the number of cycles which provides a balance closest to 100% mass conservation. However, a test which never achieves mass conservation will prove challenging to arrive at a good metallurgical projection. The options are, to repeat the test, or use the concentrate production balance procedure which may be conservative but at least yields an achievable result. Using the other balance procedures or trying to interpolate what the steady state metallurgy might be raises more questions than it answers. How close to 100% steady state is acceptable: 100.00% steady state test has never been observed. Good tests will be at 100% for weight and 100% f 2% for the metals. Any test which is >5% from 100% should be deemed as not near steady state, and thus the data viewed with caution. Any test which is >lo% from steady state should be considered a bad test and must be ignored or repeated. How Many Cycles for the Test How many cycles should be performed? We believe most tests should be conducted for a minimum of six cycles based on practical consideration. Tests should be conducted until they achieve steady state only. In a Bond grindability test, the cycle results are known prior to the next cycle, and thus, the number of cycles can be rigorously determined. This unfortunately does not apply to flotation locked cycle tests due to the long analytical turnaround time needed to assess the results of each test. Agar and Kipkie (1978) present a relatively simple numerical simulation technique that can be used to estimate the number of cycles and the potential stability for the test. This procedure provides a reasonable estimate of the test's potential for success, but is at best a pre-test estimation of the number of cycles required, and provides no indication of steady state during the test. The technique found to have the most success was the tracking of the wet filter cake weights during the test. Each of the final products are weighed and recorded during the test. Target weights are established prior to the test so that the technician(s) can gauge the success of the test. The target weights can be derived from the weights produced during the batch tests, or using a simple calculation to account for the filter paper weight and the cake moisture content. The test is deemed to be in steady state when the all the target weights are being met for at least a few cycles in succession. Carrying out a test for twenty cycles does not necessarily ensure the test comes to

152

steady state. If a test must be greater than nine cycles, then the operators are trying to artificially force the concentrate grade higher and the tailings grade lower than they naturally want to be. The simpler the ore and process, the fewer cycles should be required. A simple, monoinineralic ore with excellent liberation may only require four cycles for a good cycle test. A Cu-Zn ore of similar simplicity may require only five cycles. A complex Cu-Pb-Zn ore with poor liberation may not come to steady state after even nine cycles. It is felt that a minimum of six cycles should be planned for any locked cycle test, and the wet filter weights tracked during the test to monitor how well the test comes to steady state. The individual wet filter weights and the total wet filter weight should be tracked.

MINERALOGY FOR METALLURGICAL FLOWSHEET DEVELOPMENT The mineralogy of an ore defines the limit of any physical separation process, such as flotation. Therefore, understanding the mineralogy must underpin all stages of metallurgical flowsheet development. Mineralogy for metallurgical investigation must focus more on the textural relationships of the ore and gangue minerals (e.g. occurrence, association, grain size and liberation) than mineral identification. Understanding the textural nature of the middling particles provides more useful metallurgical information than just the degree of liberation itself. It is the nature of the middling particles that will dictate the grind and regrind targets and the metallurgical results for a sample. Quantitative and qualitative mineralogy are both valuable, but metallurgical development work should be backed up by quantitative mineralogy. Qualitative mineralogy can be used to rapidly guide a preliminary scoping program and, particularly for scoping reagent changes. Qualitative mineralogy can often be achieved using binocular microscopes located in the flotation laboratory. However, quantitative mineralogy, especially in the pre-feasibility and pilot stages allows a more detailed analysis of grindhegrind selection. A hierarchical mineralogical methodology frequently used includes:

0 0

0

Metal recovery (grade versus recovery) Size-by-size metal recovery Mineral texture recovery (for example, free sphalerite, simple binary sphalerite, tertiary and complex sphalerite binaries and pyrite recovery to a zinc concentrate) Size-by-size mineral texture recovery study (for example, the same mineral texture referred to above, but in five different size fractions, such as +lo0 mesh, -100 mesh + 200 mesh, -200 mesh + 38 micron, -38 micron +I5 micron, -15 micron).

Detailed quantitative analysis requires a statistically based mineralogical methodology. Traditionally this was accomplished using intensive point counting, but recent developments (i.e. QEM*Scan and some types of image analysis) mean this can be fully automated. Papers which discuss the use of mineralogy in mineral processing are given (Grammatikopoulos 2002; Grammatikopoulos and Roth 2002; Petruk 2000).

PILOT PLANT STUDIES Pilot plant studies are often included in feasibility studies. A feasibility study does not necessarily need a pilot plant but most flotation flowsheets should and usually include a pilot plant study. The reasons for use of a pilot plant have been fully discussed (Engle 1978; Kuestermeyer 2000; Wilson and Dawson 1978) and a synopsis is given. Rational Typically, a pilot plant study provides much more confidence on metallurgical response and results than laboratory testwork. Therefore, the use of a pilot plant campaign is to reduce the technical and financial risk that result from scale-up and operation. Given this, a decision to have or not have a pilot plant campaign should be based on the need to resolve potential process risks andlor unknowns in the proposed flowsheet. Each laboratory flowsheet should be reviewed with respect to operation in an industrial setting, as well as new technology; water supply andlor water

153

recycle; changes in the scale-up grinding criteria and the differences in product size curve of a SAG/AG grind versus a controlled laboratory grind, to highlight but a few areas. Purpose and Objectives The key objectives of a pilot plant are to: Define the concentrate quality and recovery (of desired metals) of a representative samplets) or the ore(s) and the best known metallurgical process route. Produce a ‘bankable’ document that will be appended to the full feasibility study. Provide key engineering data for the concentrator design. Reduce design safety factors and improve accuracy of capital cost estimates. Study, define, control or optimize those interactions or processes that cannot be properly studied at a laboratory scale (for example, water recycle, bleed streams, regrind positioning or gravity concentration). Provide bulk sample for downstream processing (such as concentrate, tailings and water for thickening, filtering, further processing, disposal and others). Train metallurgical and operational staff on the flowsheet operation. Provide market samples for economic assessment. Provides psychological comfort to project financiers and due diligence consultants from a successful pilot plant demonstration. Data Quality The quality of metallurgical data from a pilot plant depends on the stability of the flotation circuit at the time that it is sampled (aside from sample homogeneity and chemical analytical technique). Typically, a flotation pilot plant is operated for several hours with minor reagent adjustments to obtain the reasonably correct conditions. The circuit operation is then frozen for a few hours prior to sampling the circuit under these stable conditions. The circuit should be sampled over several hours to produce a composite sample. Various process measurements are taken concurrent with sampling of the process. Circuit stability is documented with control assays (either separately taken or through on-stream analysis) and stability in key mass-flow rates (for example, cleanerscavenger tailings). Only metallurgical results from stable pilot plant runs should be used in data analysis and conclusions. Data Evaluation A typical pilot plant result or report can include:

(i) Sample description

- chemical characterization - mineralogical identification - petrographic deportment information - liberation criteria - grindability nature

(ii) Water description

- open circuit

(iii) Baseline laboratory testwork (iv) Pilot plant testwork program

- recycle - batch flotation tests - locked cycle test(s) - single tests conducted - conditions for ‘demonstration’ or ‘final’ run

(v) Final metallurgical result

- from ‘demonstration’ run

- comparison to locked cycle test - projected gradehecovery curve for desired metal(s) (vi) Detailed pilot plant analysis

- flowsheets compared - reagents compared

154

(vii) Engineering data from the ‘demonstration’ run - grinding circuit (kWh/t, operating work index, etc.) rougher flotation (retention time) - regrind circuit (kWh/t) - I St cIeaner/cIeaner-scavenger (retention times) - other cleaner stages (retention times) - inass flow data/mass balance - water balance - assay balance - reagents / pH/ Eh - media consumption - size analyses of a variety of streams - kinetic flotation analysis (if required) - concentrate settling and filtration - tailing settling and filtration - tailing rheology (if required) - pulp rheology (if required) - column scale-up criteria (if required) (viii) Concentrate characterization - chemical (to expected smelter contracts) - mineralogical composition - liberation - size - YOmoisture - transportable moisture limit (ix) Effluent characterization - chemical - toxicological (x) Tailing characterization - chemical - mineralogical - acid-generating capability - leachate testwork - rheology (xi) Support testwork during pilot plant campaign - laboratory flotation testwork - mineralogy - liberation studies - special plant surveys (xii) Special testwork (as defined), for example: concentrate combustibility - downstream concentrate hydrometallurgy - downstream batch concentrate testwork (e.g. Cu/Mo separation) (xiii) Review of pilot plant operation - strengths - weaknesses - stability

-

A ‘demonstration’ or ‘final’ run in a pilot plant is a continuous pilot plant operation incorporating all the findings, with regard to flowsheet and reagent regime/addition rates from previous pilot plant operations, in such a way as to demonstrate the final metallurgical performance of the ore. This usually lasts 48 hours to five days, and is a continuous operation. Locked cycle flotation tests are used to indicate stable metallurgical performance at a laboratory scale (given known sample, flowsheet and reagent regime). Therefore, a locked cycle test performed on the pilot plant feed sample(s) can be compared to the final ‘demonstration’ run metallurgical result. Such a locked cycle result should give the same metallurgical performance as the pilot plant. This comparison is considered important as it provides a connection of pilot plant with previous and future laboratory-based flotation testwork. It provides a baseline so that changes to the final piloted flowsheet can be assessed in a cost effective manner.

155

Limitations The largest potential limitation of a pilot plant is sample representivity. Pilot plants can require between thirty to several hundred tonnes of material. The questions that then surround this are “how does one obtain this amount of sample(s),” ‘what does the sample represent?” and “what does one want the sample to represent.” The sample should also be linked back to historical and future laboratory-based testwork, by the use of “bridging” testwork as previously discussed. The next problem presented is the homogeneity of the pilot plant sample. Consistent head grade must be provided throughout a pilot plant campaign. Classical rod mill/ball mill grinding circuit pilot plant configuration require a fine crushed feed (- 1/2” to -1/4”) that makes it easy to homogenize a pilot plant feed pile. SAG/AG mill pilot plant grinding circuits are usually fed with a top size of six-inch material. Homogenization of this material is more difficult. Samples that contain gravity recoverable gold can also be particularly problematic because of gold concentrating in the fines of the ore pile. Practical Requirements Sample. Refer to the comments above on sample selection. There is a range of pilot plant sizes available. Some typical sample requirement sizes for pilot plants are: Throughput 1 t/h 0.5 t/h 0.15 t/h

Time 4 week campaign 4 week campaign 4 week campaign

Sample Size 400 tonnes 200 tonnes 50 tonnes

Head grade of the desired metal(s) needs to be considered when sizing pilot plant feed rates. Cleaner flotation flowrates and cell volumes guide the throughput for pilot plants. Thus, higher grade ores can use lower feed rates, while lower grade ores need higher throughputs. Also, higher throughputs (>0.25 t/h) are recommended for complex polymetallic flotation because these pilot plants are difficult to balance and control at low throughputs. Pilot plant size SAG mills consist of 5%’ to 6’ diameter mills that typically require 1-4 t/h feed rates, depending on the autogenous work index. Pilot plants can be scaled from laboratory testwork, ideally using a projected mass balance from a balanced locked cycle test (although batch testwork can also be used). The authors use a flotation retention time scale-up factor of 2 5 1 to 3:l for laboratory to pilot plant. Methodology A typical well-structured pilot plant program should potentially include: SAG/AG grinding circuit testwork

Flotation circuit commissioning Selected specific tests on flowsheetheagent - reagent optimization - flowsheet deviations (e.g. regrind location, column cells) - use of recycle water Continuous operation with selected final flowsheet without recycle water Continuous operation with selected final flowsheet with recycle water Product thickening and filtration testwork Product characterization Downstream product testwork It should be pointed out that a pilot plant should not be used to scope reagents, but rather to optimize reagent additions, points of addition and requirements, given a closed system.

156

METALLURGICAL VARIABILITY PROGRAMS Purpose and Objectives A well planned metallurgical mapping program can significantly reduce project risk and contribute to the completeness of a pre-feasibility or feasibility program. The key objectives are to: 1)

2) 3) 4) 5)

Evaluate the ore variability on the basis of head grade, mineralogy, rock type, alteration type, location and/or mining year (mine plan). Provide correlation between mineralogical testwork and metallurgical testwork. Develop model(s) that forecast flotation performance based on some criteria such as head grade, mineralogy. Assess robustness of the proposed flotation flowsheet to the established mineral variation. Provide for an optimization process for the flowsheetheagent regime that can be incorporated in a final concentrator design to make it more robust during mine life.

The use of metallurgical mapping programs are discussed in references (Winkers 2002; DiPrisco, MacDougall and Urbanoski 2000). Sample A well planned metallurgical mapping program consists of a matrix of samples that reflect the orebody. Commonly used matrices are rock type/ alteration type, and/or mineralogy; mine plan matrix (either as a stand alone matrix or combined with the above) or the simplest model can be to composite samples on a pre-established drill-core meterage. The sample selection for the metallurgical mapping program must be done by consultation between the metallurgist, geologist and mine planners. An ideal program should consist of a representative large sample base (>20 samples, preferably >loo), and the number of samples should bear some correlation to the size, value and variability of the deposit. Medium to large deposits typically have a sample for each 1-5 million tonnes of ore.

Methodology All samples should be submitted to a standard characterization program that includes head chemical analysis, mineralogical examination and a standard batch flotation test (rougher and cleaners). Data analysis needs to be at two levels: Metallurgical results need to be studied for trends in results versus matrix characterization (for example, alteration type versus metallurgical performance) and/or metallurgical result versus primary metal(s) head grade(s) and/or versus some auxiliary metal analysis (for example, “does poor metallurgy of a copper concentrate correlate to zinc contamination/zinc in the feed assay?’) 2) Dependent on the mineralogical methodology used, some attempt should also be made to relate specific mineralogical characteristics to metallurgical performance. Some examples could include the percent pyrite versus metallurgical performance or if using QEM/Scan (Winkers 2002; Sutherland, Wilkie and Johnson 1989), one could use the PSSA (Phase Specific Surface Area) factor versus metallurgical performance. I)

This type of analytical methodology calls for one number that reflects metallurgical performance.’There are various numbers that one can use to do this. Some of these are referred to in reference (MacDonald and Brison 1962). These authors refer to a selectivity index (SI), where SI = ARBj/BRAj with AR the grade of the constituent A in float, BR is the grade of constituent B in float, AJ grade of constituent A in non float an BJ is grade of constituent B in non float. An option preferred by the authors is to use a recovery number at some standardized concentrate grade. Whatever number is used, it is important to be cognizant of the assumptions and limitations behind that number.

157

CONCLUSIONS This paper has outlined concepts, philosophy, controls and limitations of laboratory and pilot plant flotation flowsheet design. Many of the themes touched on in this paper appear timeless because so many of the questions being asked today are unchanged from twenty or thirty years ago. REFERENCES Agar, G. E. and W.B. Kipkie. 1978. Predicting Locked Cycle Flotation Test Results from Batch Data, ClM Bulletin, Vol 71, No. 824, 140-147 Agar, G. E., F. Khan, B. Markovich, A. Mukherjee, B. Shea and C. Kelly. 1996. Laboratory Flotation Separation of INCO Bulk Matte, Minerals Engineering, Vol. 9, No. 12, 1215-1226 Bulatovic, S. and D.M. Wyslouzil. 1988. The Effect of Flowsheet Configuration on Metallurgy Results During the Treatment of Massive Sulfide Ores, Published in the CIM Bulletin Bulatovic, S. and D.M. Wyslouzil. 1999. Flotation Behaviour of Gold During Processing of Porphyry Copper-Gold Ores and Refractory Gold-Bearing Sulphides, 2"dInternational Gold Symposium, Lima, Peru. Bulatovic, S. and R.S. Salter. 1990. Some Aspects of Recent Improvements in Treatment and Separation of Refractory Polymetallic Ores, Presented at Salt Lake City, May. Bulatovic, S., D.M. Wyslouzil and C. Kant. 1998. Operating Practices in the Beneficiation of Major Porphyry Copper/Molybdenum Plants in Chile: Innovated Technology and Opportunities, A Review, Published in Minerals Engineering, Vol 11, No. 4, pp. 313-331, April. Bulatovic, S., D.M. Wyslouzil, and C. Kant. 1999. Effect of Clay Slimes on Copper Molybdenum Flotation from Porphyry Ores, Presented at Copper '99, Phoenix, Arizona, October 10- 13 Bulatovic,S. and D.M. Wyslouzil. 1985. Selection of Reagent Scheme to Treat Massive Sulphide Ores. Complex Sulpfides - Processing of Ore Concentrates and By-products, A.D. Zunkel et al., TMS Publications. Coleman, R.L. 1978. Metallurgical Testing Procedures, Chapter 9 in Mineral Processing Plant Design Society of Mining Engineers N.Y. Mular and Bhappu Editors DiPrisco, G., C. MacDougall and L. Urbanoski. 2000. Ore Mineral Characterization and Predictive Metallurgy of the Lady Loretta Lead-Zinc Deposit, Australia - An Integrated Team Effort for Exploration, Mining and Metallurgy, Mining Millenium 2000, Toronto, Ontario, CIMT/PDAC. Engle, L.K. 1978. The Merits of Pilot Plant and Problems of Scale-UP, Data, Designs and Decision - The Usefulness of Pilot Plant, South African Institute of Mining and Metallurgy. Grammatikopoulos, T.A. 2002, Process Mineralogy of Gold Ore Deposits, Geological and Metallurgical Implications, Mineral Wealth. Grammatikopoulos, T.A. and T. Roth. 2002, Mineralogical Characterization and Hg Deportment in Field Samples from the Polymetallic Eskay Creek Deposit, British Columbia, Canada, International Journal of Surface Mining, Reclamation and Environment. Griffith, W.A. 1962. The Design and Analysis of Flotation Experiments, D.W. Feursteau et al.., Froth Flotation, 50'" Anniversary Volume, SME. Kuestermeyer, A. 2000. Pincock, Allen and Holt, The Mining Record Volume Ill, No. 10, October, pgs. 60-6 1. MacDonald, R.D. and R.J. Brison. 1962. Applied Research in Flotation, Chapter 12, Froth Flotation, Society of Mining Engineers N.Y., Fuerstenau, D. W. Editor MacDonald, R.D., W.C. Hellyer, and R.W. Harper. 1985. Process development testing, SME Mineral Processing Handbook, N. L. Weiss Editor Mular, A. L. and J.M. Richardson. 1986. Metallurgical Balances, Chapter 39, Design and Installation of Concentration and Dewatering Circuits, Society of Mining Engineers N.Y. Mular and Anderson Editors Ounpuu, M.O. 2001. Was That Locked Cycle Test Any Good?, 2001 Canadian Mineral Processors Conference Sutherland, D.N., G. Wilkie, and C.R. Johnson. 1989. Simple Predicting of the Processing Characteristics of Ores Using QEM*SEM, Aug. IMM Sydney Branch Min Pet 89, Mineralogy - Petrology Symposium, Sydney.

158

Taggart, A.F. 1945. Handbook of Mineral Dressing 19-181, John Wiley and Sons, New York Weiss, N.L. (editor) 1985, SME Mineral Processing Handbook, AIME. Wilson, R.A. and H.A. Dawson. 1978. Metallurgical Flowsheet Development, Mineral Processing Plant Design, SME. Winkers, A.H. 2002. Metallurgical Mapping of the San Nicolas Deposit, Proceeding 2002, 34"' Annual Meeting of the Canadian Mineral Processors, J. Nasset et al.

159

Bench-scale & Pilot Plant Testwork For Gravity Concentration Circuit Design Andre‘ R. Laplante ’ and D. Erik Spiller’

ABSTRACT Successful gravity concentration is founded in adequate liberation, informed selection of unit operations, and finally the proper presentation of feed to those units. As with all physical separation, effective circuit design and equipment selection require representative samples of adequate size. Characterizing the mineralogical nature of both value and waste components is critical. This paper presents testing methods used to establish amenability including selection of appropriate unit operations, bench- and pilot- scale testing and their limitations, and scale-up for effective commercial circuit design. All manner of commercial devices are considered with particular attention to recently introduced centrifugal machines. INTRODUCTION Gravity Separation - A Versatile Technology Gravity concentration, once the stalwart technology for mineral beneficiation, has evolved and entrenched in the last 30 years to regain and maintain its commercial importance to the recovery of various mineral based resources. Precious metals, ferrous and non-ferrous metals, energy minerals (in particular coal) and numerous industrial minerals are processed all over the world using gravity separation technology. Particles as coarse as 200mm (nominally 8 inches) to as fine as lOpm are successfully beneficiated using gravity separation equipment. There are dry gravity separation systems and equipment, although the vast majority of commercial gravity concentration is performed wet. The past 20 years of technology development has been most dramatic in the application of centrifuge-based equipment for fine particle separations and for recovery of very high specific gravity particles occurring in low concentration, i.e., liberated gold in a grind circuit stream. The single most important factor in successful gravity concentration is liberation of the heavy and light components. Liberation in natural mineral systems occurs along a continuum from an essentially homogeneous state to complete physical liberation. The necessary degree of liberation coupled with correct equipment selection is fundamental to successful and cost effective gravity concentration circuits. Readers are directed to reference the chapter titled “Process Design, Scale-up and Plant Design for Gravity Concentration’’ in Mineral Processing Plant Design, 2“ Edition, 1980. This reference contains considerable relevant information on the subject. Virtually all gravity-concentrating machines tend, at times, to confuse gravity concentration with size classification so it follows that the most effective gravity separations are accomplished on narrowly size classified feeds. Of course, there is an economic and practical limit as to how much equipment and process energy is justified for size classification during the feed preparation process. This limit should be established during the design-engineering phase of study after the necessary bench and pilot studies are completed.

I

McGill University, Montreal, Quebec Canada

* OUTOKUMPU TECHNOLOGY INC. Physical Separation Division, Jacksonville, Florida USA

160

Advantages of Gravity Separation in Process applications There are a number of advantages to applying gravity concentration techniques to mineral systems. Among the most obvious are: Relatively simple equipment translates to low cost (capital and operating) Little or no reagents required (cost advantage and environmental advantage) No reaction products (no chemical change) minimizes energy cost and results in environmental advantage Can be effected at relatively coarse sizes (assuming adequate liberation of either value or gangue components) to extremely fine size materials Gravity separation, when it works, is usually the most cost effective beneficiation technology. Importantly, even in instances where complete separation cannot be achieved by gravity techniques, it is often desirable to produce a gravity pre-concentrate. This pre-concentrate product (now with less mass) can then be further concentrated or processed by more expensive, perhaps environmentally sensitive, techniques to produce the final desired product.

Definition of Testwork Scale (Bench vs. Pilot vs. Demonstration) In general terms, testing for gravity concentration amenability and application can be considered at three levels; bench, pilot, and demonstration. The bench-scale (often done more for characterization than process testing) is typically conducted on samples from 1 to 20 kg, whereas pilot-scale testing can require samples of 100 kg upwards to several tonnes. Demonstration scale testing is essentially operating the equipment at full commercial feed rates over extended time periods. Gravity concentration equipment cannot be readily miniaturized for bench-scale testing. The mechanics of the separations simply require more aredspace, higher flow rates, and more incremental time than can be accommodated by bench top units operating on small samples. For this reason bench-scale testing is often focused on characterization of the feed and desired products with an emphasis on liberation. Pilot-scale testwork for gravity concentration is commonly done on commercial equipment, albeit when possible on the smallest version of the commercial equipment. When possible, testing is done continuously in open circuit; prepared feed is presented to the test unit and timed product samples are taken at steady state. However, at commercial rates or even rates approaching commercial processes rates, feed is consumed very quickly and large quantities of concentrate and tailing products must be handled. Therefore, a modified open circuit test (commonly referred to as a batch test) is often employed wherein feed is introduced to a sump/pump arrangement from which feed to the test unit is withdrawn and presented to the concentrator unit. Products from the unit are returned to the feed sump, remixed and pumped again to the unit. When all conditions of operation are set, i.e., feed rate, feed pulp density and concentrate/tailing cut points, timed samples of all the products are taken. This procedure minimizes the need for large feed samples and the cost of preparing the product samples for analyses. Multiple tests at different conditions can often be done on one batch of prepared feed because only the amount of sample cut for the test is removed from the system. Further evolution of the batch test is the semi-continuous batch test. The objective of the semi-continuousversion is to minimize the sample requirements while providing adequate product for subsequent (cleaner or scavenger) stage testing. Here the batch test is initiated, and at proper conditions the concentrate and tailing products are diverted and collected until the desired amount is produced or until the feed being presented to the univmachine is no longer valid. This procedure can be repeated as necessary with the introduction of new feed to sump and alternate collection of products. The resulting products from semi-continuous testing can then be further tested as necessary in cleaner or scavenger stage application.

161

Demonstration (commercial scale) testwork is very common for gravity concentration. Often gravity equipment is niched into an existing operation or added onto a commercial process as a scavenging or subsequent step at the end of the flowsheet. In these circumstances, industrial, or near-industrial scale machine units are installed in the operating plant and tested for metallurgy. In this manner the equipment is not only evaluated and optimized for metallurgical performance, but also with respect to its robustness and operating cost over an extended time period. Demonstration testwork can extend for months or even longer.

Conventional and Centrifugal Based Unit Operations Centrifugal (or “enhanced” gravity) separators became a common occurrence in mineral processing plants in the eighties, when the Knelson Concentrator became the unit of choice for the recovery of gold from grinding circuits, replacing mostly spirals and jigs. The low grades of the material treated, typically between 1 and 500 g/t, make it possible to operate the unit in semi-batch mode (i.e. a tailing stream is continuously produced, but the unit must be stopped to harvest the concentrate). Soon after, other semi-batch centrifuge units became commercially available for gold recovery, the best known being Falcon’s SB Concentrator. Knelson and Falcon have also developed continuous centrifuge units, but these do not enjoy the degree of acceptance of the semi-batch units, and will only be discussed briefly. Two other continuous centrifuge units will also be discussed, Mozley’s MeGaSepMGS and the Kelsey Jig of Mineral Technologies (now a subsidiary of Roche Mining). The four continuous units are based on four very different concentration mechanisms also used in non-centrifuge units, and are generally best suited for very different applications. Further, they may be used synergistically either with each other or with conventional gravity equipment. For each centrifuge mechanism, a sole commercial unit is available, a stark contrast to flotation units or even conventional gravity units. This presentation will discuss all centrifuge units aforementioned, with emphasis on Knelson Concentrators, to reflect their much wider use. Both semi-batch and continuous concentrators require unique bench-scale testing which are atypical for other separators. The semi-batch units are generally used to process all or part of the circulating load of a grinding circuit, an application that is almost impossible to replicate at bench-scale and can be difficult to mimic at pilot-scale. The continuous separators are also difficult to test at bench-scale, because typically no bench-scale batch unit can reproduce their main concentration mechanism. For preliminary testing, typically the low-capacity version of the continuous unit is used. Feed material may be difficult to generate at the design stage, since often the unit feed is not the ground ore but rather a relatively small stream. As a result, most applications of continuous centrifuges have been retrofits in existing circuits with the bulk of the test work performed on site. Typical testing programs will be discussed in the centrifuge section. FEED CHARACTERIZATION AND EQUIPMENT (UNIT) SELECTION Characterization of the feed material is the first step in developing a successful gravity concentration flowsheet and leads directly to selecting the most effective separating equipment. Objectives of Characterization The basic objective of feed characterization is to understand the mineralogy of the ore. Once understood, especially as related to liberated particle size, then selection of the appropriate separating equipment can proceed. For‘purposes of developing a successful gravity concentration process, the mineralogical analysis is best focused on the “nature-of-occurrence’’ of both the valuable and gangue components. Determining the genesis of each mineral in the matrix is of secondary importance and is often expensive. This paper will not attempt to cover the techniques used to do a mineralogical analyses of feed ores, but rather point out the practical aspects of such analyses and what resulting information is desired.

162

Spending money on feed characterization is equally important with all other steps in flowsheet development, and it can be argued that without a proper understanding of the feed it may be impossible, excepting luck, to end up with a successful process. Understanding the nature of occurrence also prevents unnecessary testwork and the loss of precious time. A good mineralogical characterization of the feed ore will result in knowing as much about the following as time and budget will allow. The minerals of value, their chemical composition, and their specific gravity The gangue minerals, their chemical composition, and their specific gravity The grain size range of both the value and gangue minerals, including information on intergrowth occurrence in both

Sampling and Sample Requirements In a perfect world, the mineral processing engineer would have access to as much sample material as desired and the samples would be perfectly representative both individually and as a blend of all the variation (ore types) in the resource. This is rarely the situation. Furthermore, statistically valid samples based on sampling theory are also usually out of the question due to their size and to the cost of procuring multiple large samples which often require extensive drilling, adit construction, trenching, deployment of personnel, and required sample transportation in remote areas. Finally, the specific mineral resource commodity being considered also plays a role in the amount of sample required to do a complete job of characterization and testing; for example, spotty gold ores require more sample material than does coal from a continuous seam. As a practical approach, the following guidelines should be followed. Communicate with the geologist about the nature and variation of the resource. Obtain samples representing the average and reasonable extremes of the resource. Obtain samples representing a mine plan, i.e., for example, what the process plant will see the first year, the third year, and subsequent years through the economic life of the resource. Think ahead; ensure the integrity of any sample materials (splits) left behind for future requirements. Samples for bench-scale studies, including feed characterization samples, should generally be from 1 to 20 kg in size. This assumes, based on particle size and preparation procedures, a sample of such a small size can be considered valid for the purpose of the work. Pilot-scale studies require from 100 kg of material upwards depending on the mineral commodity being studied, the equipment being considered, and the amount of resultant data required. Demonstration scale studies are conducted at commercial rates most always in a commercial plant setting where there is no limit to the amount of feed sample available (either as an integral part of a flowsheet or when operated on a bleed stream in the plant).

Characterizationof Feed Resource There are a number of different approaches to developing characterization data. First, is a simple microscopic examination, second is the use of sinWfloat analyses, and third is the use of scanning electron microscope technology. An additional way of estimating the likelihood that gravity concentration could be successful is the concentration criterion calculation. There are numerous descriptions of each of these techniques in the literature. The same techniques of characterization can be used on intermediate (inter-stage) product streams to determine where they should report in a flowsheet, or if they should be rejected. They can also be used to differentiate between true middlings and particles that report to a middling stream because of mechanical misplacement.

163

Microscopic Analyses. Simply looking at crushed, ground and size-classified samples of the resource under a microscope can be enlightening. More sophisticated analyses are entire subjects to themselves and include such techniques as polish and thin section analyses and grain counting. On occasion, magnetic fractionation, such as with a Frantz Isodynamic Separator or other devise, is a useful tool to identify specific minerals in the resourse. SinWfloat Analyses. This technique is used without exception for investigating coal and mineral sand resources, and is of value in all potential gravity separation applications. Safety issues surrounding the use of organic heavy liquids mandates strict containment procedures and thus in the past discouraged their use. However, there are now commercial and user friendly substitutes for the organic liquids in the form of sodium polytungstate solutions and other similar substitutes. These solutions can produce sinWfloat separations up to 3.1 gkm’ density. Therefore, these non-toxic solutions are useful to both the coal and mineral sands industries, and directly substitute for analyses previously done in tetrabromethane. For procedures and data interpretation techniques you are referred to the many references on this subject (Mills, 1980). Scanning Electron Microscope (SEM) Based Analyses. There are commercial liberation analyzers that are based on SEM technology with various software packages attached. These units (commercially known as the QEMScan and JKMRCPhillips Mineral liberation Analyzer-MLA) are capable of producing an enormous amount of information on mineral liberation, grain size, and intergrowth of mineral species. This information, especially if coupled with sink float analysis, will provide very useful information to the investigator regarding the suitability of the feed ore to gravity concentration. Of equal value is the information it provides relative to problems encountered in gravity concentration circuitry. Concentration Criterion. Assuming that the particle size effect is eliminated or minimized, and that the value and gangue components are liberated, the remaining task is to ensure there is sufficient specific gravity differential to effect a commercial gravity based separation. Fundamentally, there must be a marked difference in the specific gravity of the liberated value and gangue components. Applying the concentration criterion (does not apply to dense medium separations) is a convenient mathematical means for determining if the specific gravity differential is sufficient. Use: CC = (Dh-D,) / (D, - D,) Where: CC = concentration criterion and :

D, = specific gravity of the heavy component D, = specific gravity of the light component

D,= specific gravity of the fluid (usually water at 1.0) When the absolute value of CC > 2.5, it is likely that some form of gravity concentration will be possible CC < 1.25, gravity concentration will be virtually impossible using commercial methods 2.5 > CC > 1.25, gravity concentration will be difficult but may be possible to some extent on narrowly sized feed and with slow and careful feed presentation

1 64

Using the guidelines offered by the concentration criterion, one can speculate about the possibility of using gravity concentration successfully on any given mineral resource system. Of course, additional criteria must be met in a gravity concentration circuit to ensure success. For instance, feed preparation and presentation are important to varying degrees depending on the equipment and minerals being beneficiated. In general, finer particle sizes require larger concentration criterion values to indicate the possibility of successful gravity concentration.

CONVENTIONAL GRAVITY SEPARATION SYSTEMS (UNIT OPERATIONS) There is an amazing array of mechanical devices that have been conceived, built, and marketed to separate minerals based on particle specific gravity differences. The guidelines presented in this paper cover gravity separation via the classical system definitions of stratification, flowing film, and shaking (see Kelly and Spottiswood 1982); density systems, i.e., dense medium separations (DSM) are not included. Added of course to these systems are the centrifugal based separating machines that have become common over the past 20 years. Conventional Equipment Alternatives and Considerations Over the years there have been a number of graphs and charts developed to aid in selecting the most suitable gravity processing equipment for a specific application. The effective feed particle size range and capacity of the various concentrating units differ substantially. Presented herein are two of the more common summaries that provide indications of equipmenufeed compatibility. Of course the information presented in the figures are generalized and reflect both experience and prejudice of the authors. Considerationsof the ore’s mineralogy, i.e., liberation characteristics, and (assuming a favorable concentration criterion) coupled with the information in the summary tables will indicate if gravity concentration is probable. Figure 1 is a graphical presentation of the size range applicability of commercial gravity separation equipment. Presented in Table 1 are again the various types of commercial gravity concentrators with information about their relative water requirements and capacity. Hindered settlers, operated as density separators via control of the teeter bed within the unit, are now common to gravity circuit design. Properly used, the units facilitate and complement gravity separation. Other authors will deal with hindered settlers in the context of gravity separation. The hindered settler is included as a gravity concentration device in Table 1.

J I G G I N G

-.-.-.-

SLUICE BOX RIECHERT CONE PINCHED SLUICE

- .- .- .S P I R A L

SHAKING TABLE BARTLES-MOZLEY

-.-.-.-

CROSS BELT PNEUMATIC JIG AIR TABLE

-.-.-.-

CENTRIFUGAL 0.01

0.1

1

10

PARTICLE SIZE

mm

Figure 1 Size Range Applicability of commercial Gravity Separation Equipment

165

Strake Spiral Shaking Table Shaking Orbital Crossbelt Spinning Bowls Centrifugal Hindered Settler Density Separator Air Dry Pneumatic Jig Air Table

+ 2.0 + 2.0 0.02 + 2.0 0.01 + 0.07 0.01 + 0.03 0.01 + 1.7 0.07 + 0.60 0.15 + 25 0.25 + 6

0.15 0.03

high medium medium

low medium

high high vew high

low low

medium none none

medium

high high

medium

low

' Spiller, 2000; Generalized; specific machine operating on site-specific materials may or may not perform to these characteristics Kelly and Spottswood, 1982 ' Relative Relative per unit; multiple units always equate to high capacity Projecting Data to Commercial Application, and Limitations Thereof In general, data developed in the laboratory and pilot plant can be used directly to select equipment and predict performance at commercial scale. This of course assumes that the samples tested represent the eventual feed to the plant or gravity concentration circuit. Care should be taken to understand the nature of recirculated middling streams in complex circuitry. For example, product middlings that are liberated and are simply mechanically mixed components will eventually respond to separation and exit the circuit as concentrate or tailing products. However, unliberated middlings will build up in the circuit and choke performance if they are recycled upstream without additional liberation. Further difficulty arises when the gravity concentrator is treating a small portion of a larger stream in the plant, i.e., a recirculating load in a grinding circuit. A discussion of this situation is presented under centrifuge based unit operations elsewhere in this paper. Proper feed preparation and presentation is essential in gravity concentration. In fact, certain machines such as shaking tables demand a very consistent feed rate. Other considerations in feed preparation are desliming and sizing to reject coarse tramp oversize. CENTRIFUGE BASED SEPARATION SYSTEMS (UNIT OPERATIONS) Objectives of Bench and Pilot-scale Test Work Semi-batch Centrifuges for Gold. The difficulties of trying to replicate at bench-scale the performance of a centrifuge semi-batch unit, processing all or a bleed of the circulating load of a grinding circuit, has already been alluded to in the introduction. Additional hurdles challenge this process, namely the inability to achieve, at bench-scale, the very high concentrate grades full-scale circuits will typically achieve, the sampling difficulties normally associated with gravity

166

recoverable gold (GRG’), and the wide range of recovery efforts’ used in full scale circuits. The typical gold gravity circuit is also unusual in that it usually works in tandem with a flotation or cyanidation circuit. Thus, most of the gold it recovers would be eventually recovered even in the absence of a gravity circuit. As a result, the optimum economic recovery can be highly variable. Where the ore is simple and the downstream circuit very effective, the net contribution of gravity recovery could be as little as 0 to 0.5% of the gold values. Contrast this with highly preg-robbing ores where the impact of gravity recovery has been measured at more than lo%, or even mills that rely entirely on gravity recovery (“extreme” gravity recovery; Van Kleek, 2001). The extent of the gravity recovery effort will therefore vary from as little as 1% of the gold in the circulating load to in excess of 20% for extreme gravity recovery. To address these issues, the objectives of benchscale testing have to be clearly understood and test design chosen accordingly. Broadly speaking, testing objectives are two-fold. Firstly, bench-scale tests will seek to generate a gravity tailing that most closely resemble the feed that the downstream cyanidation or flotation circuit would process, for bench-scale testing of flotation and cyanidation. The focus of these tests will be to generate the correct gravity tailing, with emphasis on the recovery of coarse gold that would not normally report to cyclone overflows and be directed to downstream recovery circuits. Most bench-scale tests would be of this type. Secondly, bench-scale tests should aim to generate the data needed to design, size, and predict the performance of the intended gravity circuit. Such test work should not attempt to mimic the full-scale gravity circuit, but rather to characterize the GRG content in the ore. Typically, these tests are far more detailed than the first type, but few are required. A number of ores have been characterized with a single test, from which a gravity circuit has been successfully designed. This cost-effective approach is particularly appropriate when the economic impact of gravity recovery is low, the mineralogy of the ore is simple, and the scope of the project is limited. As the size or complexity of the project or the role of gravity recovery increase, so does the number of tests required, but very rarely should it exceed three or four. Typically, two tests would be needed to characterize the effect of head grade on the GRG content. The two objectives of bench-scale work are not easily reconciled, but both may not be needed. Most of the test work should address the first objective. Typically, this is achieved by treating a 5kg or 10-kg batch of ore at final grind with a laboratory centrifuge, either a Knelson Concentrator 3MD (3-inch unit) or a Falcon SB50 (4-inch unit). The concentrate is then cleaned to a small mass, 1 to 10 grams, by hand panning, superpanner or Mozley laboratory table. The rougher and cleaner tailings are combined and used as feed for flotation or cyanidation test work. The procedure is repeated on as many head samples as deemed necessary to assess issues of ore variability, optimum grind size, and cyanidation or flotation conditions. This typically requires many tests and may in fact represent the bulk of the gravity test work. Some semi-batch centrifuge representatives claim that actual plant performance often sits between the cleaner and rougher recoveries (Peackocke, 2002). This lends credence to the ability of this experimental protocol to produce a gravity tailing that is similar to the grinding circuit product. In essence, for this type of testing gravity recovery services cyanidation or flotation test work, much in the same way that full-scale gravity circuits service downstream cyanidation or flotation circuits. As effective as the above procedure is in producing a suitable feedstock for cyanidation or flotation test work, it does not provide the information needed for optimum design of the gravity circuit and prediction of how much it will recover. This additional information can be generated only from a test specifically designed to do so. This test should have the characteristics outlined in Table 2.

’ GRG: gold present in particles in concentrations such that the behavior of the particle in gravity units is significantly affected (Laplante et al, 2001) Recovery effort: overall gravity recovery expressed as a percent of the gold in the circulating load of the grinding circuit (Laplante and Xiao, 2001).

167

Fable 2 Test Requirements for Gold 4 kavity Circuit Design Action Objective Eliminate the nugget effect associated Perform test work on representative sample size adequate as defined by sampling theory with GRG Use large feed mass to keep weight recovery low and Recover only GRG examine concentrate to determine degree of GRG liberation (Laplante, 2000) Determine screening requirements and Generate size-by-size information; recover gold as soon as liberated to avoid over-grinding by staged best possible recovery unit comminution and recovery Slurry the feed directly above the separation unit Eliminate the risk of gold traps Difficult to achieve at bench-scale - use model to predict Predict gravity recovery Derfonnance A test that fulfills the objectives of Table 2 will be described in the next section. An important observation is that the interaction of grinding, classification and recovery units would be extremely difficult to reproduce at bench-scale but is now reasonably well understood and described. It is therefore more expedient to model this behavior mathematically, using the information generated by a test that satisfies the requirements of Table 2. Piloting semi-batch centrifuge units in grinding circuits is advantageous in achieving the interaction between grinding, classification, and recovery units that defines how much gold will be recovered at full scale. Typically, the objective of such piloting is to generate a higher degree of confidence in predicted metallurgical performance, when warranted by the scale or complexity of the project. This is not often the case for typical gravity and CIP flowsheets. In some cases, piloting is used to assess head grade more accurately by processing a large representative sample of ore. Should piloting be warranted, gravity recovery can still be over-predicted if the recovery effort used is larger than what will be used at full scale. This is often the case, even when the gravity unit used in the piloting exercise is rather crude (e.g. a strake or blanket). It is essential to measure the gravity recovery effort when piloting, to either design the full circuit to use the same recovery effort, if economically justified, or downgrade pilot performance when scaling up the performance of gravity. Ideally, the pilot recovery effort should be set at that which will be used by the full-scale circuit. The usual pitfalls of bench testing with semi batch centrifuges are:

a

using too many samples that are too small to be statistically meaningful, or samples that are not representative of the ore body testing many operating conditions that generate unusable information, failing to generate size-by-size information about the natural size distribution of the GRG and producing very low grade concentrates that contain too much gold that is not gravity recoverable.

The approach described above is appropriate except when testing for recovery from flash flotation concentrates. Direct test work can then yield an accurate estimate of full-scale recovery, provided the plant unit is fed at a very conservative feed rate (Laplante and Dunne, 2002). When gravity recovery is the sole recovery method, its economic impact is increased at least tenfold, and the test program becomes similar in scope and magnitude to those used for other recovery circuits. Continuous Units: Continuous centrifuge units are typically add-ons to existing circuits. Further, unlike semi-batch centrifuges, the number of units actually used in commercial applications is low and these units are still perceived as being largely developmental. Existing applications reflect this perception, as the units are typically retrofits shoehorned into existing circuits. Typically, test work consists of a limited program of exploratory test work using a “bench-scale’’ continuous unit followed by a trial period with a rental unit. Table 3 shows examples of which units are typically used at bench/continuous and piloting/demonstration scale.

168

Examples of DemonstrationPiloting With J1800 at Labrador City for Iron of Canada (hematite), Granny Smith for Placer Dome Pacific (gold and gold carriers) Falcon’s C (continuous) series C400Falcon Concentrator With C4000 at New Celebration MGS MGS C902 or MeGaSep Mozley’s MeGaSep CVD32 (32”) Luzenac’s Penhorwood CDV6 (6”) Knelson CVD Unithlanufacturer Geologics’ Kelsey Jig

Preliminary Work J200

The objective of the preliminary test work is essentially to generate results that warrant the demonstration/piloting test work on site with rental units. The test work on site is often performed with full-scale or near full-scale units, and consists generally of identifying optimum operating conditions and performance, assess the mechanical reliability and ease of operation of the unit, determine the effect of feed variability (grade, mineralogy) and generate concentrate for further test work. Typically, a two-month program is needed, although in some cases more than 6 months of test work have been necessary.

Data Development and Expectations Semi-batch Centrifuges. When a bench-scale centrifuge unit is used to generate a feed for downstream processing, simple metallurgical balancing (i.e. how much gold reports to the final gravity concentrate) is adequate. Often the final gravity concentrate will be fully fire-assayed. The cleaner gravity tailing may or may not be sampled for its gold content prior to recombining with the rougher tailing product. Emphasis is placed, as it should, on downstream metallurgical performance. Processing a sample to characterize its gravity recoverable gold content is clearly different: the size distribution of gravity recoverable gold must be characterized, as well as the grind at which it is liberated. The McGill GRG test is a release analysis that uses a laboratory centrifuge, a Knelson Concentrator MD3 (Woodcock and Laplante, 1993; Laplante et al, 2001). Well over 100 samples have thus been characterized, many from ore bodies being processed in part by gravity recovery. The test methodology, illustrated in Figure 2, is as follows. A representative sample of 40 to 100 kg is crushed to 100% minus 850 pm for a first recovery attempt. A first concentrate of about 100 g is produced and screened from 600 pm down to 20 pm. The five coarsest size classes are further upgraded with a hydrosizer and examined with an optical microscope. The tailing is split and a 27kg sample is ground to 50-5596 minus 75 pm for a second recovery with the same centrifuge, but at a lower fluidization flow and feed rate to recover finer GRG more effectively. The tailing is processed at a final grind of 80% minus 75 pm. Except for small masses kept for mineralogical examination, the concentrates of all three stages are screened from 20 pm to top size, and the fractions are assayed to extinction. A 600-g sample of each of the three tailings is screened from 20 pm to top size for fire assaying (up to 30 g or one assay ton per fraction). Because of the redundancy of the three stages and the fractional assays, the test is remarkably resilient to typical assaying problems (e.g. broken crucibles and cupolas). Trends in concentrate and tailing assays can be used to replace failed assays and gain insight into the behavior of both GRG and non-GRG. The high initial weight for all stages minimize the problems associated with statistical representativity and the recovery of gold in gold carriers (as sulphide recovery is typically only 3 to 5%). The microscopic examination of the coarser fractions of stage 1 provides valuable insight into GRG associations and potential liberation problems.

169

Samples

(40-100kg)

Main Td

Split 27b

C1ush

Grind 45-55%-75pm

Screen 100%-850pm

Conc.

Conc.

Tailing

Tailing Strtge 3

Stage 1

Stage 2

Screen screen

850 to 20pm

Tailing

Screen

Screen

850 to 20pm Pulverizing +105pm SCWCll

Scrceii

Figure 2 Procedure for Measuring GRG Content with a Knelson Concentrator MD3 (from Xiao, 2001)

The most informative way of reporting data is a graphical representation of the cumuiative retained GRG content. Figure 3 shows three typical responses. The “High-Coarse” curve represents a highly responsive ore that can generate high gravity recoveries (more than 50%), and can be processed equally well with centrifuge and non-centrifuge units. The average response is typical of free-milling ores where gravity recoveries of 30 to 45% can be achieved. The “Low Fine” curve represents unresponsive ores where gravity recovery would be seldom used, but flash flotation can assist both GRG and base metal recovery. The flash flotation concentrate would then have a high GRG content, which can be recovered in part with semi-batch centrifuge units (Putz et al, 1993; Laplante and Dunne, 2002).

Low Fine Average High Coarse

1000

100 Particle Size, pm

Figure 3 Typical responses of the GRG Test

170

When generating the response of the GRG test or any other tests performed at bench-scale to design and predict the performance of a gravity circuit, it is critical to generate size-by-size information, as full-scale units cannot recover fine heavies as effectively as bench-scale units do. Bench-scale performance must therefore be de-rated when the amount of fine GRG recovered is high (Laplante, 2002). Continuous Units. Typically, promising results at bench-scale with any of the four units discussed here will lead to a pilot or demonstration program at mine site. Thus, the emphasis on data reliability and completeness is shifted to this second phase of the work. The typical precautions associated with piloting apply: The program should be of adequate duration to optimize unit operation, cover normal variations in ore type, and assess unit mechanical reliability and ease of operation. Data generated should mass-balanced using a least-square algorithm to assess and improve reliability. Enough samples of the feed, concentrate and tailing streams should be extracted for further studies, in particular size-by-size performance and cleaning/scavengingat bench-scale Projecting Data to Commercial Application, and Limitation Thereof Semi-batch Centrifuges. When gold is recovered by gravity from a grinding circuit, typically a portion of the circulating (as low as 5 % ) is screened and fed to the centrifuge unit, which may recover as little as 10% of the GRG in its feed. Reasonable recoveries are achieved only because of the high GRG circulating loads that arise from its slow grinding kinetics and high recovery to cyclone underflows. Consequently, the projection of bench-scale data to the recovery of a semibatch centrifuge operating in a grinding circuit must take into consideration not only the GRG content, but also the dynamic interaction between grinding, classification and recovery units. Population balance modeling (PBM) has been used with some success in the past (Laplante et al, 1995). Recently, a simple regression equation has been developed to predict how much GRG can be recovered as a function of its size distribution, the recovery effort, and two parameters representing grinding and classification behavior, respectively (Laplante and Xiao, 200 1). The effect of classification is extremely important, because GRG finer than 106 pm is unlikely to be ground into finer particles that are not gravity recoverable on account of its very slow grinding kinetics (Banisi et al, 1990). It thus, either exits the grinding circuit via the cyclone overflow or is recovered by gravity. The probability of reporting to the cyclone overflow and thus not be recovered by gravity is the mirror image of the GRG partition curve. Figure 4 shows that this probability is highly dependent on grind size, and is either much lower than or comparable to typical GRG recoveries at fine size for semi-batch centrifuge units (from 10 to 30%). Generally, full-scale recovery will be as low as 20% of the GRG content of the ore for difficult applications or poorly designed circuits, to more than 80% of the GRG content for very efficient circuits. Circuits fail to perform as predicted for reasons outlined in Table 4. Note that if some problems can be corrected, there is no correction for unrealistic expectations. Factors affecting gravity circuit performance are discussed in more detail in Laplante (2000b). The advent of commercial intensive cyanidation units to treat centrifuge unit concentrates (Gray and Katsikaros, 1999; Lethlean and Smith, 2000) is seen as very positive, as these units achieve GRG recoveries in excess of 97%, as opposed to the conventional tabling approach which can achieve GRG recoveries as low as 40 to 50%. Not only can overall gravity recovery be significantly improved, but also the uncertainty of gold room performance is eliminated.

171

100 80 U

5

s

60

.64% ~81Oh

40

20 0 10

100

1000

Particle Size, pm Figure 4 Proportion of the GRG Reporting to Cyclone Overflow for Coarse (64% minus 75 p)and Fine (81% minus 75 pn) Classification

Table 4 Common Causes of Poor Performance for Gold Gravity Circuit Problem The GRG content is based on a nonrepresentative sample It is assumed that all the GRG content will be recovered

Solution Follow an accepted sampling protocol; do not selectively choose a hinh-made sample for testwork Generally, between one third and two-thirds of the GRG is recovere GRG recovery can be estimated from gravity is recovered. circuit desien and grind size The wrong stream is bled for gravity recovery Always choose primary cyclone underflow or the’ corresDondinc!ball mill discharge Too little of the circulating load is bled to the Design screening ahead of primary recovery carefully. primary recovery circuit, or the primary Use horizontal vibrating screen of adequate surface area recovery unit fails to perform because its feed is too coarse or too dense (i.e. massive sulphides) The primary gravity concentrate is processed in Using intensive cyanidation whenever possible. If not, apply the guide-lines of Laplante (1 999) a poorly designed gold room Classification is too coarse and fails to keep Problem can be alleviated by increasing the recovery effort. GRG below 106 pm in the grinding circuit Flash flotation used in the same grinding circuit Recover GRG from the flash flotation concentrate using a semi-batch centrifuge centrifu (Laplante te and 2002) decreases gravity recovery semi-batch and Dunn, Dunn, 2002

Continuous Units. Because continuous units are relatively unproven, they are likely to be extensively tested before a final decision is made to proceed. The final test program is likely to be performed with either the full-scale unit or one only slightly smaller. As a result, data projection to the commercial application is less likely to prove disappointing. The major risks pertain to mechanical issues; a recent application of a continuous centrifuge unit was recently discontinued, despite a significant initial investment, because of wear and maintenance costs. Examples of Successful Testwork, Design, and Commercialization Semi-batch centrifuges. There are many examples of applications of semi-batch centrifuge units (Laplante et al, 1989; Darnton et al, 1993; Putz et al, 1993; Gregory et al, 1996; Hewitt, 1996; Folinsbee and Hewitt, 1997; Ritson and Tyreman, 2001; Choquenaira Bombilla and Muiioz, 2002). A number of these replaced other units with one or many semi-batch centrifuges, with little or no test work, and reported increases in gravity recoveries. In other cases, a semi-batch centrifuge gravity circuit was shoehorned into existing grinding. In virtually all cases, significant gold recoveries are achieved. Figure 5 (Laplante and Xiao, 2001) shows that GRG recovery is proportional to the logarithm of the recovery effort (i.e. the proportion of the GRG circulating load recovered into bullion). It follows that even when half the planned recovery effort is

172

r-

40 Q

a?

Q

20

,

-E- Finest GRG

I

0 1

10

100

Recovery Effort, 016

Figure 5 GRG Recovery as a Function of the Gravity Recovery Effort (Laplante and Xiao, 2001) achieved, around 12% GRG recovery is lost. For typical free milling ores, whose GRG content is about two-thirds of the total gold content, this corresponds to a relatively small drop of 8% in gravity recovery. This explains why few gold gravity circuits are deemed to perform poorly. The logarithm relationship does, however, limit the upside potential of improved gravity recovery: increasing the recovery effort follows a law of very rapidly diminishing returns. The challenge of gravity circuit design for gold, then, is to estimate the economic potential of gravity recovery and adjust the recovery effort (i.e. the capital and operating costs of the gravity circuit) accordingly. The best approach to gravity circuit design is to obtain an estimate of gravity recoverable gold content with the GRG test or other methods. The economic potential of gravity recovery is then estimated. A limited number of circuit options are then costed and their projected gravity recovery and economic impact estimated. Classical economic analysis criteria (i.e. NPV, IRR) are then used to select the best option. The procedure is illustrated in Laplante (2002). Continuous Centrifuges. A limited number of successful applications of continuous centrifuge units have been reported (Laplante, 2000). The actual number of successful applications and units on order is much higher, but this information is generally no published or difficult to confirm independently. The most successful application of the Falcon continuous concentrator is at Tanco, Manitoba (Deveau, ZOOO), for the scavenging of tantalum-bearing minerals. The gravity tailing is treated in a two-stage circuit, the concentrate of a Falcon C20 unit (now C2000) being treated by a Falcon C10 (now ClOOO) unit. This paper suggests that for relatively new units operating at high Gs, a close relationship between the supplier and client to address wear problems and creativity on the part of the user are essential for successful applications. A second application at Kettle River gold mine sees a Falcon C4000 upgrade a cyanidation feed (cyclone overflow) into a 7% yield (with a gold recovery of 1618%). The concentrate is directed to the first tank of the cyanidation train, which has a very high retention time and high cyanide additions. The tailing is combined with the discharge of the first cyanidation tank. The metallurgical benefit of the application is difficult to measure on account of the variability of the ore, but it is estimated at 0 to 5%, with an average of 1.5%. The unit has also been extensively tested for coal cleaning (Luttrel et al, 1995) The Renison application of the Kelsey jig is well documented (Beniuk et al, 1994), and is only one of the many applications for tin (cassiterite) recovery (e.g. Wyslouzil, 1990). The unit is also used for the recovery of tantalum minerals at Greenbushes (Western Australia), and the separation of zircon from kyanite in beach sand operations. Recent test work on iron ore has yielded promising results, and a circuit is being commissioned in the early 2002 in Western Australia for the scavenging of gold and gold carriers from a cyanidation tailing. The iron ore and gold applications represent a bold move to the recovery of relatively low value materials (for gold because of the low gold content of the feed treated, about 0.3 to 0.5 g/t). If these prove successful, the number of Kelsey jig users could increase significantly.

173

The Knelson CVD and Mozley's MeGaSep are relatively new comers in centrifuge separation. To date, the only commercial application of the CVD is a talc cleaning duty in Eastern Canada. The MeGaSep precursor, the MGS C902, has been used extensively in tin before, but its low capacity, typically 2 to 5 t/h, was clearly a deterrent. Other, non-publicized, applications include the sporadic cleaning of flue dust at the Horne Smelter in Noranda, QuCbec. The MeGaSep is claimed to achieve capacities of 30 to 60 t/h, which would make it viable for a much larger number of applications than its predecessor. The web sites of most suppliers has been significantly improved in the past tow years. For potential users, this is a welcome development. Independent parties have produced much of the data that can be downloaded from these sites. The address of relevant web sites is given in appendix.

REFERENCES Banisi, S., A.R. Laplante and J. Marois, 1991. A Study of the Behaviour of Gold in Industrial and Laboratory Grinding, CIM Bull,, Nov. 1991, pp. 72-78 Beniuk, V.G., C.A. Vadeikis and J.N. Enraght-Moony. 1994. Centrifugal Jigging of Gravity Concentrates and Tailings at Renison Limited. Minerals Engineering, Vol. 7(5/6), pp. 577-589 Burt, R., 1985. Gravity Concentration Technology. Elsevier, Amsterdam Choquenaira Bombilla, V. and O.A. Mufios. 2002. Cold Gravity Recovery in Copper Circuits at BHP Tintaya. Proc. of 34''Ann. Canadian Mineral Processors ConJ, Ottawa, Jan. 2002, pp. 91-104 Darnton, B., S. Lloyd and M.A. Antonioli, 1993. Gravity Concentration: Research, Design and Circuit Performance at Montana Tunnels. Randol Gold Forum Vancouver '93, pp. 137-143 Deveau, C. 2000. The Evolution of Falcon Continuous Concentrators at Tantalum Mining Corporation of Canada. Proc. of 32" Ann. Canadian Mineral Processors ConL, Ottawa, Jan. 2000, pp. 1-18 Folinsbee, J. and B. Hewitt. 1997. Gravity Concentration at Placer Dome's Campbell Cold Mine. Mining Engineering, Oct. 1997, pp. 44-47 Goulsbra, A., R. Dunne and S. McAllister. 1998. The Application of the Continuous Falcon Centrifugal Gravity Concentrator for Gold and Pyrite Recovery. Randol Gold and Silver Forum '98,Denver, April 1998, pp. 11 1-113 Gray, A.H. and N. Katsikaros. 1999. The InLine Leach Reactor - The New Art in Intensive Cyanidation of High Grade Centrifugal Gold Concentrates. Randol Gold and Silver Forum, May 1999,5 pp. Gregory, S., R. Dunne, P. Gelfi, V. Martins and A. Goulsbra, 1999. Gravity Concentration at the Telfer and New Celebration Gold Mines. Randol Gold Forum '96, Olympic Valley, April 1996, pp. 79-85 Hewitt, B., Gravity Concentration at Campbell Mine. 1996. Randol Gold Forum '96, Olympic Valley, April 1996, pp. 93-99 Hope, G.H., J. McMullen and D. Green. 1993. Process Advances at lac Minerals Ltd. Est Malartic Division. 2fhAnn.Meet. of Canadian Mineral Processors, Ottawa, Jan. 93, Paper 12, 13 pp. Kelly, E., Spottiswood, D. 1982. Introduction to Mineral Processing. John Wiley & Sons, New York. Laplante, A.R., L. Liu and A. Cauchon. 1989. Mineralogy and Flowsheet Changes at the Camchib Mines Inc. Mill, Chibougamau, Quebec. Process Mineralogy IX: Applications to Mineral Beneficiation, Metallurgy, Gold, Diamonds, Ceramics, Environment and Health, Eds. W. Petruk, R.D. Hagni, S. Pignolet-Brandomand D.M. Hausen, TMS Pub., pp. 247-258, 1989 Laplante, A.R., F. Woodcock and M. Noaparast. 1995. Predicting gravity separation gold recovery, Minerals and Metallurgical Processing J-, May 1995, pp. 74-79 Laplante, A.R., L. Huang and B.G. Harris. 1999. The Upgrading of Primary Gold Gravity Concentrates. Proc, of 31" Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 1999, pp. 21 1-226 Laplante, A.R. 2000. Centrifuge Units for Gravity Separation - An Update. Proc. of 32"dAnn. Meet. of Canadian Mineral Processors, Ottawa, Jan. 2000, pp. 475-488 Laplante, A.R., F. Woodcock and L. Huang. 2001. Laboratory procedure to characterise gravityrecoverable gold. SME Trans., Vol. 308, pp. 53-59

174

Laplante, A.R. 2000. Testing requirements and insight for gravity gold circuit design. Rand01 Cold Forum, Vancouver, April 2000, pp. 73-83 Laplante, A.R. 2000. Ten do's and don'ts of gold gravity recovery, Randol Gold Forum, Vancouver, April 2000, pp. 107-1 17 Laplante, A.R. and Z. Xiao. 2001. Optimizing Gold Gravity Recovery: the Role of the Recovery Effort. Proc. of 3YhAnn. Canadian Mineral Processors ConJ, Ottawa, Jan. 2001, pp. 371-388 Laplante, A.R. and R. Dunne. 2002. The GRG Test and Flash Flotation. Proc. of 34*Ann. Canadian Mineral Processors Cant, Ottawa, Jan. 2002, pp. 105-124 Laplante, A.R. 2002. The Selection and Design of Centrifugal Concentration Equipment. paper D-D4, in Proc. of Symposium on Mineral Processing Plant Design, Control and Practice, in print Leathlean, W. and L. Smith. 2000. Leaching Gravity Concentrates Using the Acacia Reactor. Randol Cold Forum, Vancouver, April 2000, pp. 93-100 Luttrel, G.H., R.Q. Honaker and D.I. Philips. 1995. Enhanced Gravity Separators: New Alternatives for Fine Coal Cleaning. Proceedings of 12Ih International Coal Preparation Conference, Lexington, Kentucky, pp. 28 1 -292Peacock, K. 2002. Private communication Mills, C., 1980. Process Design, Scale-up, and Plant Design for Gravity Concentration. Mineral Processing Plant Design, 2"dEdition. Edited by A.L. Mular and R.B. Bhappu, Society of Mining Engineers, New York, New York. Putz, A., A.R. Laplante and G. Ladouceur. 1993. Evaluation of a Gravity Circuit in a Canadian Gold Operation. Randol Gold Forum, Beaver Creek, September 1993. pp. 145-149 Ritson, G. and R. Tyreman. 2001, Goldcorp's Red Lake Mine - Design and Commissioning. Proc. of 3JhAnn. Canadian Mineral Processors ConJ, Ottawa, Jan. 2001, pp. 253-270 Spiller, E. 2000. A Commercialized Dry Gravity Concentrator for Gold Applications. Randol Gold & Silver Forum 2000, Vancouver, B.C. Canada. Spiller, E. 2000. Mineral Beneficiation - Gravity Concentration. Mineral Processing Handbook. to be published, Society of Mining Engineers, Littleton, Colorado. Traore, A., P. Conil, R. Houot and M. Save. 1995. An Evaluation of the Mozley MGS For Fine Particle Gravity Separation. Minerals Engineering, Vol. 8(7), pp. 767-778 Van Kleek, D.M. 2001. Knelson Concentrators Extreme Gravity, from the Knelson Concentrators Web Site (www.knelson.com),7 pp. Woodcock, F. and A.R. Laplante. 1993. A Laboratory Method for Determining the Amount of Gravity Recoverable Gold. Rand01 Gold Forum, Beaver Creek, September 1993. pp. 151-155 Wyslouzil, H. 1990. Evaluation of the Kelsey Centrifugal Jig at Rio Kemptville Tin. Proc. of 22" Annual Meeting of Canadian Mineral Processor, Ottawa, Jan. 1990, pp. 461-472 Xiao, Z. 200 1. Developing Simple Regressionsfor Predicting Gravity Recovery in Grinding Circuits, M.Eng. Thesis, McGill University, Oct. 2001, 138 pp. Websites: Falcon Concentrators: InLine Leach Reactor (Gekkos) Kelsey Jigs (Roche Mining) Knelson Concentrators MGS and MeGaSep SpiraldShakingTables/Density Separators ConedSpiraldShaking Tables

http://www.concentrators.netl http://www.gekkos.com.au/ http://www.geologics.com.au/ http://www.www.knelson.com.ca.org/

http://www.mozley.co.uWmg.htm http://www.outokumpu.com/mintec/physicalseparation.htm http://www.mdmintec.com.au

175

Bench Scale and Pilot Plant Tests for Magnetic Concentration Circuit Design Daniel A Norrgran' and Michael J. Mankosa'

ABSTRACT The application of magnetic separation in the minerals industry has traditionally been limited to strongly magnetic (ferromagnetic) minerals. The recovery of less magnetic minerals required lowcapacity, energy-intensive electromagnetic separators. In the past decade, however, there have been significant advancements in the design and application of permanent magnetic separators. This progress is predominantly a result of improvements in the properties and availability of rare earth magnet material. As a result, today's mineral processing engineer has many options for the design and application of magnetic separation circuits. The diversity of magnetic separation systems that are currently available requires that flowsheet testing and development be closely scrutinized to ensure success in plant practice. Consideration must be given to the characteristics of both the mineral and the separator. Specifically, the test work must be carried out on representative samples that account for the geological diversity of the deposit. Furthermore, the test work must be conducted in a manner that is representative of production scale processes. This implies a high level of confidence that can be directly correlated to the metallurgical performance of the production scale separator. This paper will discuss the mineral and separator characteristicsthat are requisite to magnetic separation. The objective is to provide a definitive approach for the successll design and implementation of magnetic separator circuits. INTRODUCTION There have been significant advancements in the design and application of magnetic separators in the recent past. New magnet materials and circuit designs have allowed for the manufacture of separators that operate at substantially higher field strengths. As a result, new opportunities for magnetic separation have evolved. Magnetic separation is now a viable option that has been applied to a vast array of weakly magnetic materials. These separators are essential in view of new advances in minerals and materials specifications. Increasingly, there is a demand for higher-purity mineral products and feedstocks. The diversity of magnetic separation systems and applications require that flowsheet testing be conducted to ensure the success of plant performance. Several variables affect the process of magnetic separation. The variables are either the inherent properties of the mineral or a function of the separator. MINERAL CHARACTERISTICS The concept of magnetic separation is based on the ability to magnetke a particular mineral and then physically collect it. The magnetic susceptibility of the mineral is an inherent characteristicdirectly proportional to the response of a magnetic field and is the single most important variable when addressing the characteristicsof magnetic separation. When subjected to a magnetic field, all particles will respond in a particular manner and can be classified as one of three groups: ferromagnetic,paramagnetic,or diamagnetic. Minerals that have a very high magnetic susceptibilityand are strongly induced by a magnetic field are fmomagnetic. There are a 1

Eriez, Erie, Pennsylvania

176

few common ferromagnetic minerals that respond to magnetic separation as shown in Table 1. .&er& that have a low magnetic susceptibility and a weak response to a magnetic field are termed paramagnetic. There are several common paramagnetic minerals that are respond to magnetic separation as shown in Table 2. Table 1. Common FerromagneticMinerals Magnetite Titanifmous Magnetite Monoclinic Pyrrhotite Martite

I

Table 2. Common Paramagnetic Minerals Hematite Specularite Geothite Ilmenite Monazite Staurolite Hexagonal Pyrrhotite

Columbite Tantalite Garnet Biotite Chalcopyrite Chromite Siderite

I

Minerals with a negative magnetic susoeptibility are termed diamagnetic and for all practical purposes are non-magnetic. Fmomagnetic and to a l e s s extent paramgnetic materials will become magnetked &en placed in a magnetic field. The amount of magnetization induced on the particle depends on the mass and magnetic susceptibility of the particle and the intensity of the applied magnetic field. This can be expressed as:

M =mXH Where M is the induced magnetization of the particle, rn is the mass of the particle, magnetic susceptibility,and H is the magnetic field intensity.

x

is the specific

SEPARATOR CHARACl'ERISTICS In the design of a magnetic separator, the magnetic field intensity and the magnetic field gradient are the two fjrst order variables that affect separation respome. High intensity magnetic qmators typically operate in regions over 5,000 gauss or 0.5 Tesla. Low intensity separators are commonly referenced as those separators generating a magnetic field strength of less than 2,000 gauss or 0.2 Tesla, which is the practical limitation of conventional permanent ferrite magnets. The magnetic field gradient refers to the rate of change or the convergence of the magnetic field strength. Figure 1 illustrates two different magnetic field configurations. Figure 1. Magnetic Field Configurations

B

A

dH

dH ->o

-0

dx

177

Case A has a very u n i f m pattern of flux lines without gradation. A magnetic particle entering this field will be attracted to the lines of flux and remain stationary without migrating to either pole pi-. Case B illustrates a converging pattern of flux lines displaying a high gradient. As these lines pass through a smaller area, there is a significant increase in the magnetic field intensity. A magnetic particle entering this field configuration will not only be attracted to the lies of flux but will also migrate to the region of highest flux density, which occurs at the tip of the bottom pole piece. This illustrates the methodology for magnetic separation. In simplified terms, the magnetic field intensity holds the particle while the magnetic field gradient moves the particle. From the earlier equation for magnethation, the magnetic attractive force acting on a particle is the product of the particle magnetization and the magnetic field gradient and can be expressed as:

dH dH Fm = mxHor M dx ak

Where Fm is the magnetic attractive force,and

dK is the magnetic field gradient.

Maximum magnetic dx force results only when both the magnetic field intensity and field gradient are maximized. There are two common methods for producing a magnetic gradient in a magnetic separator. The firstmethod, which is typical of magnetic circuits utilizing permanent magnets, is to concentrate the lines of flux on a steel pole piece within the circuit. This can be accomplished simply by placing a steel pole piece between two magnets. The magnetic flux will be concentrated in the steel pole piece resulting in an area of extreme magnetic field intensity. The second method involves positioning a steel matrix, such as a metal mesh, directly in a u n i f m magnetic field generated by an electromagnetic solenoid coil. This matrix consequently amplifies the magnetic field and converges the lines of flux to produce localized regions of extremely high magnetic field intensity. MAGNETIC UNITS The subject of magnetics is complicated by the lack of standardization throughout the scientific community on the units of magnetic quantities. The cgs system was once prevalent and is still commonly encountered today. There has been,however, a movement to a single comprehensive system known as Le Systhe Internationale d Unit&, abbreviated SI, which was officially adopted in 1960. Nearly all modern physics and electrical engineeringtextbooks as well as technicaljournals have adopted SI units. Unfmately, the conversion between SI units and cgs units is not straightforward. When a choice is to be made between SI units and cgs units, the choice is usually made on the basis of convenience. Since this paper will be presented to process engineers with a wide range of backgrounds, where appropriate, both systems of units will be provided. Magnetic quantities and the associated system of units are a topic in itself. Our intent is to provide the most basic understanding of the common units used in magnetic separation. By convention, magnetic separators are rated on the magnetic induction or flux density B. The magnetic induction or flux density, B, is defined by the flux passing through a unit area normal to the direction of the flux. The unit of B in the SI system is W e k h e t e ? but the term Tala (T) has now been adapted. The unit of B in cgs units is measured in gauss or the lines of flux passing through a square centimeter. Convertingthe magnetic induction or flux density from SI units to cgs units is as follows:

Hcgs = HsI(1 0") or 1 gauss = 0.0001 Tesla. Conversely 1 Tesla = 10yOOOgauss. This is undoubtedly the most important conversionto note when addressing magnetic separators. MAGNETIC FIELD GENERATION Either permanent magnets or an electromagnet generates the magnetic field on any given magnetic separator. When addressing the historical aspects of magnetic separation, a strong differentiation exists between the developments of those separators utilizing permanent magnets and those separators utilizing electromagnets. The two separators were developed independently and ofken times in competition.

178

-

Permanent Magnets Ferrite There are two distinct types of permanent magnets utilized in magnetic separators. The first type of permanent magnet is a “ferrite” magnet and is used in low-intensity magnetic separators. The formulation is SrFe12019. Ferrite is a very inexpensive material with a moderate energy product ranging up to -4.5 Million Gauss-Oerstads (MGOe). (The energy product of a permanent magnet is a relative measure of the intrinsic strength). Ferrite magnets are used primarily in drum type separators that are used for collecting ferrous material as well as magnetite and pyrrhotite. These separators generate a magnetic field strength ranging up to 2000 gauss (0.2 Tesla) Permanent Magnets - Rare Earth The development of rare earth permanent magnets has revolutionized the field of magnetic separation. The advent of rare earth magnets has allowed for the design of high-intensity magnetic circuits operating energy fke and surpassing the strength and effectivenessof electromagnets. Samarium and neodymium are the two most common elements used in the commercial manufacture of rare earth permanent magnets. The intermetallic compound Nd2Fe18 is the third generation of rare earth magnets and is most prevalent today. Neodymium magnets are experiencing tremendous growth. Much of the growth is strictly due to the economics of using a much more abundant rare earth (Nd is 10 times more abundant than Sm) coupled with inexpensive Fe and B. The energy product of the neodymium based magnets now range up to 50 MGOe. Rare Earth magnets are used in high-intensity drum and roll type separators that are effective for collecting paramagnetic minerals. Dependent on the magnetic circuit, these separators generate a magnetic field strength ranging up to 24,000 gauss (2.4 Tesla). Presented in Figure 2 is a chronology of permanent magnets illustratingthe increase in energy product.

Figure 2. Chronology of Permanent Magnets.

Electromagnets Industrial electromagnetic separators are typically designed utilizing a solenoid electromagnetic coil. Some separators utilize the bore of the solenoid coil as the separating zone. Other separators use the solenoid coil to convey the magnetic flux through a steel circuit or a “C“ frame circuit. The magnetic field in the gap, either the bore of the solenoid or the gap of the C frame, serves as the separation zone in a magnetic separator. Several operational factors have to be taken into

179

consideration with a production magnetic separator designed for continuous use. There has to be a balance between the magnetic field strength, the associated steel circuit, the power requirements, and cooling factors. Production scale electromagnetic separators always employ a steel circuit to convey the magnetic flux and to pattern the magnetic field in the separating zone. The steel circuit will tremendously enhance the magnetic field strength in the separating zone. The power input to the separator must also be designed to avoid saturating the steel circuit and overheating the electromagnet. Typically electromagnetic separators operate up to approximately 20,000 gauss (2 Tesla). At this magnetic field strength the iron circuit begins to saturate.

MAGNETIC SEPARATION APPLICATIONS Laboratory Magnetic Separation The magnetic collection of any specific material can be assessed in a straightforward manner in the laboratory. This cursory evaluation will conclusively determine if the material is a candidate for magnetic separation. As a minimum, bench scale magnetic separation testing should be conducted to determine the response of the specific mineral to a magnetic field. The selected mineral must respond appropriately to a magnetic field in order for magnetic separation to be considered a process option. It is typically the case that the magnetic collection of the mineral is measured as a function of magnetic field strength. This provides the basic information necessary to assess the potential of magnetic separation. Several laboratory separators exist for the quantification of magnetic collection of minerals. Pilot-Scale Magnetic Separation Pilot scale magnetic separation tests provide the means to predict production plant performance. Once the bench scale testing has defined the basic separation parameters, larger scale test work provides other benefits to flowsheet development. Pilot scale magnetic separation test work should closely resemble the production process and provide a higher degree of confidence for flowsheet development. There are several specific items that can be identified from pilot scale testing. These items can also be regarded as a particle characteristic or a separator characteristic. Presented in Table 3 are several mineral characteristics that can be assessed through pilot scale test work. These characteristics are functions of the mineral. Some of these characteristics, such as magnetic susceptibility, are inherent to the mineral and can only be changed with some degree of difficulty. Other characteristics, such as extent of liberation, temperature, or moisture content, are specific to the sample preparation and can be altered in a straightforward manner. Conversely, there are a few separator characteristics that can be assessed through pilot scale test work. These characteristics are presented in Table 4.

I

I

Table 3. Mineral Characteristics

Table 4. Separator Characteristics

Magnetic Susceptibility Concentration Recovery Variability of the Ore Extent of Liberation Temperature Moisture Content Interaction with Different Minerals (Overlapping Magnetic Susceptibilities)

Magnetic Field Strength Magnetic Gradient Unit Capacity

1

There are several secondary variables related to unit capacity that are separator characteristics. These relate to the design and operation of any specific separator. Many aspects of the feed parameters will affect the capacity of the separator, such as percent solids, flow rate, drum speed, or belt speed.

180

Empirical Bench Scale Separators The initial step in assessing the magnetic separation of a mineral sample is to measure the response of the mineral to a magnetic field. The most common approach is to hold a hand magnet to the mineral sample. This is simply the first indicator to the presence of ferromagnetic material. There are two different bench scale separators commonly used for the quantitative analysis of magnetic minerals. These separators are specifically designed to provide a separation based on magnetic susceptibility. Both separators provide a high level of control and selectivity. The Davis Tube tester provides a measure of ferromagnetic minerals while the Frantz Isodynamic or Barrier separator provides a measure of paramagneticminerals. A Davis Tube tester is used to quantifL the content of ferromagnetics. Figure 3. Illustrates a Davis Tube tester consisting of a 1 inch diameter glass tube which is agitated within an electromagnetic field. An electromagnetic coil on each side of the tube generates the magnetic field. The magnetic field strength in the separating zone can be varied.

Kgure 3. Davis Tube Tester for Measuring Ferromagnetic Content.

The sample to be tested is slurried and introduced to the glass tube. The magnetic field holds the ferromagnetics within the tube during agitation. Wash water flowing through the tube removes the non-magnetics. This apparatus provides essentially a perfect separation of the fmomagnetics. The variables controlling the selectivity of the separation are the magnetic field strength, wash-water volume, agitation, and slope of the glass tube. The Davis Tube tester is used extensively in measuring the performance of wet drum magnetic separators in both the iron ore industry and in heavy media recovery applications. Samples of the non-magnetic effluent are tested for magnetite content indicating recovery performance. The Davis tube will also provide a measure of product quality for magnetite concentrates. Since the Davis Tube provides essentially a perfect separation, any diluents (commonly silica and alumina) in the magnetite concentrate occur as locked particles. The Frantz Isodynamic or Barrier separator is used to separate paramagnetic minerals based on magnetic susceptibility. This instrument selectively separates minerals into products with a very narrow range of magnetic susceptibilities. Figure 4. illustrates the Frantz separator, which consists of a steel "C" f k n e circuit coupled with electromagnetic coils. The magnetic field strength in the separating zone is variable. The Frantz separator treats dry granular mineral samples. The gap or Separatingzone in the Cfiame circuit is at a downward slope as shown in Figure 4. The mineral sample is fed through the magnetic field on an incline chute running the length of the electromagnetic circuit. As the mineral grains pass through regions of high magnetic gradient, those minerals with a selected threshold magnetic susceptibility are diverted from the mineral stream. The non-magnetic mherals pass through the magnetic field unaffected. Several passes can be made through the separator at increasing

181

Figure 4. Frantz IsodynamidBarrier Separator. This unit specificallyprovides separation of paramagnetic minerals based on magnetic susceptibility.

magnetic field strength to perform a fractionation of the entire sample based on magnetic susceptibility. This particular series would yield several magnetic fractions with decreasing magnetic susceptibility. The variables controlling the selectivity of the separation are the magnetic field strength and the slope and pitch of the incline chute. This instrument is also used for assessing liberation. The various size fractions of a mineral sample are treated separately on the Frantz Isodynamic separator. A comparison of the chemical analysis of the magnetic products throughout the size range will indicate the extent of liberation. The level of diluents contained with the magnetic fractions indicates the relative amount of locking. MAGNETIC SEPARATOR SELECTION Magnetic separators can be classified into four major categories. The separator either functions as a dry type separator treating “freeff owing” ore or a wet type separator treating a slurry. A further category is the magnetic field strength generated by the separator. Low-intensity magnetic separators generally operate in the range of less than 2000 gauss (0.2 Tesla) and are effective in collecting ferromagneticminerals such as magnetite or pyrrhotite. Most low intensity separators utilize ferrite permanent magnets to generate the magnetic field. A magnetic field strength of 2000 gauss is the practical limit for this type of magnet. High-intensity magnetic separatorsgenerally operate in excess of 5000 gauss (0.5 Tesla) and are effective in collecting paramagnetic minerals such as hematite or ilmenite. High intensity magnetic separators utilize either rare earth permanent magnets or electromagnets. For all practical purposes magnetic separator selection can be reduced to the decision tree illustrated in Figure 5. The first node indicates that the mineral will be treated as either a slurry or as dry and fiee-flowing. The second node indicates the magnetic field strength necessary for effective separation.

-

Dry Drum Type Magnetic Separator (Low-Intensity) Dry drum type magnetic separators have a long nmning history. These separators, opadting with either permanent magnets or electromagnets, are used extensively for removing tramp metal to concentratingmagnetic minerals. This separator is very effective in treating a process stream containinga high level of magnetics either producing a “clean“non-magneticproduct or concentrating and recovering a magnetic product. In many applications the drum separator is unsurpassed in effectiveness and mechanical efficiency. The drum separator is used in literally hundreds of applications spanning a wide range of industries. Ferromagnetic minerals are easily collected with a low-intensity magnetic field. Drum type separators are the most common separators employed for the collection of ferromagnetic minerals. Used in either a dry process or wet slurry configuration, drum type separators provide excellent ferromagneticrecoveries

182

Figure 5. Magnetic Separation Decision Tree

Low-Intensity

Mineral Sample

DryDrum

Induced Roll

The dry magnetic drum separator consists of a stationary, shaft-mounted magnetic circuit completely enclosed by a rotating drum. The magnetic circuit is typically comprised of several magnetic poles that span an arc of 120 degrees. When material is introduced to the revolving drum shell (concurrent at the 12 o'clock position), the non-magneticmaterial discharges in a natural trajectory. The magnetic material is attracted to the drum shell by the magnetic circuit and is rotated out of the nonmagnetic particle stream. The magnetic material discharges fim the drum shell when it is rotated out of the magnetic field. A schematic illustration of a drum separator is shown in Figure 6.

I Figure 6. Schematicof Dry Drum Magnetic Separator FEED

1 TRAJECTORY

STAINLESS STEEL DRUM SHELL

ADJUSTABLE SPLITTER MAGNETIC TRAJECTORY

A pilot-scale magnetic drum separator is shown in Figure 7. This particular drum is 0.6 meter (24 inches) diameter by 0.3 meter (12 inches) drum width. Note the splitter at the 4 o'clock position to segregatethe

magnetic product &om the non-magneticproduct.

183

Figure 7. Dry Drum Magnetic Separator

The magnetic attractive force exerted on a fmmagnetic mineral by a dry drum type separator is opposed by the centrifugal force. This is the case with any type of rotating separator. 'The primary variables affecting separationefficiency in rotating separators are: Magnetic field strength. The magnetic field strength provides the magnetic attractive force that counteractsthe centrifugal force. Feed Rate. in assessing the feed rate, a balance must be struck between an economic capacity, product specifications, and recovery. As the feed rate increases, the layered particle bed on the separator surface increases in height This increase in burden depth on the drum shell moves material further away fiom the magnetic field and the collection of magnetics deteriorates. Drum Speed. The linear speed of the drum is also a primary variable related to the feed rate. As the linear speed is increased, the layered particle bed decreases in height responding with an improved collection of the magnetic particles. The centrifugalforce exerted by the drum or roll surface is the critical factor in providing separation. Beyond the critical speed, the centrifusal face overcomes the magnetic attractive force and the separation efficiency deteriorates. Particle size. Particle size also affects separation efficiency independent of all other variables. Coarse particles provide a relatively high burden depth on the separator surfsce and respond with a relatively high magnetic attractive force. Coarse particles typically provide high unit capacities with high separation efficiencies. Fine particles with a relatively low mass respond detrimentallydue to electrostatic forces. As a consequence, precise magnetic separations balancing magnetic forces against cenh-ifugalforces deteriorates. An effectiveseparation requires a equilibrium among these variables.

Laboratory dry drum magnetic separators typically range fiom 0.3 to 0.9 meter 12 to 36 inches) diameter. The linear speed of the drum is variable. Most applications require a drum speed fiom 0.5 to 1.5 m/s. Coarser particles can be treated at a slower drum speed than h e particles. Fine particles (essentially -200 mesh) require higher drum speeds to impart enough centrifugal force to effectively reject the non-magnetic particles. Further, coarser particles will provide a much higher feed rate or unit capacity than h e particles. This is simply the case of coarser particles providing a higher burden depth.

184

Pilot scale test work is usually required to accurately assess the feed rate or unit capacity. The ideal case is to test on a pilot scale drum that has the same diameter as a production scale drum. This manner provides linear scale-up on only the width of the drum. There is no interpolationrequired in scaling the drum diameter. The unit capacity is the feed rate per unit length of drum width. Merminmg the unit capacity of a ferromagnetic application is straightforwardwith many applications to reference. Table 5 . provides a generalizationof unit capacity treating ferromagneticmaterials on a 0.9-meter diameter drum magnetic separator. Table 5. Typical Unit Capacities of Ferromagnetic Applications Application Unit Capacity TPwMeter Dnrm Width Coarse Magnetite Mill Feed -12 mm 35 to75 Coarse Magnetite Ore -6 mm Top Size 25 to50 Nickel Slag -6 mm 15 to 25 Fine Magnetite Concentrate-325 Mesh 10 to 15 Fine Iron Carbide Concentrate-325 Mesh I5 to20 Flyash 10 to 15 Pilot scale testing typically involves a large diameter magnetic separator incorporating many agitathg magnetic poles. The DFA or Dry-Fast-Agitating magnetic drum type separator combines an agitating element with a high drum speed. This combination provides a high capacity separation with a high level of precision. The DFA drum is 0.9-meter diameter and incorporates an axial agitating magnetic element. The magnetic element configuration consists of a series of axial poles configured with alternating polarity that spans 210 degrees. Strorttium-ferritemagnets are used for the collection and recovery of ferromagnetic materials. The high drum speed in combination with many agitatingmagnetic poles provides a high level of agitation as the fmmagnetics are transferred along the drum smface. This agitation is functional in releasing physically entrapped non-magnetics fiom the flocculated bed of magnetics. This approach r d t s in a near complete rejection of non-magnetic material subsequently producing a very clean ferromagnetic concentrate. In many cases, the drum speed may be adjusted to either reject or collect locked particles containingboth distinct magnetic and non-magnetic components. Different magnetic elements are available based on the size of the feed material. Fine particles are best treated with a magnetic element employing many small magnetic poles to provide a ‘‘shallo~ magnetic field (for the relatively shallow burden depth) and maximizing agitation. A high level of agitation is required to release the physically entrapped fine non-magnetics. Conversely, coarse particles are best treated with a magnetic element employing fewer large magnetic poles to provide a “deep” magnetic field (for the relatively deep burden depth) and minimizing agitation. Physical entrapment of coarse particles is very limited.

Dry -Rare Earth Drum Magnetic Separator (High-Intensity) Rare earth permanent magnets provide an order of magnitude higher energy product as compared to conventional ferrite magnets. This has allowed for the design of high-intensity magnetic circuits operating energy fiee and surpassingthe strength and effectiveness of electromagnets. Employing a rare earth magnetic element in a drum type separator simply umverts the magnetic field strength fiom lowintensityto high-intensity. The rare earth drum separator typically uses a “salient pole” magnetic element design. This element also consists of a series of axial poles configured with alternating polarity. Steel wedges or interpolesare place between each magnetic pole. The steel interpoles concentrate the magnetic flux producing a very high magnetic gradient at the drum surfice. The magnetic field strength on the drum surhce is approximately 7000 gauss utilizing the rare earth magnetic element. Figure 8 depicts a typical gauss chart for a rare earth drum magnetic separator providing magnetic field strength at distance fiom the drum shce. The primary variables affecting separation efficiency are the same as described for the low-intensity magnetic drum separator. The rare earth magnets provide an increase in the magnetic strength. The increase in magnetic field strength extends the range of applications. While the low-intensity magnetic

185

drum is limited to the collection of ferromagnetic materials, the rare earth drum effectively collects many paramagnetic minerals. Typical rare earth drum applications are presented in Table 6. Figure 8. Typical Gauss Chart for a Rare Earth Drum Depicting Magnetic Field Strength at Distance fiom the Drum Surface

Table 6. Typical Rare Earth Drum Applications ~~

M66ral t lmenite Ilmenite Hematite Sulfides Specular Hematite Nickel Slag Silica Sand

~~

~

~

~

ProcessApplication Recovery of ilmenite from heavy mineral concentrates Separation of “high-iron” ilmenite from “low-iron” ilmenite Concentration and recovery of hematite - iron ore Recovery of pyrrhotite Concentration of specular hematite from gravity concentrates Recovery of residual metallic nickel Removina ferromannetic and paramametic constituents

Test work was conducted on an ilmenite sample to illustrate the primary variables affecting separation efficiency. The titanium-bearing ore was obtained &om an operating heavy mineral sands plant. The sample was an electrostatic conductor product that reports to multi-stage induced roll magnetic separators. Analysis indicated that this material was -60 +I40 mesh containing 56 percent Ti@ and 7 percent Al2O3. Separation test work was conducted on the salient pole rare earth drum in an attempt to produce a h a 1 ilmenite concentrate containing less than 1 percent A 1 2 0 3 while maximkhg the Ti@ recovery. Test work was conducted on both a 40-cm diameter and 60-un diameter rare earth drum. Both drum speed and feed rate were evaluated. To normalize the effect of drum speed and feed rate, tests were conducted at a constant burden depth. In each case the feed rate was increased proportionately with the drum speed to maintain a constant burden depth. This approach allows the true effect of an increme in drum speed to be evaluated. Without increasing feed rate, the burden depth on the drum decreases with an increase in drum speed. As a result, a change in paforormance may be the result of the drum speed or the shallow burden depth on the drum. The constant burden depth approach eliminatesthis problem.

186

Next Page

A series of tests were conducted to determinethe effect of feed rate on separation efficiency. These results are shown in Figure 9 as Ti& recovery as a function of feed rate (TPWmeter of drum width) and were conducted at a constant burden depth. The 40-cm diameter drum exp~encesa drop off in performance very quickly as the as fed rate is increased. The Ti@ recovery decreases from 70 percent at a feed rate of 1.5 TPWmeter of drum width to 20 percent at rates exceeding9 TPWmeter of drum width. The 60-cm diameter drum is able to maintain a Ti@ recovery of 70 percent up to feed rates of 10 TPWmeter of drum width. The improved performance of the 60-cmdiameter drum is a result of both a deeper magnetic field generated by the larger “flatter” magnetic element and the increased retention time in the magnetic field.

The effect of drum speed on the 60-cmdrum was evaluated by determining the product recovery at 1 percent A1203 for each test. These results are shown in Figure 10 as weight recovery as a h c t i o n of drum speed. As expected, the curve passes through an optimum for each burden depth. At low drum speeds, the gangue material does not have sufficient centrifugal force to be expelled fiom the magnetic product. As drum speed is increased,more gangue material is rejected, allowing the weight recovery at 1 percent A1203 to be increased. A further increase in drum speed resulted in a loss of magnetic material to the non-magnetic hction. This is due to the high centrifugal force exerted on all particles. Figure 9. Ti02 Recovery as a Function of Feed Rate

Figure 10. Weight Recovery as a Function of Drum speed

i 100% Burden

-I_IIy 50% Burden

0

2

0

4

6 8 1 0 1 2 1 4 Feed Rate TPWMeter of CkumWCdth

30

60

90

120

150

Drum Speed MeterslMinutes

The magnetic product weight recovery as a function of A 1 2 0 3 diluent in the magnetic product for several tests conducted on both the 40-cm and the 60diameter drums is shown in Figure 11. It is interestingto note that both drums operate on the same graderecovery curve. This finding is not unusual for magnetic separators with similar circuit designs. The larger magnetic element in the 60-an diameter drum projects the magnetic field further fiom the drum surface. As a result, the 60-cm diameter drum can operate at higher capacity while maintaining the same metallurgical performance. A technique that is often used when analyzing the separation efficiency of paramagnetic minerals on a production-scale, drum-type separator is termed magnetic fiactionation. This technique is commonly applied when investigating the beneficiation of ilmenite or hematite. A multiple compartment splitter is utilized with the magnetic drum separator to provide a series of products based on magnetic susceptibility.

187

Bench-Scale and Pilot Plant Tests for Thickening and Clarification Circuit Design Benjamin K. Pocock’, Cory B. Smith’, Glenn D. Welch’

ABSTRACT Reliable design of thickening and clarification circuits has been complicated by the desire to minimize capital costs through the use of flocculants and high rate sedimentation systems. These newer and lower capital circuits require a greater degree of control and operator attention to provide for successful operation. As a result, circuit design testing strategy requires careful evaluation to improve the certainty of meeting design performance goals. Bench-scale thickening and clarification tests have been used successfully for design for nearly a century when representative process samples have been evaluated using reliable test methods. Since ore mineralogy, solids particle size and shape, feed concentration, water chemistry, oxidation state, and sample aging all impact testing, an understanding of operating variationsand conditionsshould be an integral part of the design-testing program. Although expensive, Pilot Testing is advised when a new ore type is to be processed, new process conditions are being considered, or improved design reliability is sought. Careful flocculation scale-up is particularly important to obtain reliable design. INTRODUCTION Solidliquid separation design is often a key economic factor for the successful execution of metallurgical projects. Poor design or inattention to design details can result in plant capacity bottlenecks or even plant operation failure. Risk increases from replacement of existing equipment in known operations to new processes, therefore, the solidliquid separation test and design regimen should correspond to the risk. This paper will deal with test methodology to select the proper sedimentation equipment to meet the design and economic objectives of the engineer. Initially a project is defined and a preliminary flowsheet is developed that provides a process design to meet the desired objectives. Process constraints such as water usage limitations, land considerations, environmental impacts, tailings disposal, liquid waste disposal, etc. are identified and become an integral part of the design process. Solidliquid separation steps are identified within the flowsheet along with objectives for each of the separations to be made. The physical nature of the solidlliquid mixture and process objectives dictates the equipment options to be evaluated. TEST PROGRAM STRATEGY Risk analysis is a key component in assessing test strategy for a given sedimentation application. Bench scale testing can be reliably used when the test sample is well defined or the ranges of slurry compositions to be processed are well characterized and are available for testing. Design risk is generally low when uniform particle size ranges, solids composition, and water (fluid) chemistry do not substantiallyvary. Risk increases when feed variation escalatesor the properties of the feed stream are not well defined. Testing requirements to lower risk increase as the sample or process definition decreases or the knowledge of the process is limited. Particular attention to test strategy is recommended when flocculation is utilized, high rate processing units are considered, and counter-current decantation circuits are to be employed.

1 Pocock Industrial, Inc., Salt Lake City, Utah.

20 1

Pilot plants can lower risk, but also can lead the design engineer astray. Often the thickener or clarification units utilized in the pilot plant are mismatched for the process throughput or the pilot plant is not specifically operated to obtain reliable sedimentation circuit design criteria. In any case, it is advisable to conduct off-line bench scale or continuous tests on a slipstream to provide reliable scale-up results. Process scale-up issues are also very important in assessing sedimentation circuit design risk. Changes of solids particle size and slurry characteristics that may occur during process scale-up can have substantial impact on unit capacity or counter-current decantation circuit efficiency. Undersized units will bottleneck the plant and lead to low capacity, off-specification product, or inefficient operation.

TEST METHODOLOGY The first step in the sedimentation circuit and equipment design process is to fully characterize the solid/liquid mixture to be separated. Key characteristics are size distribution of the solids, particle shape of the solids, specific gravity of the solids and liquid, concentration and viscosity of the slurry, pH, oxidation state, temperature, and chemical composition of the liquid. Since equipment-sizing methods are sample specific, substantial effort should be made to produce as representative a test sample as possible. Sample aging can be particularly troublesome and lead to erroneous results. With knowledge of the solidliquid mixture to be separated, process objectives, and process constraints, preliminary evaluation of equipment options can begin. Particle size of the solids, particularly the minus 37 micron fraction, and solids concentration has the biggest impact on selection of equipment and correspondingtest method selection. Solids particle size can be enhanced by flocculation, which is often utilized to reduce sedimentation unit size or to improve overflow liquor clarity. Sedimentationoperationscan sometimes be improved by changing upstream process conditions. Flocculant Selection Sedimentation circuit design is increasingly moving towards the use of flocculants to reduce capital cost and to provide lower solids content overflows. The range of flocculants and their specificity are dramatically increasing and are providing more options to the process engineer. Flocculant testing or screening is generally conducted on dilute slurries, i.e. 5% solids or lower using flocculant at a 0.05 to 0.1 gpl concentration. Flocculants of differing composition, molecular weight, and charge are evaluated in a sequential manner. Qualitative comparisons are made that evaluate floc structure, settling rate, and overflow clarity as a function of flocculant dosage. Comparison should include cost/performance criteria when evaluating results. Flocculant vendors, solids/liquid separation equipment companies, or reliable solidsAiquid separation laboratories can provide this service.

SEDIMENTATION EQUIPMENT OPTIONS Thickeners and clarifiers are the workhorses for separation of bulk liquids from solids contained in slurries produced in the metallurgical industry. They operate on a simple gravitationalprincipal and will perform as long as the solids have a higher density than the liquid. Thickener capacity is normally controlled by solids tonnage to the unit and thickeners are usually operated to concentrate the solids to a heavy pumpable underflow consistency that may be sent directly to a tailings pond, in the case of waste material, or to filtration for hrther dewatering, if the underflow is product. Clarifiers generally operate on low solids containing feeds and are used to reduce overflow liquor solids concentration to the 5 to 20 ppm range. Design of these units is usually hydraulically controlled. Thickeners and clarifiers are available in many design variations that attempt to improve performance for a given separation. Testing methods are specific for each general class ofequipment. The following list indicates the main characteristicsof each sedimentation type.

202

Thickeners 0 Conventional. Ease of operation; high underflow densities; large size up to 185 meter diameter, high tonnage throughput. 0 High rate. Two to eight times settling capacity when compared to conventional thickeners; efficient flocculation required; generally better overflow quality than conventional; tighter control requirement than conventional; size generally limited to 35-40 meter diameter. 0 Ultra high rate. No rake mechanism; up to four times settling capacity of high rate mechanically raked thickeners; internal feed dilution results in lower flocculant consumption; internal angle plates aid solids dewatering; 60 degree bottom conical section provides thickened underflow; size generally limited to 12-15 meter diameter. 0 High density. Much like a conventional thickener, but with higher side wall depth and increased bottom slopes to extend solids retention time; modified rake and drive design; high torque requirement limits diameter to 40-50 meters; may increase underflow density by 8-1 2 percentage points; underflow rheology may be solids concentration limiting, depending on the type of underflow pump. 0 Deep cone. Similar in design to ultra high rate units but with specialized rake design; high aspect ratios, 15 meter diameter with 15 meter high side shell plus 60 degree bottom cone; torque requirement limits diameter; may increase underflow density by 10-15 percentage points; requires energy input to exceed yield stress of pulp at withdrawal point; positive displacement pumps may be required. Clarifiers 0 Conventional. For dilute slurries without particle size enhancement; sizing in the range of 5-25 m3/m2perday; flocculation may or may not be employed; subject to upset with feed or temperature change. 0 Reactor clarifier. Internal recycle with draft tube arrangement; promotes improved settling properties of solids, flocculation required, good overflow quality; accommodates metallurgical reactions and non-scaling precipitations. 0 Inclined plate. Compact units for small flows; fast settling solids, limited underflow retention time; usually no rake mechanism, limited to non-sticking solids; construction materials can often be a problem. 0 Hopper clarifier. A rougher clarifier used ahead of polishing filters; flocculation required; fluidized solids bed operation; control difficult; no rake mechanism. THICKENER DESIGN TESTING Test methodology should provide the following basic thickener design information: 0 Unit area in m2/MT solids per day as a function of feed rate, feed solids concentration, flocculant dosage, and underflow solids concentration. 0 Underflow solids concentration as a function of solids residence time. 0 Rheological characteristics of the thickened solids verses thickened solids concentration, 0 Impact of flocculation on design characteristics of the unit. 0 Impact of feed characteristics variation, such as: particle size, temperature, and solution characteristics (pH, Eh, composition, concentration). 0 Corrosion characteristics of the slurry to be thickened including the surface air interface and mist condensation on above surface structures. Bench Scale Testing Static tests. There is a general commonalty of test methods, procedures, and data analysis. However, most companies that market solids/liquid separation equipment and/or consultants have a proprietary interpretation of design methodology. Many of the methods are variations of the Kynch analysis. it is not the purpose of this paper to debate the published methods (refer to references 1 to 7 at end of paper) or to single out any method as the “best”, but to indicate what essential information should be obtained in the testing regimen and the basics of data interpretation.

203

The majority of static thickening tests are conducted in one or two liter cylinders containing some sort of raking mechanism. Representative feed slurry is introduced into the cylinder and settling velocity data of the solids/liquid interface is taken. Tests are often conducted over an extended settling period, i.e. 24 hours. A sufficient number of tests are performed to determine the impact of feed solids concentration, flocculant dosage, and process variables on unit area, underflow density, overflow clarity, and underflow solids rheology. Tests that utilize flocculation require attention to detail as minor variations in mixing energy and test protocol can strongly influence results. The most common methods used involve vertical mixing movement of some sort of rubber stopper on a rod or cylinder inversion. There is a strong “art” component in thickening testing and substantial care should go into laboratory technique as well as in interpreting results. Determining unit area from the test (cylinder) data can be done using methods reported in the literature. Most methods require some sort of semiempirical depth adjustment factor andor safety factor to arrive at a design unit area. The use of highly experienced and reliable personnel is preferable since work consistency is the most important aspect. Dynamic tests. As noted earlier, the use of flocculation can substantially improve overflow liquor clarity and reduce size of the sedimentation unit employed, thus cutting capital costs. However, there is a premium on control and operator attention that must be considered. This is particularly the case for so called high-rate units. Although high rate unit design information can be inferred from static tests, a more reliable approach is to use small-scale continuous tests. Test equipment utilizes columns that are 8 cm or larger in diameter and are set up with feed well, overflow collar, raking mechanisms, and flocculation system. Feed, flocculant, and underflow pumps complete the arrangement. The feed-well arrangement can be set up for above and below settled solids interface level feeding. Testing regimen considers the impact of solids concentration, flocculant dosage, and flocculation method on unit capacity as a function of feed loading rate. Overflow is collected to determine clarity for (dynamic) conditions being tested. Underflow residence time may not be sufficient to obtain underflow solids concentration data verses time. To obtain additional data, static extension tests are run in raked, two-liter cylinders on underflow collected from the dynamic unit. Data Analysis Considerations Translation of bench scale data to design criteria should be concerned with providing a design that is both economic and operable. A special caution is offered when designing high capacity units. Too tight a design reduces operating degrees of freedom and limits feed variation flexibility. Unit area designs below 0.1 m2/MT solids per day should be carefully evaluated. Flocculation scale-up should also be carefully evaluated. It is sometimes difficult to duplicate laboratory flocculation conditions in a full-scale plant. This is particularly so in providing good flocculant-solids contact without excessive shear. It is important to ascertain the impact of process variables on flocculation quality during testing and to select flocculant(s) that exhibits the best flocculant performance and strength properties for the application. Perhaps the design parameter least considered and understood is the impact of slurry rheology. Process performance parameters, rake mechanism design, critical solids concentrations, underflow solids pumpability, pipeline design, and mixing design are some of the factors to be considered. A thorough understanding of slurry rheology is particularly important when considering operating in a high underflow solids concentration range or in paste thickening. Flocculation affects slurry rheology and its impact should be evaluated. In difficult cases, slurry rheology modifiers can sometimes be employed. Counter-current decantation circuits (CCD) are common in the minerals industry for washing valuable solute from leach slurry or to remove undesirable solute from the solids. Circuits can employ several stages depending upon efficiency required and water balance considerations. Since thickening properties will be affected by solute concentration, temperature, pH, and Eh changes, it is important to run tests using the conditions predicted for each stage. Reliable data for inter-stage mixing efficiency and stage underflow density is needed to predict recovery for various wash ratios. When flocculants are used, careful testing on stage requirements is particularly important. Stage over-flocculation can lead to poor design and inoperable conditions.

204

Pilot Scale Testing Pilot scale testing can substantially reduce design risk and provide more reliable information than bench scale testing when properly done. Often thickeners are an after-thought in pilot plant design and the units employed are poorly designed for the application and/or are oversized for the capacity ofthe system. The result is that little reliable thickener design information is developed. The following strategies will provide reliable design information from an operating pilot plant: Operate the pilot plant where the focus is to provide thickener design information. Maximum thickener capacity that meets overflow and underflow criteria should be determined. When operating the entire pilot plant, take periodic samples ofthickener feed and run off-line static or small-scale dynamic thickening tests. Run a smaller scale continuous thickener on a slip-stream from the operating pilot plant circuit. If flocculants are used, obtain information, such as mixing intensity and mixing residence time data, to provide scale-up information. Make sure that thickening data is taken for each process change made during the pilot plant campaign. The pilot plant is an ideal place to test instrumentation and control strategies for the thickener, particularly if high rate or other tighter control requirement systems are being considered. Larger scale slurry rheology tests (loop tests or pipeline Ap) can be run on thickener underflow to assess pumping and pipeline design criteria and limits. Pilot thickener design should emulate the cross section of an operatingthickener such that the pulp depth can operate at the 1 to 2 meter level. For example, a 0.3 meter diameter thickener would have an overall height in the 2.5 to 3 meter range. Larger units would just expand in diameter to increase capacity. Feed well design should also be carefully done to emulate depth and residence time of an operating thickener. Rakes can be top or bottom driven but should have vertical rods or pickets that extend through the pulp. If possible, it is highly instructive to put the pilot thickener on load cells. This enables the operator to determine solids inventory at any time and to monitor changes in pulp density for a given pulp height. CLARIFIER DESIGN TESTING Clarifiers are employed when the feed solids concentration is generally below the 5% solids range and the solids are in the non-hindered or “free” settling mode. Sizing of the unit is controlled by overflow rate, expressed as cubic meters/meter2per time, to achieve a target overflow solids concentration. The following information is required for correct sizing: 0 0

0 0 0

0

Bulk settling rate of the solids. Overflow solids concentration verses residence time. Underflow solids concentration verses residence time. Rheological characteristics of the thickened solids verses thickened solids concentration. Temperature variations, solution characteristics (pH, Eh, solute concentration and composition), and corrosion behavior. Impact of flocculation on the design characteristics of the unit.

Bench Scale Testing Static tests. As discussed in the thickener testing section, representative samples are required to obtain reliable test results. When coagulants or flocculantsor coagulantlflocculantcombinationsare to be evaluated, a screening program is conducted to determine the most effective reagent(s). The general approach is to conduct “jar” tests or to utilize individual cylinder tests to determine bulk settling rate of the solids and overflow clarity verses time.

205

To obtain the bulk-settling rate, feed slurry is contacted with flocculant in a cylinder orjar using a consistent mixing method, such as a rod-rubber stopper vertical mixing, cylinder inversion, or paddle mixers in the case of jar tests. The bulk rate is the settling rate (i.e. metershour) of the bulk of flocculated solids. As contrasted to the definite line seen in thickening tests, the solids-liquid interface area is usually quite difhse and several repeat tests may be required to establish a reliable rate. Determination of overflow solids concentration verses time is generally done as part of the bulksettling test. Samples of overflow are withdrawn from the cylinder(s) or jar(s) at prescribed time intervals and residual solids concentration is obtained. Underflow solids verses residence time test usually require production of enough solids so that a thickening test can be performed in a raked one or two liter cylinder. Dynamic tests. Bench scale continuous tests provide more reliable results particularly when solids recycle or reactor clarifier modes are being considered. The sample requirement is greater (200 liter drum size verses 20 liter size) than for static testing. The equipment is similar to the thickening dynamic test unit with the exception of the addition of a solids recycle pump. In the reactor clarifier mode, feed enters near the bottom of the unit and underflow is withdrawn through a dip tube in the sludge bed. Data analysis considerations. Unit area sizing of a clarifier is normally based on 50% of the measured bulk-settling rate. The next consideration is the residence time required to obtain the desired overflow solids concentration. The overflow solids verses time data can usually be plotted as a straight line log-log relationship and the residence time needed to achieve a desired overflow solids concentration ascertained. Scale-up of residence time information is difficult as translation from bench scale to commercial scale operation is affected by a number of factors that relate to flocculation and hydraulic factors within the operating unit. Usually an empirical “efficiency” factor is used in scale-up that ranges from 0.2 to 0.5 depending upon the application. Normally at high diameter to depth ratios (8 to 12) the 0.2 factor range is used and at low ratios (1 to 3) the 0.5 factor is used. Pilot Scale Testing In addition to the comments made in the thickener section, it is obviously important to use pilot test equipment that emulates the clarifier design to be evaluated. Rental units or vendor-supplied equipment that can be accurately scaled-up to commercial size can also be used for on-line or off-line testing.

REFERENCES Coe, H.S., and Clevenger, G. H. 1916 A Method for Determining the Capacities of Slime-Thickening Tanks. Trans. AIME. 55: 356-385. Kynch, G. J. 1952. A Theory of Sedimentation. Transactions of the Faraday Society 48: 166-175. Talmage, W. P., and Fitch, E. B. 1955. Determining Thickener Unit Areas. Industrial and Engineering Chemistry. 47:3 8. Wilhelm, J. H., and Naide, Y. 1981. Sizing and Operating Continuous Thickeners. Mining Engineering. 12: 17lo. Weiss, N. L. ed., 1985. SME Minerals Processing Handbook. Volume 1 Section 9. Perry, R.N., and Green, D.W., eds.1984. Chemical Engineering Handbook. 6ThEdition. New York McGraw Hill. Schweitzer, P.A. ed. 1997. Handbook of Separation Techniquesfor Chemical Engineers, 3“/Edition. New York: McGraw Hill.

206

Bench-Scale and Pilot Mant Tests For Filtration Circuit Design Tim Kmm'

ABSTRACT C m n t filtration technologiesoffervarying options for the modern mineral processing operations. Identificationof product charaMeristicsenables the approPriate filtration technologies to be tested. Detennination of individual product filtration requirements is the basis for setting testing parameters. Bench and pilot-scale testing can give results that can be used to specify full sized filters. The various filtration technologies have individual characteristics that match their respective testing methods with a varying degree of axuracy; proper use of the generated test data requires detailed knowledge of the applied filtration technology. Successful testing requires open constructive interaction between the mineral processor and the equipment vendor.

INTRODUCTION The trend in the mineral processing industry is to produce larger product volumes with less overall resource input. Proper filtration technology choice and sizing is important. The volumes processed today magnify the importance. The circuit design s h d d consider the solid-liquid separation hierarchies: classifiers,thickeners, sepamtors, filters and thermal dryers. The best filter technology should be balance capital costs, operation costs, filtration quality, capacity, size and environmental requirements. Economics favor a filter technology that can dewater the material to an acceptable level without the use of thermal dryers (Mapes, 1994). Once the best solid-liquid separation process has been identified,bench-& and pilot plants test can then be used to verify the technology choice and provide p p c r sizing. This paper will review the advantages and disadvantages of bench and pilot-scale tests, data that you will need before you begin any tests and suggestionsas to who can perform the tests. Filter Operationswbae the product is either solid or liquid are equally treated.This paper Will not cover filter type selection.

B

m AND LIMITATIONS OF BENCH- AND PILOTSCALE FILTER TESTS

As with any machine the best way to test is with the most representative sample and device. Benefits and limitations of bench-scale and pilot filter tests are listed below in Table 1.

Beneb-SerkTds Fairly accurate scalable sizing possible.

PPdphntTests Very accurate scalable sizing Requires only a small amount of material. Requires substantial amount slurry to test. Batch sample operation Continuous inline operation Inability to show long-tern effects of slurry May show medium-tern effects of slurry on the filter media. on film media.

I

Outokumpu Technology Inc, Fmglewood, Colorado

207

KNOWING THE PROCESS Knowing what is going into a filter and what is neccswy coming out of the filter is of utmost importance. The input variables (solid concentration particle size distribution, etc) and filter product characteristics (residual cake moisture, filtrate quality, etc.) will directly relate to the find word - cost of filtration per unit volume. Table 2 lists the process characteristics that must be known before proper sizing can be completed. In the case of producing dry solids the limiting factor can come from different areas. There might be a need for a particular moisture at a down stream process or minimum moisture for lower transport costs. When transporting by ship the moisture must be kept at a level where the materid acts as a solid and not a fluid mas - a dangerous situation occurs when a ships load shifts. With f i l m products the amount of allowable impurities knowledge is paramount. Knowing what chemicals can be added to aid filtration. KNOWING THE MATERIAL Product knowledge is key to choosing the correct filtration technology and determining the proper sizing. Filter vendors have a varying degree of experience with similar products. During benchscale and pilot plant test the material data will be used to set initial parameters, choose filter media, determine mineralogical compatibility and chemical filtration modifiers or filter-aids. Tables 3 and 4 go over the filtration product information that needs to be gathered. Not all of the information will be practically available.

SAFETYJSSUES Filtered material's volatility (solids and liquids) ought to be h-.P d c d a r care should be taken when utilizing vacuum systems. Filtrate inside a vacuum system can readily form vapors. Know what reactions are possible with any material that is added to the filter system. Toxic materials can necessitate the filter being placed inside a containment area. Flammable materials can be filmed with in inert environments (e.g. nitrogen purged containment). Products should be isolated from substances that they could react with. Proper safety devices (e.g. hoods, gas detection devices, inert atmospheres) should be incorporated into the filter design when necessary.

PERFORMJNGTHETESTS Testing can be done on site, in the lab or at other facilities. If testing is performed off site the difference in environmental factors, such as altitude and ambient temperature, and sample aging should be taken into account. Testing for one type of filter should not be used to size another type of filter.

-

Doing it yourself While not recommend, it can be done. Some manufactures can provide lab or pilot test equipment that can utilize to test a filter technology. Others can provide you with instruction on how to use common laboratory equipment to simulate filtration types. Typically the test kit will be sold for a fee and the testing instructions should be provided at no cost. 0 Research Facilities some research facilities and testing labs can do filter testing. Care should be taken to make sure that the testing procedures are up to date and will apply directly to the filter type. Confirm procedures with filter manuf-. This testing will be done for a fee. Ask what relationship, if any, the facility has with filter manufactures. 0 Filter Manufactures Manufacturers are experts on their own machine. They are the holders of common and proprietary knowledge on their filters. The manufacture can provide equipment lab or pilot testing, typically for a fee. Some manufactures are able to do system dewatering testing - encompassing other methods other than filtration. Depending on the testing conditions filter manufacturers can give process guaranties.

-

-

208

T.Mt2Site Conditions

Plant-

Cost of Drying Transport Moisture Limit Cost of Utilities

.. Knowledge of site conditions is necessary to properly size the filter. Ambient tempemtme can affect filtration rates and Vacuum filuation decreases with an increase in altitude. when the& drying is to be performed the per unit cost of thermal drying is necessary to evaluate capacity vs. machine size calculations TML should be known before testing The per unit cost for power, blowing air, instrument air, maintenance operation are necessary to determine capacity vs. machine capital cost calculations

T a b 3 sdid Material R o P t I ' k MineralogicalComposition ' h e solids mineralogy and chemisay can give hints how the material will act during times when the system changes: pH modification, solvent introduction. tempemure and pressure. Care Solids Concentration PdcleSizeDismbution

Slurry Temperature Solids Density Solubility

Abrasive Character

Surface Roperties Table 4 Filtrate LiQllid Chemical Composition Temperature

pWionic strength

Viscosity

Surface tension

should be taken to ensure that the product does not age before it is tested. Determine the range of solids concentration that will be reporting to the filter. Each filter technology works efficiently within a limited particle size range. Rocess temperamre variations can greatly affect filtration rates. Solids density is very importaat in detennining the specific filtration capacity. Filtering slurries that are at or near Sanuation can lead to precipitation within the filter media and filtrate discharge system. Solubility is also important for cake washing considerations. (Wake& and Tarleton) It is important to know for p r o p materials of construction. Highly abrasive materials can quickly wear through the improperly chosen material. Surface charge affects the state of particle dispersion and cake formation. Surface charge is PHdependent.

Liquid chemical composition affects separability of the particles. (Wakeman and Tarleton) Flocculent and dispersants can have strong effects on filter operation. Affects surface tension, viscosity, reactivity of chemicals in solution, and has influence on the materials of construction for the filter. (Wakeman and Tarleton) Plays an important role in determining comsive nature of liquid, which has influence on the materials of construction for the filter. (Wakeman and Tarleton) Give knowledge about elecmcal charge on the particle surfaces. Controls the rate of filtration. Higher filtration rate are obtained at lower viscosities. V i t y lowers with a rise in temperature. Surface tension in combination with the particlefliquid contact angle, conmls the moisture content of dewatered filter cake. (Wakeman and Tarleton)

209

-

Testing companies There are a number of companies that will do testing for a fee. Make sure that the test results will apply to all types of fltecs that are being examined. Confirm testing pn>cedures with the filter manufacturers. Some companies can do more dewatering testing other than just filtration. 0

WHERETOFINDA FILTERVENDOR It is not always the easies thing to find the proper filter manufacturers. Ones that have the necessary mineral process experience. Table 5 briefly lists several sources that can be used to find filter manufacturers.

Table 5 Findiag a Bilttr nmmbchmr Internet

SME American Filtration Society Other Recess Operators

Using popular search engines to find filter references. Visit professional society pages. Examine filter manufactures sites. Search www.amazon.com for reference material on filtration. The equipment w l d o n portions tbe SME Mineral Processing Handbook. AFS provides a yearly guide for filter manufactures. The guide is not complete and has many manufactures that will not possess experience in mining areas. Find out what the current technology is and who is the purveyor. Otherwise known as networking.

CONCLUSIONS The performance of bench-scale and pilot tests will continue to be the method of confuming mineral processing operations. Knowledge of process is key to Setting test parameters and choosing the test method. Roper preparation will help ensure drat testing will produce viable results. ACKNOWLEDGMENTS I am continually grateful to Mr.Simo Ruonamaa, Application and Process Manager, Outokumpu Technology GmbH, and Mr.Kalle pukki, Customer Service Manager. Outokumpu Mintec Oy, for their continuing support and assistance with this endeavor and all my other hare-brained schemes. REFERENCES C.B. Mapes, Development and Application of Ceramic Media Disk Filter Technology at Phelps Dodge Morenci, Inc., SME National Convention, February 1994, Albuquerque, New Mexico, P 1. R.J. Wakeman, E.S.Tarleton, Filtration: Equipment Selection, Modelling and Process Simulation, 1998, Elsevier Science Ltd, pp 341-342.

210

Gold Roasting, Autoclaving or Bio-Oxidation Process Selection Based on Bench-Scale and Pilot Plant Test Work and Costs Jacques McMuIIen’ and Kenneth G. Thomas2

ABSTRACT Most refractory gold ores require a pre-oxidation step such as roasting, pressure oxidation, or biooxidation to enhance gold recovery with conventional cyanidation. To select the most appropriate pretreatment process and to understand process chemistry, the use of bench and pilot scale test work is required. Once in operation, these laboratory techniques can also assist in selecting the optimum operating parameters for the large-scale plant. While laboratory techniques will facilitate process understanding, in order to ensure that all objectives are met, process selection must be performed recognizing other constraints of the project. An overview of generic key criteria will be discussed. Test limitations and shortfalls will be highlighted. Ore types and project specifics must also be considered in the process selection.

INTRODUCTION Generally speaking, gold ores are classified as refractory when the gold recovery is less than 80% when conventional leaching techniques are applied. These refractory ores require a pretreatment stage prior to cyanidation to enhance gold recovery. Selection of the pretreatment stage(s)‘’*2’can also influence the design of the rest of the processing flowsheet. For example, whole ore roasting would most likely require the selection of a dry grinding circuit. The selection of the optimum pretreatment process is therefore on the critical path of the overall process flowsheet definition. The objective of this publication is to present the know-how associated with the development and interpretation of the experimental program required to elect a pre-oxidation process flowsheet. To properly assess test results, a thorough understanding of the process chemistry is also required, Additionally, ore reserve characterization, sample selection, and process costs are discussed in the context of refractory gold ores. To illustrate examples, Barrick’s experience at Goldstrike (where both pressure oxidation and roasting are used to pretreat ore prior to cyanidation) is presented. Process selection for gold refractory ores is achieved through specialized laboratory test work using various ore characterization and testing techniques, which vary from bench top experiments to pilot plant testing. The use of a reputable laboratory with significant experience, track record and database is a fundamental requirement. The same laboratory techniques will also be required once the plant is in operation to optimize the operation and develop the life of mine metal plan. This paper will focus on three pre-oxidation methods: roasting, pressure oxidation, and biooxidation. Not included within the scope of this paper are other pretreatment methods‘3*4, such as chlorinati~n,’~*~) and ultra fine grinding(’). BACKGROUND Refractory gold ores can be classified as single or double refractory. The first refractory component refers to the nature of the gold occurrence being either gold in solid solution in the sulfide crystal lattice or encapsulated within sulfides, typically pyrite.‘*’ Pyrite morphology varies,

’ Barrick Gold Corporation, Toronto, Ontario, Canada Hatch and Associates, Mississauga, Ontario, Canada

21 1

but the pyritic gold carrier is often enriched in arsenic in refractory gold ores. The second refractory component refers to the preg-robbing nature of the ore, which is mostly a function of the total carbonaceous matter (TCM) or organic carbon content of the ores. Roasting,(” pressure oxidation“” and bio-oxidation are three recognized pretreatment methods for gold refractory ores. All of these pretreatment alternatives can be considered for whole ores. However, if it is possible to produce a flotation concentrate with good recovery, significant capital savings can result. As a result, flotation screening tests should be performed in the early stages of process definition. As part of the investigation, the characteristics and mode of gold occurrence in the flotation tailing should be evaluated. For example, the preg-robbing material may report to the concentrate‘*’and as a result, some of the gold in the flotation tailing could be processed through a straight cyanidation approach. Carbonates may be rejected during flotation, which could also negate the pre-acidulation step if recombined with the autoclave slurry discharge or the biooxidation slurry discharge. Increasing the sulfide content above the auto-thermal limit would negate the requirement for pre-heating autoclave feed. Pretreatment processes are generally chemically and physically aggressive. In addition to being more mechanically demanding on the downstream equipment, selection of specific materials of construction is often required; selection must also consider chemical compounds generated by the process. These factors are of equal importance to gold recovery, as they can severely impact the ability to operate the plant. Ensuring representative sampling of the orebody is critical. Due to the complexity of preoxidation processes and environmental considerations, such sampling is essential during the early stages of a project. The unknown occurrence of elemental mercury in an ore could result in air quality permit non-compliance for a roaster’s dry grinding circuit. Minor element deportment must be viewed as elements of risk when considered from an environmental perspective.(”) Most of the time, the process development tests will be performed in specialized third party laboratories with extensive experience with such procedures. Each pilot plant is different, and the procedures must be thoroughly detailed for each test. Feedback from the technicians must be carefully reviewed to ensure that procedures have not been modified without justifiable reasons, to avoid proceeding with unrecognized changes. Depending on the scope of the program, seconding an engineer to the research program should be considered and is strongly recommended.

ORE RESERVE SAMPLING A common statement during difficult plant startups is: “The ore has changed.” Process selection must incorporate the required flexibility and robustness to be able to process, within acceptable economic boundaries, the various ore types. Understanding the ore reserve is critical. The ore reserve is, by definition, the zone that can be mined economically. The geological reserve and the mining reserve should be reconciled prior to initiating the metallurgical test work. Ore variability is generally initially assessed from a geological standpoint. From a processing perspective, this reserve evaluation may be incomplete. Often, refractory gold ore reserves have more variability in relation to ore characteristics such as distribution and speciation of the sulfide minerals, carbonates and/or carbonaceous content. As well, refractory ore variability must also account for minor element occurrences such as antimony, arsenic, mercury etc. In that regard, process sensitivity due to unpredicted or uncontrolled side reactions could significantly affect process performance if not accounted for in the process development. If such conditions occur, they could have significant impact on future project viability in terms of post-commissioning additional capital, higher operating charges, lower metal recoveries, or environmental issues. Accordingly, the following ore variability issues must be considered: 0

Understand the ore deposit’s geology (including sequence and alteration events), mineralogy, and chemistry in terms of both process and environmental concerns. Include the effects of dilution.

212

Understand metal deportment (both payable metals and minor elements) and refractory nature as a function of mineralogy. Understand scale forming minerals, clay content and mineralogy. Address permitting issues before initiating the metallurgical tests; this may add to assay requirements during the program. Such issues can affect capital and operating costs. Understand drill core losses as they may hide unexpected mineralogy or bias the block model and the metallurgical test work. Evaluate the effect of core aging when the test work extends over a long-term period (for example: pyrrhotite oxidation). Evaluate future ore inventories, and evaluate the necessity of ore blending. Define blending criteria such as carbonate levels, sulfide levels, and grade, and integrate blend planning into the mine plan as early as possible. Perform ore hardness distribution per rock type and for the different geological settings, i.e. lithologies and alterations. Understanding metal deportment and mineralogy is important enough to be considered separately. Such characterization should include studies of the following: 0 0

0

0 0

Detailed analysis of payable metal(s) grade, mineralogical deportment, and variability. Thorough understanding of the ore’s refractory nature, be it encapsulation (either silica or sulfide), lattice substitution, or preg-robbing. Grade and speciation of critical elements such as sulfide sulfur, carbonates, total carbonaceous matters (TCM). Minor elements of interest and their mineral carriers i.e. Cu, Sb, As, Hg, Te, Se, Zn The effect of specific minerals such as orpiment or realgar (which can substantially degrade gold leaching kinetics)

Understanding these factors may result in a different processing scheme for various ore types.

As metallurgical performance is linked to the mineralogy, a solid understanding of the ore mineralogy can be used as a metallurgical performance predictor. Also, understanding an ore bodies’ variability may drive mine plans, as well as blending and stockpiling decisions. Conducting a literature and operations’ review for similar ores is often invaluable. The ore variability issue raises the following question: “Should metallurgical tests be performed on individual samples or composites?” It is our opinion that composites can only be used only after sufficient geo-metallurgical understanding of the ore is acquired. Bulk samples are necessary for the pilot plant stage but their selection is critical. When possible, samples should be selected after the ore reserve is understood sufficiently to generate a block model integrating all the key metallurgical components affecting plant performance. For example, under a roasting flowsheet where in-situ SO2fwation is desired, a high carbonate ore having a gold grade less than the cut-off could become ore. Such understanding could result in the modification of the mining sequence. The metal plan then becomes optimized on the basis of the net revenues per mining block, and the likelihood of unrealistically “boosting” the metal plan is reduced. With refractory ores, advanced mineralogical characterization techniques may be warranted on both the fresh ore and on key processed products. The use of microbeam techniques“2)can assist in understanding hndamental process mineralogy.(13*14* Is* Because interpretation of analytical results leads to process selection, the analytical techniques used to characterize ore and assess metallurgical performance should be carefully evaluated. The accuracy and reliability of these assays is often taken for granted. Analytical techniques essential to refractory gold ore processing include diagnostic leaching, Leco assays, and preg-robbing tests; these tests are discussed in the following sections.

213

Diagnostic Leach Test The gold industry extensively uses the diagnostic leach test“”2’ to infer gold metallurgy; these techniques complement advanced mineralogical investigations. Such techniques are welldoc~mented(~~*’~’ and should be used at the various stages of metallurgical investigations. Of interest in that area, General Mining Central Research Labs‘2o’has developed a very quick leach test for bio-oxidation amenability. The ore or concentrate is digested in nitric acid and then neutralized before cyanidation testing. If the gold can be liberated for leaching by nitric acid this indicates that the ore is likely amenable to bio-oxidation. Leco Assaying Leco assays for sulfide sulfur, total carbonaceous matters (TCM) and carbonates are critical to refractory ore processing and ore characterization. The Leco method essentially consists of burning sulfur and carbon to SO2 and CO2 respectively, and analyzing the off-gas for these species. Different methods can be used and must be thoroughly reviewed to ensure that they are precise and accurate. Three Leco assay suites used at Barrick’s Goldstrike operation are the short, long and long insoluble. (”) The Leco short method can be used for blast holes, exploration samples, mill feeds, roaster feeds and ores that have not been chemically treated. In this method, a sample of the ore is roasted at 650°C to bum off the sulfide sulfur leaving behind the sulfate sulfur. The roasted sample is then run on a Leco for total sulfur and carbon, along with an unroasted sample. The difference between the total sulfur of the as-received sample and the total sulfur of the roasted sample is the sulfide sulfur. (If there is finely encapsulated pyrite it may not be fully roasted, and will then report as sulfate sulfur in the Leco analysis on the roasted ore.) The reliability of the method, however, is sensitive to the carbonate content of the sample. At Goldstrike, the use of the Leco short method on autoclave discharge samples presented a high bias on the residual sulfide sulfur as some sulfates generated during autoclaving volatilized during the roast procedure. The Leco long method is normally used for autoclave discharge and CIL tails. In this method a portion of the sample is digested in sodium carbonate solution to remove sulfates. A portion of the digested sample is then run on the Leco. The total sulfur value from the Leco bum of the carbonate treated ore consists of the sulfide and insoluble sulfur. Barite does not dissolve in the carbonate solution and will appear as sulfide sulfur. If sulfide minerals, such as orpiment or realgar, are soluble in carbonate solution they would dissolve and the measured sulfide value would be lower than actual. The Leco long insoluble (sulfate) method is run on autoclaved ores when the presence of barite or aluminum sulfate are suspected. It corrects for the sulfur in the barite reporting as sulfide sulfur being insoluble during the sodium carbonate digestion. All samples submitted for either a short or long method Leco analysis are run as is for total carbon (carbonate + TCM) and total sulfur (sulfate sulfur + sulfide sulfur). All samples submitted for any Leco suite have a HCI digestion step followed by Leco bum to determine carbonate carbon by dissolving out the carbonates, i.e., calcite, dolomite and siderite, in the ore and then determining the total carbonaceous matter remaining (TCM). This acid leach step is important as it has been discovered that carbonates will fuc sulfur on the calcine during the roast which will bias low the sulfide sulfur. The HCI will also dissolve all sulfates except barite but should not dissolve sulfides. The total sulfur value measured by the Leco on the burn of the HCI digested material should then theoretically be the sulfide sulfur. Preg-Robbing Ore Tests The presence of preg-robbing is a major factor in directing ores to a pressure oxidation process or a roasting process. Techniques to detect preg-robbing can be simple or complex; the use of an electron microprobe to perform gold concentratiodanalysis of the carbonaceous grains, after having contacted them with an auro-cyanide solution, has been proposed.‘**’Preg-robbing behavior is not straightforward. Pressure oxidation could, for example, result in activating carbonaceous

214

matters. Graphitic carbon can be activated during roasting.(2)However, it should be noted that the carbon can also be destroyed by pressure oxidation or roasting in addition to processing with chlorine.(23’Accurately evaluating the ores’ preg-robbing content is therefore important. In most tests, the carbon activity or preg-robbing level assessment will be performed at normal teqxx&rss and pressure. As mentioned before, autoclaving can also result in increasing the activity level of the TCM’s and this possible effect should therefore be verified. Barrick uses a series of specific characterization techniques: the standard preg-robbing test, the spike CIL test, and the Bleach Leach test.

Standard Preg Test (SPRT). The standard preg-rob test measures the ability of an ore to pregrob cyanide soluble gold. A gold-bearing spike is added to the slurry of ore and cyanide solution. The ratio of gold remaining in solution determines the preg-robbing level. A ratio of 1 would indicate no preg-robbing, while 0 would mean a high preg-robber. This is a sensitive and conservative test, with two drawbacks: it does not account for the counteractive effect of the carbon in the CIL circuit, and it does not account for cyanide soluble gold present in the ore. Even refractory ore typically contains some cyanide soluble gold; such gold will continue to leach in the test and distort results. This explains why sometimes more gold than the added spike is recovered. Therefore, even if all of the gold in spike is recovered, it does not entirely rule out the pregrobbing potential of the ore. For Barrick Goldstrike ores, it has been determined that the pregrobbing is also a function of the gold spike concentration. The occurrence of cyanicides in the ore can also bias the test as secondary gold loss may result. Therefore, the standard preg-rob test should be considered as an indicative measure of preg-robbing. Hausen and B ~ c k m a n ‘described ~~’ it as semi-quantitative and relative, not quantitative and absolute. Spike CIL Test. Processing ore in a CIL circuit will be capable of counteracting some of the effects of certain preg-robbing ores. As long as the CIL carbon exhibits more activity for gold adsorption than naturally occurring carbon, the gold-CIL recovery process should yield acceptable results. Similarly, the spike CIL test is like the standard preg-rob test, except that carbon is added to the ore slurry. This allows the CIL carbon to compete with the naturally occurring ore TCM. As per the standard preg-rob test, any gold that will naturally leach will have the ability to be either preg-robbed or adsorbed onto the CIL carbon. A full suite of tests can be performed such as straight leach vs straight leach with carbon, CIL vs. standard preg-rob and CIL vs. spike CIL. The interpretation of these test results can assist in further characterizing the preg-robbing effect. The use of different carbon concentrations and /or carbon activities will change the results. Through this test, some ores which have been defined as high preg-robbers by the standard preg-rob procedure, will be reclassified as non-preg robbers. While the test better characterizes the ore pregrobbing ability, it is more time consuming to perform. Barrick Bleach Leach Test (BLORT- Bleach Leach Ore Routing Tool). Chlorination is a recognized full-scale technique used to de-activate the carbonaceous content of the ores. The bleach leach test utilizes that character is ti^.'^^' This test is used at Goldstrike as a rapid test to assist in ore routing at the full-scale mining operation level. The ore slurry is first contacted with a diluted hypochlorite solution. Depending on the chosen concentration of hypochlorite and the ore, TCM de-activation is achieved. Following that step, a standard preg-rob test is performed. Being ore specific, the hypochlorite concentration and contact time should be selected and calibrated to the results of the spike-CIL test. The hypochlorite will, however, also oxidize part of some sulfides, which could be gold carriers. The “free” gold portion content of the ore is therefore increased. Recovering well in excess of the spike is therefore not uncommon. At Goldstrike, ores presenting BLORT values in excess of 100, defined as 100% recovery of the spike, are considered autoclavable, ores below the 85 benchmark are routed to the roaster. For the ores presenting BLORT values between 100 and 8 5 , the standard preg-rob value is used. When the standard preg-rob value is greater than 40%, the ore will be routed to the roaster.

215

In summary, carehl consideration should be given to sampling and ore characterization. While studying composite samples instead of discrete samples can reduce both cost and length of the schedule during the flowsheet development phase, such an approach can mask the effects of certain ore types. While compositing or blending will occur in the regular mining operation due to the mining sequence, failure to understand the orebody can result in less than optimum plant performance in terms of metal recovery and operating costs. Common sense must be applied to determine how many samples have to be tested. This is a risk assessment exercise. As the ore reserve increases in complexity from a geological point of view, more samples will be required. However, it can be re-assessed when a rather comprehensive geological block model is available. Refractory deposits are expensive to develop; this initial step is equivalent to purchasing an insurance policy on the anticipated production.

PROCESS SELECTION BASED ON COSTS Process selection cannot be considered in isolation from the payable metal recovery perspective, and should aim at optimizing the net revenues inclusive of the capital expenditures, operating costs,(26)and recognizing project risks, such as process robustness, environmental compliance, etc. The financial elements discussed below can be used to fiuther assist in the preliminary analysis of process selection. As a general rule, selection of known and proven technologies will reduce risk. This section will discuss the costs of roasting, pressure oxidation and bio-oxidation. Biooxidation has been developed as a low-cost process for treating sulfidic concentrates or whole ore under a heap leach approach. It competes with autoclaving and roasting up to 2,000-3,000 tonnes per day whole ore or concentrate. Above this tonnage, the bio-oxidation slurry tanks become too large and expensive. However, bio-oxidation offers advantages in less developed countries where skilled personnel are scarce for operating autoclaves. Bio-oxidation can be coupled with autoclaving as a pre-oxidation step to improve autoclave utilization. The Ge~coat‘~’’ approach, a heap leach bio-oxidation approach, is an interesting emerging technology that may offer the potential to fiuther reduce operating and capital cost. Being ore dependent, bio-oxidation could yield comparable gold recovery to acidic autoclaving. Heap bio-oxidation would most likely yield less recovery. Gold recovery by alkaline autoclaving can be as much as 10% lower than acidic autoclaving. The reason for this relates to gold entrapment in the oxidation products and encapsulation of unoxidized sulfides. In acid autoclaving the product of oxidation, ferrous sulfate, is soluble in acid solution. It will diffuse away from the reacting pyrite surface, and precipitate as hematite. In alkaline autoclaving the hematite forms at the oxidizing surface, entrapping gold and thereby reducing gold recovery. For non-carbonaceous ores, higher gold recoveries can usually be achieved in the autoclave than from roaster calcines. The difference can be as much as 5%. Despite lower recovery, alkaline autoclaving should be considered wherever possible. Materials of constructionare more conventional, and plants have reduced capital and operating costs. Operating Cost for Roasting and Pressure-Oxidation A comparison between the operating costs for the Barrick Goldstrike roaster and acidic autoclave is presented in Table 1. These costs are all-inclusive for pretreatment and integrate costs from crushing to tailing disposal. Costs for cyanidation, strip circuit, ore handlingblending, cyanide detoxification, tailing disposal, or overhead are not included.

216

Table 1 Operating cost roaster versus acidic autoelaving for year 2001 Won Roaster Autoclave Manpower 2.20 3.15 Maintenance 2.40 4.20 Reagents 2.00 3.20 Power 4.10 3.65 Grinding 1.35 0.85 TOTAL 12.05 15.05

Difference (0.95) (1.SO) (1.20) 0.45

0.50 3.00

Manpower. Barrick’s roasting unit labor costs are less because the roaster plant has fewer units, and a higher per line capacity as compared to Goldstrike’s autoclaves. This comparison is somewhat distorted, as the Goldstrike autoclave was developed in successive phases. A green field approach would allow rationalization on the number of grinding units, which would reduce the difference. The fact will always remain, however, that the roaster unit can process for Goldstrike about twice the amount of tons as one autoclave. Maintenance. These costs include crushing, grinding and oxidation. The roasting plant is a brand new facility. It is to be expected that these charges will grow somewhat over the years, but it is not anticipated to exceed those of the autoclave and in fact should remain below. Reagents. The comparison in Table 2 is truly comparative base, and includes all reagents related to the oxidation of the sulfides. For roasting, it includes all of the reagents associated with the gas train cleaning, and lime associated with in bed SO;! fixation and carbon neutralization. For the autoclave, costs relate to acidulation, neutralization, and propane charges for temperature control. Reagent costs at the Goldstrike autoclave are high, reflecting the acid required to remove carbonates prior to autoclaving and the lime required for subsequent neutralization of the autoclave discharge. The cost of oxygen strictly relates to operation and maintenance of the oxygen plant, assuming that the plant is owned by the mining company. The oxygen demand is a function of the amount of sulfide and carbonaceous matter to be oxidized. Oxygen for autoclaving (high-pressure) costs about $25/MT, a11 inclusive (power, operation and maintenance). The practical limit for oxygen utilization under acidic autoclaving conditions is the 80 to 85% range. Roaster oxygen utilization is expected to reach 80 to 85%. The average combustion for sulfide has been 96.6% and 82.7% for TCMs. The oxygen cost per ton (low-pressure) amounts to $18.50/MT. Power. Includes all connected horsepower from crushing to tailings disposal, inclusive of oxygen plant. A significant proportion of the power relates to the oxygen plant for both facilities. The oxygen plant demand relates to the ore oxidation requirements. Grinding. Comparison between dry and wet grinding, excluding power, presents wet grinding as being cheaper. The main element that makes dry grinding more expensive is ore drying using propane, costing about %0.75lton. When power is included, both grinding techniques are cost comparable. Overall. Roasting unit operating costs are less than those for autoclaving and the significant gains relate to manpower, maintenance and reagents. At Goldstrike’s current throughput levels, the expected benefit of using roasting is in the range of 10 to 20%. Acid versus Alkaline Pressure Oxidation For whole ore pressure oxidation, from an operating cost perspective (at a typical sulfide sulfur content of 2% and carbonate content higher than lo%), alkaline autoclaving would be more economical. Alkaline versus acid autoclaving can display marked differences in operating costs, with the difference primarily in consumables. Manpower should be equivalent or slightly less as there is no acidulation or neutralization circuit. A practical range to use is $1.50 to $2.00/T. At equivalent sulfide sulfur oxidation, power and requirements should be equivalent, though the oxygen utilization (relating to the venting requirement associated with the C02 formation) is lower for alkaline autoclaving. Propane costs will be greater for alkaline autoclaving. Maintenance costs should be less for the alkaline process where a range of $2.75 to $3.25/T could be used.

217

Bio-Oxidation versus Pressure Oxidation Relative costs for autoclave and bio-oxidation are shown in Table 2.

-

Table 2 Comparison of autoclaving and bio-oxidation - operating costs Yo shown Item Whole Ore Autoclaving/CIL Concentrate Bio-Oxidation Wages 11 5 Reagents 48 10 14 50 Power 24 30 Maintenance 3 5 Other 100 100 Total The large power cost in bio-oxidation is related to compressed air required for oxidation. The high maintenance cost in bio-oxidation is proportionally equivalent to autoclaving. In autoclaving, oxygen for sulfide oxidation is the next most significant cost, as over 50% of the electric power costs are associated with its production. For a flotation concentrate processing using bio-oxidation, as a rule of thumb, about 1 g/ton of gold in the concentrate is required per each percent of sulfide sulfur for the process to be economically viable. Capital Costs For a double refractory ore, carbonaceous and sulfidic, the process of choice is oxygenated roasting. Whether an oxygenated roaster or acid autoclaving process is selected for a sulfidic ore depends on mineralogy. Oxygenated roasting, generally, will yield on average 2.5% less recovery than autoclaving. For a 10,900 mtpd oxygenated roasting or acid autoclaving facility, inclusive of comminution and gold recovery, capital costs are similar at US$330 million. Costs are summarized in Table 3; roaster costs are based on the Goldstrike roaster, with autoclave costs normalized based on the actual construction costs of two Goldstrike acid autoclaves (rated for a total capacity of 7,000 tonneslday). (**) Alkaline versus acid autoclaves are US $50 million versus US $66 million respectively. Alkaline costs are lower because there are no lined acidulation tanks before autoclaving, no neutralization tanks after autoclaving, and significantly less exotic materials of construction are used. Process Selection and Financial Analysis Integrating capital costs, operating costs, process recovery, and expected revenues in a decision model will allow selection of the best economic process. The decision process must consider risk and process robustness. The financial model robustness will be directly proportional to the robustness of the block by block geological-metallurgical model and the associated mining plan. These will provide the basis to generate a cash flow model that can then be used to perform discounted cash flow analysis, internal rate of return, or payback period. Confidence weighting on the geological-metallurgical economic model parameter is therefore crucial and all of the assumptions that are used should be listed as sometimes it is not possible to address all of the issues in the development program.

218

Table 3 Capital cost comparison between roaster and autoclave Design Capacity 10,900 mtpd Roaster Project Costs Autoclave Plant Estimate (US$ '000)

Site General Site Facilities (Warehouse) Plant Utilities Electrical Utilities Primary Crushing Secondary Crushing Coarse Ore Reclaim Grinding Roasting Plant Roaster Gas Handling Neutralization & Thickening Autoclave plant Thickening & Acidulation Neutralization Boiler facility CIL Plant Carbon Handling & Refining Tailings Reagents Oxygen Plant Indirect Costs Total

$2,226 $ 834 $6,368 $7,339 $7,286 $12,798 $87,695 $49,554 $23,424 $7,394

$14,966 $1,878 $5,702 $3,024 $28,320 $67,364 $326,171

(US$ '000)

$2,226 $834 $6,368 $7,339 $7,286 $3,500 $29,000

$120,000 $8,000 $3,500 $10,000 $14,966 $1,878 $5,702 $7,000 $30,000 $67,364 $324,963

ROASTING CHEMISTRY AND TESTS For roasting of whole ores or concentrates, the use of a two-stage fl~id-bed'~~' roaster either with air or oxygen enriched as per the Freeport McMoran (FMC) process can be used. Another roasting technique relates to the use of the Circulating Fluid Bed (CFB) techn~logy.'~~' Barrick Goldstrike's Roaster flowsheet is discussed more filly in Thomas et. al."" The objective of roasting is to oxidize auriferous sulfides or preg-robbing carbonaceous material to oxides, and allow subsequent recovery of gold by cyanidation. Roasting is done in the presence of aidoxygen, but below the fusion points of the constituent minerals. The product of roasting must be a porous calcine that allows permeability during cyanidation. The success associated with this conversion process is dependent on the ore mineralogy and the process operating conditions. Fundamental process differences will exist if the roasting process is designed for pyrite or arsenic-bearing ores(32),or for removal of carbonaceous matter. Some ores will be much more temperature-sensitive than others. These temperature ranges are to be defined by test work. Furthermore, the ore sensitivity to temperature can have a bearing on the type of furnace selected for roasting, or alternatively, how the temperature control must be performed. Ore mineralogy and detailed understanding of the associated relevant roasting chemistry will dictate and assist in developing the most appropriate experimental design for the tests to be executed. ( 3 3 . 3 4 3 3 ) Roasting results in mineral and phase transformations,which is a diffusion-controlled process. Good process control (in terms of temperature, atmosphere, and residence time) is required to produce controlled oxidation and produce a calcine that is amenable to cyanidation.

219

Roasting Pyrite Chemistry It is well accepted that roasting of sulfide requires carefbl control to produce a calcine with optimum gold exposure. The thermal decomposition of pyrite produces sulfur dioxide and iron oxides (hematite and/or magnetite), and is an exothermic reaction. At low temperature 45OOC (842”F), and when pyrite is roasted in air, direct conversion to hematite occurs as in Equation 1. 4 FeS2(pyrite) + 1102 + 2 Fez03 (hematitelred)+ 8 SO2

EQ 1

When the roasting atmosphere is less oxidizing (by limiting the oxygen supply) and the roasting temperature is higher, 500°C (932”F),magnetite will be produced as shown in Equation 2. 3 FeS2(pyrite) + 8 O2 + Fe304(magnetitehlack)+ 6 SO2

EQ 2

Although direct conversion of sulfide to oxide can proceed as postulated by Equations 1 and 2, a different pathway has also been ~uggested.‘~’’This pathway of oxidation, via a pyrrhotite intermediate,appears to result in better gold recovery during subsequent cyanidation. Pyrite roasted in excess air at temperatures above 500°C (932°F) can develop oxide rims due to direct conversion to magnetite or hematite (Equations 1 and 2). Once a localized sulfur dioxide atmosphere develops (>5 per cent SO2),pyrite first converts to pyrrhotite as per Equation 3. (3s*14* 15)

nFeS2+ Fe,, Sn+l+ (n-1) S 2 FeS + O2+ 2 FeO + 2 S

EQ 3 EQ 4

The elemental sulfbr atoms of Equation 4, are rapidly oxidized to sulfbr dioxide. A characteristic pattern of concentric oxide layers(39’are frequently observed in the calcine and results from the cracking of the protective rim, which is followed by the oxidation of the fresh underlying sulfides. These rims have been characterized as being impervious to gas. Favored by the rich SO2 environment of the first-stage, pyrite would roast mostly through the pymhotite intermediate. It would ultimately result in a spongy porous hematite, with only minor dense structures that would be indicative of direct conversion to oxide. The next oxidation stage is the magnetite step where pyrrhotite oxidizes to magnetite as shown by Equation 5:

When most of the sulfides have been oxidized, the following magnetite conversions occur: 6 Fe2O3 + 2 SO2 + 4 Fe304 + 2 SO3 6 FeO + 0 2 + 2 Fe304 FeO + Fe2O3 3 Fe304

EQ 7 EQ 8 EQ 9

Note that Equation 9 is a reaction between solids and it cannot therefore be expected that all the hematite can be converted into magnetite even though the reactants are finely divided. This has in fact been found to be the case in practice. Finally, when the oxidation is complete, the hematite step proceeds: 4 Fe304+ 0

2

-+ 6 Fez03

EQ 10

220

Some ferrous sulfate could also be formed during the various roasting steps, Equations 11 and 13. Such formation is considered generally undesirable because of the associated elevated subsequent cyanide consumption. The ferrous sulfate formation is most likely to occur if the finishing roasting temperature is low: FeS2 + 3 02 -+ FeS04+ SO2 2 SO2+ O2+ catalyst + 2 SO3 FeO + SO3+ FeS04 Fez03 + 3 SO3+Fe2 (S04)3

EQ 11 EQ 12 EQ 13 EQ 14

Roasting temperatures for pyrite-bearing materials range from 500°C (932°F) to 705OC (1300°F). Most pyrites start to oxidize between 425OC (797°F) and 500°C (932OF). For a proper oxidizing roast, excess oxygen should be available above theoretical oxygen requirements for oxidation of sulfide sulfur and carbon. Typically about 110% of theoretical oxygen demand is used both to generate a calcine of about 75-80% hematite and 20-25% magnetite. Sometimes it is referred to as a 'chocolate brown' calcine, however, the colour of the calcine is ore specific. In some operations, the color of the product has been used as a process control. A red calcine indicates hematite, possibly from too high a temperature combined with too much excess oxygen. A dark calcine indicates larger proportions of magnetite resulting from insufficient oxygen or too low a temperature. This color guide is not valid if the calcines contain some quantity of un-reacted organic or graphitic carbon, which will tend to darken the color of the calcines. Over-roasting conditions may occur above 75OOC (1382OF) but are ore or concentrate dependent. Over-roasting is defined as the point at which a dramatic decrease in gold extraction occurs when the roaster calcine is cyanide leached. There is a generally accepted mechanism that occurs when the over-roast temperature is exceeded. The structure of the metallic oxide, such as hematite, produced during roasting, collapses, re-crystallizes and encapsulates the gold. This encapsulation prevents cyanide solution from contacting the gold; decreases of up to 50% extraction has been observed if material is over-roasted. Since the over-roast temperature varies widely for different feed materials, only actual roasting tests can determine this temperature. For example, in the roasting of flotation concentrates, localized high temperatures from the intense exothermic reaction of the sulfides are likely to occur and result in over-roasting conditions. Since not all of the particles are at equilibrium during the roasting process, a mixture of the reaction products might be obtained in the final calcine. That sequence of events will be critical to the achievement of a representative test at the laboratory scale. Roasting Arsenopyrite Chemistry The roasting of an arsenopyrite". 41* 42. 43, 44) concentrate or arsenic-bearing ores such as arsenian pyrite typical of the Carlin trend in Nevada is considerably more difficult and complex than pyrite roasting, due to the detrimental influence of femc arsenate (FeAs04 or As20S. FezO3). If this compound is allowed to form via high temperature and fully oxidizing conditions in the roast, gold extraction will be seriously inhibited. The formation of complex non-porous iron arsenates (depicted in EQ 15) can occlude gold rather than exposing them to cyanide attack, and/or blocks the pores of the calcine, inhibiting cyanide access.

..

2 FeAsS + 6 0

2

+ 2 FeAs04 + 2 SO2

EQ 15

In order to produce a suitable calcine from arsenopyrite concentrates or arsenic-bearing ores, a two-stage roast is used. The initial stage of roasting is done in a slightly oxygen-deficient atmosphere to allow formation of volatile trivalent arsenic either in the oxide or sulfide form. Then a second stage of roasting under oxidizing conditions is required to produce the desired calcine. The reactions can be generally expressed as follows:

22 1

2 FeAsS + 5 O2+ Fez03 + As203+ 2 SO2 12 FeAsS + 29 O2+ 4 Fe304+ 3 As406+ 12 SO2

EQ 16 EQ 17

During the first stage of roasting in a fluidized bed reactor, the typical temperature range is between 4OOOC to 575°C (752-1067T). Although the temperature range is wide, Spen~e'~''has determined that higher temperature results in higher gold losses. To obtain good results, arsenic must be preferentially volatilized at rather low temperatures. It is often recommended that the available oxygen during the first-stage be set at approximately 80 to 85% of the stoichiometric requirements for optimum arsenic removal. The roasting temperature to some extent controls the volatilization of arsenic and sulfur and has very marked effects on the porosity of the calcine. As for pyrite, two oxidation pathways appear to exist. henopyrite heated in air loses both arsenic and sulfur to the gas phase, and removal appears simultaneously at 45OoC (842T), resulting in a layer of iron oxides (either hematite or magr~etite).'~*~') At about 5OOOC (932T), it would appear that arsenic is removed first, to yield pyrrhotite which then progressively oxidizes to magnetite and hematite."'. 47) The volatile arsenic species can either be sulfide (e.g. As2S3 (g)) or oxide (As406 (g)), with the oxide predominating under normal air roasting conditions. The lower temperature reaction results in direct oxidation conditions and can either be expressed by:

At higher temperatures pyrrhotite forms first, in a sequence that can be represented by: 4 FeAsS + 3 O2(g) + 4 "FeS" + As406(g) 3 "FeS" + 5 02 (g) + Fe304+ 3 SO2(g)

EQ 20 EQ 21

The pyrrhotite mechanism is likely to involve an endothermic decomposition within a particle to form pyrite with the release of arsenic vapor. The vapor diffuses to the outer perimeter where it reacts with oxygen to form As406 (g) and releases heat. The consumption of arsenic at the perimeter drives the internal diffusion process, and the endothermic decomposition is maintained by the transfer of combustion heat back into the particle towards the cooler centre. Assuming pyrrhotite production in the first stage, complete arsenic removal can be achieved since the retention of arsenic, in the calcine, is due to the reaction of arsenic compounds in the gaseous phase with iron oxides. The following reactions have also been proposed to indicate that, in the early stages of roasting, arsenic elimination takes place under reducing conditions which is essential if the formation of ferric arsenate is to be avoided. FeAsS 3 As + FeS 4 FeAs2-+7 As + F e d s 2 FeAsS + 2 FeS2 + 4 FeS + As&

EQ 22 EQ 23 EQ 24

The above reactions are then followed by two reactions, which show the oxidation of arsenic to arsenious oxide. This also provides SOz, which is required to form a reducing atmosphere inside and at the near surface of the grain. EQ 25 EQ 26

222

Once the arsenopyrite is completely decomposed, the oxidation of pyrrhotite begins. As indicated for pyrite roasting,(473 I J ) arsenopyrite can oxidize directly to magnetite and then hematite, which results in an apparently dense structure. When the roasting process is via pyrrhotite, a coarse porous structure is generated. Arriagada and Osseo-Asare‘&’ have reported magnetite as the only roast product for temperatures up to 525OC (977°F). This is important since directly formed magnetite seems to retain arsenic and appears to be a potential source of gold lock-up. Such evidence has been encountered at our Goldstrike roasting operation. Swash and Ellis(’” claimed that submicroscopic gold will also not coalesce during direct magnetite formation and will remain in the iron oxide lattice and be inaccessible to leaching. A maximum surface area occurs when secondary magnetite forms and superimposes a fine porosity on relic pyrrhotite. These pores are developed with the evolution of sulfur dioxide gas, and are likely to interconnect, so as to provide maximum porosity. However, when hematite subsequently forms without gas evolution, isolated pores are more likely to develop which will then increase gold losses. Despite that, it is usually desirable to form hematite, since magnetite is a conductor and can passivate gold during the dissolution process.(48’Fortunately, the likelihood of substantial gold loss is minimal below the hematite re-crystallization temperature of around 727°C (1340T). Furthermore, for the two-stage fluid beds application, in the second stage, where any remaining sulfur and carbon is removed, it can be operated typically at temperatures up to 604°C (1 120°F). That allows the magnetite to be transformed to hematite under oxidizing conditions.

The Arsenic Retention Mechanism It is commonly reported that arsenic can be retained in the calcine as either iron arsenate or arsenic pentoxide and that this retention is highly detrimental to gold recovery. For arsenic-bearing materials, this mechanism should be well understood. Equation 15 is generally accepted and will occur if arsenopyrite is roasted in a highly oxidizing atmosphere. It should be kept in mind that the roasting mechanism outlined for arsenopyrite will also be applicable to “arsenian pyrite” or other arsenic-bearing gold ores. The harmful effect of iron arsenate in the calcine on the recovery of gold has been extensively studied in the literature. Chakraborti and Lynch(4’)speciated arsenic retention as femc or calcium arsenate. The associated volume increase can potentially block pores and inhibit gold extraction. Djinghen~ian‘~’) reported that iron arsenate locks gold, while M ~ r t i m e r ‘reported ~~) the formation of fusible arsenates which block the pores of the calcine. Zhuchkov et al. (50) reported that the combined presence of iron arsenate with pyrrhotite leads to the formation of “dense” low porosity structures. The Barrick Goldstrike data further reinforces the above as it indicates that as the arsenic concentration in the calcine increases, unleached gold concentration in the tails increases. According to Lindkvist and Homlstr~rn,(~” the major factors affecting arsenic removal are residence time, partial pressure of oxygen and temperature. Interestingly, if roasting is carried out under reducing conditions, the final content of arsenic in the calcine does not depend on the initial content of the feed. Landsberg et al.‘52’stressed the importance of sulfur presence for preventing the formation of inter-metallic compounds and hence for more effective de-arsenification. Of great interest are experimental test results which state that under a normal operating pressure of one atmosphere, femc arsenate (FeAs04) can theoretically form as a reaction between hematite and As406(g),and never between magnetite and As406(g). This situation was found to prevail over the complete range of roasting temperatures. Since all roasting of arsenopyrite is envisaged to proceed through a magnetite stage, the implication is that arsenic cannot be retained by h-situ oxidation of arsenopyrite, but rather through a reaction of As406(g) with hematite. That condition would be more easily encountered at the outer perimeter of the grains. If the formation of hematite begins on the outer particle surface before the interior decomposition of arsenopyrite is complete, any escaping As406(g) could be trapped as iron arsenate around the outside rim. This process may account for the distinctive arsenic rims previously observed on hematite particles from a two-stage roaster.

223

Some have proposed that an addition of pulverized coal or coke toward the end of the arsenic elimination step might aid in its completion by reducing arsenates and providing a gaseous carrier. This was found to be the case, however, the advantage gained was not great enough to offset the cost of coal and its preparation. Similarly, while never discussed in literature, the contribution of naturally occurring carbon in the ore may play a role in removing arsenic, but it will be a fbnction of temperature. It is generally accepted that reducing conditions should be maintained in the early stages of the roast to ensure elimination of the arsenic in the arsenious state. Good arsenic removal requires mass transfer of the arsenious oxide as it is produced. Provided this has been accomplished, the finishing stages of the roast can be done under oxidizing conditions.

Gold Coalescence in Calcine As stated before, gold can be lost through occlusions in un-roasted sulfides, in dense iron oxides or even in porous calcines if uncoalesced submicroscopic gold is present. The latter possibility is especially relevant for arsenic-bearing materials that tend to be refractory by virtue of their high levels of submicroscopic gold. Graham et aI.JH) and Swash and Ellis"S) have experimental evidence that indicates that submicroscopic gold in arsenopyrite will coalesce during the pyrrhotite stage of roasting. It has been suggested by Swash and Ellis"s' that roasting via pyrrhotite is a prerequisite for the successful extraction of gold from arsenopyrite. Furthermore, gold coalescence is brought to completion by a reducing roast during which the pyrite and arsenopyrite pass through an intermediate pyrrhotite stage before the sulfides in the roaster charge have completely oxidized. As previously explained, in a reducing roast atmosphere, which is the prevailing condition inside the grains initially, low pressure SO2 prevails, and arsenic or sulfur which are the loosely held atoms in the arsenopyrite and pyrite structure, diffuse to the surface of the grain. As the sulfur levels in the particles are reduced, the localized reducing atmosphere is replaced by a more oxygen-rich one. Thus pyrrhotite is converted to iron oxide (magnetite). The oxidation initially occurs along grain boundaries and on pore walls. This fine, delicate, iron oxide network tends to form in a relatively rigid structure. Another important textural change observed for the ores, where the gold is invisible or in solid solution, relates to the formation and presence of very small equi-dimensional gold particles (smaller than 0.1 micrometers) within roasted auriferous arsenopyrite grains breaking down as described above. The formation of such particles appem to be somehow related to the removal of arsenic from the arsenopyrite lattice. Such gold grains are always found associated with pyrrhotite after arsenopyrite has been roasted in inert or in partially oxidizing atmospheres. Gold, according to the phase diagram, is readily soluble in arsenic and liquid sulfur and, since the liquid is at confining pressure, any flaw or weakness in the lattice will be a point at which these pressures can be released. Eventually, the arsenic and sulfur move out completely and the gold remains behind in the form of a bleb situated adjacent to the pore in the calcine microstructure. As indicated earlier, a strong correlation can be established between residual arsenic and gold grade within these FeOx phases. As de-arsenification performance reduces, higher gold tails are to be expected.'42) Providing that gold coalescence has been performed through adequate de-arsenification, quenching of the hot calcine and fine re-grinding will promote the formation of fractures and cracks, which will assist in opening up isolated pores. These steps increase the surface area of the calcine particles and should enhance gold recovery. Some investigators have tried to identify correlations between porosity and recovery. The relationship did not appear to be straightforward, most likely because of the effect of simultaneous different mechanisms which all have to be wellcontrolled in order to deliver an optimum gold recovery. The adequate control of each one of these mechanisms is further complicated by the ever-changing conditions of the ores, which do require specific rather than generic processing conditions.

224

Total Carbonaceous Matter (TCM) Oxidation For most feed materials the sulfide oxidation ranges from 80% to almost loo%, and carbon oxidation from 60 to 75%. Complete oxidation of carbon to eliminate any preg-robbing can be difficult because carbon oxidation is often a slow reaction. Thus, using CIL leaching for the calcine to counteract any residual preg-robbing is preferred over complete carbon oxidation. Also, there are cases where roasting has activated what was otherwise inactive carbon.‘2’ Equation 27 represents organic or elemental carbon oxidizing to carbon dioxide (C02). This reaction is dependent on the type of carbon present and the roasting temperature. Organic carbon with a fairly low ignition temperature will oxidize readily to COz whereas graphitic-type carbons with high ignition temperatures may not react during roasting or may present incomplete combustion characteristics.

c+o2+co2

EQ 27

Increased second-stage temperature improves carbonaceous matter bum performance. As long as this is done without re-crystallizing hematite, recovery improvement should result. To control second-stage temperature, adjustment to the ore blend is generally the practical alternative. This is primarily achieved by increasing the carbonaceous content of the ore feed, which tends to bum in the second-stage. The use of fuel guns in the second-stage is possible, but this use would be of limited effect in its ability to significantly raise second-stage temperature. These guns should be rather seen as “temperature trimming” devices.

Roasting of Carbonate Ores Calcium and magnesium carbonates decompose to calcium and magnesium oxide. This reaction is dependent upon the type of carbonates present and roasting temperature. Magnesium carbonate decomposes at much lower temperature than calcium carbonate. CaCO3/MgCO3+ CaO/MgO + C02

EQ 28

The reaction between CaO or MgO (either generated from the ore carbonates or by addition of lime) and SO2during roasting will produce anhydrite calcium or magnesium sulfates. This reaction is referred to as fixation of S02. This reaction is generally desirable, since it reduces the quantity of SO2 in the roaster off-gas, unless test work shows that gold extraction is reduced. CaOMgO + SO2 + % O2+ CaSOflgS04

EQ 29

The amount of SOz, which can be fixed without addition of lime is dependent on the content and the types of carbonates present in the ore or concentrates and the temperature of the first-stage roast. In some high carbonate feed materials, up to 60% of the SO2 generated during roasting can be fixed as calcium or magnesium sulfates as long as the first-stage roasting temperature is sufficiently high. If lime is added to the feed, SO2 fixation can reach 75%.

-

Roasting Test Procedures Bench Roasting can be conducted using air, oxygen enriched air or pure oxygen. Typically, for double refractory ores, two-stage roasting is recommended. Standard variables to investigate include the effect of roast temperature (in both stages if applicable), retention time, grind, minor element deportment (such as As, Sb, and Pb), oxygen concentration, ore type, and carbonaceous matter content. Sulfides are typically oxidized in the first stage while carbonaceous matter is burnt in the second stage.

225

Laboratory-scale roasting testing can be performed by using a bench-scale rotary kiln or a small fluid bed assembly (<5cm in diameter). Other methods, such as using a dish in a muffle fiunace, are not generally reliable techniques. Such static tests can result in formation of ‘hotspots’, material fusion, effects due to localized atmospheres, and poor mass transfer. Regardless of technique, control of temperature and atmosphere are important. The use of bench top rotary k i h to perform initially scoping tests is an acceptable technique as long as its limitations are well understood and recognized. Such units allow control of the atmosphere, albeit only above the charge. Primary temperature control of kiln is by the electrical elements linked to a thermocouple control; cooling can be accomplished by manual lifiinglclosing of the lid. The industrial-scale technique of using water guns to trim the temperature of the bed cannot be duplicated in bench-scale kiln approach. At the Barrick Goldstrike facility(55’small scale tests are conducted using a batch rotary kiln (Bench-Top Roaster, BTR). Figure 1 presents the general arrangementof the BTR unit.

Figure 1 Bench Top Roasting Unit. Photo courtesy of Barrick Goldstrike Mines Ltd. The furnace has a heating chamber in which the temperature can be varied fkom ambient up to 1100°C. Quartz glass vessels, with a nominal charge of 500 grams, are placed into the furnace for the,roasting tests. Rotation of the vessel is provided by an electric motor with variable speed control. A built-in flowmeter allows adjustment of gas flow to the inlet side of the quartz vessel, with the outlet port directed to a scrubber system to remove sulfur dioxide. Both internal and external temperatures are measured and recorded during the test. A programmer allows for varied temperature profiles to be evaluated. Both the temperature ramp rate (how fast the temperature rises in the heating chamber - in OC/minute) and dwell time (how long it stays at a specified temperature- in minutes) can be adjusted. A standard test procedure to emulate the full-scale twostage roast has been developed; the test is conducted using standard times, temperatures, and atmospheres to simulate a two-stage roast. (55) The test atmosphere concerns, outlined above for the dish test, must also be considered for the BTR test. Depending on the ore type studied, this may have a significant bearing on the results. Typically, for high arsenic ores the BTR recovery will tend to outperform both the full-scale plant and the pilot plant results and the effect is proportional i.e., the higher the arsenic concentration will be, more significant the bias will be. While the BTR method can be calibrated to either the pilot or the full-scale plant after commissioning,at the development stage of a project, it would be unacceptable to extrapolate these results for industrial performance. To perform a BTR test, there

226

is a series of operating variables that have to be selected that will influence the test results, (initial heating rate being a good example). Because the test is conducted in a kiln, the charge is not fluidized and is rather in a rocking mode inside the tube. The exposure of all of the particles to the “superficial” roasting atmosphere is therefore low. This mode indirectly allows the core of the charge to be in a “neutral to reducing” environment despite the pure oxygen gas feeding the unit. Consequently, the gas action acts more as a superficial gas flow to the charge. Such reducing conditions favour arsenic removal, which promotes gold recovery. For a grass roots process definition, having no benchmark, piloting is the preferred alternative, albeit that it would be acceptable to attempt calibrating the BTR tests to the pilot plant results. The BTR technique could be used as an indicator for ore variability and the metallurgical performance results must be interpreted as relative and not absolute. Furthermore, these results must be interpreted in light of the mineralogical occurrences and the associated variability.

Pilot Roasting Test Procedure. When conducting a refractory gold ore roasting evaluation, the feed material must be characterized in terms of chemical and mineralogical analyses, particle size determinations, and cold flow fluidization tests. Fines elutriation should be monitored as they could by-pass the roasting requirements. Cold flow fluidization tests are also important to estimate and ensure minimum superficial velocities in full-scale design. The Barrick Goldstrike test work indicated 0.65 Wsec was minimum and the plant was designed for the second stage at 0.70 Wsec minimum. Depending on the analytical results, a thermodynamic model that calculates the heat balance and assists in identifying the ranges of operating conditions should be evaluated. (56) Great care must be practiced in selecting the samples for testing, as has been said over and over, samples must be as representative of the reserve as possible. This is critical to the success of pilot program as well as future commercial units. Understanding of the mineralogy is key to designing a successful program. Different pilot-scale units are available; typically four and six-inch column units can be used. Because of reduced wall effects and a lower risk of bridging, the six-inch unit is typically favored over the 4-inch. Goldstrike uses a six-inch-diameter unit, with a height of 22 feet, followed by a 26-inch extension spool connecting to the cyclone inlet port. The overall length of reaction zone, measured from the fluidizing-air distribution plate to the roaster exhaust port, is approximately 23 feet. The system is equipped with numerous instruments to obtain operating data: 0

0 0

0

Thermocouples linked to a multi-point recorder measure and record temperatures in the roaster and exhaust gas system. A large proportion of these thermocouples are located in the reaction vessel. A data acquisition system continuously monitors and records the gas composition. Gauges to measure pressures throughout the system including the direct pressure in the windbox and roaster, and the differential pressure across the fluidized bed, cyclone, and baghouse. Flowmeters to measure and control the various gases to the system.

The furnace shell is externally heated by either a propane-fired pre-heater or electric heating elements. The bed is fluidized by the gases passing through a windbox attached to a perforated distribution plate. An electric gas pre-heater is used to heat the fluidizing gas prior to its entering the windbox. A triaxial designed nozzle is at an elevation of one inch above the distribution plate. A nozzle installed through the side of the fluidized bed vessel is used as a bed-pressure tap. The ore is fed to the roaster using a variable-speed two-inch-diameter screw with an open hopper. Process gases generated from the roasting process pass through the vessel reaction zone and discharge to an external cyclone and bag house, where entrained calcine particles are captured. The gases leaving the bag house enter a packed-tower caustic scrubber circuit for acid gas removal. The scrubbed gas is vented to the atmosphere through an air-induced draft fan, which is used to control the pressure within the roaster. Calcine is collected continuously in a bed overflow canister.

227

First-stage and second-stage roasting tests are de-coupled. During first-stage roasting, the canister is submerged in a water bath to quench the temperature of the calcine and stop the oxidation reaction. For second-stage operation, the water bath is not used; calcine collected in the canister is discharged (hot) into a bucket of water (about 20 liters) to wet quench the calcine. The process exhaust gas is sampled at the bag house outlet and analyzed continuously for 0 2 , C02, CO, SO2, NOx, and THC. The desired percentage of excess oxygen is maintained by adjusting the oxygen, carbon dioxide, and air rotameters for the fluidizing gas stream. The retention time is varied by changes in feed rate. The retention time is calculated by dividing the contained bed mass by the feed rate for a particular test. The bed mass is determined at the end of each operating testing day by discharging and weighing the bed material. With the above system, the achieved retention time accuracy could be off by as much as 25% of the targeted retention time. This is a key element to be monitored. The total hourly product rate divided by the average feed rate determines the mass accountability. Products include bed overflow, cyclone underflow and bag house fines. The calculated mass accountability is typically within the range of 5 to 10% of 100%. In a typical two-stage fluid bed application, the fluidizing gas for first-stage processing is the process gas exhausted from the second-stage roast. During pilot-scale operations, the composition of this gas is emulated by adjusting the 02,C 0 2 and air rotameters for the fluidizing gas stream to provide the desired input concentrations. However, the solid to gas ratio between the industrial unit and the pilot plant are not comparable. On the basis of the same ore processed in both units (pilot vs. industrial), Barrick test work has revealed that this condition could create a significant off-set in the test results interpretation. The Barrick roaster is typically operated so that the roaster off-gas is in the 20 - 40% by volume on a wet basis range for oxygen concentration. In order for the pilot unit to perform metallurgically as per the industrial unit using the same ore (a relatively high arsenic concentration of about 3,000 ppm), different gas composition regimes had to be used in the pilot unit when compared to industrial operation. For example, when operating the pilot plant at the same SO2 concentration as in the industrial unit, averaging about 5% in the off-gas along with similar oxygen concentration range as the industrial plant, averaging about 40% in the off-gas, the resultant SO2 concentration in the pilot plant was in the range of 2.5%. In order for the pilot plant to produce an SO2 concentration of 5.5%, the off-gas oxygen concentration in the pilot plant had to be reduced to the range of 8%. That regime yielded equivalent gold recoveries in the pilot plant when compared to the industrial unit. The abovementioned oxygen concentration, for Goldstrike ores, is applicable for ores which carry arsenic concentrations up to 2,200 ppm. Further reducing oxygen to 1 to 2% yielded metallurgical performance better than the full-scale plant for high arsenic concentration (3,000 ppm). While it can be argued that it would not be practical to consider 1% oxygen concentration in the industrial off-gas, it should not be interpreted as such. As demonstrated above by the chemistry review, the SO2 concentration plays an important role. The literature indicates that as the SO2 concentration increases, the formation of hematite is retarded and iron sulfates are preferentially formed. As mentioned earlier, the occurrence of hematite must be avoided to remove arsenic efficiently. Optimum gold recovery can only be obtained if adequate arsenic removal is first achieved. The differences between between pilot- and full-scale can be partially explained by differences in the gas to solid ratio. As the pilot ratio is higher, the result is a gas composition regime that affects metallurgical performance when arsenic occurs at sufficient levels. Furthermore, the industrial units typically have a limited number of feed pipes and they are physically grouped together to avoid being close to the bed overflow in order to limit the potential of short-circuiting. This concentration of feed material possibly creates a specific gas environment within the industrial unit that differs from the pilot unit. Consequently, while the pilot plant test work indicates that reduced oxygen concentration favours recovery (as does the literature), it does not mean that the full-scale roasting facility has to operate at that low oxygen concentration in the offgas to potentially see the added benefits. This perhaps indicates that reducing the oxygen concentration, while maintaining minimum concentration SO2 in the first-stage, should enhance

228

gold recovery. When scale-up evaluations of the off-gas are performed, they must be closely analyzed and not done using face value of the test results. Doing so could significantly jeopardize the gas train design. In the case of Barrick Goldstrike, the SO2 and O2would have been acceptable but NO,, CO, C02 would have resulted in further complications. Comparison to other commercial units has proved to be more acceptable. Generally the total sulfur and TCM oxidation performed in the pilot plant is a good indicator for what is to be expected from the industrial roaster. However, Bamck test work and comparison to the large unit revealed that the calcine from the first-stage pilot plant generally presents as much as 10% less oxidation of TCM’s. An operating parameter for the Barrick Goldstrike operating conditions is to prepare a feed blend for which the arsenic concentration will not exceed 1,800 ppm. Consequently, during the test work phase, consider performing a diagnostic of the resultant mineralogical phases, which contain the gold in the leached residues. This will greatly assist in the data interpretation. Another critical element of the pilot plant test work relates to the control of the second-stage temperature, which is artificially maintained via fuel addition. While the test work may show that recovery benefits by maintaining a higher second-stage temperature, it is important to recognize that in practice the ore fuel value components (sulfide sulfur and carbon) will determine the actual heat balance unless external heat input to the second stage is integrated into the design. Fuel guns in the second stage can assist in trimming the temperature, but can only contribute roughly 25 to 50°C. It is recommended to perform differential thermal analysis @TA) on the feed and first-stage calcine to assist calibrating the heat and mass balance model. Some plants do operate allowing the temperature of the second-stage to float, which is easier for the operator, but again such requirement or lack of control is ore specific and should be well-defined at the design stage. Similarly, when evaluating pilot-plant data for off-gas environmental compliance, differences between pilot- and full-scale gas:solids ratios must be considered. Primary elements to monitor are particulates, acid mist/SO3 which will impact opacity, mercury vapor and particulates, S02, CO and NO,. Again, pilot plant results may only be indicative of full-scale performance. The above outlined test work procedures are considered adequate to determine the effect of temperature and retention time per stage. Special attention should be given to the arsenic deportment, balance, and environmental stability of the formed species. The pH of the calcine will determine the alkali demand for the neutralization stage. Various type of alkali can be considered. The selection should not be strictly based on its effect on gold recovery but also on the ability to practically handle gypsum formation. It is believed that rapid quenching of the calcine increases its porosity and thereby improves the gold dissolution during subsequent cyanidation. Some plants in the past have considered calcine re-grinding as a means to further improve recovery. In-bed SO2fixation will need to be quantified along with determining if it should be enhanced through alkali addition to the fresh feed. Once the optimum operating conditions have been selected and the ore reserve is well understood, with regard to carbonate distribution, the test work can focus on this cost element of the operation. Often it will be debated as to whether an acid plant should be integrated into the gas train rather than SO2 neutralization. Again, understanding the above elements will be critical in assessing that selection. Confirmatory pilot plant studies can also be conducted in either a 15-inch or a 24-inch-diameter fluidized-bed. These reactors would typically be direct-fired. Again, caution is recommended; experience showed that the relationship developed in the pilot plant varied significantly from that in the commercial unit. Mineralogy applied to the roasting products will reveal avenues for optimization.”” Ba~m(’**~~) has summarized a series of findings: 0

Typical alteration minerals of refractory sulfide ores as well as slime-forming gangue may generate large amounts of detrimental fines. These fines can retard oxidation efficiency, impact reaction kinetics, cause sintering and scale buildup or create problems in the roaster gas train.

229

Next Page

Although graphite is largely inactive carbon, it may become extremely pre-robbing after roasting. Carbonaceous matter is frequently associated with clays or sericitic alteration minerals. This mineralogical occurrence may prevent complete oxidation and may significantly increase the preg-robbing potential of the carbon. Complex sulfides (i.e., sulfosalts) containing Pb, As, Sb, Bi, Cu and Zn may induce severe refractorinessof the gold and silver during roasting. It is of importance to determine if the roaster feed (concentrate or whole ore) contains coarse gold (actual gold particle size > 200 mesh). During roasting, very coarse gold particles may incur considerable surface contamination (coatings, alloys, sinter agglomeration). This may make the readily recoverable gold severely refractory to leaching. Metal and/or gangue sulfate minerals in the feed increase the amounts of scale buildup in the roaster gas train. Slimes are a major reason for agglomerationand clinker formation in fluid bed roasting. The presence of antimony sulfides (stibnite) and/or antimony-bearing sulfo-salts (e.g. tetrahedrite) may cause considerable particle fusion problems in fluid bed roasting. The good cleavage of these minerals causes them to report to the slimes during grinding and may thereby increase caking of the bed. Antimony minerals increase the volatilizationof the gold. The presence of fine-grained carbonates (calcite, dolomite, siderite, ankerite, magnesite) inter-grown with sulfides and/or siliceous gangue is frequently overlooked. This is a major reason for poor roasting due to COz formation and promotes the buildup of anhydrite or gypsum in roaster equipment. Minor differences in the whole ore/concentrates composition can be the reason@) for considerably different gold recoveries. Ferreira et alj6O'report the loss of gold due to a soluble gold species (Au-As-sulfur mineral) in the calcine quench water. Dorr-Oliver has been doing test work on fluid bed application since the 1940's. The scale-up of laboratory data is complex, and based to a high level on each company's proprietary data base. Stanley Bunk of Technip/Dorr-Oliver(61) offers the following comments: First a potential commercial flowsheet with heat and mass must be generated. Important parameters must be defined. For refractory gold ore (after it has been decided to use fluid bed roasting) important decisions must be made, such as: whole ore versus concentrate feed; "bubbling" versus Circulating fluid bed; single versus multistage; etc. Each of these decision branches has other secondary decisions. Examples are oxygen versus air roasting for whole ore and turbulent versus "fast" for CFB. Assume a representative ore sample is available and "bubbling" fluid bed using the two-stage oxygen roasting system similar to the Bamck Goldstrike mine is chosen. Test rig size is a h c t i o n of particle size to limit the bias of test results due to wall effects. A minimum diameter is about 4 inches internal diameter. Important parameters include temperature, solids' residence time and oxidizing potential. Since a multi-stage process has been chosen it must be decided if each stage can be tested individually or if the entire system must be tested in an integrated fashion. The original test work for Freeport McMoran was an integrated two-stage roasting test. Experience has shown that integrated testing is generally not required.

230

Bench-scale and Pilot Plant Tests for Cyanide Leach Circuit Design Gene E. McClelland and Jack S. McPartland’

ABSTRACT Cyanide processing has been used for precious metals recovery for over 100 years. Since the 1960’s, many improvementsto cyanide leach circuitshave been made, including carbon adsorption/desorption, heap leaching, agglomerationheap leaching,CIP and CIL cyanidation,pressure oxidationkyanidation, and biooxidationkyanidation.This paper will discuss bench and pilot-scale testing procedures for al I these process options with particular attention given to objectives for various phases of testing, utility oftest procedures and data, and limitations and pitfalls for bench and pilot-scale testing protocols and data. INTRODUCTION Metallurgical test data are invaluable to a project where cyanidation is used as a primary or a secondary recovery process for precious metal bearing ore deposits. Each deposit is unique and specifically designed metallurgical testing programs are required to: 1) make a production decision; 2 ) develop the most economical processing sequence; 3 )provide design criteria for all processing unit operations;4) identify potential environmentalimpacts;5 ) determine if such impacts can be prevented during processing or require mitigation on closure; and 6) ensure an acceptable return on investment through closure of the project. The extent ofmetallurgical testingrequired for a given project/deposit is influenced by: 1 ) project capitalization; 2 ) a level of production risk minimization acceptable to the mining company; 3) the variability of rock types, oxidation states and ore grades characteristic of the deposit; and 4) the complexity of the proposedselected processing sequence for the ore deposit. It is true that no bench or pilot-scale testing program can exactly simulate site or ore specific conditions. However, decades of comparative laboratory data and commercial production data have established relationships or “rules-of-thumb” which provide confidence that comprehensive metallurgical test data will accurately predict commercial production performance. Most metallurgical testing programs are conducted in phases. Lower cost preliminary phases (bottle roll, mechanically agitated, and small column percolation cyanidation tests) allow evaluation of essentially all the varied rock types, oxidation states, and ore grades. Preliminary test data will establish metallurgical similarities and differences between ores in the deposit. The next phase of testing can then be conducted on ore composites based on those established similarities or differences to minimize the number of tests (or test phases) required to select an economical processing sequence and to make a production decision. The ore type, oxidation state, and ore grade considered most representative of the entire deposit is typically selected for evaluation during the second phase of testing. Other rock type, oxidation state, and ore grade composites can then be evaluated using optimum process conditions established for the most representative ore type to predict the overall production performance of the entire ore deposit. It is advisable to evaluate separately ore to be mined in the first one to three years of operation to ensure adequate cash flow during the capital amortization I McClelland Laboratories, Inc., Sparks, Nevada.

251

period. Heap leach cyanidation projects usually require more testing phases, through various pilot-scale options, than do milling/cyanidation projects. Recommended testing phases for milling and heap leach processes are described in this chapter. Other chapters in this multi-volume publication discuss various other process options like flotation and gravity beneficiation, pressure oxidation, roasting, and biooxidation of sulfide ores. Consequently, discussions of those process options are not included in this chapter, but cyanidation processing of appropriate process streams from those options are discussed. TYPES OF SAMPLE APPROPRIATE FOR PHASES OF METALLURGICAL TESTING Types of sample usually available during project development phases include RC cuttings, drill core, and bulk surface or bulk mined ore samples. RC cuttings are normally available during early exploration drilling. Cuttings intervals are used for assaying to determine ore grade and determine overall dimensions of the ore deposit. Cuttings samples are appropriate for preliminary metallurgical evaluation. Cuttings samples are not appropriate for grind size optimization for a milling/cyanidation processing approach because of ore particle size reduction which occurs during drilling. Diamond drill core samples or composites are appropriate for all phases of metallurgical testing for heap leach and milling cyanidation processing approaches. Generally, drill core is obtained later in project development phases and consequently, is available for the more detailed metallurgical testing phases. Drill core can be obtained throughout the ore deposit to ensure representative samples of each rock type, oxidation state, and ore grade in the deposit. Drill core is the favored type of ore sample/composite for heap leach evaluation if crushing is required to achieve targeted precious metal recovery and for milling/cyanidation testing phases. Large diameter core (6 inch/l50 mm) would be required for evaluation of ore deposits which are amenable to heap leaching treatment at run-of-mine (ROM) or primary crush feed sizes. Obtaining large diameter drill core is costly, but may be necessary to properly evaluate heap leach performance for coarse ore feed sizes. In most cases, project personnel use expensive drill core for purposes other than just metallurgical testing. Other purposes (not listed here) are important, but may render the core unsuitable for detailed heap leach evaluation. Typically the core is sawn or split. A 1/4 split is used for assay and another 1/4 split is typically archived (posterity sample), and only 1/2 the core is available for metallurgical evaluation. This scenario is acceptable for milling/cyanidation process development testing, but may not be acceptable for heap leach evaluation. It is recommended that whole core be made available for any detailed metallurgical evaluation, but if this is not practical for the project, the following “rule-ofthumb” is offered as an alternative to dedicated metallurgical core: The coarsest crush size available for heap leach evaluation would be one-half the diameter of the core submitted. The outer surface of the drill core is “plugged” with drilling chemicals and the sawn core surface is “plugged” with fines resulting from sawing. Consequently, the core diameter must be reduced by crushing (1/2 the diameter of core received) to create new, “non-plugged” surface area to restore the natural permeability of the ore. Drillers should be cautioned to not use any drilling chemical, especially petroleum based, which may cause a “preg-robbing” effect during metallurgical tests. One way to avoid the problem discussed above is to obtain bulk (ROM) ore sample (surface or mined) for heap leach evaluation of coarse ores. Bulk ore samples are normally obtained from near surface, from old underground workings or from constructing a decline to the heart of the ore deposit. Bulk ore samples are appropriate for all phases of metallurgical testing regardless of processing approach, but bulk ore samples will likely not be as representative as drill core samples/composites. If bulk ore sampling is practical for the project, it must be confirmed that bulk samples are representative by conducting confirmatory metallurgical tests on subsequent core samples/composites later in the metallurgical development phase. Geological and metallurgical project professionals should interface to ensure that different ore types are properly identified, sampled, and evaluated. Remember that metallurgical test data is only as good as the sample provided and how that sample is prepared before evaluation.

252

METALLURGICAL TESTING PHASES FOR HEAP LEACH CYANIDATION EVALUATION The first step/phase of metallurgical evaluation occurs at the assay laboratory. In most cases, all or a certain percentage of drill intervals submitted for ore grade determination are subjected to a cyanide solubility (“cyanide shake”) test to obtain a “first look” at metallurgical behavior. Cyanide solubility tests are generally (can vary between assay laboratories) conducted on 10 to 30 gram splits from assay pulps (- 150 mesh) at 30 percent solids slurry density. The tests are conducted hot (95°C) for one to two hours or cold (ambient temperature) for 24 hours using cyanide solution at a concentration of 5 to 10 gNaCN/L and a pH of 1 1 to 12 (NaOH). After shaking, the slurry is filtered and the filtrate (pregnant solution) is analyzed for precious metal content. Slurry solids are not typically assayed for residual precious metal content. The assayed grade of the assay pulp and extracted values from the cyanide solubility test are used to back calculate precious metal recovery (in percent) for the drill sample (interval). Cyanide solubility test data indicate amenability to direct cyanidation treatment for a finely divided (ground) sample, but does not necessarily indicate amenability for a coarser sample. Developing an historical data base of cyanide solubility results for an ore deposit provides a large advantage during actual mining and processing. During mining, blasthole samples are assayed in the site laboratory, and if cyanide solubility tests are conducted, that data can be compared with the data base to identify potentially different metallurgical behavior of the ore being mined. This early identification will clue project metallurgiststo conduct more detailed metallurgical tests on mined ore zones which may be problematic. An historical data base for bottle roll cyanidation data from an ore deposit is important for the same reason. An added benefit to a strong bottle roll test data base is that shorter term bottle roll tests, rather than longer term column leach tests, can be used by minesite metallurgists to further delineate problems associated with metallurgically dissimilar mined ore. The same cyanide solubility test procedure and the same bottle roll test procedure should be used throughout the project to obtain directly comparative historical data. Preliminary Metallurgical Testing Phase Bottle roll cyanidationtests are the next phase of a preliminary metallurgical testing program, and are conducted by a metallurgical testing laboratory. Bottle roll tests for preliminary heap leach amenability evaluation are typically conducted on RC cuttings samples/compositesat the as received cuttings feed size (nominal 6.3mm, 114 inch). Bottle roll tests are less costly than column percolation heap leach amenability tests, and each rock type, oxidation state, and ore grade in the deposit should be evaluated using bottle roll test procedures for three principal reasons: 1) development of an historical data base for the deposit; 2) to establish metallurgical similarities or differences between rock types, oxidation states, and ore grades; and 3 ) based on those similarities or differences, preparation of subsequent ore composites to minimize the number of samples/compositesrequired for more detailed, more costly metallurgical tests. Bottle roll tests are conducted on a minimum of 1 kg of cuttings sample/composite at the cuttings feed size (nom. 6.3mm). Solids are mixed with water (tap water or minesite water) to achieve a slurry density of 40 weight percent solids. Natural slurry pH is measured. Lime (or NaOH) is added slowly (= 2 hours) to adjust the pH of the slurry to between pH 10.5 to 1 1 .O before adding the cyanide. Sodium cyanide, equivalent to 1 .O g/L (2.0 Ibs/ton of solution) is then added to the alkaline slurry. Leaching is conducted by rolling the slurry in a bottle on the laboratory rolls for 96 hours (or longer if extraction is progressing at reasonable rate). Rolling is suspended briefly after 2,6, 24, 48, and 72 hours so samples of pregnant solution can be taken for Au, Ag, pH, and NaCN analysis. The volume of sample taken at each sampling interval must be recorded to account for extracted values and NaCN removed from the test to obtain a mass balance for the test. Make-up water, equivalent to that withdrawn, is added to the slurry. Cyanide concentration is restored to the initial value. Lime (or NaOH) is added when necessary to maintain leaching pH at the desired level (pH 10.5-1 1 .O). Rolling is then resumed. After 96 hours (or longer), the slurry is filtered to separate liquids and solids. Final pregnant

253

solution volume is measured and sampled for Au, Ag, pH, andNaCN analysis. The residue solids must be thoroughly washed (1 repulp, 5 displacements), dried, weighed, and assayed in triplicate (or screen assayed) to determine residual precious metal content. Calculated head grade from the bottle roll test is used to calculate precious metal recovery. Precious metal recovery, recovery rate, and reagent requirement data are obtained froin preliminary bottle roll tests. Cyanide and lime requirement data typically better represent those experienced in commercial heap production than requirements indicated by column leach test data. Recoveries indicated by bottle roll tests are usually lower (= 10%) than recoveries achieved froin column leach tests conducted on a similar feed size of ore. Preliminary bottle roll test data can also indicate: 1) amenability to heap leach cyanidation processing, 2) presence of coarse (visible) gold particles and gold bearing sulfide minerals, 3) clay content which may indicate the need for agglomeration before heap leaching, 4) a “preg-robbing” character of the ore, and 5 ) if a residue screen analysis is conducted, the degree to which values are liberated from the various particle sizes. The principal inadequacy of bottle roll test data is the small portion of ore (1 to 5 kg) used for the test. This inadequacy can be minimized somewhat by careful sample preparation, blending, and splitting procedures. Bottle roll tests can be conducted on varied feed sizes up to 2 inch (50mm), if drill core or bulk sample is available, to indicate optimum heap leach crush size. Crush size sensitivity data, however, may not be accurate because autogenous grinding of coarser feeds may occur during bottle rolling. Column percolation leach tests can also be conducted during apreliminary heap leach amenability evaluation. Drill cuttings samples are usually the only type of sample available for preliminary testwork. Data obtained from column tests on cuttings sampledcomposites is only slightly more definitive than data from preliminary bottle roll tests. Consequently, column leach tests on cuttings is not typically recommended. More Detailed Metallurgical Testing Phase If metallurgical data from the preliminary phase is encouraging, project personnel will obtain appropriate samples for second phase metallurgical evaluation. Drill core and/or representative bulk ore samples/composites are required for this phase of evaluation. The principal objective for the second phase is to optimize heap leach crush size. To accomplish the principle objective, column percolation leach tests are usually conducted on various crush sizes (normally to simulate the various crushing stages) to provide actual recovery versus crush size data. Crush size optimization is confirmed by conducting head (feed ore) and tail (column leachedlrinsed residue) screen analyses on each crush size evaluated. Recovery versus particle size data can be calculated from respective head and tail screen analysis data. Other objectives and necessary design data can and should be obtained during the more detailed phase of metallurgical testing and include, but not limited to: 1) determine the need for agglomeration pretreatment; 2) obtain crusher sizing and wear part abrasion design information; 3) identify any potential metallurgical and environmental concerns; 4) confirm metallurgical performance of the various rock types, oxidation states and grades of mineable ore; 5 ) determine the best solution recovery system (carbon circuit or zinc precipitation); 6 ) establish permeability under load for predicting ultimate commercial heap height and other geotechnical considerations; 7) optimize agglomerating conditions if agglomeration pretreatment is necessary; 8) determine neutralization rinsing character of leached residue; 9) determine if a pulp agglomeration (higher grade ground ore used to agglomerate crushed heap leach grade ore) processing sequence is economically feasible; 10) establish carbon loading capacity and kinetics and carbon type for a carbon-in-column (CIC) adsorption circuit or zinc requirements and solution chemistry for a zinc precipitation circuit; and 1 1) obtain other design criteria data for a prefeasibility study. Optimization of leach solution application rate and cyanide concentration are not typically required during heap leach evaluation. Space restriction precludes procedural description of all types of tests normally conducted in this

254

second phase of heap leach evaluation. Fairly detailed procedures for column percolation leach tests and agglomeration condition optimization tests are, however, discussed. Carbon-in-column (CIC) loading capacity and rate tests as well as considerations and procedures for pulp agglomeration are discussed in the milling/cyanidation evaluation section later in this chapter. Testing to obtain crusher sizing and wear part abrasion design information is outside the scope of this paper. Column percolation leach tests are normally conducted in columns (usually PVC) of varied diameter and height. Column diameter must be six times the diameter ofthe coarsest ore particle size leached to minimize column wall effect (solution flowing down the wall of the column rather than percolating through the ore charge). Column height is somewhat arbitrary, but should be 10 to 20 feet (3 to 6 meters) to minimize the solution tons to ore tons ratio for each day leached. Each ore charge is mixed dry with the appropriate quantity of lime (determined from bottle roll tests) or agglomerated using optimum agglomerating conditions before being loaded into the leaching column. Ore charges are loaded in a manner to minimize particle segregation and compaction. Some type of inert material is placed on the loaded surface of the ore charge for even leach solution distribution. A multi-stage carbon circuit should always be used during column leaching so leach barren can be recycled to the ore charge throughout the leach cycle. Loaded carbons must be assayed (after leaching and rinsing) to obtain an additional metallurgical balance for respective column leach tests. Dry ore charge weight and apparent bulk density is obtained before and after leaching. Saturation and residual moisture contents are determined to aid commercial water balance calculations. Leaching is conducted by applying sodium cyanide solution (concentration determined from bottle roll test consumption) over the ore charge at a rate of between 0.003 and 0.005 gpm/ft’ (0.12 to 0.2 Lpm/m’) of column cross-sectional area. Pregnant solution is collected in acovered vessel and volume is accurately measured daily. Daily pregnant solution is sampled and analyzed for Au, Ag, pH, NaCN, and other metals (if required). Daily pregnant solution is pumped through the carbon circuit (3 to 5 stages of attrited carbon) for adsorption of dissolved values. Daily barren solution volume is measured and sampled for Au, Ag, pH, and NaCN analysis. Make-up water (preferably minesite water) and reagent are added, and barren solution is recycled to the ore charge daily. If value breakthrough to barren occurs, carbon from the first column is pulled and assayed, succeeding carbon columns are advanced, and the fresh, attrited carbon is placed into the trailing carbon column. All loaded carbons are dried, weighed, and assayed after leaching and rinsing. Frequent carbon changes can be required if the Ag:Au ratio in pregnant solutions is greater than lo:]. The daily procedure should continue until pregnant solution grades approach the lower detection limits of the analytical equipment used. At that point, rest cycles (= 2 to 4 weeks), where the ore charge is allowed to stand idle in contact with residual cyanide solution, should be allowed to increase pregnant solution grade when active leaching is re-initiated. Several rest cycles may be needed during the leach cycle to ensure accurate pregnant solution analysis results. Column leach tests, with rest cycles, should continue until no additional gold, and possibly silver, is detected in daily pregnant solution to ensure that maximum precious metal recovery is achieved. After leaching, column residue is rinsed (with carbon circuit intact) to recover dissolved values and to obtain preliminary neutralization rinsing character of the residue. Barren solution, rather than fresh water, should be used for the rinse cycle. After rinsing, residue is removed from the column, and is blended and split (moist) to obtain sample(s) for residual moisture content determination. Remaining residue is then air dried, blended and split to obtain samples for multiple direct tail assay and sample (at least 1/2 of the residue) for tail screen analysis. Three metallurgical balances (solution extracted value/taiI grade, loaded carbon valuehail grade, and head gradehail grade) are obtained from the recommended column percolation leach test procedure described above. Solution and carbon balances should be about 95% or better precision to ensure that column test data are accurate and meaningful. Silver precision can be about 90%. Second phase column leach test precious metal recoveries, if column tests are run to completion,

255

are accurate and should not be discounted when doing economic evaluation of the proposed project. Precious metal recovery rate data are also accurate if rate data is calculated on a recovery versus tons leach solution applied per ton of ore leached basis rather than on a recovery versus time basis. Recovery versus ton/ton rate data can be used to project commercial recovery rate data even for a multi-lift commercial heap leach operation. Multi-lift column leach test advantages will be discussed in the pilot-scale heap leach testing section ofthis chapter. Recovery versus crush size data, along with head and tail screen analysis and resulting recovery by particle size data are accurate for determining optimum heap leach crush size. Column leach test data does not accurately predict NaCN consumption for commercial production. There are several reasons for column test NaCN consumptions being higher than for actual commercial consumption, but space constraints disallow an adequate discussion of each. Typically, column test consumptions are about four times higher than experienced in commercial production. That “rule-of-thumb” does not apply to ores containing significant quantities of cyanicides. Bottle roll test NaCN consumption better represents actual commercial consumption. The need for agglomeration pretreatment before conventional heap leach processing is determined in the preliminary heap leach evaluation phase. Ores containing significant quantities of clayey minerals and ores where significant quantities of fines are generated by crushing usually require agglomeration pretreatment. If more than 10 weight percent minus 150 mesh (1 06 pm) fines or clays are present in the crushed feed, the need for agglomeration pretreatment is indicated. Optimum agglomerating conditions are determined from a bench scale test procedure referred to as agglomerate strength and stability tests. Submersion tests are a measure of agglomerate green strength. “Jigging” tests are a measure of agglomerate stability. Both test procedures are detailed in the following paragraphs. These procedures have been used for successfully determining optimum commercial agglomerating conditions for heap leach operations throughout the world. The three main agglomeration parameters are binder (cement, lime or a combination) addition, agglomeration moisture and cure time. Usually, only binder addition requires optimization. The minimum cure time is 24 hours, but cure times measured in days or weeks are acceptable. In commercial practice, cure time is dictated by the time required to produce a sufficient agglomerated heap surface area to initiate leach solution application (usually 72 hours or longer). Consequently, the cure time does not require optimization. Agglomeration moisture is optimized visually during testing and commercial agglomeration and, in turn, does not require separate testing. Agglomeration moisture dictates the agglomerate growth mechanism (layering andor coalescence). A properly prepared agglomerate will contain a coarse particle nucleus with adhering layered and coalesced fines and medium coarse particles and will appear moist, but not “shiny” or “slimy” wet. While separate testing is typically not required for optimization of agglomeration moisture, a correct moisture addition during agglomeration is critical for the preparation of strong and stable agglomerates. Excessive moisture addition, agglomeration of ore with too high a “natural” (before agglomeration) moisture content, and excessive drying of agglomerates during the cure time can all have detrimental effects on agglomerate strength and stability. Agglomerate strength and stability tests (Agg S&S) are conducted by hand, so the entire procedure should be performed by the same technician to minimize the human factor. Preparation of the feed to be agglomerated is critically important to insure the same quantity of plus 10 mesh material and minus 10 mesh material is contained in each charge to be agglomerated for the test series. A sufficient quantity of ore at the optimum heap leach crush size is screened on a 10 mesh screen and the f10 mesh weight distribution is calculated. Plus and minus 10 mesh material are each blended and split to obtain the appropriate weight for each ore charge. Several one to two kilogram charges (usually 7- 10) with the same natural crushed ore rt 10 mesh weight distribution are reconstituted for the Agg S&S test series. Prepared ore charges (1 -2 kg) are agglomerated with different binder additions (typically 0, 1.25, 2.5,3.75,5.0,6.25,7.5, and 10.0 kg binderlmt ore) by adding the binder, and then wetting with water

256

to the optimum moisture content (determined visually) while manually tumbling in a small bucket. The volume of water added to affect agglomeration, and the initial moisture content of the ore must be accurately measured. Agglomerated charges are placed into sealed containers and cured for 72 or more hours (consistency is required) before conducting the ‘tjigging” and submersion tests. Agglomerated charges, after curing, are placed onto a 10 mesh screen. Two typical agglomerates are removed, weighed and submerged in separate beakers of water. The degree of agglomerate degradation is measured over time. An agglomerate of sufficient green strength will not degrade over 24 or more hours of submersion. An agglomerate of insufficient green strength will completely degrade in less than 1 hour. Agglomerates remaining on the 10 mesh screen after two are selected for submersion testing are manually “jigged” in and out of a container of water ten times in a 30 second period. Agglomerates retained on the screen are dried and weighed. The increase in weight percentage retained over that of the natural (dry screened) weight distribution is calculated. Stable agglomerates will usually contain at least 25 weight percent more plus 10 mesh material after “jigging” than the natural crushed plus I0 mesh material. Weight percent retained is plotted against binder addition. Optimum binder addition is determined by the point at which near maximum weight percentage retained was achieved, and by the binder addition where no agglomerate degradation occurred on 24 or more hours of complete submersion. Establishing heap permeability underthe predicted compressive loadings encounteredduring heap leaching is conducted as part of the agglomeration optimization. Geotechnicaltesting of agglomerates using load cells can be used to confirm the indicated optimum agglomeratingconditions. Additionally, these load versus permeability tests should be conducted on select simulated heap residues generated during this phase of testing (column residue) or during pilot testing. Conventional testing procedures for the 11 item list of objectives offered above provide sufficiently accurate data to meet test objectives and provide design data for a prefeasibility study. Sufficient data is usually obtained from the second phase of metallurgical heap leach evaluation to allow a production decision for the project. At this point a decision may be made to conduct pilotscale heap leach cyanidation testing to resolve any questions raised from second phase data and to bolster confidence in predicting commercial heap production performance. Pilot-scale Heap Leach Evaluation Options Pilot-scale heap leach cyanidationtests are nearly always necessary for a proposed ROM commercial heap, but are not always necessary for heap leachable ores requiring two or more stages of crushing. Acquiring sufficient ore sample for a pilot-scale evaluation is costly. Pilot-scale ROM heap leach evaluations are conducted in either large diameter columns (4 to 6 ft; 1.2 to 1.8 m) about 20 feet (6 meters) high or by actual field heap leach trials. About 20 tons of representative ore sample (large diameter core or bulk sample) is required for each pilot column test and head screen analysis. A pilot field trial should be conducted on a minimum of 50,000 tons of ROM ore. Because of cost and time considerations,pilot column tests are much more common than field trials for a project in development phases. The pilot column percolation test procedure for a ROM feed is the same as described in the previous section of this chapter with essentially the same advantages and “pitfalls.” A large column is required to meet the column diameter to coarsest ore particle size ratio. A beneficial pilot-scale heap leach test for crushed ores is a heap height simulation column percolation leach test procedure. Heap heights of up to 400 feet (1 22 meters) and crush sizes up to a primary crusher discharge product (nominal 4 inches; 100 mm) have been evaluated using this procedure. A 4 inch feed would require a series of 24 inch (0.6 m) I.D. x 20’ (6 m) columns. Finer crush sizes can be evaluated in smaller diameter columns, but probably no less than 8 inch (200 mm) diameter. The heap height simulation pilot column test procedure is the same as described earlier, but a

257

series of 20 ft (6 m) high column charges are leached to simulate the projected commercial heap height. New lifts (additional 20 ft columns) can be added to the column charge series in accordance with proposed lift heights and new lift construction cycle planned for commercial multi-lift heap leach production. Principal objectives and advantages for conducting heap height simulation pilot column tests are; 1) cost and expediency; 2) determination of “lag-time” in precious metal recovery as new lifts are added; 3) prediction of overall commercial leach cycle required to achieve targeted recovery, on a tons solution applied per ton of ore leached basis; 4) evaluation of a very representative composite of mineable ore; 5) opportunity to conduct a heap height simulation rinse test for closure planning; 6) generate representative sample for characterization of process solution for minesite disposal and solid heap residue sample for other environmental characterization tests including revegetation and plant uptake studies; 7) obtain thorough drain down rate and volume data for modeling to predict actual commercial drain volume and long term drain down rate; and 8) obtain all final design data for the final feasibility study and design of the commercial operation. Some testing pitfalls might include: 1) cost for sample acquisition; 2) ensuring a representative samplekomposite of mineable ore; and 3) the inadequacy of the procedure to simulate actual load at ultimate commercial heap height. This load information can be obtained, however, by conducting appropriate “outside the column” geotechnical tests. Items discussed above for heap leach cyanidation testing phases are not exhaustive. Testing programs and protocols can and should be tailored to specific project objectives, needs, and, where possible, to minesite specific conditions. Heap leach cyanidation test phases accurately predict commercial heap performance if mined ore is prepared (crush size, agglomeration conditions,etc.) to specificationsdetermined during the testing phases. METALLURGICAL TESTING PHASES FOR MILLLNGKYANIDATION EVALUATION Exploration development of an ore deposit at times identify significant tonnages of higher grade ore which may be amenable to primary or secondary millingkyanidation processing. Milling refers to the physical process of grinding an ore to a finely divided (finer than 48 mesh, 300 pm) feed size to achieve liberation of contained precious metal values. Primary cyanidationprocessing refers to direct cyanidation of milled whole ore. Secondary cyanidationprocessing refers to cyanidation of a process stream(s)resulting from pretreatment ofmilled ore to affect concentration ofvalues (gravity, flotation) or pre-oxidation of carbonaceous or sulfide minerals (roasting, pressure oxidation, bio-oxidation). Various milled ore pretreatment technologiessuch as mentioned above are discussed elsewhere in this multi-volume publication. This chapter discusses only cyanidation processes for milled primary or secondary feeds. Preliminary Metallurgical Testing Phase Cuttings sampleskomposites obtained during the exploration developmentphase of an ore deposit are typically availablefor the preliminary phase ofmillingkyanidation metallurgical evaluation. Drill core and bulk ore samples/compositesare acceptable for preliminary evaluation if available. Preliminary whole ore cyanidationtests are conducted on appropriate samples/compositesmilled to a practical feed size (P,,200 mesh, 75 pm) using either bottle roll or mechanically agitated direct cyanidation test procedures. The need for pretreatment of milled whole ore for concentration or preoxidation is usually determined in this preliminary testing phase. Secondarycyanidation metallurgical tests are normally conducted in the next phase of millingkyanidation metallurgical evaluation. Preliminary milling/cyanidationtests should be conducted on ore sampleskomposites of varied rock type, oxidation state (oxide, transitional, sulfide), and ore grade to establish metallurgical similarities and differences. Subsequent composites based on similarities or differences can then be prepared to minimize the number of composites for evaluation in the next and more detailed phase of metallurgical testing.

258

Objectives for preliminary milling cyanidation evaluation are to: 1) determine amenability to millingkyanidation treatment, including precious metal recovery, recovery rate, and reagent requirements; 2) indicate optimum grind size with tail screen analysis data; 3) determine if pre-concentration or pre-oxidation treatment is necessary to improve overall precious metal recovery and recovery rate; 4) indicate selection of the most feasible solution recovery process (carbon or zinc precipitation); and 5) identify potential liquid solid separation and pulp rheology concerns. The bottle roll test procedure discussed earlier, and milling/cyanidationbottle roll test procedures are similar except that: 1) ore charges are stage ground (not pulverized) to the desired feed size using a laboratory stainless steel ball mill; 2) the duration of the test will likely be less than 96 hours, but ground feeds should be leached to completion; 3) interim pregnant solution sampling is more frequent the first 24 to 36 hours of the leach cycle; 4) slurry density may be higher than 40 weight percent solids; and 5) leached residues are usually screen assayed to determineresidual precious metal content and distribution to obtain preliminary indication of optimum grind size. Mechanically agitated cyanidation rather than bottle roll tests are recommended even for the preliminary amenability testing phase to: 1) better simulate a commercial agitated leach tank; 2) allow air or oxygen sparging;3) facilitateobservation of pulp rheology; and 4) allow interim pulp sampling without interrupting agitated leaching. The laboratory mechanically agitated leaching apparatus should be constructed to simulate commercial impellorto tank diameter ratio, tank diameterto height ratio, and baffling with appropriate dimensions. The test procedure is essentially the same as the bottle roll test procedure except that: 1 ) leaching (agitation) is not interrupted during interim pulp sampling; 2) pulp sample is filtered to separate pregnant liquor from partially leached solids; 3) partially leached, filtered solids are returned immediately to the leach vessel; and 4) pregnant liquor is analyzed for Au, Ag, pH, and NaCN as well as dissolved oxygen content. Preliminary mechanically agitated cyanidation tests can be conducted on milled feed charges of up to 15 kg (33 Ibs). Preliminary test data can meet stated objectives and provide information required to delineate a scope of work for the next phase of milling/cyanidation evaluation. The authors are not aware of any significant inadequacies or “pitfalls” associated with preliminary milling/cyanidationtest procedures. Optimization Metallurgical Testing Phase If preliminary milling/cyanidation test data is encouraging, project personnel will obtain appropriate samples/compositesfor the optimization (second) phase of milling/cyanidationevaluation. Drill core is the preferred type of ore sample for this phase of testing. Bulk ore sample is appropriate if representative of mineable ore. Cuttings samples are not appropriate for this phase because of size reduction resulting from the drilling method. Whole ore and pretreatment (gravity, flotation, preoxidation) process streams are evaluated in this phase. Principal objectives for this phase of metallurgical evaluation are to: 1) confirm the processing method and sequence indicated by preliminary data; 2) optimize grind size, retention time, and reagent addition; 3) confirm solution recovery system and efficiency; 4) obtain all necessary crushing and grinding circuit design criteria; 5) obtain all necessary liquid solid separation and pulp rheology data for plant design; 6) complete geotechnicalevaluation for tailings impoundment design; 7) characterize process streams to prevent or mitigate potential environmental impacts for mine waste rock and impounded tailings; 8) provide all data necessary to make a production decision and conduct feasibility studies; and 9) identify unit operations which may require pilot-scale evaluation. Space constraints allow detailed discussion of only objectives 1 through 3 in this chapter. Appropriate samples and/or process streams are generated during this phase to conduct specialized tests to accomplish objectives 4 through 7. These specialized tests are conducted by personnel with capability and expertise in those specific technical disciplines. Combined test data (objectives 1 through 7) is necessary to accomplish objective 8. Portions of objective 9 are discussed later in this chapter.

259

Most common cyanidation processes for recovering values from higher grade whole ore and pretreatment products (streams) are carbon-in-pulp (CIP), carbon-in-leach (CIL), and counter current decantation (CCD) for feeds requiring zinc precipitation as a solution recovery system. Mechanically agitated cyanidation tests are used for the three most common processing sequences. CIP and CCD processing evaluations are conducted using a direct cyanidation test method followed by either CIP or zinc precipitation circuit optimization tests. CIL processing is similar, but activated carbon is added to the alkaline slurry at the same time NaCN is added. Extracted values are determined by loaded carbon assay results. Precious metal recovery rates are determined from back calculation from interim tail assay results. Confirmatory processing sequence testwork should be conducted on the rock type, oxidation state, and ore grade representing the major portion ofthe mineable ore. Once optimum processing conditions are established for the major ore type, other ore types can be evaluated using optimum conditions to confirm metallurgical behavior and to determine if commercial processing conditions should be modified to ensure maximum production from the minor ore types. Grind size optimization is highest priority regardless ofthe processing sequence. Staged grinding or batch grinding procedures should be used to prepare feeds for grind versus recovery cyanidation tests (CIP, CIL, CCD). Staged grinding procedures are conducted by placing the feed for subsequent cyanidation into the laboratory ball mill, adding water to achieve 55 to 65 percent solids, grinding for a specified time, screening the partially ground feed on the appropriate screen size (i.e., 100, 150,200, 270 M), regrinding oversize for a designated period and again screening to obtain oversize product. The stage grinding procedure continues until the desired 80 percent passing size (Ps0) is achieved. At least three grind sizes should be prepared for subsequent grind versus recovery cyanidation tests. Batch grinding procedures are also effective in producing desired grind sizes for subsequent grind versus recovery cyanidation tests. A grind time curve must be generated, by separate batch grind tests, to establish the grind time required to achieve desired grind sizes. About five grind times, using separate ore charges, are evaluated to establish a grind time curve. These ground pulps are not used for subsequent cyanidation tests. Feeds for subsequent grind versus recovery cyanidation tests are ground in separate batches for specified times (determined from the grind time curve) to achieve at least three desired grind sizes. Cyanided tails from grind versus recovery tests are screen assayed to determine residual precious metal content and distribution and to confirm ground ore (feed) particle size distribution. Tail screen analysis data along with cyanidation test data from the various grind sizes evaluated are sufficient to optimize grind size for the ore. Retention time is also optimized during the grind versus recovery test series if sufficient interim pregnant solution samples are taken and analyzed during the leach cycle. Cyanide leach solution concentration optimization tests are the next step in the optimization test series. Ore charges at optimum grind size should be leached separately using at least three cyanide solution concentrations. Optimum cyanide concentration is determined by the observed effect on retention time (recovery rate) and NaCN consumption. Optimization of leaching pH is not always necessary, but if pH affects cyanide consumption, an optimization cyanidation test series should be conducted. Evaluation of leaching pH is restricted to a fairly narrow pH range. A minimum pH of 10.3 is required for protective alkalinity (minimize HCN votalization) and a pH much higher than 1 1.2 can adversely affect precious metal dissolution rate. Suggested leaching pH’s for pH optimization are pH 10.3, 10.6, and 11.O. Once grind size, retention time and reagent additions are optimized, a larger scale confirmatory mechanically agitated cyanidation test should be conducted on the ore using selected optimum leaching conditions. Leached pulp or decanted pregnant solution is available from this larger confirmatory test for detailed CIP kinetic tests (pulp) or zinc precipitation tests (decant). Testwork for optimization of a zinc precipitation circuit is discussed in a different chapter of this publication. A discussion of carbon kinetic studies to optimize a carbon adsorption circuit design for CIP and CIL processing options is provided in this chapter. The carbon circuit for recovery of

260

dissolved values from heap leach pregnant solution (carbon-in-column - CIC) can also be optimized using the kinetic test procedure described below. CIP and CIL carbon adsorption capacity and adsorption rate tests are conducted on leached pulp while CIC tests are conducted on pregnant solution. Test procedures are somewhat different for the CIL process, but details of those differences, in the interest of brevity, are not presented here. There are several carbon kinetic test procedures which provide sufficient data, along with computer modeling of that data, for design of a commercial carbon adsorption plant. Most of those procedures are variations of the more common procedure described in the following paragraphs. Activated carbon of appropriate hardness (attrition tests) and activity (loading capacity tests) should be selected as optimum before conducting detailed carbon loading capacity and rate tests. Equilibrium carbon adsorption capacity tests are conducted on several identical aliquots of leach pulp (usually 6) or pregnant solution to determine maximum carbon loading and, combined with adsorption rate data, to determine forward and reverse loading constants. Each aliquot of pulp (or solution) is contacted with a different weight of coarse (6x12 m or 6x16 m) activated carbon (pre-attrited) and agitated for 24 hours to establish equilibrium conditions. Recommended carbon weights for a typical preg grade (5 ppm Au or less) are 0.1,0.2,0.5, 1.O, 2.0, and 5.0 g carbon/liter of pulp (or solution). After 24 hours, loaded carbon is screened from the pulp (or solution) and barren solution is analyzed for Au, Ag, pH, and NaCN. Pregnant and barren solution concentrations and carbon weights are used in a mass balance equation to calculate carbon loadings. Loaded carbons from the 1 .O, 2.0, and 5.0 g carbon/liter tests are assayed to provide a check on calculated carbon loadings. Data is used to generate an adsorption capacity curve, plotting carbon loadings against barren solution concentration, to be used with rate data to determine the number of carbon contactors and retention time required to achieve desired barren solution concentration. The adsorption rate test is conducted by contacting at least four liters of leached pulp (or solution) with 1 .O g carboniliter and agitating for 24 hours. Pulp or solution samples are taken at 20, 40, 60 minutes, and 2, 5, 8, 10, and 24 hours and are analyzed for Au, Ag, pH, and NaCN to establish adsorption rate. The mass balance equation is used to calculate carbon loading. The loaded carbon is assayed at the end of the test to check calculated loading. The 1.O g carbon/liter charge is used to ensure that adsorption rate is not too rapid. At least eight hours of contact time should be required before a low grade barren is achieved. This test, run for 24 hours, is a metallurgical check on the adsorption capacity test using the 1.O g/L carbon density. Adsorption capacity and rate data are usually computer modeled to provide commercial carbon adsorption circuit design criteria. Quite often, an ore deposit amenable to heap leaching contains significant tonnages of selectively mineable higher-grade ore, but not sufficient tonnages to warrant a separate and complete milling/cyanidation plant. Recovery of ounces from higher-grade ore zones can be critically important to project economics, and recoverability of those ounces is normally enhanced significantly by grinding for liberation of values. A technology termed pulp-agglomeration was developed in 1989 to improve the economics of heap leachable ore deposits with higher-grade zones or high-grade satellite deposits. The technology has been used commercially in several operations in different parts of the United States. The concept, simply stated, is to mill the higher-grade ore to the optimum grind size and use the ground pulp as agglomeration moisture for agglomeration pretreatment of the crushed heap leach grade ore. Some dewatering of the ground pulp (thickening, filtration) may be necessary to provide optimum agglomeration moisture for the heap leach grade ore. The extent of dewatering required is dependent on the high-grade/heap grade ore ratio most practical, and on optimum moisture for agglomeration of the heap grade ore. Typically, higher-grade ore is closed circuit milled in cyanide solution, and cyclone overflow is fed to a grind thickener. This provides sufficient time for dissolution of a significant portion of the contained values. Thickener overflow is fed to the CIC circuit and thickener underflow is either

26 1

filtered (depending on agglomeration moisture requirement) or fed directly to the agglomeration circuit. Process alternatives include milling of the higher grade ore followed by concentration using gravity or flotation methods. The resulting tailings are then used for agglomeration pretreatment of the heap leach grade ore. Test procedures described earlier are used effectively to provide design data for the pulp agglomeration processing circuit most economically feasible for the ore deposit. Items requiring optimization include: 1) grind size for higher-grade ore; 2) pulp agglomerating condition; 3) higher grade/leach grade ore ratios; 4) dewatering methods; and 5 ) milling/dewatering retention time for acceptable value recovery from the higher grade ore. The principle advantagesfor pulp agglomeration are: 1) improved recovery from higher grade ore which would otherwise be heap leached; 2) stronger and more stable agglomeratesbecause of a higher fines content of the ground pulpheap leach grade ore placed on the heap; 3) lower capital cost than for a separate and complete milling/cyanidation circuit; and most importantly 4) elimination of the need for a tailings impoundment. The only real disadvantage is a somewhat delayed complete value recovery from the higher grade ore, as compared to from a separate and complete millingkyanidation circuit. Values not recovered in the milling/dewatering circuit are recovered over a longer term with values from the heap leach grade ore. Optimization test phase data will accomplish the nine objectives stated previously for the samples/composites representing the major portion of mineable ore. Again, minor ore types should be evaluated separately using all optimum conditions determined for the major ore to ensure effective production by the selected processing sequence. Metallurgical test data from the bench scale optimization phase “scales up” well to commercial equipment planned for use in commercial production. Consequently, pilot-scale evaluation may not be necessary for the cyanidation circuit and solution recovery system. Pilot-scale evaluation may be necessary to further minimize project financial risk and to resolve any metallurgical questions left unanswered during the optimization testing phase. Pilot-scale MillingKyanidation Evaluation Summary Pilot-scale evaluations for pretreatment processes (gravity, flotation, pre-oxidation) are usually necessary to obtain sufficient plant design data. Those pilot-scale evaluations are discussed in various other chapters of this publication. Pilot-scale CIP, CIL, and CCD evaluation procedures and equipment are available should project personnel deem pilot evaluation necessary. CIP and CIL pilot tests can be conducted using a continuous, cascading leach circuit. A continuous grinding circuit may be used, but batch grinding of a sufficient quantity of feed to conduct continuous CIP or CIL pilot test is usually acceptable. A semi-continuous/batchlarge bottle roll test CIP or CIL pilot-scale test procedure has also been used effectively to generate design data for a commercial cyanidation circuit. Detailed pilot-scale test procedures cannot be adequately discussed in this chapter because of space constraints.Advantagesof athorough pilot-scale evaluation are: 1) evaluation of a large quantity of representativemineable ore; 2) identification and resolve of any metallurgical or process problems not identified during the optimization phase; 3) generation of any design criteria not obtained during previous testing phases; and 4) production of sufficient process stream sample for specialized equipment selection and sizing tests summarized earlier (objectives 4-7). Disadvantages of pilot-scale milling/cyanidationevaluation may include: 1 ) cost for acquisition of appropriate sample quantity and type; 2) cost and time required for pilot-scale evaluation; and 3) likely inability to adequately evaluate all ore types in the mineable reserve. Advantages of pilot-scale evaluation usually “outweigh” disadvantages if required information was not obtained during previous testing phases, or if “novel” processing methods are included in the

262

proposed flow sheet. CONCLUSION Properly designed and conductedmetallurgical evaluationsthrough the various phases, along with well reasoned interpretation of the resulting data, are invaluable to mining companies for making production decisions and designing and operating commercial cyanidation circuits. The quality of data produced during such metallurgical studies is dependent first on the quality and preparation of the ore samples evaluated. Decades ofcomparative laboratory and commercial production data have established relationships which provide confidence that comprehensive metallurgical test data accurately predict production performance in commercial cyanidation circuits. Decisions concerning the level of testing detail and scale required for design of a commercial cyanidation circuit are typically based as much on project financial risk considerations as on metallurgical considerations.

While outside the scope of this paper, site specific conditions such as terrain and weather will often dictate the list of possible commercially successfulprocessing options. These conditions should be given careful consideration before embarking on a detailed metallurgical testing program. While metallurgical testing related to various other process unit operations are discussed in other papers contained in this volume, well planned coordination between the various entities responsible for the different types of testing is critical to a projects success. Careful evaluation of data after each stage of testing is important to ensure that the metallurgical evaluation proceeds “down the correct path,” and that any required changes in the planned scope of testing are made in a timely manner.

263

Bench-scale and Pilot Plant Work for Gold- and Copperrecovery Circuit Design Dave Thompson'

ABSTRACT Laboratory and pilot plant p r o g r a ~to~ establish ~~ design criteria for plants recovering gold from cyanide solutions and copper from acid leach solutions are discussed. The gold recovery methods include both activated carbon adsorption and zinc dust cementation. The copper recovery method is solvent extraction followed by eleclrowinning. Although these technologies are mature, some experimental work is still often needed to avoid unexpected problems in the commercial plant.

INTRODUCTION The use of activated carbon to adsorb gold from cyanide leach solutions was researched and reported extensively by the U.S. Bureau of Mints and others starting in the late 1960s. The first large commercial plant based totally on heap leaching followed by carbon adsorption was commissioned in 1976 at Round Mountain,Nevada. Process development work for the plant had been done at Mountain States Research and Development, mostly building on concepts published by the Bureau. Gold recovery by zinc dust cementation, known as the Merrill-Crowe process, dates back to 1918. Origmally patented by the Memll Company. it was subsequently marketed by Don-Oliver and is commonly associated with their name. Although now largely replaced by activated carbon systems. Menill-Crowe can still be the gold recovery method of choice in certain situations. Solvent extraction of copper was first applied commercially in 1968 at the Ranchers Bluebird Mine near Miami,Arizona, based on experimental work conducted primarily at Hazen Research, Inc. It has since almost completely replaced iron cementation (at kast in North America) for recovery of copper from acid heap and dump leach solutions and has made possible the exploitation of many previously uneconomical low-grade copper resources. These technologies are mature, well-known, and h@ly successful. Once the subjects of extensive laboratoq, pilot plant, and semi-commercial demonstration programs, they are now sufficiently understood so that commercial plants can usually be confidently designed based on far less experimental work or none at all. Some testing, however, is pNdent in many situations. This paper discusses what those situations might be and the types of tests that should be performed. We owe a great deal not only to the technical people who originally developed these processes, but particularly to the owners and managers who supportbd their efforts. It's not a trivial thing to sponsor years of research and then risk further money and reputations building a firstsf-its-bd metallurgical plant. Few have ever been willing to do so, but without those few our industry would be poor indeed.

GOLD RECOVERY BY ACTIVATED CARBON ADSORPTION Coverage of this subject will be limited to ttcovery from solutions rather than slurries. Recovery from slurries by carbon-in-pulp or carboa-in-leach techniques are so entangled with leaching itself that they are more effectively described togetha with leaching. Gene McClelland has most ably done this elsewhere in this volume (Mcclelland 2002).

' Hazen Research Inc., Golden, CO

264

The most common version of the carbon ldsorption process for gold cyanide solutions employs multi-stage, expanded-bed carbon columns. Pregnant solution is pumped upward through an orifice plate. It expands (fluidizes) the bed of granular carbon, overflows the top of the column, and is pumped back up through the next column. Five stages are typical. The carbon is advanced countercurrently by a system of hydraulic eductors, loaded carbon being removed from the fmt stage and stripped carbon being added to the last (from which the barren solution exits). For 840 x 595 pm (10 x 28 mesh) carbon. the flow rate in the column is about 5.1 Umin/M’ (15 gpmlft2). In the early days of gold ore heap leaching,extensive experimental programs were conducted on this system to establish carbon loadings, stage requirements, adsorption kinetics, attrition losses, degradation of carbon performance over time, and the buildup of unwanted contaminants. The effects of many variables were determined, and dozens of M e r e n t kinds of carbon were compared. These experiments were usually done on a relatively small scale including laboratory batch tests. bench-scale circuits, and the occasional pilot plant. Sometimes when field heap leaching tests and demonstrations were conducted, as they often were, a fairly large carbon adsorptloo system was included and became part of the program. To establish loading capacities, small samples of the carbon could be agitated with portions of a given pregnant solution until equilibrium was reached. After assaying the solutions and carbons, a type of McCabe-Thiele diagram was constructed to determine the number of stages and carbontesolution ratios necessary to produce the desired barren solution concentrations and carbon loading capacities. Another method involved passing pregnant solution through a small static bed of carbon and assaying the effluent periodically until no more adsorption occurred. Both the ultimate loading and the loading rate could thus be ascertained. It was a relatively simple way to compare one carbon with another and to determine the effects of various ion concentrations, temperatures, and the llke. There were no “standard” procedures for any of this; each lab seemed to develop its own set of empirid tests. This was acceptable as long as sound engineering principles were followed in both design of experiments and interpretation of results. The rate at which carbon lost its activity was observed by subjecting it to many cycles of loading and stripping, using a representative sample of pregnant liquor from the actual ore to be processed. Buildup of lime scale almost always occun-ed to some extent, and it could be overcome by periodically washing the carbon with dilute nitric acid. Organics and other impurities in the solution could foul the carbon, and this was dealt with by thermal treatment in a reactivation kiln. If any mercury was present it could build up on the carbon and would have to be recovered in a special retort. Requirements for these “carbon maintenance” systems were readily determined by interpretation of the laboratory or bench-scale circuit results, scaleup being straightfoxward. Large pilot plants were usually not needed. At present, even the basic laboratory tests are not always done for design of commercial plants, because so much experience and confidence has been gamed in carbon adsorption and carbon maintenance systems over the past 25 years. Now most of the experimental work is done on leaching of the ore, not carbon adsorption of the gold. It is by leaching tests that metals recoveries and major reagent consumptions are determined under various conditions of particle size, leach time, etc. These are the “big ticket” parameters in the economics of heap leaching plants. When pilot-scale column leach tests are conducted, it is common to remove gold from the effluent by passing it through one or more small containers of activated carbon. The solution is then recycled to the top of the column after adjustment of the cyanide strength. Later, the carbon is assayed for the purpose of obtaining a material balance around the leach. If the carbon produces satisfactory barrens and nodung out of the ordinary is observed, linther experimental work on the adsorption is often not performed. Exceptions, of course, do arise. Perhaps carbon from an unfamiliar source is contemplated for use in the commercial plant. P e w the pregnant liquor contains an unusually high concentration of one or more known contaminants, either from the ore or from the local water to be used in the process. In such cases, some adsorption tests like those done in the early days are advisable. Their results may suggest carbon-tegold ratios, frequency or intensity of acid washing and thermal regeneration, etc., that are different from typical plant design and practice.

265

The most common deleterious impurity encountered in cyanide leach solutions is probably copper. In sufficient concentration it can occupy active sites (although not irreversibly) on the carbon and seriously reduce the gold and silver loading capacity. 'Ihis situation dictates changes in the plant design and operating strategy in ways that can be derived from laboratory test results. Stripping of the loaded carbon is accomplished in several different ways, usually with some combin&on of sodium cyanide and sodium hydroxide, with or without additives such as alcohol. It may be done just below the boiling point or (more likely in mwer plants) above the boiling point while under pressure. The method of choice for a particular plant depends largely on economics and on designer or operator preference. No testing is normally required. Likewise recovery of gold from the s l i p solution, mostly by electrowinning, is wellestablished technology. Mild steel or stainless steel wool are the most common cathode materials, each having its advantages and disadvantages. Ihe cells and accessories can be designed from strip solution compositionsand flow rates. Experimental procedures and techniques for inkzpretation of results for carbon adsorption and stripping systems are presented in more detail in Heinen, Peterson, and Lindstrom 1978.

GOLD RECOVERY BY ZINC DUST CEMENTATION On overall cost considerations, the Merrill-Crowe process may be selected over activated carbon when the plant is very small, when the pregnant leach solution is very high grade, or when it contains large amounts of silver. As an old deaf-thumb, if the ratio of silver to gold in the solution is less than 5:l. carbon is preferred. If it is greater than 101,Memll-Crowe is preferred. Carbon Circuits would be impraCticabIy large and expensive in order to accommodate that much silver. If the ratio is between 5 1 and 101,a detailed cost cornperison is warranted. Most U. S. gold ores contain relatively little silver, so carbon is nearly always selected. In Mexico and 0 t h locations where ores with high silver are common, Memll-Crowe installations are more often seen. Feed solution for Menill-Crowe plants must contain less than about 10ppm suspended solids, so the first step is a clarifying filter. either presswe or vacuum. 'Ibe next step is deaeration to remove dissolved oxygen that would otherwise cause excessive zinc consumption, interfere with precipitation, and risk redissolution of precipitated metals. The target oxygen concentration is 2 ppm or less. m e n zinc dust is added in the amount of about one Unit per unit of silver plus gold. A small amount of lead nitrate is often also added to "activate" the Zinc. actually forming a leadzinc couple on its surface. The p i p i t a t e and excess zinc are then removed using filters precoated with diatomaceous earth. Precipitation is so fast that it happens in the pipe and on the filter itself. no holding tank being required. Bamn solution concentrations of 0.01 g/t gold (0.0003odst) are typical. Ihe precipitates. containing between 20 and 80% precious metals, are dried,mixed with fluxes, and smelted to d d metal. F+riorto about 1980, Menill-Crowe plants required significant operator skill and attention. They now lend themselves to automatic control with computer monitoring of pressure drops across filters, oxygen content after deaention. and other parameters. Experimental work is not normally needed for design of a M d l - C r o w e plant, all items being experientially sized based on solution flow rate, solution composition (gold, silver, copper, cyanide), and site altitude (for vacuum system design). It may be wise, however, to perform laboratory filtration tests on the pregnant solution as a check on sizing and performance of clarifying filters if there is any reason to suspect problems. Inadequate clarification is the principal cause of trouble in the system, not just from fine ore particles but from precipitated hydrates of iron, magnesium, and aluminum, which can coat on the zinc and seriously interfere with p i o u s metals precipitation. Excessive pressure drops across the zinc filter can also result from improperly clarified feed solutions. Temperaoures, flux additions, etc. for the smelting step are detetmined by experience and by analyses of the precipitates. No testing should be necasary unless a highly unusual precipitate presents itself. A more detailed description of the Merrill-Crowe process and its issues appears in Potter (1980).

266

COPPER RECOVERY BY SOLVENT EXTRACTION AND ELECTROWINNING Before the first solvent e x d o n and electrowinning installation at Ranchers Bluebird mine, copper had baen recovered from acid leach solutions by cementation on scrap iron, often s W d & food and beverage cans. It was a somewhat inelegant procedure that caused iron to build up in rechulating solutions and did not produce commodity-grade copper metal without further processing. Behind the Ranchers plant were hundreds of laboratory shakeout tests. at least two continuous bench-scale circuits, numerous special tests utilizing ad hoc procedures, and a pilot plant at the site. The majority of the work was done by Hazen Research (Golden, Colorado), with significant technical contributions by the extractant manufacturer, General Mills. Solvent extraction, as the term is used in extractive metallurgy, means contacting an aqueous leach solution with an immiscible organic solvent containing a chemical called an extractant. The metal or metals of interest are transferred from the aqueous into the organic phase, and the phases are allowed to separate by gravity. The metal is then stripped off the solvent by contact with a different aqueous solution, and the solvent is returned to extract more. In addition to the ex-mt, the solvent contains a diluent such as kerosene and possibly modifiers that enhance phase separation or otherwise improve performance. "he varied and complex issues in solvent exrxaction of copper are what necessitated such extensive experimental work before the first commercial plant could be designed. The solvent's affinity for copper (expressed as an extraction coefficient or distribution coefficient) needed to be sufficient so that huge numbers of countercurrent stages were not required. The solvent had to have a low affinity for impurities. Practical extraction kinetics were necessary. The aqueous and organic phases had to wparate cleanly and in reasonable amounts of time. Interfacial sludge ("crud'') buildup needed to be held to tolerable levels. Solvent losses and solvent degradation had to be affordable. The copper had to be strippable with a solution recycled from an electrowinning cell, and this cell needed to produce marketable coppa cathock competitive with those from electrorefining plants. The simplest and most useful laboratory @um in solvent extraction was and still is the shakeout test. Quantities of solvent and aqueous arc added to a seperatory funnel and manually shaken until equilibrium is achieved. The phases are then analyzed. Results are plotted as copper concentration in the organic phase vs. that in the aqueous phase, and the number of stages and organic-tcwpeous ratio wxssary to produce the desired barrens and loadings are determined using a McCabe-Thiele diagram reminisCent of that described earlier for adsorption of gold on ,-et copper and sulfuric acid activated carbon. Some of the important variables include concentrations in the feed solution, extractant type and concentration, and the effects of a nearlyinterminable list of impurities. Once the solvent in a shakeout test has been fully loadad. the aqueous phase can be drained offand replaced by a portion of strip solution. The funnel is shaken to equilibrium again, and the m e type of information is obtained for the stripping circuit. By adjustment of phase ratios, the copper is not only transferred to the strip solution but concentrated to make good electrowinning feed. A rough idea of extraction and stripping Linetics and phase separation behaviors can also be gathered in shakeout tests, but more accurate numbers along with several other measurements and observations are available only by operating a continuous c h u k In the early days of SX-EW process development, nobody knew how the solvent would hold up through hundreds of cycles or what types of impurities would be concentrated and where they would accumulate. Crud formation was always a scary and unpredictable phenomenon, and it was found to be a function not only of the chemistry of the system but of mixer dynamics. Also,continuous as opposed to batch mixing and settling led to short ckcuithg that affected the number of stages and he settler areasrequired. Electrowinning of copper from sulfuric acid solutions was somewhat better understood than solvent exmaion in the late 60s and early 70s when initial SX-EWwork was being done, because of abundant electrochemical literaavt and also experience with electrmfining of copper. It remained to be d e m o n s d , however, that the electrowon cathode purities would be competitive with those from electrorefining. This was very difficult to do on a small scale; full-size

267

commercial cathodes were necessary for total credibility. Small electrowinning cells were occasionally used, however, to close the circuit in solvent extraction pilot plants. Electrowon copper was eventually shown to be superior to electrorehed coppa. Ruities of 99.999% (“five nines”) are routinely achievable today, and in at least one case six nines has been reported. After the first commercial plant was consQuctbd and opaated. the principal need for improvement of the system was development of better extractants. Solvent extraction of uranium and other metals was well established. and the engineering of plants was less of an issue than the relatively poor perFormance of the copper extractants then available. Comparatively speaking, the copper reagents exhibited slow extraction kinetics, poor phase Separation characteristics, and a high sensitivity to the acid concentdon in the prepant leach liquor. Much reagent development has since been done, and the COQper extractants available commercially now are far superior. Other important advances have included improved mixer design to reduce solvent entrainment losses in the raffinate (barren aqueous solution). improved settler design to reduce area requirements, and (as mentioned above) superb cathode quality. Many copper solvent extraction plants have been built, opaatad,and modified all over the world since 1968, resulting in huge data banks. Today the circuits can be designed from careful analysis of representative leach solutions and the specification of a flow rate. As for gold ores, most experimental work is directed t o w d the leaching step rather than the recovery from solution. A few shakeout tests, however, are almost always done to confirm satisfactory extraction and stripping. If unusual impurities are present, or if there is any reason to believe the situation is atypical, some phase disengagement tests or even operation of a small continuous circuit to check solvent degradation is advisable. This work is now commonly done in the labomorks of those who make the extractants, a g n k (formerly General Mills) and Avecia (Acorga) in the United States. Occasionally, copper leaching and solvent extraction pilot plants are erected at the mine site for the purpose of training operators. This has been fwnd worthwhile in parts of the world where the requisite skills are not available in the local labor market. Another interesting reason for performing laboratory or even pilot plant tests of copper solvent extraction systems today is credibility of the company. A small, unknown group might need the additional asmance that their plant will work in order to raise money to build it or to obtain lumpsum bids and performance guarantees fromengineering contractors. A major copper company, on the other hand, can be comedy assumed to be thoroughly familiar with its ores and with the appropriate metallurgical techniques to treat them. ’Ihe principles of solvent extraction, dong with detaikd experimental and design considerations, are contained in Hazen 1985.

SUMMARY AND CONCLUSIONS When contemplating the use of mature technologies like those discussed in this paper, mining companies now typically look to the computer programs in engiaaaing offices rather than to the people and equipment in research laboratories. It is important, however, that knowledge of the original strategies and procedures used to develop the technologies still be understood. One of the first principles of mineral procesSing and extractive nictallurgy is that every ore is different. This is also true for every process water, every leach solution, and every set of plant site environmental constraints. The prudent pject developer appreciates that it is far better to sponsor some relatively inexpensive tests yielding no surprises than to be later presented with really expensive surprises in the commercial plant.

ACKNOWLEDGEMENTS The principal sources of information for this paper other than those above were telephone conversations with employees and former employees of Avecia Metal Extraction Products; Cognis Corporation; Hazen Research, Inc.; Kappes Cassiday and Associates; Kvaerner Metals * ; Minerals Advisory Group; Engheerhg and Construction Division; McClelland Laboratones Mountain States Mineral Enterprises; and T.P.McNulty and Associates. Sincere thanks to them all. Other sources were the author’s personal files, notes,and remembrance.

268

REFERENCES Hazen, Wayne C. 1985. Solvent ExaaCtion. In SME Mineral Processing Handbook, ed. N. L. Weiss, 13-38 to 13-44. New York: Society of Mining Fingkers. Heinen, H. J., Peterson. D. G., and Lindstrom, R E. 1978. Processing Gold Ores Using Heap Leah-Carbon Adsoprion Methads. United States Deparmnent of 'Ihe Interior Bureau of M i Information Circular 8770. McCleUand, Gene. 2002. Bench-Scale and Pilot Plant Tests for Cyanide Leach Circuit Design. In M i n e d Processing PJMt Design Proceedings. Littleton: Society for Mining, Metallurgy, and Exploration. potter, G.orge M. 1980. Mem.ll-Crowe Precipitution of Precwus Metuls by Zinc Dust. presented at The Hydrometallurgy Division Spring Meeting, Arizona Conference, Society of Mining Engineers of AIME Tucson, Arizona,May 10.

269

Guiding Process Developments by using Automated Mineralogical Analysis David Sutherlund and Ying Gu2

ABSTRACT Effective process development of an ore body depends on an understanding of the mineral * 'cs of the deposit. Ore chanrcttristics, such as mineral abundance and ore texture, control the liberation and concenlration of values, and thus tbe economics of the deposit. Quantification of these parameters is now readily obtainable using modem image analysis systems. The paper discusses the applications of automated mineralogical analysis by considering a number of case studies, These include: 0 orecharacterisation 0 mineral process optimisation 0 treatment of tailings

-C

INTRODUCTION The effective developmnt of a minerals treatment process depends on a thorough u & n t a d h g of the characteristics of the materials being treated. If the properties of ores are better understood and reliably measured, then efficient processes can be designed and the m i d treatment can be optimized. Quantitative mcastnes of orc and mineral properties coupled with an understanding of the influence of the measured parameters on process operation can greatly improve the speed and efficiency with which a resource can be exploited. Improved characterization has a number of aspects. It can mean better ways of measuring the mineral characteristics of ores and process products using modcII1, automated equipment (e.g. LEO QemSCAN (2001) or JK Mineral Libaation Analyscr (Gu,1998)). It can also meen a bemr understaoding of the influeace of these measured parameters on process performance (e.g. Sutherlaad and Goalieb, 1991). Finally it can mxn improved methods for manipulating and presenting these mineralogical data (e.g. Gonlieb ct al. 2000). Considerable advances have been made in each of these aspects of characterization. Mineralogy is maditionally applied during periods of exploration and ore characterization but is often largely neglected at later stages in process developmnt. Nevertheless there are oppoatunities to improve the process at all stages through to process optimization. The effectiveness of testwork must be judged on the basis of mineral separations and the reasons for unsatisfactory proctss performance can often be identified only by a full mineralogical C h a n r c t e *n u u'on l of the process feed and products. Unclers-ng the liberation of values is basic as is knowledge of the distribution of values between the minerals within the ore. Tracking of losses in tailings and contamination in concentrates must all be done from a basic understanding of the mineralogy involved. In a similar way process developmnt and optimization based on operating plants uses these sanx techniques. A thorough audit of an OpMaCing plant should also include a mineralogical balance. Development of tailings meatment systems must account for the mineral characteristics of the material being treated and their long-term stability in a tailings

' CSIRO, Pinjarra Hills, QLD 2

JKMRC, Indooroopilly,QLD

270

impoundment. Mineralogy is needed to guide decisions at all these stages of process development. The following paper reviews all these aspects. It outlines the information now available from automated mineralogical equipment. the use that can be made of such information for process development and gives several case studies to demonstratethe techniques.

MINERALOGICAL DATA Henley (1983) has given an excellent and compnhensive review of traditional ore dressing mineralogy before the significant developments were made with automated systems. This covers much of the important basic information on sampling and measurement accuracy that is common to most applications. As with all metallurgical experiments the most fundamental thing to consider is the collection and subsequent treamrent of the samples. Needless to say it is critical that the sample studied is representative of the material being investigated. Mineralogical sampling is no different in this, but the requirements are particularly stringent since the amount of material finally measured is oftcn very small indeed. The collection and splittingof representative samples will not be discussad here, but is now well understood. ' b r e are particular sampling matters relevsnt to automated mineralogy using image analysis techniques. Tbe extraction of reliable quantitative data assumes that the image to be analysui is a random section in space from the mineral assemblage. Tbe usual approach to this is to randomly mix particulate material in an epoxy resin then cut and polish the resin block after hardening. The exposad surfaces of the particle sections will be random sections so long as the particles are randomly positioned in the block both in location and orientation. The major difficulty arises from the fact that the mixing and setting of the epoxy is done in a gravity field. This may lead to sedimentation, segregation and preferred orientations for multiphase particles. particular care is needed to ensure that a random section is analysed. A simple check to conreliability of the sample Preperation is to compare the true chemical assay against a calculated assay estimated by assuming chemical compositions of the mineral phases identified by volume in a modal analysis (LEO QemscAN (2001)). The basic mineralogical information that can be measwed is:

0

Mineralmodalanalysis Mineral grain size, shape, surface anxi, etc. Mineral associations and ore ''texture" Particle liberation

0

Rare/Tracemineraloccurrence

0 0 0

Modal analysis is the simplest and most reliable masunmnt provided that the mineral identificationis sound. What is seen on the Section truly represents the volume and estimates of mineral proportion can be done by summing points, lengths of line sections, or areas of area sections. These estimates are unbiased which means that in principle by increasing the measurement time estimates of inmascd accuracy can be made. Confidence limits for these estimates are also available (Henley 1983). Estimates of particle and mineral surface area are also u n b i i and generally reliable. The steraological basis for such estimates is sound and based on lengths of transects in line sections or ratio of perimeter to area for area Sections. From a practical viewpoint there is an added problem though, since such estimates are more sensitive to phase mis-identificationat a pixel point. Occasional e r r o m identifidons always a factor when using an automatic system - though they have negligible impact on the modal analysis can bias area measurements. The reason is that such errors when they occur will cut line intersections thus increasing the estimated surf= area for the phase in question. Identification problems are also more likely to occut at phase boundaries when the automatic instrument attempts to identify a mixed phase. This becomes less important if the x-ray points are moved away from phase boundaries as guided by BSE image analysis.

-

27 I

Si estimates are less swaightfonvard.The problem is the irregular shape of the particles and mineral occurrences. Volumes may be reliabiy estimated,but some model of the particle is needed to translate this into a linear size. Many possibilities exist, sometimes based on an assumption of a simple geometric shape or else on the empmd use of measured shapes for actual particles. (e.g. Barbery 1991, pp 27 - 34). The more general question of mineral "texture" is much more complex again. A type of covariance function for the different phases contains all the relevant information on the geomtrical relationships between the phases, but it is not easy fix the metallurgist to interpret or use (e.g. Barbery 1991, pp 37 - 51). A simple subset of the full association information between the phases is given by the estimation of touching phases. This is reliably measured using s w k e area estimates, but is sensitive to phase bomdary identifications, always an area of concern for an automated system. Texture quantification remains an active area for research. Particle liberation meamemem are of particular interest to tbe metallurgist since they determine the mineral seperation that can be achieved. Image analysis provides the only method of measuring liberation for complex particles. What is q u i r e d is an estimate of the disbibution of particle concentration. A fwthex problem arises with liberation due to the stereological bias introduced by the sectioning. The observed libemtion in the section is always greater than the me volumetric value (e.g. Spencer and Sutherland, 2000). Methods have been developed for estimaaing the extent of the bias but it remains an active area of research, particularly for multiphase systems. A special type of liberation measurement relates to qwIi6cation of d r r a c e mineral occurrence. It is a challenging task to collect statistically s-ti data on the liberation Charectensb. . 'cs of low-grade minerals, such as gold and platinum group minerals in the feed and other process saeams. Sophisticated image analysis is used to first locate possible candidate panicles and then perform normal libemtion analysis on them. IMPORTANTSTAGES IN PROClES DEvEulpMENT The efficient exploitation of a mineral deposit pmcecds through a number of stages: 0 0 0

0

Ore characterizationat exploration and later during production planning Ore testing for process development and plant d e s i i Plant control, plant audits and optimization Tailingstreatment

At each stage automated process mineralogy can provide data to guide the development. Initially it permits ore types to be defined and estimetsd in a more quantitative manner. Poor ptospects may be rejected at an d i e x stage and e f f m directed towards more likely targets.

Similarly once an economic lode is established the proddon schedules may be soundly based on a more thorough knowledge of the characteristics of the plant feed on a daily basis. The basic task of optimising the efficiency of mineral Separation, whether during ~aboratory testing or plant investigation relies on knowledge of the important metal ocamences in the different mineral species. This has several aspects: 0 0

0

0 0 0

Are the important elements present in one or more minerals? Are there elemental associations within a single mineral phase that preclude a separation? Are there mineral associations within particles that preclude a separation (poor liberation)? How nearly is the Separation to "ideal"?

Whataretheformsofthetailingslosses? What are the forms of the contamination of the concentrates?

272

Questions like these have to be addressad in order to optimize the process. Automated process mineralogy aims to give quantitative answers to such questions.

CASE STUDIES Some case studies will be discussed briefly to demonstrate some of these aspects of process development. orechu8cte~tion

During the explorationand assessment of a new deposit a great deal of time and effort is spent obtaining samples, generally h m drill cms, that are representative of the overall resource. It is important to gain as much information as possible fkom these samples. The total metal content is important but it is the mineral compositionand textural relationships that will determine whether or not the deposit can be developed economically.Textural factors distinguish an ore that is easy to treat (e.g. Broken Hill-type deposit ) from one that is very difficult(like McArthur River - see The Sir Maurice Mawby Memorid Volume, 1993). Differencescan be seen by the mineralogistduring qualitativeassescrment, but quantification of such characteristics can be very useful to put mon confidence on these more subjective m a s . A simple study was made which related the performance of a range of Australian mineral concentratorsto the mineralogy of the feed ore (Sutheriand et al. 1991). Samples of the crushed fead material were taken for e neralgrain size analysis and the concentrator performance monitored. Reasonablecorrelations were then obtabcd between the major processing parameters, such as grade, recovery and flotation concentme particle size and the grain size measurement.This has led to a simple way to rank ores based on grain size measurementsof the major minerals present. The measurements also give modal analysis of the ores and mineral associations thus enabling the elemental deportmentto be dehed. This can be impOrtaat with copper and nickel ores where several minerals contain the target metal but the ease with which these differentminerals can be recovcrtd may be very variable. The assay gives the total metal content but the mineralogy teils how this metal occurs in the ore. Such ore characterisation sndies have been CBIfiCd out for many ores. An example of the detailed investigation of a complex mineralogical system is given in the study of silver mineralogy of some silver ores (Wilkie & Bojcevski, 1996). Tbe deporanent of silver between a range of mineral phases was important in assessing the ease of recovery of silver from the differentore types.

. .

M i n e r 8 l p r p a s s ~ When ore testing and plant design are underway, it is important that detailed mineralogical studies are also made to minimise the amount of testhgrequbed. dre h e needed and the risk involved in the resulting technical decisions. Ore charactensatl * 'on is the basis on which the final selection of ore types and samples is made, but then more detailed testing is needed to give the quantitative answers for detailed design. Lsboratory testing of samples requires mineralogical examinationto establish optimum conditions for processing. The first point to establish is the flotation feed size required to give suitable liberation. Early estimatesof grain size will give guidance for initial testing, but liberation prediction is extremely difficult so that liberation measurements at a range of flotation feed sizes are generally needed to establish best operating conditions. Similarly after flotation the questions of dilution in concentrates and losses in tailings genedy require mineralogicalinvestigation. Liberation measurements can quickly determine whether losses are due to poor liberation or poor Separation. Further testing will be guided towards a change in grinding conditions or alterations to the reagent regime.

273

lEum#idgrade recovery curves can be generated early in the testing stages to assess the efficiency of later separations.In a review of process mineralogy,Batteirham et al. (1992) gave results of tests where the liberation measurementsat a range of flotation feed sizes have been used to calculate theoretical gmdekmvery curves for the ore. These were then compared with results f h m laboratory flotationtests with good agreement,The liberation results are used to define the curve that represents the best possible Separation, which can be obtained from this material when it is ground. The curve is calculated by accumulating particles progressively from the liberated end of the particle concentrationspectrum. This result can only be achieved with a perfect separation. while any inefficiencyin a real process will lead to a worse result. Stemlogical effect may also con@ibuteto the Mence between the theoretical and measured grade recovery curves.

..

are still difficult and remain an area of active research so that experimentationis generally needed. Liberation measurements at Merent flotation feed sizes are used to estimate the optimum flotation feed size for a test sampk.In practice a single grinding test can be used to generate a range of size tiactions that, as a first approximation, can be used to represent liberation of the matgisl at these sizes. The assumption made here is that liberation is a function of particle size only and daes not depend on the breakage path followed to reach that size. If selective breakage occurs there will be some deviations from this assumption but in general the deviations are not significantenough to be a problem. For, example, a non-preferentialmodel (Morrell et al. 1993a. 1993b)can successfully describe the grinding of a gold ore and fine grinding of a zinc ore.

-P-

T a t o p t h d l on of am operat& phnt is the final objective in d l u r g i c a l terms. The aim is always an increase in profitability either through improvementsin throughput or conumration efficiency. Usual targets for investigation are the losses of valuable minerals in the tailing and the dilution of concentrate by unwantedgangue minerals. In both cases detailed mineralogy of products is required in reaching the correct solution to a problem. An example of the analysis of a copper regrind circuit is given on the QemSCAN web page (LEO QemSCAN (2001). The results showed clearly that to improve overaU recovery, it was important to improve the recovery of the finer size fractions.This would involve improving flotation efficiency for very fine liberated size fractions. Furhemme, since most of the losses were liberated material, the results suggest that a review of the re-griad procedure should be made in an attempt to reduce over grinding. "lw aesc of diiution ot a -trate is often a simple question of poor liberation or else of poor sekctive flotation. Simple mineralogicalmtesItceMLtscan generally determine this in a straightforwardway. Locking of values with gangue or fkc parricla of gangue in the concentrate reveal the story. But other more subtle effects may be at work. Coating or rimming of one mineral by another can have a profound effect on the separation, which can be achieved in a flotation operation. Flotation is a surface process and flontability can be affected by surprisingly small concentions of a floatablecomponent.Exposure of trace quantities of floatable minerals when they pferentially occur on the particle surfaEe can lead to poor separation (Gottlieb & Adair 1991).

Tk aidllJaPtiaaoflosses in tailing and the optimisation of liberation and flotation feed size are common aims in plant studies. A typical study by Frew and b v e y (1993) was undertaken for a complex sulphide ore plant. A careful Sampling campaign with metallurgical balancing on a size-by-size basis including liberation is the starhg point for an assessment of the performance of the circuit. This is a very significant task but essential for a sound analysis of operaring conditions. The study shows how a liberation balance for a circuit can be used to aack the flow of composite particle and assess the flotation behaviour of the different particle classes. This base data is needed before changes to the circuit can be proposed to address problem areas. Once again, qtmntitativem i d @ c a l measurements are hnportant in guiding the decisions.

274

’IberegIlhrrrscofrmincnkgyis an esstntial part of the recording process for plant operation. As well as using the information in a predictive way wbcn ore changes are expected, itcan be usedintheaudikcording process to give a long-term view of plant operations and plant changes (seeLEO QcmSCAN(2001). The change in ore types with time c a be ~ observed ~ by consi&ring the modal analysis and liberation over time.Changes to the plant performance can then be assessed as due to opaational causes or ore changes. Too often a change of ore type is given as a rcllson where regular monitoring of the operation can pinpoint the real cause. Such regular use of mineralogical analysis on monthly period samples is also useful for assessing the longer-effects of major plant modific8tons.Griffin et al. (1 993)show the effects of a change to autogewus grinding in the copper concentrator at Mount Isa Mines Limited. The measured liberation obtaioed firom the grinding circuit over about two years was shown against what was predicted. Such effectsare not easily assessed unless there is a sound history of plant operation for reference. ~

t

i

o

n

o

t

~

The behaviour of tailings over a long period is of the utmost importance when a disposal system is designed. Initially the mineral composition of the minerals must be determined so that the expected leaching characteristics of the tailings can be assessed. Mineralogy will determine whether acid will be generated or neutralized and whether toxic elements are present in a leachable form. The ease with which tailings can be stabiised and the potential for problems with leaching out of unwanted materials h m dams is also dependent the fineness of the texture. As processing technology improves to enable very fine-grained ores to be treated, the problem of stabilisation and rehabilitation for the very finely ground tailings becomes ever more important. Sutherland and Richards (1997) have attempted to consider the impact of such fine grinding on the economics of tailings lrcatment. The possible effects of very fine texture on processing and ultimate disposal prompted this sndy. The conclusion reached was that the fine grinding needed for finely textwed ores did significantly raise the difficulty and cost of establishing tailing dams, but that the costs were well below the extra revenue generated from the added recovery of values. It was important to average this cost out over the life of the mine rather than leave it entirely until the end of the mining cycle. A further use for quantitative mineralogy is the “forensic” s a d y of effects from previous mining operations. Disposal of tailings into rivers and coastal systtms can sometimes be traoed through a careful examination of sediments, even centuries after the mining has ceased (Perry 1999). Similarly the impact of ocean dumping has been studied (Riley et al. 1989). Mineralogical study at the design stage can be useful in planning for the overall life of a mine. The composition and quality of the tailing can be calculated and the mineralogical status of residual heavy metals used to estimate the likely stability of important elements. The intgaction of the process with the m i d is again critical. Tao often tbese factors are not considerad until very late in the mining cycle whereas they could be assessed early on the basis of measured properties of the feed ore and the proposed process.

DISCUSSION Mineralogical analysis is critically important for ptocess developmnt. Its benefit is widely recognised in the mineral processing industry. However, its application has been limited in the past, largely because of the cost and slowness of acquiring data. Recent advances in the development of automated mineral analysis systems are bringing significant improvements. Modem systems will not only generate data faster, but also at higher accuracy and resolution. Advanced image analysis methods are used to automate the de-agglomeration procedure (i.e separating particle sections touching each other). and x-ray psttern recognition and neuronetwork technologies are used to identify mineral species. Specialised software packages have now developed to allow metallurgists to view and to internogate mineralogical data at ease.

275

Active research is also under way to accurately model the liberation during the grinding process and to model the flotation process based on liberation data There is no doubt that future process development will be greatly assisted by these developments.

CONCLUSIONS The characteristics of the m;nerals in an ore control the processing characteristics required for efficient extraction. Modem automated measurement systems arc now available to provide this information. If care is taken with sampling, sample preparation and measurement, accurate and reliable data can be obtained. Some techniques have been developed for the manipulation and presentation of this data to assist the metallurgist obtain useful results in process development. More work is neaded in this area to extend our understanding of the effectsof mineralogicaldata on process behaviour.

ACKNOWLEDGEMENTS The authors would like to acknowledge the conmbutions of their many colleagues in CSIRO and JKMRC who have contributed to the development of the systems, carried out the case studies and advanced the science of Process M i o g y .

REFERENCES LEO QemSCAN h m J / w w w . a.e m accessedOct29,2001 Gu, Y.1998 “Rapid Mineral Liberation Analysis Using the JKMRUPhilips MLA, Mineralogy for Mineral Processing Engineers Workshop, M i s Processing’98, Cape

Town. Gu, Y.,Gay, S.,Guerney, P.and Napier-Mum, T. 1998 “Measuring and Modelling Mineral Liberation with the JKMRUPhilipsMLA”,Proc Minerals Pnxxssing’98 Conf. Cape Town,1998.0R26. Sutherland, D N and Cottlieb. P, 1991. “Application of automated quantitative mineralogy in mineral processing,” Minerals Engineering,4 pp. 753 - 762. Gottlieb, P,Butcher, A R,Ho Tun,E and SutherW D N,2000 “Applications of automated process mineralogy,” Proaxdm * gs ICAM2000, Wttingen, Germany, Vol 1 p. 13-06. Henley K J, 1983 “Ore dressing mineralogy - a review of techniques, applications and developments”ICAM 81, Spec.Publ.Geo.S.Afr., 7 pp 175 200. Barbery, G,1991. “Mineral Liberation” (Les Editions GB). Spencer, S and Sutherland, D, 2000. “Stere.ologiCal correction of mineral liberation grade dismbutionsestimated by single sectioning of particles” Image Anal Stereol, 19 pp 175 182. Australasian Mining and Metallurgy: The Sir Maurice Mawby Memorial Volume (Second Edition) 1993, ed J T Woodcock and J K Hamilton (AusIMM) Vol I pp 507 & 591. Sutherland, D N, Gottlieb, P,Wilkie, G and Johnson, C R, 1991. “Assessmentof ore processing characteristics using automatad mineralogy,” MI International Mineral Processing Congress, Dreden, pp 353-362. Wilkie, G and Bojcevski, D. 1996. “Applied mineralogy as a diagnostic metallurgical tool for silver bearing base metal sulphide ores,” Sixth MU Operams’ Conference, M e h n g , pp 199 203 (Ausrralasian Institute of Mining and Metallurgy). Banerham,R J, Grant,R M and Moodie, J P, 1992. “A perspectiveon process mineralogy and mineral processing,” in Proceedings of rhe First International Conference on Mudern Process Mineralogy and M i n e ~Processing, l Beijing, pp 3-12 crhe Nonferrous Metals Society of China, Internatonal Acsdemic Publishers). Morrell, S., Stems, U.J. and Weller, K.R. 1993a‘lkapplication of population balance models to very fine grinding in tower mills”, proceedings of XVIII International Mineral Recessing Congress, Sydney, pp61-66.

-

-

-

276

Morrell, S.,Dunne, R.and Finch, W.1993b“Ihe fibemion performanctof a grinding circuit treating gold-bearing on”,Rocaadings of Xwr International Mineral Processing Congress,Sydney, 1993, pp197-202. Gotdieb, P and Adair, I, 1991. ‘Quantification of talc rimming of c b m i t e s in Bushveld ores using QEM*SEM,” Proc. ICAM ‘91, Retoria Fmw, J A and Davey, K J, 1993. “‘Effect of liberation on flotation pedormance of a complex ore,” XVIII Inremational Mineral Processing Congress. Sydney, pp 905 - 9 11 (Australasian Institute of Mining and Metallurgy). Griffin, L K,Hart, S,Espinosa-Gomez, R and Johnson,N W,1993. “Chalcopyrite flotation and liberation charamristicsbefore and after autogenous grinding at Mount Isa Mines Limited,”XVIII InternatioMl M i n e d Processing Congress, Sydney, pp 913-922 (Australasian Institute of Mining and Metallurgy). Sutherland, D N and Richards, B R, 1997. ‘Ihe impact of millingpractice on tailings rehabilitation,” XX International Mineral Processing Congress, Aachen, Vol5, pp 59-70 Perry, D,private communication. Riley, S J, Creelman, R A, Warner, R F,Greenwood-Smith, R and Jackson, B R, 1989 “The potential in geomorphology of a new mineral identificationtecbaology (QEM*SEM)”, Hy&Obiok@ 1761177 p~ 509 - 524.

277

Guidelines To Feasibility Studies John Scott’, P. Eng., and Brian Johnston2, P. Eng.

ABSTRACT Feasibility studies are used for project evaluation at all stages of project development, from initial exploration to arranging project financing. The content and format for scoping studies, preliminary feasibility studies and final (bankable) studies are presented and discussed. The use of risk analysis in assessing project robustness is also addressed, particularly for bankable studies.

INTRODUCTION All mining projects pass through a series of stages over the project life: 0

0 0 0 0 0

Exploration Discovery Investigation Development Production Abandonment

Project evaluation normally accompanies all the earlier stages of a mining project and provides the basis and justification for continued investment of money for the proposed development. The evaluation process culminates formally in a final feasibility study, which will demonstrate the economic feasibility of the project with the degree of certainty required to make the investment decision to bring the mine into production. The final feasibility study is preceded by other evaluation studies, sometimes called “prefeasibility studies”, that reflect the increasing level of technical and economic knowledge gained at the various earlier stages. To be more exact in the description and naming of project evaluation studies, they should be called: 0

0 0

Preliminary Evaluations Preliminary Feasibility Studies (Intermediate Technical-Economic Studies) Final Feasibility Studies

Effectively, preliminary evaluations of a potential mine will commence prior to the initiation of the exploration program leading to the discovery of an economic orebody. The establishment of target values for the grades and tonnages of ore required to make a project attractive for exploratioddeveiopment for specific metals in specific areas of the world is based on the preliminary evaluation procedure, and is the basis for committing risk capital to an exploration program, whether “grass roots” or connected with an existing mine.

’ Director, Technology, Fluor Daniel Wright Ltd., Vancouver, B.C., Canada

Manager of Financial Analysis, Fluor Daniel Wright Ltd., Vancouver, B.C., Canada

28 1

Following the discovery of a potential orebody, a more specific preliminary evaluation of the deposit will be carried out to justify continued exploration. The exploration or investigation stage will continue until sufficient information is available to undertake a preliminary feasibility, or intermediate technical-economic, study. The preliminary feasibility study will normally consider a number of project development alternatives related to mining methods, production rates, processing schemes and infrastructure needs. The basic economic feasibility of the project will be demonstrated at this phase, and specific requirements for further definition drilling, metallurgical testwork and site investigations to support a final feasibility based on the most attractive alternative will be determined. The investment required to proceed with the final feasibility study will be justified and quantified at this stage so that an appropriate budget and schedule can be established. The final feasibility study will lead to a production decision, project budget approval, appropriation of funds or arrangement of project financing, and final design and construction. It must be emphasized that the preliminary evaluations and preliminary feasibility studies are iterative in nature and are related to the direction and continuity of the investigation of the deposit. Each evaluation or study will be based on the information available at the time the study is carried out and will of necessity examine and reexamine the same basic information: 0 0 0 0

0

the geology, mineralogy, geography and infrastructure of the deposit the technical aspects of mining, processing, transporting and selling products the marketing plan for the proposed products and the determination of a net selling price the economic evaluation of the proposed project at an appropriate level, including capital and operating costs, revenues, taxes and royalties the current assessment of the attractiveness of the project in terms of sufficient potential profitability to justify continuing the development process.

Clearly the final feasibility study and the decision to proceed to production are really, then, the last (and most detailed) of a long series of economic evaluations. In the following sections, each level of study is defined more fully and the estimating procedures and techniques normally used for each type of study are presented. In general the discussions reflect a typical new project being developed from the discovery stage, although the principles and procedures are equally valid and applicable to evaluating expansion projects or for reviewing or auditing other reports.

PRELIMINARY EVALUATIONS Preliminary evaluations are usually carried out very early in project planning and may in fact be used to develop exploration targets and strategies. Normally the first evaluation study of a deposit is used to justify further expenditures on exploration drilling, geophysical or geo-chemical programs, or perhaps for acquiring an option on a prospect held by someone else. At this early stage the amount of investment committed by a positive decision is not usually too large. The purpose of the ensuing work is generally to gather additional geological information for another preliminary evaluation based on a better definition of the mineral resource. Considering that these first evaluations will be used primarily to justify a further exploration effort or to identify and choose between competing exploration programs or property acquisitions, it is apparent that most of the evaluation will be based on assumptions. Information Base As noted in the introduction, the information available for the earliest of preliminary evaluations is normally minimal, but presumably there will be some tangible indication of a potential orebody. While this indication can be as “intangible” as the exploration manager’s “hunch”, based on an experienced assessment of the exploration potential of the area, more often some specific information such as a gravity survey, satellite photographs, outcrop samples and assays,

282

geochemical assays, sample trenches, geological mapping, core from between one and ten drill holes (exceptionally), or even a metallurgical test, will exist. The general nature of the area will be known with respect to the geography and the existing infrastructure such as roads, railways, towns or cities. The taxation, environmental and other governmental regimes and laws for the particular country and region will be known in general terms, but specific issues relating to land tenure, ownership, labor laws, etc., will not have been investigated in detail. The ore reserves for the deposit will not be known; however, a potential tonnage will be assumed, which will form the basis for assuming a mining production rate and estimating the quantity of final product to be produced. Similarly, the metallurgical characteristics of the ore will generally be unknown, and therefore a typical flowsheet and metallurgical balance will be assumed. The range of selling prices for the concentrates or metals produced will be known, and an assumption will be made that the product from the mine can be marketed within reasonable expectation.

Cost Estimation for Preliminary Evaluations In common with all other levels of preliminary and final feasibility studies, the basic purpose of the preliminary evaluation is to assess the potential economics and thus the value of a particular project. To permit any kind of economic analysis, some estimate of the capital and operating costs of the mine development has to be made. There are two basic methods of generating such costs, namely, comparison with similar projects and direct estimation. Comparative cost estimates normally rely heavily on rules of thumb and the extrapolation of cost information from similar projects to the one in question. A number of problems are associated with this particular kind of estimate, however, due to the unique aspects of mine location, the unreliable effects of inflation on reported costs and the inherent differences in projects that may appear to be very similar. The most reliable method of estimating costs, even at the earliest stage, is to use a combination of comparative costs and specifically estimated costs. Clearly, as the project advances more and more of the costs will be estimated directly for the particular job. Capital cost estimates are generally understood as being of various levels, or classes, based on the fraction of total engineering completed and the availability of project-specific cost information. These classes of cost estimate are well defined in the industry and for the purposes of this presentation are listed as follows: 0

0

0 0

Class I Preliminary Evaluation: “Order-of-MagnitudeEstimate” “Capacity Factor Estimate” Class I1 Intermediate Economic Study: Preliminary Feasibility Study, “Equipment Factor Estimate” Class I11 Final Feasibility Study: Bankable Feasibility Study, “Forced Detailed Estimate” Class IV Project Control Estimate: Definitive Cost Estimate, “Fall-out Detailed Estimate”

A “capacity factor” estimate is generally the kind of comparative estimate that would be used for the earliest technical-economic appraisal of a project. Once the major equipment is well enough defined to permit meaningful pricing, an “equipment factor” estimate can be derived for a preliminary feasibility study. The “forced detailed” estimate in most cases would be adequate for a final or bankable feasibility study. A “fall-out detailed” estimate is normally performed once project approval has been received and the detailed engineering is 30% to 40% complete. It is then used as a project control estimate for final construction or as a basis for soliciting fixed price or turnkey bids for a project. The normal engineering requirements for the various levels of estimate are listed on Table 1, with the related time phasing for these levels shown graphically in Figure 1. Table 1 and Figure 1 are applicable to all types of industrial/chemical/metallurgicalprojects and are somewhat general

283

in nature. A more specific list of the information normally available for a mining/metallurgical type of project and the engineering requirements for the various cost estimate levels is provided in Table 2 (McQuat, 1992).

Table 1 Information requirements for capital cost estimates

GENERAL Project Scope Product, capacity, location Project schedule

X X

ENGINEERING Process flow diagrams Mechanical flow diagrams Process data sheets Plot Plan Equipment list Design criteria Line list Equipment specificationsand vessel sketches Utility requirements Soils data Design specifications(all accounts) Sewer and paving layouts Concrete, steel & building drawings/sketches Piping drawings/sketches:alloy large dia. C.S., special fabrication Motor list Single-line wiring diagrams Area classification (electrical) Electrical equipment specifications Conduithable schedules Electrical design drawingsllayouts Instrument list Insulation schedules (equipment and piping) CONSTRUCTION Labor productivity Labor wage and benefit rates Temporary facility requirements Field staff requirements Equipment/tool requirements

X X

X X

X

P'

X

X

X

X

X X X

P P P P

X P P X P P P P P

P P

P P P X P

X P

P P P

P P

P

X X X X

X

X

X X X X X X X X

X X X X X X X X X X X X X X X X

Table 2 also summarizes the basis for capital and operating cost estimates and financial analysis for the different levels. Figure 2 provides a correlation between estimate accuracy and project definition. It can be seen from Figures 1 and 2 that only 1% or 2% of the total engineering effort for a project is required for Class I preliminary project evaluation estimates. These can range from a I

P = Preliminary

284

true “order-of-magnitude” capacity factor estimate, based on only a project scope and the assumed products, capacity and location, to a combined type of estimate based on some equipment factor and some takeoff costs, depending on the level of information available. At the preliminary evaluation stage, operating cost estimates are almost invariably factored from costs at other operations, with the factors adjusted for the location, local wages, taxes and transportation costs. The capital and operating cost estimates for a preliminary evaluation are normally prepared by experienced senior engineers with specialized knowledge of similar projects.

100

90 80 70

zm

& 50 El

a 8

40 30 20 10

0

: T l I c I .I 10

20

30

40

50

60

I

I

I

I

70

80

90

100

% Engineering & Design Duration

Figure 1: Time phasing of estimates

Order of magnitude estimate

x

?i

.u c 0

s

40/

30

0” 20

Ea

4

3

0

2 a

Preliminary Feasibility Estimate Bankable Standard Definitive Estimate Mechanical Completion

10

0 Project Definition

Figure 2: Estimated accuracy versus project definition

285

Table 2 Engineering requirements by type of estimate (McQuat, 1992) Item Site Plant capacity Geographical location Maps and surveys Soil and foundations tests Site visits by project team Process Process flowsheets Bench-scale tests Pilot ulant tests Energy and material balances Facilities Design Nature of facilities Equipment selection General arrangements, mech. General arrangements, struct. General arrangements, other Piping drawings Electrical drawings Specifications Basis for Capital Cost Estimates prepared by Vendor quotations Civil work Mechanical work Structural work Piping and instrumentation Electrical work Indirect costs ContingencJ Operating Cost Labor rates Labor burden Power costs Fuel costs Expendable supplies Reagents PaaS Economic Analysis Use of Estimates

Class I Order of Magnitude

Class H Preliminary Feasibility

Class HI Bankable Standard

Class IV Definitive

Assumed Assumed None None Possibly

Preliminary General If available None Recommended

Optimized Approximate Available Preliminary Essential

Finalid Specific Detailed Final Essential

Assumed If available Not needed Not essential

Preliminary Recommended Recommended Preliminary

Optimized Essential Recommended optimized

Finalized Essential Essential Finalized

Conceptual Hypothetical Sometimes None None None None None

Possible Preliminary Minimum Outline Minimum None None Performance

Probable Optimized Preliminary Outline Outline One-line One-line Gedmajor equip.

Actual Finalized Complete Preliminary Preliminary Some detail Some detail detailed

Project Engineer Previous Rough sketchlprev. % of machinery Rough sketcwprev. % of machinery $ per kwlprev. % of total

Sr.Estimators Single source Drawing est. 76 of machinery Prelim drwgs 76 of machinery S per kW 76 of total

Sr. Estimators Multiple Drawing est. ManMtonne Take-off/tonne Take-off Take-off Calculated

Est. Dept. Competitive Take-offs ManMtonne' Take-off/tonne' Take-off' Take-off' Calculated

20-2576'

15-20%'

15%'

10762

Assumed Assumed Assumed Assumed Assumed Assumed Assumed If meaningful Comparisodreject. Preliminary eval.

Investigate Calculated Actual Verbal quote Verbal quote Verbal quote Verbal quote If required Preliminary feasibility

Get contracts Calculated Actual Letter quote Letter quote Letter quote Letter quote Required Final feasibility

Get contracts' calculate' Contrac? Contract) Contrac? Contrac? Letter quote If requested Project control

Content and Structure of Preliminary Evaluation Reports The preliminary evaluation is carried out when there is little or no concrete information available with respect to the quantities of ore reserves and other important factors. The evaluation considers what could be with respect to the project, rather than what is actually known, and the fundamental purpose of such an evaluation is to determine whether it is worth pursuing what could be. As the basis of evaluation is largely assumption, one would normally tend to be somewhat optimistic, but not unreasonably so. If despite the inherent optimism the economic evaluation is Often subject to subcontract bids

* In this definition the percentage assigned to contingencies is a judgement factor and is not to be interpreted as meaning that estimates are necessarily accurate within this percentage range, nor is there an implied reference to any order of accuracy. Contracts can be solicited if project is near-term.

286

unsatisfactory, then work on the exploration program, option agreement or further study can be stopped. The report itself must be concise, readable and understandable by non-technical people. Reports of this type will normally have a simple table of contents and be well indexed. A typical table of contents for a preliminary evaluation report is as follows:

-

Preliminary Evaluation Report Typical Table of Contents SECTION 1 - INTRODUCTION SECTION 2 - SUMMARY SECTION 3 - MINERAL RESOURCE SECTION 4 - MINING AND PRODUCTION SCHEDULE SECTION 5 - METALLURGY SECTION 6 - CAPITAL COST ESTIMATE SECTION 7 - OPERATING COST ESTIMATE SECTION 8 - FINANCIAL ANALYSIS The content of each section of the report is briefly outlined: Introduction. Describes the location and scope of the project, what is proposed to be done, how it is to be done and the potential market for the products. The purpose of the report is also defined, whether for an exploration budget expenditure, for acquiring an option or to justify a more detailed study. Summary. Summarizes the rest of the report, chiefly in tables and figures. Normally would specifically include any qualifications, reservations about reliability of information and, finally, recommendations for further work. Mineral Resource. Includes the following information: history of the property, geology and structure, geophysics, geochemistry or other exploration methods, mineralization and mineral deposits, sampling methods and types of samples obtained (if any). Mineral resource quantification will normally be a very preliminary estimate of the potential volume of both a geological and a mineable resource based on estimates of the size and grade of the deposit and sensitivity to various cutoff grades. Mining and Production Schedule. Describes the mining method, equipment selection, mining schedule (ore and waste) and any potential expansion. Metallurgy. Briefly describes the process, together with preliminary flowsheet and metallurgical balance. Also provides estimated product grades and tonnages. Capital Cost Estimate. Tabulates the cash cost of mining, milling, transportation, ancillary services, general and administration costs and other treatment charges. Financial Analysis. Provides a description of the type of analysis used (normally a discounted cash flow method), the tax structure, depreciation, financing assumptions and revenue estimates. The project cash flows will be presented on a year-by-year basis, and a variety of investment indicators such as internal rate of return, payback and net present value will be calculated.

Staffrng for Study Preparation The preparation of a preliminary evaluation report will normally involve work of at least three experienced engineers: the explorationist, a mining engineer and a metallurgist. In this report, most emphasis will be on mining, geology and ore reserve estimation. For medium-to large size mining companies, this type of study is almost always done in-house by staff engineers. Smaller companies or junior resource companies will often contract independent consultants or, in some cases, recognized engineering firms.

287

Unique Characteristics Risks. Decisions made at this stage of a typical project will not normally lead to the commitment of a large amount of money. The major risk would be the inappropriate abandonment of a serious mining project. For this reason, it cannot be overemphasized that competent, experienced people should be involved in the evaluation, for both formulating the assumptions and finally preparing recommendations for review and discussions with senior decision makers. Reversed Studies. Most studies rest on physical facts and work from that basis toward cash flow figures. In contrast, it may be helpful, particularly in the early stages of property exploration, to reverse the process. This approach may be of benefit if, for example, the general grade of the deposit is becoming evident, but not the potential tonnage. Or, more commonly, the geology may give scope for wide manipulation of both grade and tonnage by selective mining or cutoff control. In such situations a reversed study may offer useful guidance, usually by helping to define a desirable exploration target. A range of suitably sized operations is assumed, each being required to satisfy some pre-stated financial requirement. From rough estimates of capital and working costs and presumptions of operating methods and recoveries, the annual minimum required revenues can be derived, and on this basis equivalent combinations of tonnage and grade at various metal prices can be calculated. This exercise provides a reference scale against which exploration results can be compared and checked almost directly. A different and simpler type of reversed study is common at the intermediate valuation stage when tonnage, grade and other features are reasonably established. This is the determination of the minimum selling prices for metal or product needed to achieve specified financial results. PRELIMINARY FEASIBILITY STUDIES (IntermediateEconomic Evaluations) Most projects will pass through at least two preliminary feasibility study stages before a final feasibility study is undertaken. The purpose of these preliminary studies is to justify further commitment of money to the ongoing development of the project, either for more exploration or for a final study. In general terms there are three reasons for carrying out a preliminary feasibility study, namely, to: 0

0

0

justify a major ongoing exploration program following a successful initial exploration effort. This could involve committing funds to, for example, a 30,000-metre drilling program, followed (potentially)by an underground exploration and development effort. If more than one competing project were under consideration, this study would help in setting priorities and as well as providing the justification for ongoing exploration. provide the basis for proceeding to a final feasibility study. This preliminary feasibility would be carried out on completion of the major exploration effort, when the orebody was sufficiently well defined that all the vital areas of the project that require further study and refinement for a final feasibility could be identified. This study would normally be sufficient to demonstrate the economic feasibility of the project and would identify and critically examine the alternatives for the major elements of the project, namely, - Mining - Processing - Environmental issues - Reclamation - Infrastructure to attract a joint venture investor or a buyer for the entire project, or to provide the basis for raising additional risk capital by way of a major underwriting. This type of preliminary feasibility work may also be carried out by potential purchasers as part of their due diligence process before “buying in” to the project.

288

-

Preliminary Feasibility Study Main Features 1. Mine design based on actual exploration results. 2. Three principal alternatives examined. 3. Preliminary siting, geotechnical and environmental studies complete. 4. Bench scale metallurgical tests. 5. Equipment factor cost estimate, written quotations. 6. Contingency level 15% to 25%. 7. Best alternative selected. 8. Basis for final feasibility developed. Information Base The information base for a preliminary feasibility study is generally fairly substantial compared to that for an earlier preliminary evaluation. However, less data will be available for a study that is used primarily to justify an ongoing exploration effort than for a more advanced study carried out to justify committing funds to the final feasibility study. The critical information required relates primarily to the establishment of geological and mineable ore reserves. At a minimum, the available information will normally consist of drill hole and assay data sufficient to establish the shape and attitude of the mineralization and to permit preliminary ore reserve calculations. For later studies, the drilling program will be essentially complete with respect to defining the size of the resource. Further drilling or underground sampling may be required at the final feasibility stage, but this would normally be to increase confidence in the already established reserve or more closely estimate ore grades and waste to ore ratios, rather than to increase the reserve tonnage. Metallurgically, a fairly complete program of mineralogical and bench scale metallurgical testwork will have been carried out. For ores that are identified as being particularly complex, difficult to treat, refractory or unique, some locked cycle testing or even small-scale pilot tests would normally be done at this point. Cost Estimation for Preliminary Feasibility Studies As outlined before, a preliminary feasibility study would usually contain an “equipment factor” capital cost estimate. Once again, as much direct estimation as possible should be done to provide the best reliability. Due to the limited technical information available at this stage, the cost estimate is unlikely to be more accurate than a nominal +20%.This level of accuracy would not be a sound basis for project financing, but would normally provide sufficient confidence that the project is profitable that moneys could be approved to proceed with a final feasibility study. Basis f Estimate. The major elements of a capital cost estimate at the preliminary feasibility level for a mining project are as follows:

P

0 0

0 0 0 0

Access road and site preparation Mine equipment Mine pre-production Process plant Ancillaries Indirect costs Reclamation and abandonment

Depending on whether the project will involve an open pit or an underground mining method, the mine equipment and pre-production costs can vary significantly.

289

For an underground mine, typical equipment requirements include: Air compressors Hoists and headframe LHD units (trackless) Locomotives Ore cars Ventilation fans Dewatering pumps Air, water and power reticulation Underground crusher Miscellaneous mobile equipment Pre-production development to support the first few years of production from a typical underground mine will include shaft sinking and hoist installation, haulage and skiploading development, ore passes, ventilation raises, drifts, crosscuts, ramps and substations. Stope development for a minimum of six months of mill feed is also necessary. For an open pit mine project, the mine equipment will usually include: Drills Trucks Shovels Dozers Loaders Graders Fueling system Miscellaneous mobile equipment Pre-production development for an open pit mine consists of clearing and grubbing, stripping overburden and waste, establishing waste dumps and stockpiles, constructing haul roads and installing communications and pit electrification systems. Construction of the truck shop, tire shop and pit warehouse would accompany this development. At the early preliminary feasibility stage, costs for the underground mine equipment would normally be factored from the estimated mine capacity, with pre-production development costs (shaft sinking, drifting, etc.) estimated on a cost per foot or metre basis, adjusted for shaft area and heading size. For an open pit mine, drills, trucks and shovels would normally be sized and delivered costs then estimated directly from similar recent purchases. The cost of overburden removal and pre-production waste stripping would be estimated from preliminary mine schedules on a per tonne basis. The process plant costs would be estimated from a combination of equipment factoring and takeoff based costs. A general site plan would be prepared, along with a flowsheet, equipment list and process plant general arrangement drawings. Piping and electrical costs would be factored from delivered and installed equipment costs. The costs of ancillaries, such as power supply and distribution, water supply, offices, shops and warehouses, laboratories, camps and/or townsite, concentrate handling and shipping and tailings disposal, would normally be estimated from a combination of preliminary design, takeoffs for earthmoving, square foot or metre factors for buildings and factored direct equipment costs. The indirect costs for a project will be substantial and include: Engineering General site costs Construction management

290

0 0 0

Administration Working capital Insurance

At the preliminary feasibility level, these costs will normally be developed as a percentage of estimated capital cost, with allowances for specific project location and complexity. Working capital will normally be set at 10 weeks of operating costs plus the cost of the spare parts inventory.

Content and Structure of Preliminary Feasibility Studies: Preliminary feasibility studies are carried out when there is sufficient information about the potential orebody to: 0 0

0

justify a study to support a decision to cany out the major definitive exploration program following this exploration program, justify a study to define economic feasibility and the allocation of funds to a final feasibility study adequately support the preparation of a “selling document” This level of study is based on known information about the orebody and is concerned with defining deficiencies in what is known in order to plan and organize further work. The conceptual design of the project will be reasonably certain, although a few assumptions may still be made with respect to such areas as marketing and minor metallurgical details.

This report will have a more-or-less complete table of contents, with subconsultant and laboratory reports appended. A typical table of contents for a preliminary feasibility study is presented on the following pages. This table of contents is general in nature but contains all the normal elements. Certain areas could be expanded, reduced or (rarely) deleted, depending on the particular project at hand. It is recommended that a complete table of contents always be considered, as this will force the study team to consider all aspects of the project, provide allinclusive costs and identify areas for further study. The table of contents for the final feasibility report will be very much the same, differing mainly in level of detail:

-

Preliminary Feasibility Study Typical Table of Contents LETTER OF TRANSMITTAL SECTION 1 - INTRODUCTION 1.1 GENERAL 1.2 TERMS AND CONDITIONS 1.3 STUDY EXECUTION SECTION 2 - SUMMARY AND CONCLUSIONS 2.1 PROJECT CONCEPT DESCRIPTION 2.2 FINANCIAL ANALYSIS SUMMARY 2.3 CAPITAL COST SUMMARY 2.4 OPERATING COST SUMMARY 2.5 CONCLUSIONS 2.6 RECOMMENDATIONS FOR FUTURE WORK SECTION 3 - PROPERTY DESCRIPTION AND LOCATION 3.1 LOCATION AND ACCESS 3.2 PROPERTY DESCRIPTION 3.3 OWNERSHIP 3.4 HISTORY AND PREVIOUS WORK 3.5 CLIMATE 3.6 LOCAL RESOURCES

29 1

-

Preliminary Feasibility Study Typical Table of Contents

SECTION 4 - GEOLOGY AND MINERAL RESERVES 4.1 REGIONAL GEOLOGY 4.2 LOCAL GEOLOGY 4.3 MINERAL RESERVES 4.4 EXPLORATION POTENTIAL SECTION 5 - MINING 5.1 GENERAL MINING METHODS AND MINE DESIGN 5.2 5.3 MINEABLE RESERVES AND GRADE MINE PLANNING AND PRODUCTION SCHEDULE 5.4 5.5 EQUIPMENT SELECTION 5.6 WASTE ROCK DISPOSAL SITES 5.7 MANPOWER REQUIREMENTS 5.8 DISCUSSION SECTION 6 - METALLURGY 6.1 GENERAL 6.2 METALLURGICAL TEST RESULTS 6.3 FLOWSHEET DEVELOPMENT 6.4 DESIGN CRITERIA 6.5 EQUIPMENT SELECTION 6.6 PROCESS PLANT DESCRIPTION 6.7 MANPOWER REQUIREMENTS 6.8 DISCUSSION SECTION 7 - ANCILLARY SERVICES AND FACILITIES 7.1 WATER SUPPLY 7.2 TAILINGS DISPOSAL 7.3 POWER SUPPLY AND DISTRIBUTION 7.4 SHOPSlWAREHOUSES 7.5 COMMUNICATIONS 7.6 OFFICE FACILITIES 7.7 EMPLOYEE HOUSING 7.8 DISCUSSION SECTION 8 - TRANSPORTATION 8.1 ACCESS ROAD 8.2 OPERATING SUPPLIES 8.3 CONCENTRATE HANDLING 8.4 MANPOWER TRANSPORT SECTION 9 - ENVIRONMENTAL CONSIDERATIONS 9.1 PERMITS 9.2 BASELINE STUDIES 9.3 RECLAMATION 9.4 DISCUSSION SECTION 10 - CONSTRUCTION PLAN 10.1 GENERAL 10.2 SCHEDULE SECTION 11 - CAPITAL COST 11.1 BASIS OF ESTIMATE 11.2 MINING 11.3 MILLING 11.4 ANCILLARY SERVICES 1 1.5 CONCENTRATE HANDLING 1 1.6 INDIRECTS

292

-

Preliminary Feasibility Study Typical Table of Contents 11.7 CLOSURE COSTS SECTION 12 - OPERATING COST 12.1 MINING 12.2 MILLING 12.3 PLANT AND SERVICES 12.4 GENERAL AND ADMINISTRATION SECTION 13 - FINANCIAL ANALYSIS 13.1 SALES AND REVENUES 13.2 CASH FLOW PROJECTIONS 13.3 SENSITIVITIES 13.4 DATA SUMMARIES APPENDICES APPENDIX 1 SUBCONSULTANT REPORT APPENDIX 2 SUBCONSULTANT REPORT LIST OF DRAWINGS LIST OF FIGURES LIST OF TABLES The content of each section of the report will be similar to that described for a preliminary evaluation, but will contain significantly more detail, particularly for a late stage preliminary feasibility study. Once again, the critical area of this study will be the mineable ore reserve estimate. Sufficient drilling data should be available to confidently classify the ore reserve as proven and probable. No possible ore can be considered in the reserve tonnage estimate; however, if minor in-fill drilling could upgrade these reserves, then a recommendation would be made for the work to be done in the final feasibility stage. Some of the critical additional information required for a preliminary feasibility report is summarized on a section-by-sectionbasis: Introduction. At this stage the study may often be undertaken in whole or part by a recognized consulting engineering firm. It is important that the terms of reference for the study and its envisaged specific end use be completely and clearly explained. In addition, the assumptions that have been made should be specifically qualified. Summary and Conclusions. As the critical areas for a study of this level are the mineable ore reserves and the project financial analysis, the summary should report concisely on these areas plus provide tabular summaries of the capital and operating cost estimates. The heart of the study will be the conclusions and recommendations, which will indicate economic feasibility and clearly identify the specific areas in which additional work is required prior to commencing a final feasibility. A preliminary construction schedule will have been prepared, which will provide management with a funding and cash drawdown timetable to drive the financing program. Property Description and Location. A description of the claims, concessions and other property needed for plant sites, access roads, power and water right-of-ways is required. Topographic maps of the mine and plant site area should be included for estimation purposes, and a legal survey of the claims should be in hand. The ownership of all such land should be ascertained and steps taken to gain control if necessary. A reasonable detailed review of the history of the property (in the mining sense) and previous work by others should be included. A brief review of the local resources and potential socioeconomic impact of the project will be adequate. Geology and Mineral Reserves. For an early preliminary feasibility study intended to justify definitive exploration drilling, only the regional geology and perhaps detailed surface mapping of the local geology will be available. The geological model of the orebody will be developed as the drilling information base expands. This model should be well developed and confirmed prior to undertaking the ore reserve estimate for a later preliminary feasibility study. Careful attention must be paid to the exploration potential both at the potential mine site and regionally.

293

Mining. The determination of a mineable ore reserve based on the proven and probable ore categories will provide the necessary confidence to determine economic feasibility and thus support a final feasibility study. For this reason a well-chosen mining method is critical; normally the alternatives will have been considered and a best case selection made. Equipment will have been sized and selected based on realistic production schedules for both ore and waste. Waste disposal concepts will be will developed, with preliminary consideration given to recontouring, reclamation and abandonment. Metallurgy. Sufficient mineralogical and bench scale testwork will have been done to determine the probable process flowsheet and material balance for the process plant. Based on plant throughput, equipment will have been sized and selected, with preliminary specifications prepared for major equipment. Product specifications will have been set following preliminary discussions with potential buyers. In the case of novel or unproven technology, a mini-pilot plant program may have been carried out. A series of plant site visits will have been made to assist in determining plant site and tailings facility locations and to assess access, power and water supply. Conceptual general arrangement drawings of the process plant will be prepared to provide the basis for rough takeoff cost estimation. Ancillary Services and Facilities. Critical areas such as power and water supply and the tailings management scheme will have been investigated and developed to the conceptual design phase. Sufficient drawings will be prepared to enable preliminary quantify takeoffs for major earthworks and ongoing dam construction. Alternate routes for power and water supply will have been considered, trade-off studies carried out and the preferred alternatives selected. Environmental Considerations. This important area will have been thoroughly studied. A specialist consultant is almost invariably required to fully define all permitting and regulatory matters, to help coordinate the necessary environmental studies and to assist in the permitting process. The baseline studies for an environmental impact statement will be underway, and the status of these studies should be reviewed in the report. Construction Plan. A preliminary construction plan and schedule will be prepared. Critical aspects such as road access, construction seasons, construction labor availability and camp requirements will have been assessed. A conceptual project execution plan and contracting strategy will be prepared. A basic constructability review of the proposed process plant and ancillary facilities will be included. Capital and Operating Costs. The elements and nature of these cost estimates were described more fully in the section “Cost Estimation”. At the preliminary feasibility stage the capital cost estimate still reflects a conceptual installation that could be built rather than the actual installation that yiJ be built. Operating costs will be developed from manning and equipment lists and estimated reagent and consumable usage. In general these estimates will still carry contingencies in the range of 20% - adequate for determining probable economic feasibility but not for budget allocation or project financing. Financial Analysis. This final item on the table of contents is the most critical to informed decision making. The economic feasibility information required will consist of the projected revenues based on forecasts of metal prices and operating costs, combined with capital cost and financing charges. All taxes, royalties and financing assumptions must be very clearly and fully described. Sensitivities to capital and operating costs, metal prices and production rates will all be included, and the expected project economics will be analyzed. Marketing projections for concentrates or metal products will be based on existing operations and assumed prices, not on specifically negotiated terms with particular buyers. Staffing for Preliminary Feasibility Studies Preliminary feasibility studies can be prepared completely in-house by mining houses with significant engineering capability. This is more common in South Africa or the U.K., however, whereas a larger mining company in North America generally will organize the preliminary feasibility study and prepare an internal report, but will assign a number of the tasks to

294

independent consultants, coordinating their activities to ensure consistency and a common format. Smaller companies with few internal resources will normally contract a recognized engineering company to manage the study on their behalf, with specialist subconsultants contracted as necessary. For a preliminary feasibility study the key personnel will include the explorationist, the mining engineer, the metallurgist and an experienced project engineer who serves as the study manager. The cost estimate will normally be carried out by the engineers involved, with a professional estimator assembling the final capital costs under the supervision of the project manager.

Unique Characteristics The preliminary feasibility study is an extremely important document. It is the first true test of a project’s economic feasibility based on more factual information than assumption and as such serves as the real decision point for many projects. The report not only provides an evaluation of the project but also becomes the primary planning document for the final feasibility and the source of schedule and cost data relied on by project and corporate management. It is essential that highly experienced people be involved in preparing the study, as the document produced will represent the “life or death” of a project and is the foundation for all future activities. FINAL FEASIBILITY STUDIES The final feasibility study for a mining project is the culmination of all the exploration, investigation and engineering work that has gone into defining the geological, mining, metallurgical and ancillary aspects. This study emphasizes what is known without a reasonable doubt; it presents and reviews the relevant information and references all the supporting data. Normally based on the most attractive alternative for the project as identified by the preliminary feasibility study, the final feasibility study is prepared for three purposes: 0

0

0

to demonstrate conclusively that the project is technically sound and economically viable, and is constructible and operable to present sufficient supporting information to become a bankable document acceptable to banks or other sources of finance for the project development expenses to provide a firm design basis for the project, suitable for proceeding to detailed design and construction with only relatively minor changes. 1. 2.

3. 4.

5.

6. 7.

-

Final Feasibility Study Main Features Ore reserves calculated and verified to be proven and probable with sufficient confidence. Mine design based on second-stage predevelopment and test mining data if required. Most attractive alternative developed in sufficient detail to provide the basis for project approval. Engineering work advanced to the point where a design basis document can be prepared and a reliable detailed cost estimate generated. Final environmental impact documents and permit applications presented. Contingency level 10% to 15%. Appropriation of funds for design and construction justified.

Information Base The information base for a bankable final feasibility study will be exhaustive. All aspects of the project will have been studied in enough detail to support a production decision and to ensure that no unwelcome surprises appear.

295

General. The information base with respect to general matters consists of three major areas: 0

0 0

Legal and regulatory considerations Topographical maps and surveys Socioeconomic and work force aspects

Legal and regulatory considerations will normally have been thoroughly investigated by an established legal firm, which will present a professional opinion on land ownership, mining claims, surface rights, water rights, royalties to previous owners (if applicable) and other legal matters, backed up by the relevant deeds, legal agreements or government decrees. A second aspect of regulatory requirements relates to construction, operation, environmental and abandonment permits, all of which should be in place or in the final processes of application, with certainty of being awarded. Topographical maps and surveys are an often underestimated element of the project development. Topographical maps of the area surrounding the project sites are normally available through governmental agencies at a scale suitable for general planning and location of sites, roads and tailings disposal area. Aerial photographs are now commonly used both for stereographic examination of road, power line and water line routes and also as the basis for photogrammetry for detailed topographic maps. Such maps should be prepared for the project area at a contour interval of about two metres. Plant site and mine site surveys at a minimum one metre contour interval are necessary for the design drawings for the mine site, plant site, townsite and other specific installations. In addition, road, water and power line routes will usually have been surveyed for profile and estimation of cut-and-fill quantities. An information base will have been developed with respect to the socioeconomic aspects of the project, work force requirements, labor relations and employee matters. These studies may be tied to permitting requirements, but are primarily needed as a basis for manpower planning and training and to assist in estimating operating costs. Geology and Ore Reserves. Sufficient drilling information will be available to prepare an ore reserve estimate demonstrating adequate proven ore reserves to support the planned mine life. A program of check assays, twinned holes and independent reviews will have been carried out and documented to provide the basis for mineable ore reserve estimation. Local surface mapping will be complete: regional mapping may have been available previously. Mining. The geotechnical aspects of the proposed mine will have been studied and a report prepared. The proposed mining method and mine design will be appropriate for the nature of the rocks in the mining area. A computer model of the proposed mine will be prepared, incorporating ore grades, lithologies, alterations and geological structures. Metallurgy. A database consisting of bench scale tests, pilot scale grinding and concentration tests, and mineralogical reports on ore and concentrates will be available. In addition, detailed chemical analysis of all ore types, concentrates or bullion products and tailings will have been completed, and the acid generation potential of the ore, waste and tailings will have been tested. The shipping moisture limits for concentrates will be known, as well as any pyrophoric tendencies for high sulfide concentrations in the concentrate. Ancillaries. A major element in this area requiring a detailed information base is the project water supply. Hydrology and groundwater studies will have been carried out, full-scale pumping tests conducted and a potential well field identified. The tailings dam area will have been drilled for geotechnical information related to dam stability as well as percolation and seepage rates. Power, water and tailings lines routes will have been mapped and surveyed. Financial Analysis. In order to prepare the proper financial analysis, a database containing all applicable federal, state, county and municipal taxes, depreciation rules, royalties, agreements and other relevant items will have been compiled.

296

COST ESTIMATION FOR FINAL FEASIBILITY STUDIES A final feasibility study cost will be prepared to the level of a “forced detailed” capital cost estimate as described in the section “Cost Estimation”. The information required for this level of estimate is listed in Table 2. The forced detailed cost estimate requires sufficient engineering effort to generate complete equipment lists for all areas of the project as well as general arrangement drawings to a level suitable for obtaining material takeoffs that will support a 10%to 15% contingency factor. Normally, major equipment will be specified, and written competitive quotations for supply and delivery will have been obtained. Major earthworks would also be bid out in order to obtain realistic unit prices for this area. The cost estimate would be prepared by professional estimators supervised by an experienced project engineer, based on the written quotations. Basis of Estimate The basis of estimate for a final feasibility study includes all the same areas described for the preliminary feasibility study, but there are some major differences in the level of detail that must be developed for certain items. The following particular areas should be noted: Access Road. Cut-and-fill quantity takeoffs based on survey information will be used to generate material quantities. Differentiation between drill and blast versus rippable versus dozable material will be made. Unit rates from local contractors capable of performing the work will be solicited, with written quotations required. For particularly complex road alignments, competitive bids may be called for at the feasibility phase. Mine Equipment. The major mine equipment will be sized and data sheet preliminary specifications prepared. Written quotations from at least two suppliers will be obtained. Written specifications would be prepared for any unusual requirements (i.e., high altitude turbochargers for trucks, special filters for particularly severe dust), followed by competitive tenders and written quotations. Process Plant. The process plant cost estimate will be based on sufficient drawings, equipment specifications and quotations to provide a forced detailed estimate. Earthworks, structural steel, foundations, building cladding etc., will all be based on material takeoffs. Preliminary piping routings and in some cases preliminary P&IDs will be prepared to assist in a combination takeoff / factored piping estimate. In particular, large diameter rubber-lined pipe or significant quantities of alloy pipe or valves will be taken off. Ancillaries. Sufficient preliminary engineering wi11 be done for all ancillary buildings and equipment that takeoff estimates can be prepared. For typical pre-engineered building applications a factored square foot or metre estimate may be used. Indirect Costs. These costs form a substantial fraction of the capital cost of any project. For more preliminary studies, indirect costs are usually factored as a percentage of installed capital costs. For a final feasibility study, however, every effort should be made to estimate these costs directly, as follows: Engineering: The engineering firm preparing the cost estimate and feasibility study should prepare an estimate of the engineering required for the project. This will include a drawing list, specification list and estimate of workhours by discipline, and a schedule and estimates of disbursements and consumables for the engineering. Field engineering services should be included, as well as procurement and inspection services. Construction Management: An estimate of construction management costs based on the construction schedule and reasonable work force requirements should be prepared. Direct field cost such as camp, transportation, home leave and bonuses should be included. General Site Costs: These will include insurance, the capital and operating costs of the construction camp, owner’s site representatives and project managers, construction power costs and similar items.

297

0

0

0

Administration: A project requires management by the owner or owners. It is necessary to estimate the costs for people seconded to the project as well as the administration fee that would normally be charged for management services associated with financing and directing the work. Working Capital: A properly prepared estimate of working capital is required, based on site operating costs plus insurance and shipping costs plus insurance and shipping costs for the length of time it will take to obtain payment for final product. Generally this should be a minimum of 10 weeks. Spare parts and consumables must also be considered as part of working capital. Start-up Costs: These should be considered an indirect cost to the project and as such should be carefully estimated. They include such items as vendors’ representatives, engineering assistance and training and an allowance for reduced production during the first year of operation.

CONTENT AND STRUCTURE OF FINAL FEASIBILITY STUDIES Typical Table of Contents The final feasibility study for a project provides the basis for financing and proceeding with construction of a project. As noted previously, this study contains all relevant information generated during the course of the study, with supporting documents provided in appendices, as necessary. This study finally and conclusively demonstrates the economic and technical feasibility of the project. It is based on established and verifiable facts, and in large measure describes what will be built following the financing phase. A typical table of contents for a final feasibility study is presented on the following pages. It should be noted that this will be a large report requiring multiple volumes to contain and properly display the information. As the feasibility report will be reviewed by experts in various specialties who need access to only certain information, attention should be paid to how the information is organized within these volumes. A suggested grouping is presented at the end of the typical table of contents.

-

Final Feasibility Study Typical Table of Contents LETTER OF TRANSMITTAL SECTION 1 -EXECUTIVE SUMMARY 1.1 SYNOPSIS, CONCLUSIONS AND RECOMMENDATIONS 1.2 PROJECT DESCRIPTION 1.3 FINANCIAL ANALYSIS GEOLOGY, MINERALIZATION AND ORE RESERVES 1.4 1.5 MINE PLAN 1.6 MILL AND PROCESS PLANT 1.7 TAILINGS DISPOSAL 1.8 INFRASTRUCTURE ENVIRONMENTAL IMPACTS AND MITIGATION MEASURES 1.9 1.10 DESIGN FOR CLOSURE 1.11 CAPITAL COST AND CONSTRUCTION SCHEDULE 1.12 OPERATING COST SECTION 2- INTRODUCTION AND TERMS OF REFERENCE 2.1 INTRODUCTION 2.2 STUDY OBJECTIVES 2.3 STUDY EXECUTION SECTION 3- FINANCIAL ANALYSIS 3.1 INTRODUCTION 3.2 DATA AND ASSUMPTIONS

298

-

Final Feasibility Study Typical Table of Contents 3.3 RESULTS SECTION 4 - GEOLOGY. MINERALIZATION AND ORE RESERVES 4.1 REGIONAL GEOLOGIC SETTINGS 4.2 GEOLOGY AND MINERALIZATION OF THE PROPERTY 4.3 ORE RESERVES AND ORE TYPES 4.4 ASSAY QUALITY ANALYSIS 4.5 POTENTIAL FOR RESERVE EXPANSION SECTION 5- MINING 5.1 INTRODUCTION 5.2 MINING SCHEDULE 5.3 MINING MINE WASTE STOCKPILING AND DRAINAGE 5.4 5.5 ENVIRONMENTAL IMPACTS OF MINING SECTION 6- SITE SELECTION, ACCESS ROUTES AND ORE HAULAGE 6.1 INTRODUCTION 6.2 PLANT SITE LAYOUT 6.3 SITE ACCESS SECTION 7- SITE SELECTION, ACCESS ROUTES AND ORE HAULAGE 7.1 INTRODUCTION 7.2 METALLURGICAL TEST PROGRAMS AND RESULTS 7.3 PROCESS PLANT DESIGN CRITERIA 7.4 FLOWSHEETS AND PLANT DESIGN 7.5 PROCESS CONTROL SECTION 8 TAILINGS DISPOSAL 8.1 INTRODUCTION 8.2 DESIGN OBJECTIVES AND ASSUMPTIONS SUMMARY OF LABORATORY TESTS ON TAILINGS 8.3 GENERAL LAYOUT AND STAGED DEVELOPMENT 8.4 8.5 STORAGE CAPACITY 8.6 DEPOSITION STRATEGY 8.7 TAILINGS PIPEWORK 8.8 PROCESS WATER MANAGEMENT SYSTEM 8.9 INSTRUMENTATION SECTION 9- WATER MANAGEMENT 9.1 REQUIREMENTS AND SOURCES PROCESS, TAILINGS AND RECLAIM SYSTEMS 9.2 9.3 MINE AND WASTE DUMP POTABLE SUPPLY AND SANITARY DISPOSAL 9.4 9.5 SURFACE RUNOFF 9.6 ENVIRONMENTAL REPORTLNG SECTION 10 ANCILLARY FACILITIES AND SERVICES 10.1 MILL BUILDING 10.2 ANCILLARY FACILITIES REFUSE DISPOSAL AND SEWAGE TREATMENT 10.3 10.4 FRESH WATER SUPPLY 10.5 FUEL STORAGE 10.6 ADMINISTRATION OFFICES 10.7 WAREHOUSING 10.8 SECURITY SECTION 11- POWER SUPPLY AND DISTRIBUTION OVERVIEW AND POWER REQUIREMENTS 11.1 11.2 HYRDROELECTRIC

299

-

Final Feasibility Study Typical Table of Contents 11.3 GENERAL DESIGN PARAMETERS 11.4 COMPONENT DESIGN DETAILS 11.5 ELECTRICAL DISTRIBUTION SECTION 12 -COMMUNICATIONSAND INFRASTRUCTURE 12.1 TELECOMMUNICATIONS 12.2 TRANSPORTATION 12.3 HOUSING SECTION 13 -SOCIOECONOMICIMPACT AND HUMAN RESOURCES 13.1 SOCIOECONOMICCONSIDERATIONS 13.2 MANPOWER TRAINING PLAN 13.3 TECHNOLOGY TRANSFER 13.4 COMMUNITY IMPACTS SECTION 14 -ENVIRONMENTAL 14.1 GENERAL 14.2 PERMI'ITING PROCESS 14.3 STATUS OF PERMITS 14.4 ENVIRONMENTAL AND HEALTH LEGISLATION SECTION 15- DESIGN FOR CLOSURE MINE AND WASTE STOCKPILE AREAS 15.1 15.2 PLANT SITE AND INFRASTRUCTURE 15.3 TAILINGS 15.4 WATER QUALITY 15.5 POST-CLOSURE MONITORING SECTION 16 -CONSTRUCTION PLAN 16.1 GENERAL 16.2 ACCESS AND HAUL ROADS 16.3 MILLSITE 16.4 TAILINGS STORAGE 16.5 HYDRO POWER 16.6 ENVIRONMENTAL IMPACTS OF CONSTRUCTION 16.7 SCHEDULE SECTION 1'7-CAPITAL COST 17.1 SUMMARY 17.2 SCOPE 17.3 ESTIMATE BASIS 17.4 ASSUMPTIONS 17.1 EXCLUSIONS SECTION 18-OPERATINGCOST 18.1 SUMMARY 18.2 MINE 18.3 MILL AND PROCESS PLANT 18.4 TAILINGS STORAGE FACILITY 18.5 PLANT AND ADMINISTRATION 18.6 ENVIRONMENTALMONITORING AND CLOSURE SECTION 19-CLOSURECOST 19.1 SUMMARY 19.2 HUMAN RESOURCES 19.3 MINE AND WASTE STOCKPILES 19.4 MILL AND PROCESS PLANT 19.5 TAILINGS 19.6 POST-CLOSURE MONITORING

300

-

Final Feasibility Study Typical Table of Contents SECTION 20 - MARKETING 20.1 INDUSTRY STUCTURE HISTORICAL AND PROJECTED SUPPLY / DEMAND RELATIONSHIP 20.2 DEMAND, SUPPLY AND PRICING TRENDS 20.3 20.4 COMPETITION 20.5 MARKETING CONTRACTS 20.6 PROJECT ADVANTAGES SECTION 21 -RIGHTS, OWNERSHIP AND LEGAL MATTERS 2 1.1 PROPERTY OWNERSHIP 2 1.2 MINING CLAIMS 21.3 SURFACE RIGHTS 2 1.4 WATER RIGHTS 21.5 PERMIT REQUIREMENTS 21.6 TAXATION APPENDICES APPENDIX 1 INCOME TAXES AND ROYALTIES APPENDIX 2 CASH FLOWS CAPITAL COST ESTIMATE DETAIL APPENDIX 3 APPENDIX 4 MAPS AND DRAWINGS LIST OF REFERENCES AND PREVIOUS REPORTS APPENDIX 5 LIST OF DRAWINGS LIST OF FIGURES LIST OF TABLES Final Feasibility Study -Organizationby Volumes Volume 1 - Executive Summary Volume 2 - Financial Analysis Volume 3 - Geology, Ore Reserves and Mining Volume 4 - Metallurgy and Process Plant Volume 5 - Socioeconomic and Environmental Impact Volume 6 - Capital and Operating Cost Estimates Volume 7 - Legal, Finance and Marketing Volume 8 - Appendices Incremental Information The final feasibility study will contain more information than the preliminary study on which it is based. In particular, the selected “best” alternative will be effectively restudied in detail to provide all the necessary information for cost estimation and design. Some of the more critical requirements are noted below. Executive Summary. The executive summary should be a stand-alone document that concisely presents the entire project. This volume will be the primary document for distribution to financing institutions and interested agencies. The main elements are as follows: 0 0

0 0

Stand-alone document. Emphasis on project economics and schedule. May include significant information on project sponsor if financing is sought. Must be positive, but completely supported.

301

Financial Analysis. The financial and economic analysis of the project is the key result of a final feasibility study. This must be based on a financial analysis model that properly includes capital costs, operating costs, cash flows, metal prices, depreciation allowances, taxes (and tax holidays if applicable), production rates and smelter or refinery schedules. The analysis will be based on mine and mill production schedules and a year-by-year projection of the amount of final products. The net revenue at the mine is calculated after considering all mine site and off-site operating costs, shipping and freight costs for products, deductions for smelter and refining charges, insurance and analysis costs. Cash flow schedules on a year-by-year basis for the life of the mine, considering depreciation and taxes, should be presented, along with the key economic indicators such as internal rate of return, payback period, net present value, and debt:equity assumptions. Sensitivity analysis to capital and operating cost changes, metal prices and debt:equity ratios will be carried out and presented in graphical form. The key elements are as follows: 0 0 0

Fully documented basis for taxes, financial model, expected revenues and costs. Adequate and realistic sensitivity analysis. Inflation, escalation and currency exchange properly considered.

Geology, Mineralization and Ore Reserves. 0 0 0 0

Detailed methodology of reserve calculation. Correlation of bulk sampling, metallurgical testing, pilot plant drilling information. Statistical reliability analysis. Reserve classification basis.

Mining. 0 0 0 0

0

Optimized mine plan and production schedule. Finalized equipment selection. Geotechnical and rock mechanics information complete. Equipment replacement schedule. Detailed manpower schedule. Pre-production and start-up learning curve effects.

Site Selection, Access Routes and Ore Haulage. 0 0 0 0

Detailed site survey and mapping complete. Access routes finalized. Ore haulage profiles studied and matched to truck capability. Stockpile access and management plan.

Metallurgy. 0 0

0 0 0 0

Optimized bench scale testwork. Continuous process and grinding pilot plant results. Final metallurgical process design criteria. Product marketing acceptability results. Detailed flowsheet, mass and water balance. Metallurgical balance and metal production forecast.

302

0 0

Optimized plant site selection. General arrangement drawings, plans and sections for stockpiles, grinding and flotation, dewatering and stripping.

Tailings Disposal. 0

0 0 0

0

Final geotechnical and hydrological investigations for design of tailings embankment and water reclaim. Tailings line route surveyed and profile optimized. Ongoing embankment construction schedule. Dam capacity and filling curve. Reclaim water and process water management strategy. Environmental and permitting constraints.

Ancillary Facilities and Services. 0 0 0

Potable water supply, refuse disposal and sewage treatment concepts well developed. Fuel storage and fuelling facilities for both mine equipment and site mobile equipment. Administration offices, shops and drys designed to reflect personnel lists and shift schedules.

Power Supply and Distribution. 0 0

0

Letter of intent with firm pricing for power supply. Power supply line route survey. Rights-of-way and land ownership issues resolved. Substations, site and mine electrification schemes.

ConstructionPlan. 0 0

General construction contracting strategy developed. Construction services requirements known. Contractor and labor resource survey completed. Critical path project schedule prepared, including financing period, basic and detailed engineering, procurement and construction.

Environmental. 0

0

Baseline studies complete. Environmental Impact Statement (EIS) submitted and critical permitting processes in final phase. All applicable environmental and health legislation considered. Total permit requirements listed and compliance in place.

Design for Closure. 0 0

Negotiations for closure bonding initiated. Closure plan requirements for permits in hand. Post-closure monitoring program developed and costs estimated.

303

Capital Cost Estimate. Work breakdown structure developed for estimation work package. Written quotations in hand for all major equipment, including provisional delivery schedules. Quantity take-offs for excavation, concrete, structural steel, platforms, cladding, piping, etc. Construction labor cost worked up. Written budget quotes for unit rates from capable local contractors (if possible). Freight costs calculated, not factored. Construction equipment costed. Contingency allowance calculated on same basis as work packages. Financing costs considered. On-going capital requirements for equipment replacement, dam building, etc. estimated. Operating Cost Estimate. Detailed personnel lists and pay scales for all areas including mine, mill, site administration,head office. Fringe benefits, social legislation and housing costs considered. All measurable consumables estimated - drill bits, explosives, lubricants, grinding media, mill liners, power, reagents, maintenance materials, laboratory supplies, office supplies, etc. Mobile equipment operating cost estimate. Fixed administration costs, rent, taxes, insurances, communications, donations, etc., estimated. Marketing. Product specifications,transportation, marketing regulations or restrictions. Market analysis and forecast of metal prices. Effect of hedging. Likely purchasers. Smelting refining and insurance costs. Letter of intent for smelting contract, contract duration, amendments, cost escalation. Requirements for sampling, assaying, weighing, umpiring and dispute resolution. Rights, Ownership and Legal Matters. An independent legal opinion should be obtained regarding property ownership, mining claims and rights, surface and water rights, and road, water pipeline and power line rights-of-way. All licenses and permits required for construction and operation should be listed, along with a plan and schedule for obtaining them. Employment laws for local and expatriate workers. Any legal requirements regarding local and federal taxes, currency control, corporate responsibility and registration should be noted.

304

ORGANIZATION AND STAFFING FOR FINAL FEASIBILITY STUDIES Organization The final feasibility study is a large and complex document requiring the work of many technical specialists and diverse groups of people. The study will take a number of months or even years to complete, particularly since the final exploration efforts and metallurgical pilot testing programs are considered part of the study. The costs of such a study can easily be more than $1 million for engineering services alone, and may be as much as $10 million if all drilling, assaying, metallurgical testing and specialized consultant costs are considered. Proper management of such a complex study to meet schedules and budgets is essential and has historically been carried out in two general ways: Internal Studies. A very large and technically capable mining company will appoint an internal project manager who will then organize the study and assemble the final feasibility report. Various tasks and specialized contributions to the report will be subcontracted to outside consultants as follows. 0

0 0 0 0

Exploration drilling. Environmental baseline studies and investigations Possibly metallurgical testwork and pilot plant operation. Detailed design drawings and material take-offs Specialized geotechnical investigations.

The company would do its own geological assessment and modeling, mine design and planning, production scheduling, flowsheet development and estimating of both capital and operating costs. The precedence network could be used to define all the tasks required, and then a decision would be made as to which tasks could be carried out with internal resources. These internal tasks will require geologists, mining engineers, civil and electrical engineers, metallurgists, lawyers, accountants, personnel and labor experts, and purchasing, construction and marketing experts. A formal project organization should be set up, with the necessary internal people assigned responsibilities for budgets, deliverables and schedules. All externally contracted parts of the study should have a very well defined scope and definition of work, including the contractual basis for carrying out the work and the required dates for completion. Externally Contracted Studies. Smaller mining companies or joint ventures lacking the necessary technical and managerial expertise will need to contract out most of the feasibility study. A bankable feasibility study may require the independent opinion of a major engineering company to provide project credibility. A larger mining company may also find it advantageous to contract out such a study for a number of reasons: it may have multiple projects underway, need an independent opinion or require specialized process or mining expertise not available in-house. The mining company, regardless of its size, will almost always retain the responsibility for “owner’s” concerns - property titles, legal matters, financing arrangements and product marketing. Essentially the same work as previously described will be done to prepare the study, except that the engineering company will act as the “prime contractor”, supervise all the subconsultants and take responsibility for assembling and preparing the final report, ensuring that schedules and budgets are adhered to. Typically, subconsultants would be required for the following work 0 0 0

0 0 0

Geotechnical studies. Ore reserve calculations Metallurgical testwork and pilot plant operation. Tailings dam design. Environmental studies. Hydrological investigations.

305

The engineering company will normally assign a senior and highly experienced project manager to lead the study. The cost estimation will be carried out by a team of professional estimators with recent experience in similar projects. Financial analysis will be carried out by the principal consultant, utilizing client-specified metal prices, cash flows and taxation regimes. The most effective organization for such a study normally involves a “pepper and salt” approach, with some specialized mining company personnel being seconded to the project team. As with all activities of this nature, various combinations of internal and external resources are possible. For a totally internally funded project there will probably be little need for any independent study. On the other hand, for a large project requiring significant financing, an independent study certified by a recognized engineering firm will be mandatory for the study to be regarded as “bankable”.

Unique Characteristics A “bankable” feasibility study is one that will be suitable to enable the project owner or owners to negotiate project financing from typical lending sources. The bankable document will satisfactorily provide all the technical/economic information and auditing necessary for a banker (and the banker’s independent engineer) to determine that the project risks are acceptable and that the project is indeed viable on a stand-alone project financing basis. Risk Ideally a bank or financing / lending institution would like to see a clear and concise review of the project risks associated with any proposed mining project. Risk categories with respect to a mining project are classified below.

Project Risks Technical Schedule cost Completion

Company Risks Economic Market Political Foreign exchange Environmental

Bank Risks Sponsor

Standard methodologies and specialist organizations exist, that can coordinate a review and analyze the results. The methodology is fairly subjective, but the modeling procedure can produce a result in economic terms in many cases, and the process identifies those areas where mitigation and control are required. For the project, cost, schedule and specific risks are evaluated. For the larger corporate picture, such items as commercial and economic risk, production loss and safety issues as well as political and environmental risk may be evaluated. The results are combined in an overall risk program. For a large project, a team is required, led by an experienced facilitator, who is not involved in the project, and can remain unbiased. The project manager, and other company managers are involved in the process and are responsible for implementing control measures. Expertise is required from the company - operations, maintenance and technical, and from the engineer - both design and field. Methodology. The following methodology relates to the risks associated with the Feasibility study - Engineering through construction. It applies to the larger corporate picture as well. The process starts by determining the parameters - areas to be examined, criteria for evaluation, methodology for follow up. Evaluation criteria can be presented in the form of a risk matrix shown in Table 3. Criteria can be subjective or monetary and are project specific. Each possible risk is identified in a review process - what the risk is and under what circumstances it can occur. Following this, each risk is evaluated on the basis of the previously established criteria. Risks are categorized and prioritized in the process - high, medium, low.

306

Certain identified risks will be addressed simply on the basis of unacceptability. In general, the results are focused on items with significant impact. Significant items will be addressed, even if unlikely. Items in the mid-range are evaluated on a cost-risk basis, using standard economic techniques.

Table 3 Risk Evaluation Criteria Category Severe Major Moderate Minor Slight

cost >$20 million $5-20 million $2-5 million $0.5-2 million <0.5 million

>90% H H H H M Will occur

6590% H H H M M Should occur

Probability 3565% 10-35% H M M M M M M L L L May Could occur occur

40% M M L

L L Should not occur

Probability and cost elements allow results to be analyzed for economic impact. A model is created and Monte Carlo simulation used to quantify the probability and range of effects Treatment options are identified and evaluated. The results are distilled into a plan that is then implemented by the appropriate personnel. High and medium risk items require mitigation and control. High risk items may require additional investigation and analysis. Low risk items are managed with the ongoing work. Anticipated results are fed back into the model. During implementation, the program is reviewed on an appropriate time basis and updated based on identified changes to risk and new risk items identified. In a similar manner, criteria are created for schedule impacts and the outlined methodology employed to manage these risks. Contingency for the capital cost estimate is calculated by creating a probability forecast for each cost item and using Monte Carlo simulation to arrive at a not to exceed cost, generally at the 80% confidence level. The difference between the not to exceed cost and the forecast becomes the contingency. Contingency measures only the risk associated with the procurement and construction items. Other items become part of the overall risk response plan. Recent trends are to minimize cost. It is no longer expected, or even acceptable that new plant will exceed design throughputs by 20% or more. The result of “close to the line” design and pressure on the cost and the schedule have made risk analysis and mitigation an important part, if a neglected part, of the Feasibility Study. Competent engineering and a standard process will reduce these risks, but each new project carries its own risks, which must be evaluated and controlled. Even a modest program can provide significant benefits.

Bank Criteria The bank views the project loan on an entirely different basis than the company views the project. Banks are not rewarded for excess risk taken by the sponsor which may earn him return and so will limit their risks to those considered appropriate for them. Even so, the bank will analyze all the project risks using internal resources and outside consultants, including risks they don’t carry in theory, but which might affect them regardless. Worth (1991) and Halupka (2002, this volume) have expostulated this process from the bank’s viewpoint. A partial list of rules of thumb for mitigating or managing risk is as follows. Reserve Risk. Careful review of the data and analysis can provide comfort in the accuracy and consistency of the estimate. The mine life should be twice the loan life with a high proportion of proven reserves. Technical Risk. For newer technologies it can be difficult to assess the eventual outcome of the effort, although experience has shown that achieving production in a reasonable time frame (if

307

ever) can be problematic. McNulty (1998) has examined the characteristics of new projects with innovative technology that resulted in serious shortfalls. The potential disruption of cash flow causes banks to be conservative in this area. Economic. Banks have evaluation parameters that they apply. Along with mine life constraints and debt to equity ratio, Worth (1991) stated these as follows. “Protection Ratio: Defined as the present value of the net operating cash flow divided by the outstanding loan balance at any time. The ratio should be equal to or greater than 1.5: 1. Debt service ratio: Defined as the annual surplus cashflow after operating costs divided by principal plus interest payments. The ratio should be greater than 1.4:1.” Beyond this, cash production costs should be in the lower half of industry cost curves, ideally the lower quartile. Firm sales contracts should cover most or all of the debt service (after operating). Premium product quality is important to ensure sales in weak markets. Political Risk. Insurance may be necessary. Participation by the IMF or export development banks or the National Government or may provide some comfort. Currency hedging can mitigate exchange rate risk. Sponsor Risk. All sponsors are not equally bankable. Strong management and an operating history of similar technologies are important for credibility. Equity contribution and credit worthiness are important. Completion Risk. Banks don’t accept completion risk. For non-recourse financing, a completion test will be required to prove the technical and economic success of the project. This will demonstrate to the bank that no substantial changes have occurred to the project as detailed in the feasibility study. The completion test can include such items as: Minimum current reserve estimate Achievement of operating stability Achievement of design production Achievement of design capacity Achievement of design recovery Specified operating cost Specified product quality Cash flow coverage of debt service Debt ceiling (Debvequity ratio) Marketing. Beyond the specific areas noted above, banks in general would like to see a more complete section on product marketing than is normally provided in a feasibility study document. The price projections for the particular metal in question as well as an assessment of the competitive position of the project in the market are essential. A brief checklist is as follows:

0 0 0 0

Industry structure and competition Historical and projected demand-supply curves Consumption outlook Technological changes or potential substitutions Project competitiveness in relation to operating mines Specific marketing advantages J disadvantages

Sponsors. A detailed description of the sponsors is required, including the following: 0 0

Previous project experience in this type of project Experience in the geographical area Present operations and their financial performance

308

0 0

Current management strengths Specific form of association of sponsors (joint venture, partnership, etc.)

Financing plan. In conjunction with the description of the project sponsor, a preliminary financing plan should also have been prepared. This will show the major sources of financing, as follows: 0 0

0 0

0

Equity contribution from each partner Disbursements schedules "Bank" loans from governmental agencies, international or commercial banks Overrun financing Completion guarantees

The importance of having the final feasibility study prepared by a reputable and independent consulting engineering firm in the mining field should not be underestimated. The work carried out by the sponsor should always be clearly identified as such to forestall any potential misunderstandings.

ACKNOWLEDGMENTS We wish to thank Fluor Daniel Wright Ltd. for permission to publish this paper. REFERENCES McQuat, 1992, Preliminary Feasibility Studies, Presented at the CIM Mineral Economic Society Symposium, January, Vancouver, B .C. Worth, 1991, How Banks Assess Mineral Properties and Companies, Presented at the CIM Mineral Economic Society Symposium, January, Vancouver, B.C. Halupka, 2002, Mining Project Finance Explained, Proceedings Mineral Processing Plant Design, Control and Practice McNulty, 1998, Developing Innovative Technology, SME Mining Engineering, October

309

Major Mineral Processing Equipment Costs and Preliminary Capital Cost Estimations

'

Andrew L. Mular, P. Eng., Distinguished Member SME-AIME, Fellow CIM

ABSTRACT Estimating the cost of major equipment is unnecessary when a firm price has been established by the manufacturer/supplier. However, estimating prices of equipment is important for a number of reasons discussed herein, where Cost Indices are employed to adjust from any historical base to current or future values. Capital Cost estimates, depending upon required accuracy, are grouped into three classes by the American National Standards Institute (ANSI Standard Z94.2), although five major types of fixed capital cost estimates have been recognized by the AACE in past years. Estimation via order of magnitude methods (Type 1 or Class I) and factored or ratio methods which rely on major equipment costs and some prior information (Type 2 grading between Class I and 11) provide rapid preliminary estimates and are discussed in some detail. Feasibility estimates (Class I1 with up to 40 percent or more of engineering completed and Class 111) are discussed in another chapter (Scott and Johnston, 2002) of the Financial and Feasibility Studies section of this volume. INTRODUCTION A variety of cost estimation publications are available. Certified cost engineers, who are members of the American Association of Cost Engineers, have ready access to the Cost Engineers Notebook. For our purposes several text-like sources relevant to the mining industry are listed under REFERENCES for convenience. This paper is adapted from previous papers (Mular, 1978; Humphreys and Mular, 1982) and from a recent handbook (Mular and Poulin, 1998) entitled CAPCOSTS which is useful for estimating costs of major mining and mineral processing equipment, for estimating capital expenditures, and for evaluating mineral projects. CAPCOSTS was published as CIM Special Volume 47 by the Canadian Institute of Mining, Metallurgy and Petroleum and is an update of CIM Special Volume 25 (Mular, 1982) entitled Mining and Mineral Processing Equipment Costs and Preliminary Capital Cost Estimations. MAJOR MINERAL PROCESSING EQUIPMENT COSTS Usefulness Typically, we wish to obtain with minimal effort an estimate of the cost of major equipment when (1) a procedure is being used for capital cost estimation that requires estimates of major equipment costs, (2) the cost of a standard item is being compared with an item of similar function but of different design, (3) an existing circuit is being expanded, so that additional equipment must be purchased, (4)a new plant is being designed and several alternative circuits, which provide similar gradeshcoveries, must be compared, ( 5 ) existing equipment is worn out and must be replaced and (6) you are registered in a mining and mineral processing plant design course that requires capital and operating cost estimates for reports. With time an important factor, "quickie" equipment costs may be needed. Suppliers Chances are high that we have had contact with several of the major suppliers of mineral processing equipment. However, from time to time, we may search for suppliers of new and/or 'Mineral Processing Professor Emeritus, Dept. of Mining, University of B.C., Vancouver, B. C., Canada

310

unfamiliar equipment. Fortunately, in North America, mining and mineral processing equipment suppliers and manufacturers advertise in journals such as:

Mining Engineering, CIM Bulletin, Engineering and Mining Journal, Canadian Mining Journal and Chemical Engineering. Each year, the latter three will print equipment catalogs that have proven useful. In addition, most libraries house Frasers USA (or Canadian) Trade Directory, which is an extensive compilation of virtually all manufacturers/suppliers. Alternatively, the Thomas Register (Thomas Publishing Company, Rexdale, Ontario) reportedly locates the most qualified suppliers in all of North America with ease. Perhaps the best source for our purposes is the Canadian Mining Journal's Mining Sourcebook published each year. The Sourcebook contains an up-to-date Buyers' Guide in two parts: the first is a Products and Services Listing and the second is a List of Suppliers of corresponding products/services of interest. Lists are updated each year and reflect name changes due to acquisitions and expansions. The Sourcebook can be found in most libraries and most mining companies purchase it every year. Importance of Specifications The cost of equipment depends on various factors such as the basic item and its mass and/or volume, the complexity of its design, the materials of construction, the choice of accessories and the nature and complexity of the drive train and motor. Specifications must be clearly stated. Thus a lubrication system, if not carefully specified, may be different from previous ones which were more reliable and less noisy. The supplier might choose to supply something less costly to permit of a lower bid. Unless specifications are clearly stated, supplierdmanufacturs are faced with providing quality components subject to, in most cases, bidding competition. It is of interest that a buyer need not send out for bids and a choice may depend on factors other than price (such as availability). In choosing a supplier, the buyer might first ask for the delivery date, perhaps because he has construction deadlines to meet in order to satisfy financial agreements concerning startup dates. In most situations, buyers will purchase that which their experience has shown to be reliable. When choices are possible (generally the case) cost, availability, size-to-capacity and other factors become of prime importance. Cost Indexes Cost indexes are ratios used to estimate current prices of equipment from obsolete prices. Where the price of an item at some time in the past is known, the current price is estimated from:

There are a large number of indexes available. The American Association of Cost Engineers (AACE) Notebook lists 13 building cost indexes, 9 general constructiodequipment indexes, 7 plant constructiodequipment indexes and 7 miscellaneous ones. Obvious questions are: What is a cost index? Where does it come from? How do you use it? A cost index is a ratio of costs at a particular time to costs at a specified base year. Where the price of an item at some time in the past is known, the current price can be estimated in several ways by application of a cost index. Such methodology has been in use since the early 1900's. Cost indexes are based on mean costs over a period of time. Their accuracy varies widely for individual items of equipment but in some cases can be as high as f 10 percent. Indexes can be very inaccurate after about 5 or 6 years. In consequence, cost indexes should be restricted to order-of-magnitudeand factored estimate costing methods. Of the various cost indexes available, the following are commonly employed in the process industries: (1) Engineering News-Record construction index (ENR), (2) Marshall and Swift cost

31 1

index (M&S), (3) Chemical Engineering plant construction cost index (CE) and (4) Nelson Refinery construction cost index (NR).Each index is based on certain specific information (see Cost Engineering, October, 1968). It has been argued that it does not matter which of the four indexes is used, because the difference between them is within the accuracy of factored and orderof-magnitude cost estimation methods. If the accuracy of these two methods is improved in a particular industry, the selection of an index will be important. The M&S index for mining and milling, denoted by M&S (Minemill) or M&S(M/M), is recommended herein. Current values of cost indexes and other economic indicators are published in various journals such as Chemical Engineering, Oil and Gas Journal and Engineering News-RecordMagazine. Figure 1 shows the variation of the M&S(M/M) Index versus year from 1982 to 1996 (Mular and Poulin, 1998). As of September, 2001 the index was approximately 1132 (for further information contact www.che.com or refer to Chemical Engineering, September, 2001).

1150

A

750

1981 1983 1985 1987 1989 1991 1993 1995 1997

YEAR Figure 1: M&S(M/M) Index versus year (after Mular & Poulin, 1998) Example of Use of Cost Index. A grinding mill was purchased for $5,000,000 in May of 1997 when the M&S(M/M) Index was 1087. What is the approximate cost of this unit today? Letting “today” be September, 2001, the M&S(MinelMill) Index is 1132. Using Equation (1):

= [$5,000,000[~] 1087 = $5,207,000

If the mill had been purchased in December of 1997, a more representative index would be the average for the years 1997 and 1998. Clearly, the application of a cost index is an averaging procedure. If an index for each item of major equipment was estimated in some way and

31 2

compared yearly, significant differences would be apparent between items and from year to year. Accuracy is lost over the years for many reasons. For example, significant equipment changes and modifications may take place because of the tendency to reduce costs and to build equipment with large capacity to size ratios. Costs will change accordingly. Quickie Equipment Costs Several ways to obtain quickie equipment costs are to phone suppliers, to use a cost index with file data, to use a cost index in combination with the 0.6 or 0.7 rule and to use a cost index in combination with cost/ parameter relations developed from old and/or new cost data. Phone Suppliers. This is an obvious way to obtain the approximate cost of an item of equipment. However, there is a financial penalty for both the potential buyer (often he only needs a cost and never buys) and the supplier in that phone calls are necessary and some manhours are consumed. Most likely, the price cannot be provided immediately, so that a delay is involved. Clearly, individuals, both buyers and suppliers, may invest a substantial time to cost estimations. Rapid techniques are certainly called for, unless the project is beyond the preliminary stage. Cost Index With Minimal File Data. It is possible that an item of equipment identical to one of interest was purchased several years ago. Because there is a record of the transaction, the specifications and the cost are in your files. The current cost can be estimated by means of Equation (1). Cost Index With Cost Versus Parameter to 0.6 or 0.7 Power. Average costs of major equipment have been observed to roughly follow an expression written as: Cost = aparameterp7

(2)

The exponent 0.7 has been replaced by anything from 0.6 to 0.667 to 0.7. The choice of exponent is dependent on the experience of the cost estimator for a particular industry and upon the degree of conservatism exhibited by the estimator. The parameter can be mass, volume, capacity, dimensions, area, power draw and any other, including any combinations (such as capacity times dimension) that work. Mass or size or capacity are first guesses. A useful expression can be derived from Equation (2), namely:

(3)

For example, suppose that the cost of an item of equipment is sensitive to its capacity. If the cost is $265,000 at a capacity of 400 stph, what is the cost of a similar unit having a capacity of 500 stph? From Equation (3):

’”I:[

(Cost), = (265,000) -

= 309,800

If the exponent is 0.6, this becomes 303,000 rounded to the first 3 digits. Cost Index With Cost Versus Parameter Relation Found From Data. When cost data (either from old files or from new quotes or other sources) are available for specific items of equipment it is possible to find an equipment parameter to which the cost is sensitive. Data must be placed onto a common index basis, so that the value of the index associated with each cost must be known. This cost is then converted to a common index using Equation (1). For example, suppose

313

that file data show that in June of 1991, when the M&S (Minemill) index = 959, the cost of a ball mill of standard specification is $800,000. To place this onto a common index basis of 1400, determine:

(Cost),,

[-

= (800,000)9,,

]:!:l

= 1,167,900

The choice of a common base index of 1400 is arbitrary, but convenient. With each cost on a common basis, graphs of (Cost)14~versus an equipment parameter can be plotted on log-log paper in search of straight lines. Care must be exercised to employ costs of an equipment item of comparable specifications. Thus, if costs are available for 8 different ball mills, 6 of which have single pinion drive trains and 2 have dual pinion drives, the latter two must be analyzed separately for obvious reasons. Specifications are important to the accuracy of an estimate! When data appear to fall onto a straight line, a non-linear least squares method can be employed to fit the equation:

(cost),,,

= a[Parameterp

(4)

Here a is a constant that depends on a variety of factors and b is a constant that varies with basic equipment type, structural features, design, efficiency and other things. Coefficient a can be viewed as the "intercept" of a log-log line (at Parameter = l), while coefficient b is the slope. If the cost versus parameter data are not linear on log-log paper, alternative equations can be employed, although this additional complexity can be avoided by establishing ranges over which the data appear to give straight lines. The above equation is then fitted to each range. Of course, a and b will likely vary from range to range. After (Cost),a is estimated from the cost vs parameter equation, it must then be converted to the current cost index. Suppose (Cost)lm was determined to depend upon the diameter. At a diameter of 15 ft. the (Cost)14~is $1,170,000. What is the cost of this item when the M&S (MineNill) index is 1132? Using Equation (1):

[:3:]

(Cost), 132 = ($170,000)- = $946,029 This methodology has been used extensively in CAPCOSTS, where the Marshall and Swift Cost Index (Mining & Milling) is employed for updating equipment costs that have been adjusted to a common base value of 1400. Current Index values are found in the journal, Chemical Engineering, McGraw-Hill, 1221 Avenue of the Americas, New York, NY, 10020.

PRELIMINARY CAPITAL COST ESTIMATION Terminology Capital cost estimation is important to many facets of mining and mineral process engineering, where decisions must be supported by financial analyses. The total capital investment in a mine or a concentrator (mill) consists of a m e d capital portion and a working capital portion. Fixed capital is the total amount of money needed to purchase the necessary equipment, buildings and such auxiliaries as site preparation, preproduction development and utilities. Working capital represents cash that must be available to begin the operation. Of major importance is the fact that the total capital investment is a sum that is separate from the totalproduction cost (total operating cost) paid out per unit time from gross income per unit time. Total product costs include operating costs such as direct production costs, fixed charges and plant overhead, and general expenses such as administrative costs, distribution and marketing costs, research and development costs and gross earnings expenses.

314

Purpose of Estimates Preliminary capital cost estimates are useful to engineers who are involved in selection and design. Since most companies have a limit to available capital funds and/or lines of credit, an immediate assessment of initial cash requirements is provided. In situations where capital funds must be borrowed, the less borrowed the better. In effect, a preliminary estimate may serve as the basis for a “go or no-go” decision. In situations where the total product costs of alternative, but technically feasible, processes are similar, the one with smallest capital expenditure is favored. Here the preliminary estimate serves as a way to assess processing alternatives. Obviously, depending upon their degree of accuracy, capital cost estimates will serve many other purposes. These include participation in feasibility studies, plant expansion, and presentation of bids. Types of Fixed Capital Cost Estimates Five major types of fixed capital cost estimates have been proposed by the American Association of Cost Engineers. These are:

(1)

Order-of-Magnitude estimate based on previous cost data and minimal knowledge. It is a ratio estimate with confidence limits that exceed f 30 percent. (2) Factored or ratio estimate based on major equipment costs with a probable accuracy within 30 percent. (3) Budget authorization (preliminary) estimate based on sufficient data to permit. fund procurement and budgeting. The probable accuracy is within f 20 percent. (4) Definitive (project control) estimate based on almost complete data (some specifications missing and drawings not complete). Probable accuracy is within f 10 percent. (5) Detailed (contractors) estimate based on complete engineering drawings, specifications and site surveys. The probable accuracy is f 5 percent. A Type (1) estimate does not have the flexibility of the Type (2) factored estimate; the latter allows personal judgments to be made. A Type (3) estimate becomes more time consuming and expensive compared to Types (1) and (2). Types (4) and (5) involve substantial time and money. Their extra accuracy is usually not justified when the preliminary feasibility of a project is still under evaluation. In more recent years, the American National Standards Institute has grouped estimates into three classes (ANSI Standard 294.2). Class I is an order-of-magnifude estimate with an accuracy of +50 to -30 percent; Class II is a preliminary (budget) estimate with an accuracy of +30 to -15 percent; Class Ill is a definitive estimate with an accuracy of +15 to -5 percent. Accuracy can be increased as more and more information on the basic mining and processing method, capacity, equipment specifications, flowsheets, civil, structural, electrical, mechanical drawings, arrangement drawings, piping and instrumentation diagrams, site layouts, and the like are acquired. Figure 2, summarizes the accuracy of various classes. Accuracy is well within the range of ANSI specifications.

315

’

30 20

10 0

10 20

30

I OEPENOIMG ON PROJECT SIZE VARIES FROY SEVERAL OAVS EFFORT TO SEVERAL WEEKS

I

OEPENOIYG OM PROJECT SIZE VARIES FROM SEVERAL WEEW TO SEVERAL MONTHS

I

-

35 45 % OF ENOINEERIMG COM?LETE PLUS F I R M 8lOS ON MAJOR EOUIPYEMT AN0 ACTIVITY STARTED IM FIELD

Figure 2: Accuracy Ranges For Various Estimates (after McKellar, SME Preprint 75-B-23) In this paper, estimates are expected to have the accuracy of Type (1) or Type (2)which serve as a rapid check and may lead to further investigation of design and layout decisions. In general, the cost of preparing an estimate will increase with the desire to increase its accuracy. Estimation costs can be as high as 2 percent of total project cost!

Useful Prior Information Regardless of the method employed to estimate capital expenditures, prior technical information will be necessary. Specific details on the orebody and ore types, mining method (s), the basic flowsheet for milling, materiavenergy balances, major equipment necessary, approximate equipment sizeshapacities, infrastructure envisaged and other details may be necessary. Working Capital Cost Estimation Working capital cost may be estimated from the following, provided the corresponding information is available: Raw materials inventory (1 month supply at cost) (1) (2) Materials-in-processinventory (1 month supply at cost) Product inventory (1 month at manufactured cost) (3) (4) Accounts receivable (1 month at selling price) (5) Available cash (to meet expenses of wages, raw material, utilities, supplies for I month at manufactured cost) (6) Working capital = (1)+(2)+(3)+(4)+(5) where the Canadian Mining Journal's Mining Sourcebook may be helpful for estimating (1) through (5).

31 6

OHara (see "Quick Guides to the Evaluation of Orebodies", CIM Bulletin, February, 1980) recommends that working capital be equivalent to 4 months of estimated operating costs on a full production basis. Another alternative has been to estimate working capital required as a percentage of the fixed capital investment. Anywhere from 10 to 20 percent of fixed capital will likely be necessary, with 12 to 15 percent being reasonable. Some other methods have been reviewed in CAPCOSTS (Mular and Poulin, 1998). When concentrators (or mills) are constructed with 2 or more processing circuits in parallel, it is possible that one of them will be operational before final construction and start up of the others (one after another is brought into production). The idea is to generate income as rapidly as possible.

Fixed Capital Cost Estimation Via O'Hara Method For Processing Plants Many operating mines supply ore for their own concentrators (or mills) and ship their concentrates to treatment plants such as smelters. For this reason, estimators have developed techniques to estimate the capital cost of mine-mill complexes, where estimation methods for open pit mines with mills differ from those for underground mines with mills---see CAPCOSTS (Mular and Poulin, 1998) which deals with mining and mineral processing equipment costs and reviews preliminary capital cost estimating for mines, mills and mine-mill complexes. In contrast, this paper only reviews typical preliminary capital cost estimation strategies for mineral processing plants, as typified by the O'Hara method (O'Hara, 1980) and the popular ratio and factored estimation techniques (Balfour and Papucciyan, 1972). O'Hara has itemized mineral processing fixed capital costs which were divided into the following categories: (1) (2) (3) (4) (5) (6) (7) (8) (9)

Plant-Site Clearing and Mass Excavation Concrete Foundations and Detailed Excavations Crushing Plant, Coarse Ore Storage, Conveyors Concentrator Building Grinding Section, Fine Ore Storage Flotation and/or Processing Section Thickening and Filtering Section Concentrate Storage and Loading Tailings Storage (10) Electric Power Supply and Distribution (Minemill) (1 1) Water Supply (MineMill) (12) General Plant Services; Access Road to Main Thoroughfare; Townsite and Housing estimated as Infrastructure (13) Feasibility Studies, Design Engineering, Technical Planning ( 14) Project Supervision, Contract Management, Expediting and General Construction Facilities including Camp Costs ( 15) Administration, Accounting, Legal, Pre-Production Employment of Key Operating Staff Cost-Parameter equations were developed for all items other than items (11) to (13) which are calculated as percentages of the others. These are summarized in Table 1.

317

Table 1: Summary of Mineral Processing Plant Capital Cost Estimation* Graph cost Cost Item Parameter Range Eauation Comment 1) Clear/excav. T=capacity, stpd 500-7000 C1 = 86924 Fs 9 . 3 Fs = site factor 2) Foundation

T=capacity, stpd 500-7000 C2 = 43463 Fc f l . 5

3) Crush/conv.

T=capacity, stpd 500-7000

C3 = 97790 f l . 5

4) Mill bldg.

T=capacity, stpd 500-7000

C4 = 65193 Fw 9 . 5 Fw = climate factor

Fc = rock factor

5 ) Grindktorage T=capacity, stpd 500-7000 C5 = 17386 Fg 9 . 7

Fg = grind factor

6) Flotation/etc. T=capacity, stpd 500-7000 C6 = 5433 Fp

Fp = processing factor

f l a 7

7) Thicken/filt. T=capacity, stpd 500-7000 C7 = 10866 Ft 8) Con. storage TC=con., stpd 20-500 C8 = 8693(Tc)0.8 9) Tail pond 10) Power, lines

T=capacity, stpd P=peak load kw

500-700

Cg = 6520 f l . 5

2000-3oooO C l o l = 99963 Po.6 C102 = 9780 f l . 6

M=miles of lines =O if paid by utility 11) Water

Q = water IGPM 500-6500 L = miles pipe

Ft = thickening factor Dam; flat terrain

coal-fired generator diesel generator

C103 = 761Po-8+ 13038M utility substation C i a = 1305 fi.8

low volt lines

C111 = 761 LQo.9

pipe costs

C 112 = 4999 Qo.6

fresh water pumps

C 113 = 6520 Qo.6 reclaim water pumps 12) Plant services, access roads, townsite, housing estimated as part of infrastructure. Item 10) and Item 11) include mine requirements as well. 13) Feasibility [4 to 6% of ((1)+(2))] + [6 to 8% of (sum of items (3) to (ll))] plan,design 14) Supervisekamp 8 to 10% of sum of items (1) to (1 1) 15) Admin, staff 4 to 7% of sum of items (1) to (11) *Adapted from O’Hara (CIM Bulletin, 1980) To use the above table, water requirements, peak power loads and O’Hara factors must be found from the table below. O’Hara factors are obtained from Table 2 on the next page.

-

From: Application 0,IGPM plentiful supply; 1 mile away Q = 12 To6 Fresh Water Fresh Water scarce supply; open pit, hi tonnage Q = 2.5 To6 reclaim when fresh supply scarce Reclaim Water Q = 0.26 TI.’ Peak power loads, P, for open pit/mill complexes, are found from P = 136 For underground mine/mill complexes use P = 27 To7 where T is mill capacity in stpd. Concentrate tonnage is the mill capacity (stpd) multiplied by fractional recovery and the ratio of feed grade to concentrate grade.

318

Factor Fs = site factor Fc = rock factor F, = climate factor Fg = grind factor Fp = process factor Ft = process factor

Table 2: O’Hara Factors for Table 1 ARRlication Value flat sites; less than 10 ft of overburden 1.o moderate slopes; some blasting required 1.5 steep slopes; extensive blasting required 2.5 solid rock for foundation support 1.o gravelhand as support 1.8 moist soil as support; piled foundations 3.5 mild climate 1.o cold climate 1.8 severe climate 2.5 soft ores; 55% -200 mesh; work index under 12 1.o medium ores; 70% -200 mesh; work index = 15 1.5 1.8 hard ores; 80% -200 mesh; work index = 17 1.o Au ores; cyanidation 1.2 flotation; coarse low grade Cu ores 1.6 flotation; hi grade CdZn ores 2.0 selective flotation; complex base metal ores 3.0 complex Au ores; float, roast, cyanide 5 .O gravity concentration 1.o low grade Cu ores 1.6 hi grade Cu/Zn ores 2.0 complex Pb/Zn/Ag or Cu/Zn/Pb ores 3 .O cyanided Au ores

-

It should be noted that item (12) in Table 1 is estimated as part of infrastructure and that Items (10) and (1 1) include mine requirements as well. Infrastructure costs involve General Plant Services; Access Road to Main Thoroughfare; Townsite and Housing; Feasibility, Planning and Design; Supervision and Camp; Administration, Accounting, Legal and Key Staff. These are estimated from Table 3 below. Table 3: Summary of Infrastructure Costs For O’Hara Method Cost Item Parameter * . ; Comment 1) Plant services N = # of employees 2) Access road R = miles of road C131= 65 1930 R Road b = bridge length, ft (2132 = 283 b’.’ 3) Townsite housing N = # employees C141= 119520 N Family Townsite Bunkhouses C142 = 43463 N [4 to 6% of item (2)] + [6 to 8% of (items (1) + (3))] 4) Feasibility, Plan, Design 5 ) Supervision,Camp 8 to 10% of (items (1) + (2) + (3)) 6) Admin, staff 4 to 7% of (items (1) + (2) + (3)) Items (1) and (3) depend upon an estimate of the number of employees at the mine/mill complex; item (2) depends on estimates of the lengths of roads and bridges. Items (4) to (6) reflect a portion of these costs that are associated with infrastructure and previously estimated as part of the processing plant. Equipment-Tool dependent cost items (1) through (1 1) of Table 1, that are at least somewhat sensitive to a cost index, are first calculated as shown. Then each must be multiplied by the ratio of the current M&S(M/M) Index divided by 1400. This converts each cost equation from the reference index (converted to the reference from the original data base) to a current index basis.

319

The same applies to items (l), (2) and (3) in Table 3, although these costs may be more sensitive to a wage-dependent index. Fixed Capital Cost Estimation For Processing Plants Via Cost Ratio Methods Four methods to estimate fixed capital costs are: (1) Six-Tenths Rule, (2) Plant Cost Ratio Method, (3) Equipment Cost Ratio Method and (4) Plant Component Cost Ratio Method. To use methods (2), (3) and (4) the following minimum information should be available: a rough flowsheet showing major items of equipment and their corresponding sizes along with sufficient information to estimate plant size and complexity. Material balances -- especially for recycle streams -- and energy balances should be obtainable. Methods (3) and (4) are critically dependent upon cost information files and records of actual costs. Type 2 accuracy is obtainable, provided factors employed are accurate. When factors are determined from file data, the accuracy can range between Class I and Class 11. Six-Tenths Rule. To use the six-tenths rule, the fixed capital cost of a plant of known capacity must be available. In such cases Equation (3) can be rewritten: (Cost)* (Cost),

- [(Parameter),

(5)

(Parameter),

Suppose a concentrator treating 10,OOO stpd cost $60 x lo6to build in 1989, when the M&S(M/M) Index is 9 11. What is the cost today of a similar plant that will treat 20,000 stpd. From the above equation: 0.6

(Cost), = ( 6 0 ~ 1 0 ~

= $90.94~10~

at an M&S(M/M) Index of 91 1. To estimate the cost today, the current cost index of , say, 1132, can be employed. Thus Equation (1) can be written as: (Cost),,,

= (9o.94X1o6

[z]

= $113x1O6

The exponent 0.6 is an average and depends on the type of plant. Some estimators prefer to use a 0.7 rule, since factors such as type of site, economic conditions prevalent, geographic location and regional productivity can be responsible for substantial variation. Plant Cost Ratio Method. This method requires an estimate of the cost of major items of delivered major process equipment. If this cost is equal to N (often inflated to account for auxiliary items) then: Cost of solid process plant Cost of solid-fluid process plant Cost of fluid process plant

= 3.10 (N) = 3.63 (N) = 4.74 (N)

The multipliers 3.10, 3.63 and 4.74 are known as Lang factors and can be broken down into parts as follows: MuitiDiier c/B OH 3.10 1.43 1.10 1.50 1.31 3.63 1.43 1.25 1.50 1.35 4.74 1.43 1.64 1.50 1.35

320

where F/L represents foundations, supports, chutes, installation; P represents piping costs; C/B represents construction, engineering, building; OH represents overhead. The procedure assumes that reasonably accurate estimates of equipment costs are available and nothing is "abnormal". Delivered costs have been estimated as 1.03 times purchased costs, while installed costs are sometimes estimated as Q times delivered costs. The value Q is 1.45 for solids plants; 1.39 for solids-fluids plants; 1.47 for fluids plants (note: average is 1.43). An example of the plant cost ratio method: The current cost of delivered equipment items for a 20,000 stpd primary crushing station happens to be $1.4 x 106, so that the cost of the installation is estimated as: Cost = 3.10 (1.4 x 106) = $4.34 x lo6

Equipment Cost Ratio Method. Compared with the plant cost ratio method, more accuracy results from multiplying categories of equipment of similar nature by corresponding ratio factors and then calculating the resulting sum. Let the cost of the ith major process unit be Ci . Then a "plant" cost associated with item i is equal to Fi Ci where Fi is an appropriate factor. The cost of the total plant is then: n

Plant Cost = C FiCi i=l

(5)

for a plant containing n items of major equipment and key auxiliaries. Table 4 below shows typical Fi values. Such tables may be expanded after several different projects have been analyzed.

Table 4: Equipment Cost Ratios (after Balfour and Papucciyan, 1972) Eauipment Categorv Factor, Fi Bucket Elevator 2.0 Mixer 2.0 Furnace 2.1 Drum Dryer 2.2 2.2 Kiln Conveyor 2.3 2.3 Compressor Electrostatic Precipitator 2.5 2.5 Blower, Fan Refrigeration Unit 2.5 Boiler 2.8 Mills 3.O Vacuum Rotary Dryer 3.2 Dry Dust Collector 3.5 Storage Tank 3.5 Crusher 3.5 Process Tank 411 Instrumentation 4.1 4.8 Heat Exchanger Wet Dust Collector 6.0 5.8 Pump Electric Motor 8.5 Plant Component Cost Ratio Method. This method provides considerable flexibility and involves a breakdown of fixed capital costs into plant components whose costs are a ratio of major equipment costs. The ratios are sometimes referred to as factors.

321

Some case examples of the plant component cost ratio method are shown in Table 5 (Balfour and Papucciyan, 1972) for various types of processing plants. It is of interest to note from the table that the fixed capital cost of an asbestos plant, a zinc refinery and a copper refinery respectively is 2.43,2.28 and 3.24 times the corresponding major equipment costs. A generalized version of Table 5 is given in Table 6, which shows ranges for cost ratios. This table is typical of most plant component cost ratio methods which can be extremely flexible. Components such as geographic location (desert versus far north), type of industry (Cuversus Au versus Cu-Ni), type of labor pool and other components are readily introduced, when cost files of the correct type are available for evaluation. Table 5: Some Plant Component Cost Ratios For Processing Plants Asbesios Plant Zinc Refinery Coppir Refinery Cost Relative Cost Relative Cost Relative to Eauipment to Eauipment Plant ComDonent to Eauipment 1.OO 1.00 Equipment 1.00 0.22 0.2 1 0.19 Equipment Installation 0.17 0.05 0.03 Process Piping 0.32 0.20 0.10 Electrical 0.11 0.06 0.03 Instrumentation 0.62 0.37 0.34 Process Buildings 0.09 0.07 0.10 Auxiliary Buildings 0.15 0.05 0.09 Plant Services 0.13 0.02 0.07 Site Improvements 0.09 0.07 0.13 Field Indirects 0.37 0.24 0.26 Project Management TOTAL,:

2.28

2.43

3.24

Table 6: Generalized Plant Component Cost Ratio Method 1. Delivered equipment costs from references and on current cost index basis. (if delivery costs unavailable use 1.03 times purchased equipment costs). 2. Equipment installation. (0.17 to 0.25 times Item 1). .................................... 3. Piping, material and labor, excluding service piping. (0.07 to 0.25 times Item 1). .................................... 4. Electrical, material and labor, excluding building lighting. (0.13 to 0.25 times Item 1) ..................................... 5 . Instrumentation. (0.03 to 0.12 times Item 1) ..................................... 6. Process buildings, including mechanical services and lighting. (0.33 to 0.50 times Item 1 ) . .................................... 7. Auxiliary buildings, including mechanical services and lighting. (0.07 to 0.15 times Item 1). ..................................... 8. Plant services such as fresh water systems, sewers, compressed air, etc. (0.07 to 0.15 times Item 1). .................................... 9. Site improvements such as fences, roads, railroads, etc. (0.03 to0.18 timesItem1) ..................................... 10. Field expenses related to construction management. (0.10 to 0.12 times Item 1) ..................................... 11. Project management including engineering and construction. (0.30 to 0.33 times Item 1). .................................... 12. Fixed capital costs = 1+2+3+4+5+6+7+8+9+10+11.

322

................

$000,000

$o0o,o0o $o0o,o0o $OO0,000

$00O,OOO

$OOo,o0o $000,0o0 $OOO,OOO $000,000 $000,000

$o0o,oO0 $OOO,OOO

Factored Capital Cost Estimate Guide. Table 7 shows a tabular guide (Vilbrandt and Dryden, 1959) for estimating capital costs by means of alternative components that depend upon strong subjectivity and prior experience. The table and corresponding updated versions have been used over the years by Chemical Engineering estimators for chemical plants of various kinds. Note that the table gives the estimator a substantial amount of flexibility, which implies that it can be employed for mineral processing plant cost estimation if suitable data have been acquired for selection of factors. Other Preliminary Capital Cost Estimation Methods For Processing Plants Additional methods relevant to plant cost estimation in North America have been developed for mining and milling by estimators at the US Bureau of Mines (Stebbins, 1987; Camm, 1991) and CANMET (J. S. Redpath Limited, 1986). These have been reviewed in CAPCOSTS. For Small Placer Mines. The Cost Estimation Handbook for Small Placer Mines (Stebbins, 1987) was published by the US Bureau of Mines as IC 9170. Capital and Operating cost equations, which may involve one or more multiplying factors, were developed. Methodology to update costs is described in the manual. The Stebbins handbook has filled a gap, in that small placer mines are relatively unique. Mining and Processing costs therein are representative of operations in the Western US and Alaska and should be appropriate for Canadian placer operations in the Klondike and Yukon Territories. Simplified CapitaYOperating Cost Estimation Models. The US Bureau of Mines Information Circular 9298 (Camm, 1991) presents quickie estimates of the cost to develop mineral deposits such as gold in the southwest. Costs are based on average 1989 US dollars, where a number of building cost indexes, as well as the M&S(M/M) Index are employed. The methods are adaptable to most deposits. Cost-Capacity equations must be updated before using. The costing procedure incorporates open pit mine models, underground mine models and mill models that include gold processing circuits such as C L E W , CIP-EW, CCD-MC, Autoclave-CIL-EW and Solvent Extraction-EW. Suitably updated, the method permits the estimation of capital and operating costs with minimum effort. Preproduction and Operating Costs of Small Underground Deposits. To estimate costs of mining and processing small underground deposits, J. S. Redpath Ltd prepared a manual which was published as SP 86-llE by CANMET, Ottawa, Ontario, Canada. Both capital and operating costs were estimated in some detail for mining, although associated processing costs are very general (a cost-capacity curve for the concentrator and another for the tailings disposal area). Methodology can be useful. A sensible calculation procedure, which uses a series of blank forms to be filled out via computational procedures presented in the body of the text, was recommended. CONCLUSIONS The estimation of mineral processing equipment costs is important for a variety of reasons. In particular, preliminary capital cost estimation methods may involve cost components (factors) that are proportions of total equipment cost. Common preliminary capital costing procedures have been reviewed. For complete Class I1 and Class I11 accuracy, the reader is referred to: Scott, John and Brian Johnston, 2002. Guidelines to Feasibility Studies, in this volume of Mineral Processing Plant Design Practice and Control. It is recommended that the REWRENCE list be consulted before attempting an estimation.

323

Table 7: Factored Capital Cost Estimate Guide ( patterned after Vilbrandt, Frank C. and Dryden, Charles E, 1959. Chemical Engineering Plant Design,., McGraw Hill, N.Y.) 1. Purchased equipment costs from references and on current index basis ......................................................................................... $000,000 2. Installed equipment costs a. From references; current index basis................................................................. $OOO,OOO b. Item 1 multiplied by 1.43.................................................................................. $000,000 3. Process piping............................................................................................................ $ ~ Percent of Item 2: Type plant: Solid 7-10 Solid-Fluid 10-30 Fluid 30-60 4. Instrumentation.......................................................................................................... .$000,000 Percent of Item 2: Amount of automatic control: None 2-5 Some 5-10 Extensive 10-15 5 . Buildings and site development.................................................................................. $000,OOO Outdoor 5-20 Outdoor-Indoor 20-60 Indoor 60-100 6. Auxiliaries (e.g., electric power) ............................................................................... $OOO,OOO Extent: Percent of Item 2: Existing 0 Minor additions 0-5 Major additions 5-25 New facilities 25-100 7. Outside lines............................................................................................................... $000,OOO Percent of Item 2: Average length: Short 0-5 Intermediate 5-15 Long 15-25 8. Total physical plant costs = Sum of Items 2+3+4+5+6+7.......................................... $000,000 9. Engineering and construction .................................................................................... $OOO,OOO Complexity: Percent of Item 8: Simple 20-35 Difficult 35-60 10. Contingencies........................................................................................................... $000,OOO Percent of Item 8: Type process: Firm 10-20 Subject to change 20-30 Speculative 30-50 30 Average 11 .Size factor................................................................................................................. $000,~O Percent of Item 8: Size plant: Large commercial 0-5 Small commercial 5-15 Pilot plant 15-35 $000,000 12. Fixed capital costs = Sum of Items 8+9+10+11.....................................................

324

,

~

REFERENCES (1) Mular, A. L., 1978. The Estimation of Preliminary Capital Costs, in Mineral Processing Plant Design, Eds. Andrew L. Mular and Roshan B. Bhappu, Society of Mining Engineers of AIME, Littleton, CO. (2) Humphreys, K. K. and Andrew L. Mular, 1982. Capital and Operating Cost Estimation, in Design and Installation of Comminution Circuits, Eds. Andrew L. Mular and Gerald V. Jergensen 11, Society of Mining Engineers of AIME, Littleton, CO. (3) Mular, Andrew L. and Richard Poulin, 1998. CAPCOSTS: A Handbook For Estimating Mining and Mineral Processing Equipment Costs and Capital Expenditures and aiding Mineral Project Evaluations, CIM Special Volume 47, CIM, Montreal, Quebec, Canada. (4) Mular, Andrew L., 1982. Mineral Processing Equipment Costs and Preliminary Capital Cost Estimations, CIM Special Volume 25, CIM, Montreal, Quebec, Canada. (5) O'Hara, T. Allan, 1980. Quick Guides to The Evaluation of Orebodies, CIM Bulletin, February, pp 87-99. (6) Balfour, R. J. and T. L. Papucciyan, 1972. Capital Cost Estimating For Mineral Process Plants, Proceedings of the 4th Annual Meeting of the Canadian Mineral Processors, CIM, Ottawa, Ontario, Canada. (7) Stebbins, Scott A., 1987. Cost Estimation Handbook for Small Placer Mines, IC 9170, United States Department of the Interior, US Bureau of Mines, Washington, D. C. (8) Camm, Thomas W., 1991. Simplified Cost Models For Prefeasibility Mineral Evaluations, US Bureau of Mines, IC 9298, Department of the Interior, Washington, D. C. (9) J. S. Redpath Limited, 1991. Underground Metal Mining: Estimating Preproduction and Operating Costs of Small Underground Deposits, CANMET SP 86-1 lE, Ottawa, Canada.

Additional text-like cost estimation sources relevant to the mining industry are: (a) Guthrie, Kenneth M., 1974. Process Plant Estimating, Evaluaton and Control, Craftsman Book Company of America, Solona Beach, CA., ISBN 0-910460-5-1. (b) STRAAM Engineers, Inc., 1979. Capital and Operating Cost Estimating System Handbook Mining and Beneficiation, US Bureau of Mines, OFR 10-78. (c) Hoskins, J. R. and W. Green (Eds), 1982. Mineral Industry Costs, Northwest Mining Association, Spokane, WA,, ISBN 0-931986-02-6,248pages. (d) Woods, Donald R., 1983. Cost Estimation For The Process Industries, McMaster University Bookstore, McMaster University, Hamilton, Ontario. (e) __________ , 1987. Bureau of Mines Cost Estimating System Handbook: Part 1. Surface and Underground Mining, IC 9142; Part 2. Mineral Processing, IC 9143; United States Department of the Interior, US Bureau of Mines, Washington, D. C. (f) Ruhmer, W. T., 1991. Handbook On The estimation of Metallurgical Process Costs, 2nd Edition, MINTEK Publication 14, Randburg, South Africa. (g) Noakes, Michael and Terry Lanz, Eds., 1993, Cost Estimation Handbook For the Australian Mining Industry, ISBN 0 949106 87 9, Monograph 20, Aus. IMM, Parkville, Victoria, Australia. , 1994. Mine and Mill Equipment Costs - An Estimator's Guide, Western (h) Mine Engineering, Inc., Spokane, Washington.

_________

325

Process Operating Costs with Applications in Mine Planning and Risk Analysis Doug Habe' and T.J. SmolikL

ABSTRACT This paper discusses techniques for estimating treatment plant operating

costs, including identification of high-impact cost areas and expected key cost variations year-by-year. Application of this idormation to cost development in the mine block model is presented, followed by a discussion of the risk uncertainty and cost-impact sensitivity in operating cost estimates. An example is carried from the initial base cost estimate and year-by-year annual process cost variations through the use of Excel spreadsheetsand Monte Car10 simulation.

INTRODUCTION The question that must be raised when preparing an estimate of a process operating cost is this: What will the cost estimate be used for and what level of accuracy is required in its development? It is also necessary to realize that operating costs will change during the term of an operation's life. The estimator should identify the key parameters that drive these changes. This text will start with development of various types of proms cost estimates, outline procedures for developing these costs, and then outline some methods of examiningthe impacts of major variable changes over the term of the projected mine's operational life. For this paper the battery limits for mill operation will commence with the crusher feed pocket and end with facilities for loading of the final product for transportation off-site. Costs for transportation, smelting, and refining of product are excluded, but are interrelated with other mill costs and should be included in any economic study. Project mining and admimish.ation costs, as well as taxes, amortization, depletion, depreciation, and related items will not be included. Costs are expressed in non-inflated US dollars, and most units, includii tonnes, (t) are metric. The format for cost estimation used in this paper is very similar to that used by most plants for budgeting of on-going operations. Most of the costs for a new milling operation can be estimated quite accurately, given a good base of test data and sufficient effort to obtain accurate wage and price idormation. Thus, an experienced metallurgist should be able to estimateoperating costs using detailed quantitative data and appropriate geological/metallurgical information to within about 1P?of 'final observed values Estimation of operating cost can be required for a number of uses, and we will discuss each of these separately in this paper:

Project evaluation effort with little process basic idormation Advanced studies including feasibility study applications Ore reserve optimization with process costs used in block models Process cost projections, risk areas and economic sensitivities This paper will use the terminology used by most engineering companies working in the mining area (from Pincock Allen Q Holt, 1998):

Doug Hdbe Consuhant P.C., Salt Lake City, Utah

* TJS Enterprises LLC, Blaine, Washington

326

0

0

0

Conceptual study: Based on sufficient drilling to define a resource, but flowsheet development and cost estimation are often based on limited testwork and engineering design Prefeasibility study: Based on a higher level of testwork and engineering; and typically of the order of +/- 30% accuracy. Feasibility study. Sufficient detail and aauracy to be used for a positive “go” decision and financing purposes, with an accuracy of +/- 20% or better.

EARLY-STAGE ESTIMATES ORDER OF MAGNITUDE COSTS In many cases it is necessary to make a very rough preliminary estimate of both capital and operating costs with very little information available. This is the case where exploration geologists have identified an interesting target and need enough idormation to decide whether firther expenditure is warranted. Frequently the only information available will be an estimate of the likely maximum tonnage and a very preliminary head grade, based on the results of a few drill holes. There will be no test data, nor, at this point, any reason to carry out any testwork to develop a flowsheet, reagent quantities, or recoveries. This calls for what is truly referred to as a “back-ofthe-envelope” estimate. Total time expended on looking at some core and making a cost estimate might be a few hours; or, as more information becomes available and a more carehl estimate is wammed, a few days. The initial decision required here by the exploration department is “Yes, this looks, fiom a metallurgical point of View, like the economics might be good; so thther drilling appears warranted,”or “No,the deposit it too small, and it’s located in Alaska 500 km fiom the nearest road, and this is not sphalerite, it’s high-iron marmatite.” Many companies maintain a database of information on costs of their operation and of available cost information of other operations.This infimnaton is typically plotted on a graph of total mill operating cost versus milling capacity in tonnes per day. Figure 1 is such a curve, per day of mill capacity. The data includes showing the total milling costs plotted versus t o ~ m flotation, cyanidation and those using both mills, and is taken directly from the Canadian Mining Journal 2001 Mining Sourcebook The shape of the curve is typical and the relative scatter around the fitted curve is surprisingly small. Curves of this sort can give a very rough estimate of milling costs, but no more, since large variations in reagent costs, ore hardness, power and labor costs and most other fhctorscan exist between similar deposits. Another, more accurate, method hquently used for preliminary estimates is the use of an estimator’s guide. Western Mine Engineering (WME) provides, on a subscription basis, such a guide that includes cost models for various types of mining and milling operations, for both capital and operating costs. Mill operating costs include both cyanidation and flotation plants, with a number of throughput capacities ranging fiom 100 to 80,000 tonnes per day (tpd). An advantage of this system is that,as more idormation becomes available it is possible to improve the ~ccuracyof the cost. A significant amount of detail is provided with each set of costs in the WME guide, so that it is possible to intelligently adjust these costs to better fit an individual prospect. separate costs are given for each model for supplies and materials, labor, administration, and sundry items. These are accompanied by detailed manning charts, and costs for electricity, grinding and wear steel, reagents, and other items. Adjustment of power costs, for example, fiom the WME base of $0.072kWh to the expected project power costs is a simple arithmetic calculation Thus it is possible to periodically improve the aauracy of the estimate as the prospect is developed, but before the stage at which a more detailed estimate is requried or wamnted. Frequently throughout this paper we will refer to the WME Mining Cost Service. This is not intended as an advertisement for WME. It is simply that we have found this an excellent source of usefil information for cost estimating for mining projects. Another good source of information on costs of actual operating mines is World Mine Cost Data Exchange (www.minecost.com). This is a database of shared information fiom mining analysts that gives reasonably detailed cost informationfor a large number of working mines.

327

Figure 1 Total Mining Costs - Canadian Mills: Hotation and/or Cyanidation(CMJ 2001)

ADVANCED STUDnS PREFEASI6ILrrYANDFEASIBILITYSTUDIES As a new project continues to demonstrate positive economics, it reaches a point where a preliminary or prefeasibility study should be carried out; and this is frequently done internally by a mining company. The initial prefeasibility study not only provides the first systematic estimate of capital and operating costs, but also serves to highlight any project areas that have been overlooked, or areas where data is lacking or weak. The basic information needed for the development of prefeasibility level process operating costs is as follows: Metallurgical balance (typically for the lifesfmine average grade) Life of mine production plan Lab andor pilot plant operating data and results Bond ball mill work indices, SAG mill indices, abrasion index data Flowsheet with material balances for each unit operation Basic design criteria Equipment list (key equipment or detailed, depending on the level of the estimate) with motor sizes Water balance Unit costs for labor and conmmables (including fuel and energy) Operating costs are generated initially for average or typical life-of-mine conditions. Development of more accwate costs for a feasibility study varies only in the amount of supporting information and detail required. For a feasibility study the key cost drivers need to be evaluated and adjusted year by year, as discussed later in this paper. For this discussion we will discuss the development of basic, preliminary operating costs in each of the major cost areas of milling, and include in each section some comments and suggestions for further improving the accuracy.

328

Salaries and Wages The cost of labor (supervision, operating, maintenance) can range from as high as 40% of total milling costs for a small, 1,OOO tonne per day plant to perhaps 1oo/o in a large, 50,000 tpd concentrator. Thus, at least for preliminary studies, the accuracy with which this number needs to be estimated will vary with the plant size. However, an experienced metallurgist armed with a flowsheet and design criteria should be able to prepare an organization chart and manning schedule. A less experienced metallurgist might wish to use the manning tables fiom the WME cost engineering notebook for reference (and these can provide a useful check even for an experienced operator). For operating companies, wage and salary costs are available from existing company operation costs, with information on similar operations fiequently available fiom a database in the personnel department. If the new project is located in an area with other mines, or similar industrial operations, wage rates can usually be obtained from this source. State and country governmental agencies may be able to provide typical wage information. Several private services provide such information. A convenient source for detailed mining information is the WME Mining Cost Service, which provides very detailed wage and beneffi costs for a large number of mines, by state and province in the United States and Canada, both union and nonunion. Wage and salary information is convenientlypresented in a form such as that in Table 1. This lists each position, the number of people required in each position, the wage rate or salary, and annual cost. Points to consider:

0 0

0

Fringes or payroll burdens must be included, and are usually included in the wage and salary information. For North American operations these range fiom 35-40% of the base wagdsalary. It is important to ensure that you know exactly what is included and excluded from the burden. Provide additional people to cover for vacations, illness, dumped shifts, and training. Estimate the amount of overtime needed. The level of supervision and manning required will vary significantly by country. Contract labor may be used, and should be inducded in this section.

The data in this and subsequent cost tables has been developed for a hypothetical 5,000 tpd (1,825,000 tonnes per annum) carbon-in-pulp (CIP) plant.

Power nnd Utilities Power, like labor, is one of the major costs of milling operations. Power as a percentage of total operating costs in the Canadian Mining Journal data plotted earlier ranged from 5 to 32%, with the majority of the values falling between 15 and 30%. It is possible to estimate the power requirements and resulting operating costs for a plant with a reasonable degree of accuracy if suf€icient design information is available. Table 2 shows an abridged typical spreadsheet for estimating power costs. This requires an equipment list with associated motors. The largest consumers of power will be the grinding mills, and the power for these mills can be calculated reasonably well with Bond ball mill and SAG mill work indices. For a preliminary estimate this list will probably include only the major items of equipment. Some provision must be made for the additional minor equipment, typically by adding an allowance to the overall load list. It is also necessary to estimate the hours per day of operation of each motor, and the expected power draw. A Illy-loaded and correctly engineered ball mill will draw close to its designated power rating for 24 hours per day. A sump pump might operate only one hour per day, at half or less of its motor power. The totals from the motor list information (excluding standby's) can be used to calculate total kilowatt hour demand, kilowatt-hour usage per day and per tonne of mill feed. For feasibility studies, installed kilowatts, loads, and all related power consumption information will be available from studies by the electrical engineering group. Power rates can be determined by inquiry with the local company supplying power. If sufficient power is not already available at the mill site, then a power line will have to be run. If this is done by a power company, the cost may be included by them in their determinationof rates. 329

Table 1 Wages and salaries Category Salaried

Hourly- opns

Houriy-maint

Job title Mill superintendent General foreman Shift foreman Mill maint engineer Electrical engineer Maint foreman Maintenance planner Metallurgist Met technician Gold room foreman Chief assayer Instrument technician Clerk Total salaried Control room optor Crusher operator Grinding operator Cyanidation operator Gold room crew Day crew Assayers Samplers Laborers Total op'ns - hourly Mechanics Mechanics' helpers Electricians Electricians' helpers Total maint- hourly

No.

$/hr (Note 1)

2

i

17 4 2 4 4 2 2 2 2

z

26.00 21.00 21.00 21.00 26.00 20.00 24.00 20.00 18.00

1

24.00 20.00 26.00 22.00

24 5 3 2 11 52

Total employees Total $/yr Total $ft (on 1,825,000 tpa) Notes: 1. Includes a burden of 40%. For conversion to $/yr, multiply by 2080 (40 hrs/wk x 52 weeks/yr) and by 110% to allow for 10% overtime. 2. For salaried personnel, includes a burden of 38%.

$/yr (Note 2) 100,000 80,000 70,000 90,000 90,000 80,000 50,000 70,000 50,000 75,000 60,000 50,000 40,000

Total */vr 100,000 80,000 280,000 90,000 90,000 80,000 50,000 140,000 50,000 75,000 60,000 50,000 40.000 1,185,000 237,952 96,096 192,192 192,192 118,976 91,520 109,824 91,520 82.368 1,212,640 274,560 137,280 118,976 50.336 581,152 2,978,792 1.63

If it is necessary to generate power on site, suppliers of power generation equipment can provide data for estimating operating costs. For preliminary estimates the WME estimating book contains power cost data by state and province in the United States and Canada. Both energy and demand changes need to be calculated. Demand charges can be quite large and their influence becomes more significant where daily hours are low for high kilowatt equipment, or where the plant itself does not operate on a continuous basis. Total kW (excluding stand-bys) is shown on the equipment list - power cost table (Table 2). Once the monthly demand charge is determined, it can be calculated back to an equivalent energy charge and added to the energy charge, so a single rate can then be applied. If the step of listing equipment kW in a power cost table is not taken, the energy charge should at least be "rounded up" a bit to allow for the demand charge. DO NOT assume, just because a power rate is available from some reference source, that the power company will be able to supply power to your operation. 330

Tmble 2 Power cost (mbridged)

FdCX

1

7

7

convpyor SAGmiU

I

15 2243

15 2243

24 24 24

0.92

125 251 49.525

IS

15

12

0.7

125

104,101

~

1

Freshwaterpump Misc. other

Subtotal - other Total kW and kWday

. .

-

0.7 0.7

I

I

I

I

I

Ii

I

I

I

I

I

I

I

I

=I

4631 I

slyear

other. wear Total, $/year (SO.07kWh)

. .

Gnndmg,sH

I

]TotalkWM (5OOO tpd) ,Notes: (1) Ex~lladesstand-bp

20.8

Heating of buildings can be a significant cost in northern US and Canadian operations, and nonexistent in the southem US. Calculation of this cost is usually outside the scope of work expected of metallurgists, but does need to be determined. (A paper by Barrat, et d.,1975, is old, but does provide some idormation that could be used by a metallurgist pressed to make some preliminary calculations.) Water is a significant concern for an operation. As with power, DO NOT assume that you will automatically find water if you drill for it, or that you can use the water fiom a nearby sourcejust because it is there. It is usually a very good idea to determine a water source (including quality) and to estimate both usage and cost even for preliminary cost studies to ensure that this critical requirement is kept at a high profile. Other than that,there are no g e n d rules here. It is necessary to develop a project water balance to determine the water consumption fbr the project (by season, wet years and dry years). Typically, water will be supplied fiom a well field some distance fiom the plant site. The pumping equipment motors should be included in the motor list, and necessary manpower for maintenance and periodic checking should be included in the manpower list. Other utility costs are typically low: fuel for the mill vehicles, propane for melting furnaces for a gold room, and similar costs. These can usually be included in a small percentage added for miscellaneousutility costs, at least for preliminary estimates.

Reagents, Wear Steel and Supplies The reagents necessary for mill operation are determined fiom lab andor pilot plant testwork. Total reagent cost, for the mills listed in the Canadian Mining Journal 2001 Mining Sourcebook range fiom 5 to 40% of total milling costs, indicating that these costs are highly project specific. There is no substitute for basic testwork. For flotation, the quantities used in laboratory tests are generally close to those expected in mill operation. Some estimators add 10 or 20% to bench-scale lab flotation and cyanidation reagent consumption, and/or decrease the amount somewhat if a large percentage of mill water is recuculated. Cyanide consumption in production heaps, however, can be 25-300/0 of that determined in lab columns (unless cyanicides are present). (Albert 2001; McClelland 2001)

331

Flocculant quantities can be determined by lab thickening tests, and can provide a reasonable estimate of usage. When these tests are run by suppliers, the test report will usually provide the supplier’s best estimate of consumption. For the base case a consumption for each reagent is estimated. As discussed later in this paper, this may vary from f d type to feed type (particularly in cyanidation plants), and this variation must be allowed for in accurate cost estimates, either periodically as the ore changes, or in the ore reserve mining block model. Pilot plants are not typically run just to determine reagent consumptions. However, if the cost of one or more key reagents is a major percentage of the total milling cost, a pilot plant run with recycle of process water, and using more than one ore type, may be required in order to determine costs to the required level of accuracy. Lock cycle flotation tests can also be used to improve the amracy of reagent use estimates. Reagent prices are best determined, even for preliminary estimates, by direct inquiry to suppliers, from actual costs of other plants in the area, or from the company’s corporate purchasing department, particularly where the company purchases on a corporatewide basis. Where this level of accuracy is not required the WME Mining Cost Service lists the cost range of a large number of “Chemicals”. The Chemzd Mmketing Reporter, found in the periodicals section of most good engineering libraries, lists the w e n t prices for many less-commonly-used chemicals. Listed reagent prices c8n vary significantly from year to year and location to location; prices for truck-load quantities or bulk reagent are usually significantly lower than those for small lots; and tieight to plant site may or may not be included in the listed price. Thus listings of prices at best provide only preliminary estimating data. Grinding media and wear steel are significant milling costs, and must be estimated. Steel costs as a percentage of total milling costs can vary significantly, but might typically be 10%. The estimating accuracy in wear steel costs, unfortunately, leaves something to be desired. For a modest fee many test labs will determine the Bond Abrasion Index (AI) for a sample (or better, several samples) of mill feed. This is a purely empirical test, and is done dry. However, metal usage depends not only on ore characteristics, as measured by this test, but on pulp chemistry, steel quality and operating practice. (The last is particularly important in SAG mills.) Empirical formulas developed by Fred Bond using the AI are available for predicting wear for crusher liners, and for wet and dry grinding mill liners and grinding media, in pounds of metal per kilowatt hour. These formulas are listed in the S M E Mineral Processing Handbook, Section 30, under Abrasiveness. (Note, however, that these formulas date back to 1963 and significant improvements in metal quality have been made since that time. Unpublished data indicates that for current high quality metallurgical steel these calculated values could be reduced as much as 50 percent.) A good procedure is to conduct AI tests to determine how the sample evaluated compares with others. With Al information it is possible to review operating data from other plants with similar conditions and AI’s, and make a reasonable estimate of expected wear. Generally the lab performing the tests will have a database of this sort of information. Engineers at the test lab or consulting engineers with extensive experience in grinding circuits can be very u s e l l here. If pulp chemistry is expected to present particular problems (e.g. high-chloride process water) this method may significantly underestimate the metal loss. A u s e l l reference in this area is the chapter “Comminution Energy Usage and Material Wear“(Charles and Gallagher) in Design and Z d W o n of Commimrtion Cirmils (Mular and Jergensen, 1982) Freight, delivered to the plant site, and taxes must be included with ~ ~ p p costs, l y unless handled in a separate overall project account. Maintenance supplies include such things as pump impellers, steel and liners for chutes, pipes and values, and replacementsparts for equipment, but exclude wear steel. This is a difficultcost to estimate. The cost of maintenance supplies in a typical flotation or cyanidation plant ranges from five to ten percent of total direct operating costs. Thus less accuracy in this estimate isless acceptable.

332

Information on similar-sized plants in similar seMce provides a g d estimate, but this information may not be available. A number of rules of thumb for estimating maintenance supply costs are available. The Australasian Institute of Mining and Metallurgy published in 1993 the Cost Estimation Handbook for ihe AWalian Mining Indistry. a 4Wpage book (Nokes and Lanz 1993) that contains detailed information and procedures on cost estimation. The authors of the section on Beneficiation Operating Cost suggest that ‘‘Five per cent per m u m [as a percentage of the purchase cost of equipment] is a reasonable figure but could be higher if the ore is known to be abrasive or the process environment is aggressive. Normal range is 3% to lP!.” This is generally consistent with other rules of thumb. It is worth noting @articularly to nonmetallurgists) that, unlike mining equipment, it is usually assumed that milling equipment is not replaced over the life of the project (i.e. capital provision) but is treated as an ongoing wear item. The category of “operating supplies” is frequently listed as a line item in milling cost budgets and estimates, and usually is about 10 to 15% of the maintenance supplies, and thus is a small number. Where warranted, it is possible to obtain from manufacturers typical wear information for most pieces of equipment and thus build up an estimate of repair part cost. Engineering companies generally maintain an in-house database of maintenance supplies costs, frequently by unit process area, and this may provide good information for similar plants. Reagents and wear steel, with individual consumption and cost, are conveniently summarized in a table such as that shown as Table 3.

Tailing and Effluent Treatment, and Environmental Costs Construction of the initial tailing disposal facility is a capital cost. Pumping of tailings to the facility, returning process water to the plant, routine monitoring of the pipelines and dam, and pump and pipeline maintenance are routine mill costs, and should be included as part of the sections above. Continuous raising of the dam using taitings or mine waste or by a contract construction firm is usually treated as an operating cost. Frequently a lift is added to the dam every several years as a separate construction project. This can be handled as a sustaining capital cost (such as the replacement of mine trucks when they wear out) or can be estimated and included as an accrued monthly milling cost. Regular monitoring of the disposal facility by outside consultants should be provided for as noted below, under consultants. Effluent treatment (for example, destruction of cyanide before pumping to a tailing impoundment) is usually handled as one of the mill’s unit operations. Power, manpower, and maintenance supplies are provided along with other mill costs. Reagent costs are projected from test data, much as flotation or cyanidation costs. In some cases an environmental department is included under the milling department. Staffing,analytical requirements (whether in-house or external), reclamation costs, use of outside firms, and supplies can be significant. It is recommended that these be estimated separately from the mill operating budget since their scope, a service group, is project-wide. For final summation this can be included in milling costs, if appropriate. Other Costs The above costs constitute the majority of normal plant operating costs. However, there are a number of other costs that may need to be included and do at least need to be considered: 0

0

0 0 0

Assaying (If this is done at the mine site the staff costs are typically included in the mill manpower listing, but supplies and utilities need to be allowed.) Charges from other departments Consultants Contractservices Mobile equipment he1 and maintenance

333

Table 3 Reagents and wear steel %of % of total Usage $ Cost per Cateaorv

Item

Reagents

Grinding bails

• • • • •

costs

59 12 3 1 1 0 0 0 4 4 8

25 5 1 1 0 0 0 0 2 2 3

fafe

M

Sodium cyanide

0.88

1.98

1.74

3,175

Lime

1.59

0.22

0.35

Activated carbon

0.04

2.01

0.08

Rocculant

0.01

429

0.04

Caustic soda

0.05

0.33

0.02

Hydrochloric acid

0.03

022

0.01

Jaw crusher

0.003

3.30

0.01

SAGmai

0.052

2.22

0.12

Ball mill

0.049

222

0.11

SAGmai

0.43

0.55

0.24

638 147 71 27 11 18 19 212 199 433

Ball mill

0.51

0.51

026

4Z2

2.97

5,420

0.01

Total

•

dir. mill

(OOP's) steel cost

ko/tfeed

Misc fluxes Wear liners

reagent &

S/vear

i

4

100

42

Occupational health and safety costs (All or part of these costs may be included in milling costs, or may be included in the indirect costs/project overheads.) Refinery staff and supplies for a gold room Royalties on patented processes Safety supplies Security Training (Mill training staff would be included in the mill manpower chart, but outside training and/or training supplies need to be considered.)

Indirect costs are not included in the scope of this paper. These include such items as project administrative and general costs, corporate office costs, insurance (including fire, casualty and liability), sales, taxes (except sales tax), depreciation, depletion, amortization, townsite operating costs, overall infrastructure maintenance and similar costs. However, it is necessary to ensure that all of these costs are covered somewhere in the feasibility study. Total Mill Direct Operating Costs Total mill costs can be summarized in one summary table, such as that shown in Table 4, which should also include all of the appropriate miscellaneous costs noted above. Uncertainties, Contingencies and Accuracy Contingencies need to be discussed. A contingency is included in a capital cost estimate to provide for "...unforeseeable elements of cost within the defined project scope; particularly important where previous experience relating estimates and actual costs has shown that unforeseeable events which will increase costs are likely to occur." (Gentry and 0*Neil 1984) Generally speaking, we do not expect "major unforeseen events" in an operating cost estimate. However, there are uncertainties, and these can be handled in several ways. The price of diesel fuel, or power, for example, could vary widely over the life of a mine project. The Bureau of Labor Statistics petroleum products index varied from 105.9 in 1981 down to 53.9 in 1988, and back up to 101 in mid 2001. Obviously the effect of a doubling of a major cost category like fuel or power could greatly affect project economics. Significant swings,particularly in these two major cost areas, are likely to occur. Uncertainty can be quantified statistically with a Monte Carlo procedure, or it can be estimated by calculating project cost changes over a range of estimated fuel or power costs.

334

A kctor that can have a significant beneficial effect is the gradual reduction in costs with increasing process knowledge and experience, and or increased use of automation. Also worth noting: with new or novel processes, the uncertainty in the cost estimate may be significantly greater.

~h SaMesandwages

'

Porn# Grinding

Soumelcak. From salaries and wages (manning) table From power cost table (motor list)

'a!!BQ?&]

s!Ea ~

HI

2,978,7921 1.631

!ZQ&

22

Mher

Oyer u t i l i and fuels

I

i

Included in above mteaarieS

01 0.001

0

Natural gadpropane Fuel mi

Diesel fuel and gasoline ReagenlSandWearsted Sodium cyanide Other reagents Wear steel, media Maintenance supplies a mtl.

Included with vehdes From reagents and wear steel table

5 O h of equip. purchase price

($18.62 miiin) Op. supplier, oil & lube, misc. 15% of maw. supplies & material 40 mill sampledday X Wsample) Alowance:outside labs, conlract

1 Ft-end loader a l l 0 0 hrslvr @S!Wtr

1,333,000 931.000

0.73 0.51

10' 7

139,650 87,600 100,000

0.08 0.05 0.05

1 1 1

43.800 16,425 55,Ooo

21.900 Tailingstmahmnt Ttliidamaccrual oddr0om~ts Total direct mill oper. costs

(at $0.35/tonne of mill feed) ($1 ,lM),OOO at end of three years) lnduded in items abave

Inflation cannot be predicted, but can be expected to occur. An estimate of inflation can be included in the final project cash flow projection by the accounting department, along with taxes, depletion, depreciation, and related costs. A simplifyingassumption, typically built into initial cost projections, is that overall cost escalation will be offset by escalation in the price of the product. While workable for initial projections, this assumption has been vastly off-base for more than a decade (see www.westemmine.com,Feb. 17,2000 News Release). Parameters of Feasibility Studies: The basic data in the prehsibility study is determined in greater detail and requires more projectspecific information for development of a bankable feasibility study. Examples of the additional information needed for the feasibility study might be: 0 0 0

More detailed flowsheets and watedmaterial balances Pilot plant campaign results Additional geological/metallurgical characterization and laboratory replicated tests

335

0 0

0

0 0

Additional sample acquisition and testing and more detailed grinding work index testing (completed in conjunction with the projected mine production ore delivery schedules) Competitivesupplier quotes on consumables and power supplies Detailed projections of annualized process costs, including ore grades, metal recoveries and geological characterization Product quality determinationsand concentrate parameters Tailing storage operating costs and possible final reclamation costs Clarificationof other significant issues or process risks Development of operating statistical data quantifjing the cost estimate accuracy and significant risk parameters.

PROCESS OPERATING COSTS FOR ORE RESERVE ESTIMATES One of the main reasons for developing operating costs is to predict net values of metal revenue and operating costs for units of ore. Mining optimization techniques create economic values for blocks of ore that are appropriately sized to match the analytical database and distinguishable volumes (blocks) of ore. Once the geological resource has been defined by grade, geological ore type, and block size, a mine reserve optimization technique can be used to quantify ore reserves. The optimization will require the use of all direct and some indirect operating costs associated with the development, breakage, mining, processing and disposal of the associated waste products. A protit matrix is used to define the net value of each block of ore. A simplified diagram from a paper by Baird and Satchwell (2001) on the application of economic parameters to pit optimization is reproduced below to illustrate the block value profit, or net revenue, that can be used in ore reserve optimization.

+

Calculate: Value @ Recovered Metal times Metal Price

I Calculate: Direct Costs Mining + Process' + G&A I *ProcessCosts are the subject of this document Value - Direct Costs - Royalty - Penalties - Treatment/Refining = Net Revenue The optimization procedure is commonly done with a Lerchs-Grossmann optimization program for open pit mining and utilizes current dollars. The processing cost, including tailing disposal and treatment, is one component of the duect unit costs associated with a project. The unit values for mining and general and administration (G&A) costs must also be quantified for an ore reserve calculation.. We will evaluate only the process costs and associated metallurgical parameters in this text. The process unit cost and the recovered metal projections must be defined by the metallurgist. While one base value is commonly developed for a major geologic type of ore, using this value of cost or revenue for all blocks of ore is not appropriate for the quantification of ore reserves. In addition to mineral grade values, the relationships of ore type and process costs will play a key role in establishing valid ore reserves and mine plans. These parameters will be interrelated and must be defined by the metallurgist for the mine optimizer. This section of text will discuss some of these relationships and provide examples of the type of calculationsthat can be developed fiom a process cost estimate.

336

Metallurgical Parameters

Net block revenues will be determined by the projected ore grade, metallurgical recovery, and metal price assumptions used for each ore block. This means that the product shipping and treatment cost will have to be defined and either included in the unit processing cost vahe for each block of ore in the optimization or calculated separately as indicated in the section above. The economic cost variations observed in the blocks may well depend upon such things as variable concentrate grades or mineral recoveries determined by geological characteristics of different ore blocks. Some of the metallurgical relationships that should be quantified from the metallurgical test program are as follows: 1. 2. 3. 4.

Ore grade relationshipsto metal recovery and concentrategrade

Grinding work indices and their relationship to geology in the mine Other mineral componentsthat can impact final concentrategrades Mineral componentsthat adversely impact processing costs, such as copper minerals that affect sodium cyanide consumption and reagent cost in gold mills. 5. Any other major costs unique to the particular ore type.

The optimizing ore reserve procedure will require that the unit operating cost established as a base for a major component of the ore body can be modified by mathematical relationships established from the metallurgical test program The key economic components must be identified and evaluated in the metallurgical testing program so these relationships can be quantified. The processing costs in the block model will be the sum of the non-varying costs plus each of those costs that has been chosen to be varied according to block parameters. For the examples below the processing costs are made up of the following: Processing cost = (power cost for grinding + cyanide cost

+ other non-variable costs)

Use of Grinding Circuit Indices to Determine Operating Costs The early metallurgical testing program should identie the main geological types and evaluate them, not only with respect to metallurgical responses and reagent consumptions but also with regard to grinding power, media and liner consumption Grinding costs always comprise a significant portion of processing costs. The Bond indices for ball mill grinding are often used to quantie throughput and grind product size, and Bond abrasion indices can be used for calculating grinding circuit metal costs. Some mines use “modified” grinding index relationships to improve the day-today performance of a plant through operational optimizations. MinnovEX (Kosick, Dobby. and Bennett 2001; Bennett, Dobby, and Kosick 2001; Dobby, Bennett and Kosic 2001) have developed a SAG Power Index (SPI) test for sizing SAG mills and generating operating costs for this portion of the grinding circuit. Combining the “operating” SPI idormation for SAG mills with the “operating work index” derived from a laboratory work index for ball mills generates the information needed for their grinding circuit operating costs. MinnovEX uses this procedure in Comminution Economic Evaluation Tools called CEET or CEET 2. Mer plant evaluations, including some testing, and characterizing of ore samples with SPI and their laboratory ball mill work index measurement, MinnovEX use an optimization procedure to assist new mines in specifjing their grinding circuit equipment sizes or to help operating mines increase their production f?om existing equipment. In a similar manner, with laboratory pendulum tests and using a data base of idormation and/or a grinding circuit analyses, the JKSimMet procedure (Kruttschnitt Centre undated) can be used to optimize the grindinglthroughput relationships in operating plants as well as develop operating cost projections for grinding circuits. Semi-autogenous grinding work indices can also and Mosher 2001) and this work index can be generated from MacPherson Tests (McKen, Me, be used to size mills and power train units for autogenous and semi-autogenous grinding mills. Most laboratory and pilot scale generated grinding work indices are converted to “operating” indices for appropriate use. The operating indices for grinding power and metal consumption

337

reflect plant equipment characteristics, improvements in controls and technology, corredions for mill feed and product sizing, and significant improvements in grinding liner design and liner and media metallurgyover the last few years. Based on the operating work indices and knowing the power input, average unit cost of electrical power, and metal linerdgrinding media consumptions, relationships can be developed and used in the block model for revisions in the grinding circuit portion of the process operating cost. A common assumption is that plant capital equipment is generally specified and installed to process the ores with the higher work index, unless the hard ore is only a minor component of the ore body. Once these grinding circuit values are determined in an appropriatetesting program, the information is available to quantiijr their impacts on process unit operating costs. If the grinding indices of the ore change significantly &om the original design projections used to size the plant equipment, then the plant throughput can incrddecrease when lowerhigher work index ore enters the mill. Combining the grinding circuit information with other data on consumables, utilities, labor and metallurgical parameters will develop the process information needed for the modified unit process cost predictions. Example 1. Base unit operating cost with a SPI value of 10 k W t and a Bond Work Index of 16 kWh/t. If some mineralized blocks have an SPI of 8 kWh/t, and a Bond Work Index of 12 kWh/t, the typical unit cost equation for the ore optimization can be developed fiom the test data. Grinding power cost (assuming the same mill feed and product S i ) would change as the ratio the work indices of the new ore block to the work indices of the of base ore block. Power cost for grinding = Base power cost for grinding [(S + 12)/( 10 + 16)] Whatever indices are used for these calculations, it is important to use consistent ProCeQres for the grinding work and abrasion indices determinations and cost projections. These unit cost relationships must be developed by the process person and provided to the mine planning engineer for the mining ore reserve calculations. Impact of Adverse Ore Constitoents Another example of a cost modifier that should be developed by the process cost estimator is the impact of adverse c o d e n t s in some ores that significantly impact ore processing cost. This adverse constituent could impact actual processing costs or impact the smelting and refining revenues. One such example might be the impact of soluble copper mineral components in some ores during the cyanidation of goldsilver ores. Example 2. Assuming that the database has been developed in the ore body sampling and metallurgical testing program, a relationship given to the ore optimizer could be based on a base value of 0.06% soluble copper and laboratory testing with various mine samples: Cyanide cost = Base cyanide cost [(% Sol Cu/O.O6)] x (sol. Cu- CN use hnction)

Revenue Changes Per Unit of Ore Pmcessed In a similar manner to the above calculations, the revenues projected fiom individual mineralized blocks must be projected by the process metallurgist and ore reserve optimizer. Two important revenue areas are (1) metal recoveries and (2) quality of concentratesproduced on-site. Example 3. Calculations for ore grade and impact on recovery. Within similar ore types, it is imperative that the metallurgical test program develop information on product recovery versus ore grade. If there are many significantly different ore types in the ore body, then enough grade variation must be tested to develop the recovery/ore grade mathematical relationship for each ore type. This data will then be used by the ore optimizer to determine recovered product multiplied by the assumed metal price, to indicate the revenue portion of the block model. The metallurgical test program data set will usually establish the mathematical relationship by regression equations.

338

Valudore block = tonnes in block x ore grade (% or g/t) x recovery-grade hnction x US$/unit of product Calcufrrtions on the quality of concentrate. Shipping costs, concentrate treatment charges, and smelter/refineryreturns will be affected by concentrate grades and impurity levels. These factors will have to be projected from the laboratory and pilot plant data generated fiom samples taken from the ore body. Concentrate grade relationships, or impurity levels, should be defined by mathematical relationships on the mineralized block’s contents of metals, impurities or other diluents (such as; pyrite, graphhe or sericitic clays). The economic affects will be project specific and with a properly defined resource database, can then be used in the optimimtion program directly to project the revenue portions of each block in the mine’s block model.

SENSlTVITlES AND RISKS ASSOCIATED WITH PROCESS OPERATING COSTS Following the basic development of the process operating costs shown in Table 4 of this text, the metallurgist should indicate the sensitivity of the basic assumptions in the cost projections to uncertainty in the world’s economic markets. One of the uncertainties will be the price projections for the mine’s products. However, that feature will be left to the economic planners for the mine. Other sensitivities important to the overall financial analyses are projections of ore quality, product returns, and all direct and indirect costs for the mine’s operation. Some of the uncertaintiesthat should be addressed when determining process operating costs are as follows: I. Price forecasts for consumables 2. Price f o r m s for energy 3. Variations in mine reserve blocks related to grinding energy or reagent consumptions. 4. Confidence in ore grade and quality, mill throughput and metal recoveries (These may be different with simple process circuits or complex and integrated circuits). 5. Others, as defined for the particular process application

The process data set for the mine’s life is usually developed on an annual basis. This data set will include the annual mill throughput, head grades, metal recoveries and the projected direct operating costs, including tailing disposal and environmental costs. These costs are to be expressed in current dollars (no escalation) and any projections of consumable prices or energy price changes also expressed in the same current dollar terms. This data set will become the process contribution and together with all of the other mining, G&A, and product treatment costs will become the basis for the mine financial analyses that will be used for determining project feasibility or economic returns. The average process cost, per tonne or per unit of metal produced, for the mine’s future operation will be derived fiom this data set. Table 5 indicates a typical annual cost with some process information indicated on a projected eight-year mine life. We will use the idormation in this table for sensitivity analyses and fbrther determinations of risk impacts on unit costs. Conventional Sensitivity Analysis One of the most common methods for evaluation of different events on a unit cost of production is to simply vary each individual factor by a certain amount. The variation amount will be determined tiom earlier test data and available consumable cost information, and then the impact on the unit process cost calculated. For example, it might be determined that a particular reagent consumption could vary as much as 30% higher than used in the basic determination and this may represent 25% of the estimated unit process cost. Therefore, the impact of this increased reagent consumption would be an effect of 7.5% increase in the estimated unit operating cost. Figure 2 illustrates a common method of presenting the effect of such changes in process variables over a range of values. In this example, using data fiom Table 5, increasing either the head grade or the throughput significantlyreduces the per ounce cost. Reducing energy or cyanide costs even as much as 15% has a much smaller effect.

339

-Head grade -Throughput -Energy costs -NaCNoosts

+10*

0

*S%

-5%

P a r a m e t e r - % change

Figure 2 Sensitivity Chart - Year 7

Table 5: Eight-year process operating costs and Metallurgical Factors Parameters Trtrounjtnutt-fpa (000) Grind emrgjy-kWh/t

Head gpanfe -oz/tt % recovery % Cu (acid sollin1e$

Y e a r It

Yearr2

Year3

Year4

YeatS

YiBarrt

Year7

1825

2000

1722

1722

1722

1610

1610

1610

16.26

18.89 ~ 119889 284 2t6

18.89

20.20 2:5

20.20

2020

2rl

16.26 384

2J5

225

90.00

02.00

92.00

85.00

88.00

0LO8

0JQH

88.00 O.07

88.00

0)06

O.08

0.08

0.09

85.00 0.09

ZB

YeaiGB

Work matter*-SAG

1IT.0

11.0

1X0

13.0

13.8

14.0

14.0

1H4Q

Work Imfterx-ttaal mill

MO

1MM)

1S\0

1IBBI6

1C0

17.0

17-0

17.0

1T.63

1/49

1.73

1.73

1.73

1.88

1,88

1.88

1|44 —

1,44

Mill op. costs *$rt Salaries and wages Power Grinding

127

fr.JS

1-M

8-19

0-17

134 050

1034

Other

050

0,10

MM 0$f

Other utilities and fuels

0J02

0)02

9>02

6i02

8l0t

Reagents and wear steel Sodium cgontde

1-74

1|.74

2]04

2c04

Other reagents Media and wear steel

040

040 0-77

Maintenance supplies & mttj.

04t

MO 0M7 047

Op. suppfes, o i t lube, mfse-

01)8

(fll07

044 4W8

Assaying Contrast
QLOS

8l04

Q)08

0)08

Mobile equipm&nt

Al06

Tailings traariflttflt

OJ.88

Taiings dam accmiai Total direct mill o p . costs G o U awod'n-axtift Mill costt-$toz

0-73

0OJ31

0i8f

6)91

6:91

0.01

iSi

2J33

2(69

2j.63

0J60

0)80

0,80

0,§0

0,80

0-77

0-77

0.83

0:§3

0.83

044 &08

GU84

0.88

0,§8

0,88

044

&•§!

0r0f

ft 08

0.08

Q)AS

s^Of - 0.06 35

0)08

(A.00

OAW

0.1)7

0)08

0)08

0*88

948


~

Q>98

Q>A8

OJ-.OJ

0-1)8

MO M9

0.08


W6

0)36

%•§§

%-SS

020 020 8.90 8\J80 826 164250 202,400 174)266 136,382 136.382 109,480 109,480 104.35 ~ 126.53 130.8S 82.17 69.09 77.32 100.60

0.20

03W

0.20

020

020

7.40

6.99

7.97

7.97

340

020

8.90 109,480 130.89

Cost Analyses using Risk Probabilitiesand Monte Carlo Analyses A well-known probabilistic method can also be used to determine process operating cost sensitivities and risk events happening simultaneously. The Monte Carlo technique can address risk probabilities of events, their simultaneous occwrence and identify which events impact operating costs significantly. The technique also permits examination and correlation of related events. Costing analyses using this type of statistical procedure have been used for many years by large industrial firms to predict cost profiles and future economics. Although there are some options available for Monte Carlo simulation programming using Microsoft’s Excel software, we will use Decisioneering’sproduct called Crystal Ball@(Decisioneering, 2001) for examples in this text. Crystal Ball@ is a commercial add-in to Ex&. This analysis can be easily adapted to our earlier unit process cost estimates. The final results will generate a statistical output that allows the impact of uncertain events to be quantified and understood. Example of risk analyses with Crystal Ball@. The initial basis of the evaluation will start with the annualized process costs developed from the earlier mill cost idormation reproduced in Table 5 . The resulting processing costs can be evaluated on the basis of either; (1) cost per tonne of ore processed or (2) cost per unit of metal product generated by the mine. The cost per unit of metal produced will bring the uncertainty factors of projected ore grades, metal recoveries and, perhaps, concentrate quality into the risk matrix for examination. This is generally the better unit cost to examine due to the interrelationship of these parameters to other procesSing unit costs and the direct relationship to the product and the mine’s revenues. The method used is up to the investigator, but we will illustrate an example that addresses the impact of a number of risk elements on both average unit costs. The initial step is to establish a matrix of cost impact (risk) events, their projected uncertainty ranges and the probability of each range’s OCCUKWX. It must be noted that significant events should be chosen for this evaluation and not & events. The matrix can be based on variabilities observed in laboratory investigations, pilot plant studies or metallurgicaVgeological sampling campaigns. Process risks involved with complex or simple plants can be addressed in the matrix as can projections fiom specialists on future commodity or energy prices. For the example in this text, a matrix of events and probabilities were chosen with the assumptions shown below in Table 6.

The assumptions of probabilities and ranges were run with Crystal Ball@2000 software (1000 simulation runs for each year). With the input of the above risk matrix and the simulation runs with Crystal Ball@ software, the operating unit cost information can be compared on an annual basis with the original unit cost estimates shown in Table 5. Summary information that can be analyzed from the simulation run is as follows, and is shown in Table 7:

-

-

-

Median or mean of the simulation results for each year’s data The probability (say 90%) thatthe unit costs will be less than a certain maximum value The probability (say 90% ) that the unit costs will be greater than a certain minimum value Sensitivity charts indicating the variables with the highest impact on costs

Figure 3 shows annual operating cost trends, in $/ounce produced, of the simulation data reproduced in Table 7.This trend plot indicates the 80% and 20% probabilii ranges’ in which SO%, or 20%, of the predicted operating cost values occur. These operating cost ranges and the variability noted in the annual data are more u d in mining economic analyses than a single, average operating cost projection. For reference purposes, the average process cost of gold produced in this Monte Carlo simulation was $114.40/oz with $4.70/oz standard deviation and with SO?? of the values occurring between $108.63/oz and $120.58/oz. Any on-going analyses should use the annualized cost information rather than the averqge cost projection for economic predictions.

341

Table 6: Statistical matrix of cost impact (risk) events 1. Mine—Geological Ore Characterization and Grinding Energy: Varied annually depending upon measured Work Indices* A normal probability distribution with a mean of 16.26 kWh/t and a standard deviation of 0.50 kWh/t used for years 1 and 2 with similar distributions around higher, grinding energy values in succeeding years. * correlated on a 1:1 negative basis with process throughput 2. Metal Recovery Predictions: Customized probability distribution correlated 1:1 with ore grade 10% probability (continuous range) that Metal Recovery is between 80% and 85% 25% probability (continuous range) that Metal Recovery is between 85% and 88% 40% probability (continuous range) that Metal Recovery is between 88% and 90% 15% probability (continuous range) that Metal Recovery is between 90% and 92% 10% probability (continuous range) that Metal Recovery is between 92% and 94% 3. Ore Grade: Correlated 1:1 with metal recovery A normal probability distribution, with a mean of 3.1 g/toime and standard deviation of 0.31 g/t Au for year 1 with similar distributions around other projected ore grade values for succeeding years. 4. Annual Plant Throughput: Customized probability distribution varied annually and correlated 1:1 with grinding energy requirement 30% probability (continuous range that the plant will process between 1,500,000 and 1,750,000tonnes/year(year 1) 50% probability (continuous range) that the plant will process between 1,750,000 and 1,850,000tonnes/yr(year 1) 20% probability (continuous range) that the plant will process between 1,850,000 and 2,000,000 tonnes/yr (year 1) * Other probability distributions used for succeeding years 5. Acid Soluble Copper Values in Ore: Correlated directly with sodium cyanide reagent consumption and annual costs A normal probability distribution with an acid soluble copper content mean of 0.06% and a standard deviation of .01% with similar distributions around higher values in succeeding years Table 7 Comparison of Crystal Ball® results with original $/tonne and S/oz unit cost Operating results: Parameter BasefromTable 5 $/t Simulation median Simulation mean - $/t Simulation std. dev-$/t Simulation 10% < $/t Simulation 10%> $/t

Yearl 7.40 7.51 7.56 0.37 7.14 8.09

Year 2 6.99 7.28 7.33 0.53 6.68 8.10

Year 3 7.97 7.96 7.98 0.41 7.45 8.52

Year 4 7.97 7.98 7.99 0.40 7.46 8.51

YearS 8.26 8.26 8.28 0.41 7.73 8.83

Year 6 8.60 8.43 8.47 0.40 7.97 9.03

Year 7 8.90 8.77 8.78 0.42 8.23 9.34

Year 8 8.90 8.73 8.77 0.43 8.25 9.36

Base from Table 5 $/oz Simulation median Simulation mean - $/oz Simulation std. dev-$/oz Simulation 10%< $/oz Simulation 10%> $/oz

82.17 85.31 86.12 8.33 76.01 96.86

69.09 74.93 75.80 7.98 66.30 86.35

77.32 81.77 82.41 7.65 72.75 92.45

100.60 100.17 101.00 10.10 88.29 114.01

104.35 103.70 104.79 10.66 91.71 119.21

126.53 119.77 121.20 13.53 105.02 139.56

130.89 123.53 124.70 13.39 108.55 143.15

130.89 122.89 124.74 13.88 108.31 142.79

342

$140.00 -.

fl

_----Median

$60.00

!

I

Figare 3 Year-by-year operating costs in dolladounce produced Figures 4 and 5 are the sensitivity charts for information derived from Year 7 Monte Carlo runs. These charts indicate the relative importance of key variables by indicating their relative contribution (0% to 100%) of the variance noted in the simulation. Figure 4 indicates that the major influence of the unit production cost per tonne of ore processed in year 7 is grinding work index and its correlated impact on plant throughput (combined variance contribution of 74%), with the efFect of acid soluble copper content changes in the ore contributing 26% to the variance. Interestingly, the ranking of variables on unit cost per ounce of gold p r o d u d in Year 7,Figure 5, is (1) ore graddmetal recovery (83% variance contribution), (2) work index/thmughput (13%) and (3) acid soluble content of the ore (4%). The unit costs, per ounce of gold produced, show considerablymore variability on an annual basis than the costs per tonne of ore processed. After the idormation has been analyzed on the above annual operating costs and risk assumptions, fkther refinement and other risk matrix simulations can be made. Rather than varying throughput with ore hardness, an alternative risk matrix might be developed whereby the throughput is maintained by allowing the ore grind circuit products to increme in particle size with ores of higher work index This requires process information quantitjing the recovery loss in values with larger ore particle sizes; infbrmation that is ot'ten readily available fiom the metallurgical testing programs. In fid, this simulation would be quite valuable for mine management to determine which operating practice should be used for ores of varying ore grade and hardness. The Monte Carlo simulation technique can help analyze operating costs by indicating the parameters where risk or uncertainty is higher and where changes produce the greatest economic effect. procesS evaluators should analyze the background metallurgical information carefidly and use the variabilities derived from replications of different orebody samples and associated metallurgical testing to establish risk probabilities and process variable ranges (Smolik, 2000). Other probability assessments and risk impacts may also apply in the analysis as the plant operators gain knowledge (positive impact) and the plant becomes older (negative factor). These possibilities, and their associated probabilities, should be examined when the risk matrices are developed. Addressing risks in statistical terms through Monte Carlo simulations is quite useful when a cost estimator views the overall economic situation and prejxues an estimate for determining a mine's viability. Financial advisors can contribute to the evaluation by providing commodity and energy pricing, in current dollars, and perhaps inflationary or currency rate projections, for the cost estimator's use in a risk matrix. The value of this economic cost analysis really depends upon

343

Sensitivity Chart Target Forecast:Year 7

- Operating C o s t

-50.0%

-100.0%

0.0%

50.0%

100.0%

Year 7: Throughput

r

Year 7: Work hdex Year 7: Acid Soluble Copper Year 7: Gold Recovery Year 7: Head Grade

-

Figure 4 Sensitivity dolladtonne milled

Sensitivity Chart TargetFomcastYear7- OddOpCost

30.0%

-100.0%

0.0%

50.0%

100.0%

Year 7: Head Grade Year 7: Gokl Recovery Year 7: Throughput Year 7: Work index

-

Figure 5 Sensitivity dolladounce produced (1) defining the key parameters, (2) using unbiased probability predictions and (3) using realistic cost impact ranges. There is a degree of subjectivity in the analysis and its effective use requires

knowledge, experience and good judgment fiom the estimators and metallurgists.

ACKNOWLEDGEMENTS Our thanks to Eric Spiller (WashingtonGroup) and John Mosher (Hazen Research) for discussions and information on media and liner wear and abrasion testing. Particular thanks to the following who reviewed the draft text and provided valuable comments: Larry Goldman, Decisioneering, Inc. Roger Sawyer, Kennecott Minerals Brian Johnston, Fluor Daniel, Inc. Otto Schumacher, Western Mine Engineering Phil Walker,Newmont Mining Ken Major, Hatch & Associates Peter MacPhail, Homestake Mining Co.

344

1. Albert, Terry.2001. Kappes, Cassiday & Associates. Personal communication 2. Anon. Undated. Mineral Processing Plant Simulation, Optimization and Design Featuring JKSimMet, the Mineral Processing Simulator, developed by the Julius

Kruttschnitt Mineral Research Centre of Brisbane, Australia. www.jksimmet.com 3. Baud, B.K., and P.C. Satchwell. 2001 Application of economic parameters and cutoffs during and aAer pit optimization.SMZMiningEngmeering,February, 33-40. 4. Barra#, J.A., P.G. Davey, G.B. Gatchalian, and W.L. Puckering. 1975. The Influence of Energy Conservationon Concentrator Design. CMBulZetin,December, 85-93. 5. Bennett, Chris, Glen Dobby, and Glenn Kosick. 2001 Benchmarking and Orebody Profiling The Keys to Effective Production Forecasting and SAG Circuit Optimization. International Aut&ms andsemi-Autogemus Grinding Technology,1:289-300. 6. Canadian Mining Journal. 2001. 2002 Mining Sourcebook 7. Charles, W.D., and AE.J.Gallagher. 1982. Comminution Energy Usage and Material Wear. Design and I.&dJation of Comminution Circuit., Editors Mular and Jergensen, chapter 16, SME, Denver, CO. 8. Decisioneering, Inc. Undated. Crystal Ball@ 2000 sohare, Denver, Colorado, www.decisioneerina.com 9. Dobby, Glenn, Chr; Bennett, and Glenn Kosick. 2001. Advances in SAG Circuit Design and Simulation Applied to Mine Block Models. International Autogenous and SemiAutogmous Gn&g Technology.N:221-234. 10. Gentry, Donald W. and Thomas J. O'Neil. 2001. Mine Znvesbmrd Ana?yas, SME, Denver, c o . 1 1. Humphreys, K.K., and A.L. Mular. 1992. Capital and Operating Cost Estimation. Design andZdllation of Commination Cireuifi Editors Mular and Jergensen, chapter 6, SME, Denver, CO. 12. Kosick, Glenn, Glen Dobby, and Chris Bennett. 2001. CEET (Comminution Economic Evaluation Tool). Presented at SME Annual Meeting,Denver, Colorado, USA, February. 13. McClelland, Gene. 2001. McClelland Laboratofies, Inc. Personal communication. 14. McKen, Andre, Hans Raabe, and John Mosher. 2001. Application of Operating Work Indices to Evaluate Individual Sections in Autogenous-SemiAutogenousand Ball Mill Circuits. International Autogemus and SemiAutogems Grinding Techno&, HI:15 1

-

-

264.

L., and Gerald V. Jergensen. 1982. Design and ZdZation of ComminutionCircuits. SME,Denver, CO. 16. Nokes, Michael, and Terry Lam. 1993. Cost Estimation Handbook for the Australian Mining Zdhsby. Monograph 20, Australasian Institute of Mining and Metallurgy 17. Pincock, Allen & Holt. 1998. Feasibility Studies - Minimum Reporting Requirements. Information Bulletin 98-1. 18. Smolik, T.J. 2000. Strategies for Metallurgical Accountability. Presented at the Northwest Mining Association Annual Meeting, December, Spokane, WA 19. Western Mine Engineering. 2001. Mining Cost &rviw. Spokane WA, www.westernmine.com. 20. World Mine Cost Data Exchange. www.minecost.com. 15. Mular, Andrew

345

Financial Analysis and Economic Optimization Lawrence Devon Smith'

ABSTRACT The paper presents an explanation of discounted cash flow (DCF) evaluation methods and their specific application to the economic evaluation and optimization of mineral projects. DISCOUNTED CASH FLOW METHOD Method of Choice There are a number of ways to evaluate mineral properties and projects. Some are applicable in the exploration stage, some in the very early evaluation stage, and some in the feasibility and operating stage. The focus of this section is metallurgical projects in the advanced stage of development, at or beyond the pre-feasibility level, all the way to steady state operation. When sufficient data is available, the discounted cash flow (DCF) method is the method of choice for economic evaluations. The method is accepted as by industry, the financial community, and regulatory bodies. Applicability of the DCF Method DCF methods are applicable for those properties where mineral reserves, costs, and recoveries can be established with a reasonable degree of confidence, and there is sufficiently detailed production and cost data available to develop an accurate year by year cash flow projection including capital costs, operating costs and revenues. This level of data is usually available from the pre-feasibility stage onward. It is not practical to use DCF methods for projects where there is insufficient data, such as in the early stages of exploration where reserves, recoveries, and costs are not well know. The estimation of data at these early stages is too subjective, and other methods of valuations may be more applicable. What is a Cash Flow? Annual cash flows are estimated for each year of the project life. Year by year tonnages and grades for ore and waste are projected, along with their associated recoveries, metal prices, revenues, mining and processing costs, smelter or refining costs, royalties, taxes, and both initial and ongoing capital costs. For a new project the cash flow will begin with the initial capital, allocated over the time period set out in the development schedule. For an operating mine the initial capital costs are past (sunk costs) and are not part of the cash flow (except as deductions for taxes).

' Lawrence Smith, Consultant, Toronto, Canada

346

A cash flow is the year by year representation of all of the project data, both technical and monetary, in the form of dollars. These will typically include the following sequence of inputs:

+ -

=

=

Gross Revenue Treatment, Refining and Freight Charges Royalties Operating Costs Net Operating Income Capital Expenditures (as Spent) Working Capital Mining & Income Taxes (Cash Taxes) CashFlow

(If the project considers debt, then the principal received from the financial institution must be added to the cash flow as an inflow to the project, and the interest and principal repayments must be deducted as costs or outflows. The examples in this section exclude debt and interest.)

Cash Flow Versus Accounting Values It is important to distinguish between accounting values and cash flow values. While cash flow data is what is desired for DCF evaluations, accounting data is generally more available. The significant differences between cash and accounting representations of a project are noted below: 0

0

0

Timing and Depreciation - The essential difference between cash and accounting values is that in a cash flow revenues and costs are recognized as they occur. In an accounting profit and loss calculation, capital items are not included in the year they are spent, but are depreciated over a prescribed time period (say 3,5,10 or more years) or are amortized over the life of the operation. The profit in each year includes a portion of the original capital costs, even though it may have been spent many years earlier. Sunk Costs - The cash flow method is forward looking. This means that past, or "sunk", costs are not included. The reason for this is that an investor is only concerned with whether the current investment will make a sufficient return to justify the investment. Good or poor investments that have gone before do not impact on whether the current investment is valid. The only place that past expenditures are relevant is in tax deductions where tax pools will continue to include past capital investments until they are fully deducted for tax purposes. In the accounting convention, the depreciation of previous capital assets continues until the asset is full depreciated. Back Calculating Cash Flow From Accounting Income - It is recommended that cash flows be calculated from the basic cash inputs. It is recommended that cash flows poJ be backcalculated from accounting Net Income statements. Although the cash flow and accounting values will match in total at the end of the life of the mine, values will not match year-byyear. (See Appendix for more discussion on this issue.)

What is a Discounted Cash Flow? The revenue and cost data are combined to calculate an after tax cash flow that can then be discounted (hence the term discounted cash flow) to reflect the impact of time and interest rates on the value of money (the time value of money). The results of a DCF calculation is typically expressed as: 0

a net present value (NPV), and/or an internal rate of return (IRR).

347

Net Present Value (NPV) & Internal Rate of Return (IRR) Net Present Value (NPV). Net Present Value (NPV) is the amount of cash today that is equivalent in value to a payment, or to a stream of payments, to be received in the future. To calculate the NPV of a series of future cash flows, each future value (negative or positive) is multiplied by a present value factor. Each annual cash flow is said to be discounted back to a present value. The NPV is the sum of these values and gives a measure of what the project is worth, in a dollar amount, if all of the future cash flows (negative and positive) could be converted to a single value today. (See Table 1 and the section entitled The Time Value of Money for a detailed explanation of the calculation.) Internal Rate of Return (IRR). Internal Rate of Return (IRR) is defined as the discount rate at which the NPV equals zero. This value is directly comparable to an interest rate, since it is the return that is earned on the investment. This return implies that the capital investment is repaid and there is return (interest) earned on it. The IRR is sometimes referred to as the discounted cash flow rate of return (DCFROR) or the discounted cash flow return on investment (DCFROI). In practice these terms are used interchangeably, although IRR is increasingly the description of preference. Applicability of NPV and IRR. The following circumstances are typical of the application of discounted cash flow methods. An NPV can be calculated for any stream of annual cash flows. The IRR requires an investment (negative) followed by a positive cash flow.

0

0

The classic project investment pattern of a series of negative cash flows followed by a series of positive cash flows. This represents an investment (negative) followed by the repayment and return (positive) generated by that investment. 0 NPV can be calculated. 0 IRR can be calculated. An operating cash flow which is a series of positive cash flows that represent the net positive cash flows from an operating facility, after the initial investment phase is past. 0 NPV can be calculated. 0 IRR cannot be calculated. A “cost flow” is a series of annual values that are made up only of costs. Although negative values, these are shown as positive values for convenience. An example would be the single capital cost of a piece of equipment followed by the annual operating costs of that equipment. (See the section entitled Comparison of Project Alternatives.) 0 NPV can be calculated. 0 IRR cannot be calculated unless a differential or incremental cash flow is calculated between two costs flows (as when determining the cost advantage of one piece of equipment over another) by subtracting one cost flow from the other to give the classic negative then positive pattern. A cash flow may have a series of values that go from negative to positive to negative. An example is in oil field development where there is a large initial investment followed by years of positive cash flow from the producing well, then another large injection of capital to revitalize the well, followed by another stream of positive cash flows. The negativepositive-negative pattern also happens in the case of a large expansion on a mining project. 0 NPV can be calculated. 0 IRR can have multiple results because each time the cash flow changes signs, from negative to positive or back, there is one more solution to the IRR equation.

Sample Calculation of NPV and IRR The year by year cash flows are discounted at an appropriate discount rate to obtain an NPV and IRR. NPV = sum of each year’s cash flow discounted to the present at the selected discount rate IRR = the discount rate at which the NPV equals zero

348

-

Table 1 Sample Discounted Cash Flow Calculation -1

1

2

3

4

5

6

7

8

9

10

Total

Net Revenue o p costs Royalties Capital -300 Working- Capital . Taxes Cash Flow -300

105 -29 -1 -1 -9

123 -35 -1 -1 -1

114 -36 -1 -1 +1

87 -35 -1 -1

86 -35 -1 -1

86 -35 -1 -1

51 -24 -1 -1 +6

85

77

88 -36 -1 -1 +1 -17 34

88 -35 -1 -1

65

97 -36 -1 -1 +2 -11 50

-18 33

-17 33

-17 32

-17 32

925 -336 -10 -310 0 -106 163

NPV @ 0.0% 10.0% 11.5%

+$163 +$14 $0

-9 22

11.5%

IRR=

-

-

Table 1A Sample Discounted Cash Flow Calculation NPV and IRR Calculations Years Project Cash NPV PVI 1.5% NPV pv15% pv 10% from Factor @ Factor @ Factor Year -1 1 2 3 4

Present N 1 2 3 4 5 6 7 8 9 10 11

-$300 $65 $85 $77 $50 $34 $33 $33 $32 $32 $22 $163

5 6 7 8 9 10

l/(l.lo)n

10%

l/(l.llSy

11.5%

l/(l.15)n

.9091 3264 .7513 .6830 .6290 .5646 S132 .4665 ,4241 .3855 .3505

-$273 $54 $64 $53 $3 1 $19 $17 $15 $14 $12 $8 $14

3969

-$269 $52 $61 $50 $29 $18 $15 $14 $12 $1 1 $7 $0

.8696 .7561 .6575 .57 18 A972 .4323 .3759 .3269 .2843 .2472 .2149

.8044

.7214 .6470 SO3 S204 .4667 .4186 .3754 .3367 .3020

NPv @ 15%

-$261 $49 $56

$44 $25 $15 $12 $1 1 $9 $8

$5 -$27

Project Cash Flow $100

($150)

,

I

1.-. - 2 - 1

"_______.__" ____.______I_____ "___

1

2

3

4

5

6

7

8

9 1 0

Figure 1 Project Cash Flow

349

1

The cash flow from Table 1 showing a typical project cash flow pattern of investment (negative) followed by positive cash flow that will both repay the investment and provide a return on it. Both NPV and IRR can be calculated.

Project NPV 81IRR $300

,

I

I

I

I

0%

5%

10%

15%

20%

The NPV of the project cash flow is calculated at a number of discount rates. As the discount rate increases, the NPV declines because the impact of the discounted positive cash flows far in the future becomes less and less significant. The IRR is the discount rate at which the NPV equals zero, around 16%.

Figure 2 - NPV and IRR Time Value of Money The DCF method is based on the concept of the time value of money. This is most easily understood as the calculation of compound interest, where an amount placed into a bank will earn interest over time. If the interest is left in the bank, the original investment (deposit) will grow and interest will be earned, not only on the original investment but also on the earned interest that is left in the account. The interest is said to be compounding, the investment will have a greater and greater value the longer it is left in the bank. In other words, the future value of the original investment will grow the longer it is left to compound interest. Time, and compounding, have added value to the investment. This is the basic concept of the time value of money. The equation for compounding interest is:

FV =I(l+r)" Where: I r n

FV -

Future Value = future value original investment interest rate number of years that interest is earned and compounded.

For example, at an interest rate of 5%, an initial investment of $lo00 invested for 6 years would have a future value of FV = $1000(1+.05)6 = $1000~ 1.340 = $1340 The application of the time value of money calculation to cash flow evaluations is based on the same concept and uses the same mathematics. However, the calculation is to determine the present value (PV) of a future cash flow. The equation for present value calculations is:

Present Value

PV=FV/(l+r)"

Using the example above, the present value of a future cash flow of $1340, to be paid 6 years from the present, $1000 at an interest rate of 5%. PV = $1340/(1+.05)6 = $1340/ 1.340 = $1000

350

It is the application of the equation of present value, applied to each annual cash flow value for a project, that gives the net present value of the project. (See Table 1.)

Selection of a Discount Rate Few subjects in DCF evaluation will prompt more opinions than the selection of the appropriate discount rate. WACC as the Discount Rate? Conventional management theory suggests that the weighted average cost of capital (WACC) for the firm be used. This is perhaps an valid starting point, but the selection of the value suggests that the project under consideration has the same risk profile and the overall company. This is seldom the case. Companies are often most heavily valued for their ongoing operations, which are less risky than new projects. If the new project is in a geographic location that has a different risk profile than the rest of the company, this too should be reflected in the selection of the discount rate. Industry Practice. It is the author's experience that, for cash flow evaluations at the feasibility study level of projects in low risk countries, mining companies use a discount rate in the region of 10%for evaluations in real dollars, at 100% equity, after tax (Smith, 1995, 2000). This is based on surveys conducted by the author, discussions with mining companies, published evaluations by mining analysts, direct experience in studies undertaken for mining companies and various published references. There seems to be a variation between metals, with gold evaluations having a lower discount rate. There is also a difference in discount rate depending on the stage of development of the project. This can be seen in Figure 3.

I

DISCOUNT RATE vs PROJECT STAGE

BASE METALS

I

0.0%

The results of an industry survey indicates that as the level of development (and certainty) of the project increases, the discount rate decreases.

4 Early Evloralion

PreFeasibility

Feasibility Shldy

Operating Mne

Figure 3 Discount Rate vs Project Stage

Risk Components In A Mineral Project A discount rate for a mineral project comprises three principL. components; the risk-free interest rate, mineral project risk, and country risk. Brief descriptions of each are given below.

0

0

Risk-Free Interest Rate - The value of the long-term, risk-free, real (no inflation) interest rate is approximately 2.5%. Long term averages range from 2.3% to 2.6%. Mineral Project Risk - Mineral project risks include risks associated with reserves (tonnage, mine life, grade), mining (mining method, mining recovery, dilution, mine layout), process (labour factors, plant availability, metallurgy, recoveries, material balances, reagent consumption), construction (costs, schedules, delays), environmental compliance, new technology, cost estimation (capital and operating), and price and market. The more advance the project, the less the level of risk. Country Risk - Country risk refers to risks that are related to country-specific political, geographical, economic and social factors:

351

Political factors; government stability, political parties, constitutional risk, quality of government, foreign ownership policy (risk of nationalization), foreign policy, government crises, taxation instability, environmental policy, environmental protectionism, land claims and protected areas Geographic factors; transportation, climate Economic factors; currency stability, foreign exchange restrictions Social factors; distribution of wealth, ethnic or religious differences within the indigenous population, literacy rate, corruption, labour relations. Using these components, it is possible to calculate a project specific discount rate. The table below gives an indication of the range of discount rates that might be encountered: Risk-Free, Long-Term Interest Rate (Real) Mining Project Risk Country Risk Project Specific Discount Rate (Real, No Debt)

Low -

Mid -

2.5% 5.0% 0.0% 7.5%

2.5% 7.5% 1.5% 11.5%

2.5% 12.5% 12.0% 27.0%

“Low” might be an operating mine in Nevada or Quebec “Mid” might be a feasibility study in Chile “High” might be a pre-feasibility study in sub-Sahara Africa

Real and Nominal Discount Rates. Caution should be exercised when selecting discount rates and interest rates to ensure that real rates are used with real dollar cash flows and nominal rates are used with nominal dollar cash flows. For example, a real discount rate of 10.0% will become 12.2% if an inflation rate of 2.0% is used in a nominal case. The relationship between the real and the nominal discount rate is described by the equation: (1+R) = (l+r) (l+i) Nominal to Real Discount Rate

where: r = discount rate with no inflation i = inflation rate R = discount rate with inflation

Conventions Used In DCF Evaluations There are a number of conventions used in DCF calculations that should be itemized for each case that one develops so that others using the values know what assumptions were included in the evaluation. Changes in these criteria can cause significant differences in the results of an evaluation, even though the evaluation may start out with the same input values. The author recommends that a “bare bones” reference case be calculated for each project on the basis of the following assumption so that cases can be compared one to another: constant metal price over the life of a project with inflation but de-flated to real dollars to express NPV and IRR in real terms all variables inflated and de-flated at the same rate nodebt after taxes

Metal price. Many evaluations use a constant metal price over the life of a project. This need not be the case and some evaluations use detailed price forecasts and metal price cycles. The

352

problem with these is that the timing of the cycle becomes critical and may over or under value a project, Most evaluations use a mid-range long term average price. Inflation. Most evaluations will include inflation. If this is done, it is recommended that the resulting cash flow be deflated to real dollars to calculate the NPV and IRR. Inflation is normally included to make the tax calculations more reflective of what will really happen. Once capital is spend, the value is set in time and it no longer increases with inflation. In tax calculations, while revenues and operating costs continue to inflate, capital costs, once spent, do not. As deductions for depreciation, they become relatively smaller and smaller with respect to the annual income. With smaller depreciation deductions the taxes payable are relatively larger. A no inflation case will therefore tend to understate taxes. Typically all variables (price, operating costs and capital costs) are inflated at the same rate. The inflated values are said to be expressed in “nominal dollars”, or “current dollars”, or “dollars of the day”. If the values are deflated they are said to be expressed in “real dollars” or “constant dollars”, or “today’s dollars”. The conversion from nominal to real dollars is achieved by dividing the nominal cash flow amount for a particular year by the compounded inflation factor for that year. Debt. It is recommended that cash flow evaluations be runs without debt. There are two reasons for this: 1.

Debt amounts and repayment schedules can be selected to favour the outcome of a project. This effect is referred to as leverage or gearing.

2. Debt and financing are more a measure of the ability of the larger corporate entity to raise funds than the value of an individual project.

Taxes. It is recommended that taxes be included in the cash flow evaluation. Taxes are a very significant cost in a project and if they are not included, a major portion of the total cash flow is missing. Projects with large capital expenditures will have larger tax depreciation pools than low capital projects and the differences in taxes may impact the final evaluation. Tax documentation is well established and are taxes fairly easily calculated. Advantages of the DCF Method Accounts for all cash inflows (revenue) and outflows (capital costs,, operating costs, royalties, income taxes, etc) and allocates them to the time period in whicih they occur. Accounts for risk, either in the assumptions in the cash flow, or in the discount rate. Accounts for inflation. Accounts for cost of money (interest). Method is forward looking (that is, past expenditures are irrelevant). Method is general in application and broadly accepted. Method is transparent. All calculations can be seen and the mathematics can be checked easily by hand. The resulting net present values (NPV) give a measure of the magnitude of the project, in dollars, and the values from multiple projects are additive. The resulting internal rate of return (IRR), if any, is intuitively comparable to an interest rate.

Disadvantagesof the DCF Method Method can be complex for the unfamiliar. Requires detailed estimations of reserves, revenue, costs. Requires the determination of a discount rate.

353

0

One must watch for inconsistencies due to inflation in the selection and use of discount rates, interest rates and equity ratios. The discounting does not recognize value of long life projects. Values beyond 15-20 years are not generally significant. The resulting IRR, if any, does not give a measure of the magnitude of the project, only the return. For example, both a tiny project and mega-project could have the same rate of return, but no hint of size is given. It is .possible, if there are large capital investments during the project, to have multiple IRR values.

DETERMINING THE OPTIMUM PROJECT There are always opportunities to optimize a mining project. In the early stages of development there is considerable scope for comparison and optimization, and whole operating concepts can be examined and changed. Although these opportunities decrease as the project is defined in the feasibility study and design stages, even in the operation phase there continue to be opportunities to select better equipment or try different approaches. There are a number of ways to approach optimization and they tend to vary with the stage of development of the project. Perhaps the most significant selections in the life of a project are the size of the operation and the processing method. Because these items are usually decided early on, they are often selected based on personal or corporate preferences, or early data that may not reflect the overall project. In both regards, the significance of scoping and ranking studies cannot be over stressed. These studies are relatively inexpensive and can narrow the range of likely size and processing solutions in preparation for the more expensive (and often irreversible) feasibility study. Targeting A targeting study is a very early review of a project. The data available is often too scarce to do a proper DCF evaluation, but a simple cash flow can often be created from minimal data on tonnage, grade, recovery, capital costs and operating costs. The cash flow calculation in Table 1 is the summary of a very detailed study, but all of that effort and detail eventually comes down to thee very few lines. Therefore, any project that is developed enough to allow these values to be estimated even as educated guesses, it is possible to calculate an order of magnitude cash flow, NPV, and IRR for targeting purposes. The concept of targeting means to determine the range of values that must be satisfied in order to make a viable project. These target values can be compared with the known data to determine the likelihood of any of them being satisfied. For example, in exploration, a small deposit may be of interest. A simple cash flow can be used to project a larger tonnage of mineable material and estimate how much material would have to be found for it to become a potentially viable project. If the exploration people feel that more mineralization is likely, then they can continue to search, but if the deposit has been fully delineated and there is no more size potential, then it may be best to walk away and not invest any more time or money. The same sort of process could be applied to metallurgical recovery, stripping ratios and processing methods as well. This is not a typical application of DCF methods, but it can be used to determine what else needs to be discovered or tested to have a viable project. Scoping The scoping study is the next stage in the evolution of a project. Its purpose is to use the known data for the project to begin to get a sense of what the final project will look like. There needs to be enough information about tonnage, grade, geometry, metallurgy, recovery and processing to allow order of magnitude estimates of capital costs, operating costs, and revenues. This is the time for brainstorming and open minded thinking, when no idea is too dumb to try. It is at this stage that mining and processing options are tested and the eventual project begins to take shape. This should occur in the early stage of a pre-feasibility study. Unfortunately, all too

354

often the final project concepts are frozen before any creative work can be done. Try to avoid this situation. Because this sort of examination involves all aspects of the project, including reserves, revenue, capital costs, and operating costs, a full cash flow calculation is undertaken. While the level of detail backing up the numbers will be more substantial than for a targeting exercise, the final cash flow itself remains as straightforward as that shown in Table 1. The purpose of the scoping study is to vary the project parameters, reflect these variations in the cash flow, and see which variation gives the cash flow with the highest NPV or IRR. This variation would then go on to more detailed examination in a pre-feasibility study.

Process Ranking There are occasions when it is possible to use a DCF calculation even though portions of the project are not defined or some project data is not available. An example of this is the ranking of process options when other aspects of the project have not been finalized. This can be done because when choosing between process options, one is not seeking an absolute NPV, but a relative one. In the case of ranking process options, it is not necessary to have the final mining plan, the details of offsite infrastructure, or any other non-processing data available. The only data that is required in detail is: annual throughput average metallurgy and grades metallurgical recovery for each process option operating costs for each process option capital costs for each process option It is best if some reasonable estimate is made of the non-processing items only so that the absolute values of the results are not foolishly high or low. However, since only the relative ranking of the process options matters, these values will be subtracted out anyway. An example is provided in Table 2 and Figure 4. Note that the DCF calculation gives a different ranking than the simple average incremental cash flow per tonne of ore. This same pattern will be borne out in the full project cash flow when the other major projects costs are eventually available. In the meantime the process can be selected and design work can begin. NPV Relative to Whole Ore Cyanidation

$0.0

g ($10.0)

6> ($20.01 p ($30.01 ($40.0)

Figure 4 Ranking of Process Options

355

Only data relating to the process needs to be known to rank these options. In spite of a very low recovery, the combined effect of recovery, capital cost, and operating cost favours whole ore cyanidation. Other options have a negative NPV relative to this process. Concentrate BIOX is the nearest competitor.

-

Table 2 Process Ranking Study c

Capital Operating Recovery

$million $/t Ore

Revenue $/t Ore Operating Costs $/t Ore Capital Costs $/t Ore Cash Flow $It Ore Incremental $/t Ore NPV @ 10% $million Relative NPV $million

$3.9 $10.50 72.6%

$32.8 $16.10 84.4%

$36.0 $17.20 87.6%

$38.1 $19.70 89.6%

$20.3 $17.20 78.5%

$19.4 $21.50 85.4%

$51.35 ($10.50) ($0.39) $40.80 $0.00 $236.6 $0.0

$59.69 ($16.10) ($3.28) $40.31 ($0.49)

$61.98 ($17.20) ($3.60) $41.18 $0.38 $227.5 ($9.1)

$63.38 ($19.70)

$55.53 ($17.20) ($2.03) $36.30 ($4.50) $203.7 ($32.8)

$60.39 ($21.50) ($1.94) $36.95 ($3.85) $207.3 ($29.3)

~~

$223.5 ($13.1)

($3.81) $39.87 ($0.93) $218.9 ($17.7)

Optimum Production Rate & Optimum Project Size The selection of the production rate is one of the most crucial decisions to be made in the development of a mineral property. This single factor will determine the capital costs, operating costs, and mine life, all of which influence the project economics and the viability of the project. It is not uncommon for the production rate to be decided arbitrarily, before any technical evaluations begin, and it is often not reconsidered when better data becomes available. This frequently results in a mining and processing facility that is inappropriately sized for the deposit. As often as not it is too large, burdening the owner and the project with costs that the deposit cannot support. Rules of Thumb. In the mining industry there is a “rule of thumb” for every occasion. The determination of a production rate is no exception. These rules can be broadly divided into two groups; those related to the physical characteristics of the deposit; and those related to the economic characteristics of the project. Some examples are listed below. They are presented for reference only and should be used with caution. Taylor’s Law: t/d = .014 (Reserves)0.75 For underground and open pit deposits, with the exception of a few structurally controlled situations. (Taylor, 1986) 500-1000 t/d for every vertical foot of orebody, for underground mines. Cash flow can repay capital twice. Mine life > 1.5 to 2 times bank loan. Mine life > 7 years. Mine life > metal price cycle. Annual production should not exceed the market demand. (Not usually a consideration for precious or base metals but is often a restraint for industrial minerals.) Most of these rules of thumb are rooted in practical experience and each has earned a following in the industry even though many contradict and conflict with each other. Nevertheless, two significant themes stand out:

356

0 0

the importance of reserves, the need to repay capital.

Maximum Value Curve. The mineral industry is unique in a number of characteristics. With respect to the determination of a production rate, three characteristics are significant: 0

0

0

The production rate for an individual mining project is not generally limited by supplydemand considerations. For most metals (particularly base metals and gold) an individual project will be a price “taker” and a single project has little effect on the metal price or the metal supply. Therefore a project can be sized based on its own economic characteristics, independent of the market. Initial capital investments tend to be very large and are directly related to the production rate. The relationship between capital costs, operating costs, and production rates is illustrated in Figure 5. There is physical limit to the size of the reserves for each mineral deposit. As soon as a production rate is selected, the life of the project is also determined. This differs from, say, a bakery, where the life of the facility is essentially perpetual. Mathematically this limit means that for each project there is a production rate that will yield a maximum economic value.

As a result of these characteristics, there is a typical maximum value curve associated with mining projects (see Figure 6). Although every project has its own unique curve, the humped shape with a steep rising slope and a less steep falling slope is typical.

Component Curves - Undiscounted and Discounted. The reason for the maximum value can be seen by examining the curves for revenue, capital costs, operating costs, and capital plus operating costs. This needs to be done for both undiscounted and discounted curves. The undiscounted curves show how costs and revenues actually change with respect to production rate. They are useful to have a sense of how the absolute values change. The discounted curves (10% discount rate is used) show how the values are affected by being nearer or further out in time. The largest impact is at the low production rates, where the life is longer and the present value of the costs or revenues from later years has a reduced impact on the NPV. However, since the NPV valuation is the method of comparison, it is important to see how the discounting impacts the component values.

0

0

0

Revenue - The undiscounted revenue curve is flat, since the value of the metal contained in the deposit is the same no matter what the production rate. When discounted at lo%, the values are much lower of the low production figures. Capital Cost - The undiscounted capital cost curve rises continually as the production rate increases, rapidly at first and then more gradually, almost linearly. (See Appendix B.) The discounted capital cost curves do not vary much form the undiscounted curves because capital costs occur in the first year or two and are not heavily impacted by the discounting. Operating Cost - The undiscounted operating cost curve decreases continually as the production rate increases, rapidly at first and then more gradually, almost linearly. (See Appendix B.) However, the impact of discounting tends to reduce this shape and can even cause it to show a rising shape for the low production rate cases. Capital + Operating Cost - The curve for the sum of the capital and operating costs is a combination of the two curves and the impact of discounting in each component affects the final shape.

357

$500

.............”

1

1

Z

0

3

$400

-

3

E

-

dc

P $200 W

u.

g

$100

Total costs over the life of the mine. Capital costs increase with the production rate. Operating costs tend to decrease. The relationship is usually exponential (See O’Hara, 1980 & 1992, & USBM, 1987).

-

s $03

Figure 5 - Capital & Operating Cost Curves The NPV vs Production Rate curve typically has hump shape, rising rapidly on the lower production side to a maximum (A), then falling more slowly as the production rate increases. The IRR curve has a similar shape and shows a maximum in the same range. .

.

5,000

0

10.000

15,000

PRODUCTION RATE (TPD)

Figure 6 Characteristic Maximum Value Cash Flow NPV Curve

4 Z

REVENUE

Q $500 Em-

2 $300

c-

1

-

OP COST + CAplTAL

\ OP COST

I 0

5,000 10,000 PRODUCTIONRATE (TPD)

Undiscounted total amounts over the mine life. The revenue curve is flat because the total revenue does not change. Operating costs (both unit costs and total costs) decrease with production rate. Capital costs increase with production rate. The distance between “Revenue” and “Op Cost-Capital” is the Cash Flow (before tax).

15,000

Figure 7 -Undiscounted Component Curves What is significant is the relationship of the capital + operating curve to the revenue curve. The distance between theses two curves is the cash flow. The distance between the versions of these two curves give the discounted cash flow (NPV) and when plotted gives the maximum value curve.

358

$450

,

$400 $350 $300 E $250

2

The time value of money (10% NPV) reduces the value of revenues and costs that occur further out in time due to longer life at low production rates. Capital is spent early and is not greatly impacted. The distance between “Revenue” and “Op CostCapital” is the Cash Flow plotted in Figure 6. Note the maximum distance at point A.

CAPITAL

g $200 &

$150

$100 $50 $0

1 0

5,000 10,Ooo PRODUCTIONRATE 1TPD)

Figure 8 Discounted Component Curves Variables Impacting the Location of the Maximum Value. There are number of variables that can potentially impact the location the maximum value. The notes below summarize the assumptions made for the calculations in this paper: Constant Reserves - Reserves are assumed not to change as operating costs change with different production rates. At higher production rates, operating costs tend to be lower and the cut-off grade would, therefore be lower which would in turn increase the reserves. For this review, this effect is assumed to be small. Prices - All calculations are done at the same price. Changes in price will impact the size of the maximum value, but not necessarily its location. Discount rates - A 10% discount r ate is used to determine NPV. The location of the maximum value increases as the discount rate increases until about a 10% discount rate, beyond which the maximum value production rate does not change appreciably. Taxation - Taxes are excluded. The location of the maximum value does not appear to be significantly affected by the tax rate. This means that preliminary evaluations can be carried out on a before tax basis. No debt, no interest. These assumptions are valid for illustrative purposes. However, in a specific project the impact of variation in these assumptions should be examined in detail. For a detailed examination of these factors, see Smith (1997). NPV Maximum Value As A “Failure” Point. This NPV maximum value point would seem to provide a straightforward solution to the optimum production rate determination. However, caution is urged. The NPV maximum value gives a very short mine life (only 5 years in Figure 6). A similar concern has also been expressed by Taylor (1986, 1996), that the over sizing of a project imposes a great risk of failure, with less time to recover from a bad start. Shorter lives also have maximum environmental disturbances and minimum local social advantages. In the author’s opinion, the NPV maximum value point should be viewed as an upper limit for the possible range of production rates, rather than an optimum rate. It is the point at which the project begins to “fail” economically; where the return begins to diminish relative to the investment. The shape of the curve in Figure 6 is reminiscent of plots of structural loading tests where a column or beam takes more and more load until it fails and begins to lose its load bearing capacity. With this analogy in mind, it may be appropriate to apply a safety factor to the NPV maximum value production rate. This would keep the rate below the maximum, to the left of the peak on the rising side of the curve, where it is always possible to increase production later.

359

Two Times Capital Curve. How far down the curve should one go for a safety factor? The banker’s rule of thumb, to be able to repay the capital twice, provides some guidance. Financial institutions will insist on certain coverage ratios for their loans, and the “two times” factor will generally ensure that these are met. If the undiscounted cash flow, which already has the capital cost deducted once, deducts the capital again (hence “two times capital”), the resulting values plot as a hump shaped curve with a maximum value at a production rate lower than the Cash Flow NPV maximum. The curve is undiscounted because the two times capital criteria is based on the undiscounted cash flow (although coverage ratios are often based on discounted values). However, if the curve for the NPV of the two-times-capital values is plotted, a similar production rate is indicated.

Point B is the undiscounted twotimes-capital maximum value. Point C is the 10% NPV twotimes-capital maximum value. The two curves indicate a similar production rate. Point A is the 10%NPV cash flow curve maximum value (the same point as in Figure 6).

DISCOUNTED

0

5,000 10,000 PRODUCTION RATE n P D l

15,000

Figure 9 Two-Times-CapitalMaximum An Optimum Range. Having two maximum values is confusing. What do they mean? How do they help? In the author’s opinion the Cash Flow NPV maximum value is the upper extreme of the production rate possibilities, a point at which the project begins to “fail” economically. Since the two times capital maximum value is always less than the cash flow maximum, the two points can be considered to represent the lower and upper limits on a range of likely production rates. Referring to Table 3, the Cash Flow NPV maximum value tends to give a very short mine life, whereas the TwoTimes-Capital values are much more comfortable in this regard.

-

Table 3 Range of “Optimum” Production Rates

Maximum

Calculation Method

Point A Point B Point C

Cash Flow 10% NPV Two-Times-Capital Undiscounted Two-Times-Capital 10% NPV

Production tpd 5,250 2,000 2,250

Life Years 4.9 12.9 11.4

The appropriate production rate will lay somewhere between Point A and Point B. Know Your Project. Typically an owner will be confronted with the problem of having selected a single production rate early in the project’s life. Even with a pre-feasibility or feasibility study the results only represent one point on the curve. But is this point on the rising side of the characteristic hump, ideally situated between the two maximum values shown in Figure 9, or is it on the falling side of the curve, with an acceptable NPV and IRR but well beyond its peak performance? The answer is that any detailed review of a project should include a calculation that

360

results in a plot similar to that in Figure 9. (See Appendix B for a discussion on factoring capital and operating costs for different production rates.)

COMPARISON OF PROJECT ALTERNATIVES It is often the case that it is necessary to select two competing alternatives. This can take a number of forms: 0

0

0

0

Project Expansion - This involves the comparison of the current operation with the projected expanded project. The preferred option the one that has the higher cash flow NPV. Trade-off Studies - Cost trade-offs usually arise in equipment selection and often involves the comparison of costs only, with no revenue. Therefore it is a comparison of the s t flows of the two options. This method is used when both options have the same operating life. The preferred option is the one that has the lower NPV. Replacement Chain Analysis - This method is used to select equipment when two options have different operating lives. The procedure involves developing a cost flow that is made up of a series, or chain, of replacements for each option until the length of the chains for both have a common operating period. The preferred option is the one that has the lower NPV. Equivalent Annual Cost - This method is used to select equipment when two options have different operating lives. The procedure involves converting the irregular costs of each option into a series of equal annual costs that has an equivalent NPV over the operating life of the option. The preferred option is the one that has the lower equivalent annual cost.

The Expansion Problem It is not unusual for the owner of an operating mine to consider an expansion to increase both production and profitability. DCF methodology can be used to determine whether or not an expansion is worthwhile investment and what size might be appropriate. The logic used to justify an expansion can usually be simplified into the following points: 0

0

0

Unit operating costs will be reduced if the production rate increases. This is often true, as can be seen in Figure 5 and the effect is an economy of scale. Not only do the unit costs for the production activities such as mining and processing reduce (see Appendix B), but the annual cost for administration will usually increase only marginally. Capital costs for expansions are often based on removing a bottle-neck. This means that production can be increased for a much smaller capital investment, proportionally, than the cost of building the original facility. Infrastructure (roads, power lines) and non-operational assets (administration building) are already in place and do not generally need to be altered. Only production facilities need to be expanded. The investment in the expanded portions of the facility is less, proportionally, than the cost of building the original facility.

The DCF evaluation will address the following concerns: 0

0

0

Is operating cost saving resulting from the expansion sufficient provide the required return on the capital invested to achieve the expansion? An expansion will move future cash flows closer to the present. This gives a higher NPV and generally supports the expansion decision. An expansion will result in a shorter life. This decision shares all of the concerns expressed for the selection of an optimum project size. Too short a life can be problematic.

361

STATUS QUO CASH FLOW

The first step in the evaluation of an expansion it to model the existing operation, the statue quo, This is the reference point for estimating other cases and for determining whether the expansion concept adds value to the property.

$40.ooo -

-

$30,000

($20,000) ~

-

630.000)

Figure 10 Status Quo Cash Flow The second step in the evaluation of an expansion it to model the expanded project. This will typically result in higher cash flows after the expansion, and a shorter remaining life. The negative value in year 1 is the net of the original cash flow less the expansion capital.

EXPANSION CASH FLOW 550,000

_

1

:

Figure 11 Expansion Cash Flow INCREMENTAL CASH FLOW .......

"

........... ............ ._..........

"

""

"

.

1

$20.000 $10,000

50 ($10,000)

2

3

4

5

6

7

8

($20.000) ($30,000) ($40.000) ($50,000) ($60.000) ($70,000) I

-

-

I

Figure 12 - Incremental Cash Flow

362

An incremental cash flow can be calculated by subtracting the status quo cash flow from the expansion cash flow. The resulting pattern shows the expansion investment followed by the net increase in cash flow due to the expansion. In the later years the increment is negative since these cash flows have been moved forward. The sign switch gives two IRR solutions.

NPV of Status Quo & Expansion $250,000

f

P

$.xw3oo

e >

4

$150,000

5

I $100.000

$50,000

0.0%

5.0%

10.0% DISCOUNT RATE

15.0%

20.0%

The NPV of each cash flow is calculated at different discount rates. Note that the curves cross at -17%, indicating that the return on the incremental cash flow is -17%. For project hurdle rates below this point, the expansion case has the higher NPV and would be selected as the preferred option.

Figure 13 - NPV of Status Quo 8z Expansion NPV of incremental Cash Flow

The NPV of the incremental cash flow is plotted for different discount rates. Note that the curve crosses the zero NPV line at -17%, indicating that the return on the incremental cash flow is -17%. The sign switch in the incremental cash flow results in a second IRR value at --6%. This result has no meaning.

$12.000 f 510.000

3

e 2z

g

$8,ooo $6,000 @-OOO $2,000

3

$0

(WaW -5.0%

0.0%

5.0% 10.0% DSCOUNT RATE

20.0%

15.0%

Figure 14 - NPV of Incremental Cash Flow Trade-off Analysis Cost Flow versus Cash Flow. Trade-off studies are usually used when choosing between two pieces of equipment where there are only capital and operating costs involved. The example illustrated here is the choice between a conveyor (high initial capital cost and low operating cost) and trucks (relatively lower initial capital cost and higher operating costs). The selection of either option has no impact on revenue and so only the “cost flow” for each option needs to be developed.

-

Table 4 Trade-off Analysis Input Data’ Capital Cost Operating Cost Equipment Life Purchase Equipment Total Cost over 10 Years Capital Operating Total

Truck $2,500 $1,OOO 5 years 2 times

Conveyor $6000 $450 10 years 1 time

$5,000 $lO,OOo $15,000

$6,000 $4,500 $10,500

These cost values are arbitrary and were selected for demonstration purposes only.

363

If each option had a different impact on revenue in any way (higher production, greater recovery, etc, then a full cash flow including revenue, would have to be calculated. The process is essentially the same but in the cost flow, the option with the lower NPV (lower cost) is selected, while in the cash flow, the option with the higher NPV (higher cash flow) is selected. Taxes. Both operating costs and capital costs are deductible for tax purposes. These amounts can account for significant deductions, and therefore significant reductions in taxes. In the example above, and in most situations, there is a noticeable difference in the total costs and therefore the total tax reduction. Therefore, it is necessary to include the impact of taxation in the comparison of tradeoff options. The logic for tax implications is straightforward, but its application can be confusing. This is because it involves a double negative and is easily confused with a triple negative. The costs (negatives) are deductions (negative) for the purposes of calculating taxes, and taxes are universally understood to be costs (negative). The calculation of tax in a cost flow results in a reduction in the cost flow by the amount of the operating cost plus depreciation in a given year, times the tax rate. For example, the truck in Table 4 has an annual operating cost of $1000 and a capital cost of $2500 that would be depreciated at $500/year over the 5 year life. Therefore there is a total annual deduction for tax purposes of $1000 +$500= $1500. At a tax rate of 40%, this would result in an annual tax reduction of $1500 x 40% = $600. A tax reduction is a cost reduction, so the tax deduction is deducted from the total annual cost (this is the double negative). Therefore, after taxes, the net cost flow of the truck in any operating year (not in a truck replacement year) is $1000 - $600 = $400.

-

Table 5 Trade-off Analysis Tax Impact After Tax Cost Flow: Operating Cost Tax Deduction Net Cost Flow Tax Deductions Operating Cost CaDital Deureciation Total Deductions After Tax at 40%

Truck

Convevor

$l,OOO -$600 $400

$450 -$420

$1,OOO $500 $1,500 $600

$450 $600 $1050 $420

$30

-

Table 6 Trade-off Analysis Cost Flows -1

TRUCKS Operating Capital TaxDeduct Cost Flow CONVEYOR Operating Capital TaxDeduct CostFlow Incremental cost Flow

1

2

3

4

5

6

7

8

9

10

Total

450

450

450

450

450

450

450

450

450

450

4500 6000

-420

-420 30

-420

-420

-420

-420

-420

-420

30

30

30

30

30

30

-420 30

-420

30

30

6300

370

370

370

370

370

2870

370

370

370

370

2700

0 2500

2500

0 6000 6000 -3500

364

Table 6A- Trade-off Analysis Cost Flows NPV and IRR Calculations Discount Rate 0% 5% 10% 15% 20%

Truck Net Cost Flow $9000 $7099 $5790 $4859 $4179

Convevor Net Cost-Flow $6300 $5935 $5622 $5348 $5105

Incremental cost Flow $2700 $1 164 $168 -$489 -$926 IRR = 11.1%

The NPVs of the truck, conveyor, and incremental cost flows are plotted for different discount rates. The truck and conveyor NPV curves cross each other at -1 1%. Similarly, the incremental cost curve crosses the zero NPV line at -11%, indicating that the return on the incremental cash flow is -11%.

Figure 15 Trade-off Study NPV Curves Figure 15 plots the values calculated in Tables 6 and 6A and presents a typical set of cost flow trade-off curves. The conveyor has the lower undiscounted total cost (NPV at O%), but as the impact of the savings in futures years becomes lees and less significant as the discount rate increases, the total cost NPV of the truck option becomes the lower value. The cross-over occurs at 11% discount rate. For project hurdle rates below this point, the conveyor case has the lower total cost NPV and would be selected as the preferred option. For hurdle rates above this point the truck option would be selected. The incremental cost flow is the result of subtracting the conveyor cost flow from the truck cost flow. In the first year it gives the incremental cost for investing in the conveyor and in the subsequent years it gives the incremental savings. This series of values has the typical pattern of a project cash flow, with an investment (negative) followed by a benefit (positive). This being the case, it is possible to calculate the return on the incremental investment to install the conveyor. The IRR of the incremental cash flow is the point where the incremental curve crosses the zero NPV line, at -11%. It is the same point where the conveyor and truck NPV curves cross each other. Therefore, the IRR on the incremental cost of investing in the conveyor is 11%. If this is above the hurdle rate then the conveyor is preferred, if not, select the truck option.

-

-

Replacement Chain Analysis The example of the conveyor versus the truck given in the trade-off is an example of a replacement chain analysis. The conveyor’s life is twice as long as the truck lives so the truck has to be replaced twice during the life of the project.

365

Equivalent Annual Costs DCF methodology can be applied to the costs for options with different lives to determine an equivalent annual cost, which is the equal annual cost amount that has the same NPV as the detailed cost flow. A sample calculation is shown in Table 7 using the truck cost figures from Table 6. (The calculation in Table 7 includes a salvage value for completeness. The impact of the salvage value is not significant in these examples but could be in actual individual cases.) The NPV at 10% of the 6-year cost flow for the truck is $3,567. The 6 years include a year -1 which is assumed to be an order and delivery period, with the truck beginning work on day 1 of year 1 and operating for 5 years from that date. Year -1 is included in keeping with the need to order large mining equipment well in advance, and the need to make intermediate installment payments, as well as the desire to keep the math comparable to the conveyor option, which requires construction. (Please note that some texts will show the first period as time = 0 and do not discount the initial purchase price.) The equivalent annual cost is the equal annual value that will give the same total NPV when discounted back to the same point in time, that is, the beginning of Year -1. The desired NFV value is $3,567. The sum of the NPV factors for the operating years 1 through 5 is 3.4462. Therefore, the equivalent annual cost is equal to: $3,567 13.4462 = $1,035 When $1,035 is substituted for the cost flow in the operating years only (Year -1 = O), the NPV is $3,567. In other words, the string of cost flows that represent the truck option can be replaced by a single equal value of $1,035 in each of the operating years. This equivalent annual cost calculation is essentially the same as a lease calculation, where the capital aspect is removed and replaced with only an operating cost that includes the amortized capital cost. It is worth noting that the conveyor provides a lower equivalent annual cost at discount rates up to -1 1%, after which the trucks have the lower equivalent annual cost. This is the same ranking that the trade-off analysis indicated. Table 7- Equivalent Annual Cost NPV Calculation Operating Capital Salvage Tax Deduction cost Flow NPV @ 10%

-1

1

2

3

4

5

Total

$0 $2500

$1000

$1000

$1000

$lo00

$1000

-

-$600 $400 $331

-$600 $400 $301

-$600 $400 $273

-$600 $400 $248

-$250 -$600 $250 $141

$5000 $2500 -$250 -$2900 $4350 $3567

$1035 $855

$1035 $778

$1035 $707

$1035 $643

$1035 $584

$5175 $3567

.9091

3264

.7513

.6830

.6209

S645

-

.8264

.7513

.6830

.6209

S645

$2500 $2273

Equivalent Annual NPV @ 10% PV Factors PV Factors during operating period

Table 7A- Equivalent Annual Cost Comparison of Truck and Shovel Options Discount Rate Truck Conveyor 10% $1035 $995 11% $1052 $1038 12% $1070 $1082 13% $1088 $1 126

366

3.4462

Appendix A Cash Flow Versus Accounting Values It is important to distinguish between accounting values and cash flow values. Cash Flow Back-Calculated from Accountin Fi ures

Cash Flow Calculated From Lhs;oc;;.Np;;q

+

Gross Revenue TC/RC/PP Royalties Ouerating Costs Net Operating Income (EBITDA) Depreciation & Amortization Deuletion Allowance Net Taxable Income (EBIT) Mining and Income Taxes Net Profit After Taxes (Earnings)

+

Net Profit After Taxes (Earnings) Depreciation & Amortization Depletion Allowance Capital Expenditures Working CaDital Accounting Cash Flow

Cash Taxes calculated separately + Net Operating Income - Cash Depreciation = Net Taxable Income = Mining & Income Taxes (Cash)

= -

=

=

+ + + =

=

=

Gross Revenue TC/RC/PP Royalties Ouerating Costs Net Operating Income (EBITDA) Capital Expenditures as Spent Working Capital Mining & Income Taxes (Cash) CashFlow

The significant differences between cash and accounting representations of a project are noted below: Timing and Depreciation The essential difference between cash and accounting values is that in a cash flow revenues and costs are recognized as they occur. In an accounting profit and loss calculation, capital items are not included in the year they are spent, but are depreciated over a prescribed time period (say 3,5,10 or more years) or are amortized over the life of the operation. For example, a capital investment of $10,000,000incurred in the year before production begins would be a included in a cash flow as -$10,000,000 in year -1, whereas the accounting figures would show $0 in year -1 and -$1,000,000 in each of years 1-10 (assuming a 10 year depreciation). Year -1

Cash Flow - Costs shown as incurred

Accounting 10 year depreciation

-$10,000,000 -$1,000,000 -$1 ,oO0,000 -$1 ,oO0,000 -$1,000,000 -$1 ,oO0,000 -$1,000,000 -$1,oO0,000

-$1 ,oO0,000 10 Total

-$1,000,000 -$1,000,000

-$10,000,000

367

-$10,000,000

The purpose of the cash flow is to reflect the timing of costs and income as they occur so that they can be assessed in terms of their time value and so that a return can be calculated. There is no attempt to have the income in each year reflect its portion of the original capital investment. In a cash flow evaluation, the entire project supports the repayment of the original investment and the return on it. The accounting representation of capital investment is used to show that the annual income bears a portion of the original capital cost until the capital is repaid. The income in each year has to reflect the fact that capital was spent to earn that income. Once the capital has been fully depreciated, there is no more depreciation remaining for this particular asset and the annual income will increase accordingly. Sunk Costs. The cash flow method is forward looking. This means that past, or “sunk’, costs are not included. The reason for this is that an investor is only concerned with whether a particular investment will make a sufficient return to justify the investment. Good or poor investments that have gone before do not impact on whether the current investment is valid. The only place that past expenditures are relevant is in tax deductions where tax pools will continue to include past capital investments until they are fully deducted for tax purposes. In the accounting convention, the depreciation of previous capital assets continues until the asset is full depreciated. Back Calculating Cash Flow From Accounting Income. It is recommended that cash flows be calculated from the basic cash inputs. It is recommended that cash flows not be back-calculated from accounting Net Income statements. Although the cash flow and accounting values will match in total at the end of the life of the mine, values will not match year-by-year. Unfortunately, accounting net income projections are often easier to find than cash flow projections and so are used as the basis of many cash flow calculations. This is done by calculating the net income after tax on an accounting basis and then undoing some of the accounting items (adding back the depreciation, amortization and depletion) and inputting the missing cash items (deducting capital expenditures as incurred and deducting Working Capital). This process gives a mixed bag of cash and accounting values, because the taxes are on an accounting basis and the rest of the calculations are essentially cash. To be more correct, the accounting taxes should also be added back to the accounting net income and then cash taxes (taxes with deductions based on actual tax legislation) deducted. By the time all of this is done, it is easier to start with the basic cash inputs and generate a cash flow from the beginning, and there is much less likelihood of error. The major difference between a back-calculated accounting “cash flow” and an actual cash flow is the timing of the taxes. This is due to the difference in the timing of the depreciation for accounting tax purposes and for actual cash tax calculations. For actual cash tax calculations, the legislation typically permits an accelerated write-off of assets so that the depreciation deductions are taken more rapidly in the early years of the mine (“front ended”), thereby reducing the cash taxes payable in the early years. Accounting depreciation tends to allocate the deductions more evenly and over a longer period so that accounting taxes are more even over the mine life and so are greater than the cash taxes in the early years. Appendix B General Equation for Capital and Operating Cost Curves Work by O’Hara (1980, 1992) and the USBM (1987) suggest that the curves for capital and operating costs can be reasonably approximated by exponential equations, with the general form: Cost = K tX where:

K = a constant specific to the particular cost = production rate in tonnes per day x = anexponent .5 to .7 = typical range for capital costs .7 to .9 = typical range for operating costs in $/year -.3 to -.1 = typical range for operating costs in $/t t

368

These equations should not be used for detailed estimating, although they can give guidance for order of magnitude estimating. However, a cost is known accurately, this relationship can be used to factor the cost up or down for differing production rates, within reasonable limits. The relationship is this: Cost at tl Costatt2 then:

=C2

= K tlx = Kt;

Cl/C2

= (K tlX)/(Kt;)

= C,

= tlx/t;

= (tl/t2)x (Because K is common it can be eliminated. Note that the value of “K’ does not need to be known.) simplified: CI/C2 = (t,/tz)” = C1 (tz/tl)” (note inversion of tl and t2 to put C2 in numerator) then: c 2 For a capital cost or annual operating cost, if C1 and tl are known, and x can be estimated from experience (.6 is a reasonable first estimate for capital costs), then C2 can be estimated for a given t2. Therefore, if the cost at 20,000 t/d is $30 million, then the cost at 25,000 t/d can be estimated to be: C2 = $30,000,000 (25,000/20,000)6 = $30,000,000 (1.1433) = $34,298,000 For a unit operating cost of $IO.OO/tonne, and an exponent of -2 the unit cost at the higher tonnage will be: C2 = $10.00 (25,000/20,000)~~2 = $10.00 (.9564) = $9.56/tonne

REFERENCES Cavender, B., 1992, “Determination of optimum lifetime of a mining project using discounted cash flow and option pricing techniques”, Mining Engineering, October, 1262-1268. Glanville, R., 1987, “The valuation of mining properties (with a case study of a computer-assisted application of the discounted cash flow (DCF) valuation method)”, United Nations Interregional Seminar On The Applications Of Electronic Data Processing Methods In Mineral Exploration And Development, Sudbury. Smith, L.D., 1995, “Discount rates and risk assessment in mineral project evaluations”, CIM Bulletin, Vol 88, No 989, pp 34-43, and Transactions of the Institution of Mining and Metallurgy, (Section A: Mining industry), Vol 103, Sept-Dec 1994, A137-154. Smith, L.D., 1997. “A critical examination of the methods and factors affecting the selection of an optimum production rate”, CIM Bulletin, February 1997, Vol90, No. 1007, p 48-53 Taylor, H.K., 1986, “Rates of working of mines - a simple rule of thumb”, Technical note published in the Transactions of the Institute of Mining and Metallurgy (Section A: Mining Industry), Vol 95, October, A203-204. Taylor, H.K., 1996, Personal correspondence.

369

USBM Information Circulars 9142 & 9143, 1987, Bureau of Mines Cost Estimating System Handbook (in two parts); 1 . Surface and underground mining; 2. Mineral processing, US Department of the Interior. Wells, H.M., 1978, “Optimization of mining engineering design in mineral valuation”, Mining Engineering, Dec, 1676-1684. Smith L.D., Discount rates and risk assessment in mineral project evaluations. Canadian Znstitute of Mining, Metallurgy and Petroleum Bulletin, Vol88, no. 989, April 1995,34-43. Smith L.D., Inflation in project evaluation. Canadian Institute of Mining Metallurgy and Petroleum Bulletin, Vol80, no. 899, March 1987, 129-133. Smith L.D. The argument for a “bare bones” base case, Canadian Institute of Mining Metallurgy and Petroleum Bulletin, Vol92, no. 1031, June 1999, 143-150. Smith L.D. Discounted cash flow analysis methodology and discount rates, Canadian Institute of Mining Metallurgy and Petroleum, Mineral Property Valuation Proceedings, Mining Millennium 2000, March 8,2000, Toronto, 85-99.

370

Mining Project Finance Explained Rob Halupka’

ABSTRACT Developers of mining projects may make more effective use of their time and resources in seeking financing if they understand issues from a lender’s perspective. While project developers and investors are naturally interested in the growth prospects of their business ventures, lenders must consider the downside possibilities. When banks and other financial institutions lend capital on a long-term basis, they do so in return for relatively modest and fixed returns; consequently, lenders willingly accept “borrower’s risk”, but avoid taking what they perceive as “equity risk”. The essential ingredients needed for launching a successful project financing are reviewed and summarized in this paper. RISK-REWARD IN CONTEXT The risk-reward balance (see Figure 1)is a simple enough concept - borrowers invest money in the expectation of earning higher returns by assuming higher risks; however, lenders “loan their balance sheet” for lower returns, commensurate with lower risk. Commercial banks earn a return on capital from fees and the spread between borrowing and lending rates, which are fixed prior to the execution of a loan agreement. Surpluses over and above that required to service project debt are typically used partly to prepay the project debt and partly to provide a return to the borrowers their reward for originating the project, investing their resources and accepting the risks associated with project construction and start-up. However, for lenders, preservation of capital is paramount; lenders must be conservative to stay in business. On the riskheward spectrum, junior exploration companies occupy the high end of the risk-reward curve.

JuniorEquity

/

/

Senior Equity

Proiect Loans I

RISK

Figure 1 - Risk-Reward in Context ~~

* Vice President, Global Mining & Metals, RBC Capital Markets, Toronto, Ontario

371

Investors in these ventures know (or should know) that their investments are speculative; the analogy of buying lottery tickets may not be far off the mark. From time to time, however, a junior will make a bonanza discovery, at which point the junior’s shareholders are richly rewarded (e.g. Arequipa Resources’ discovery of the Pierina gold deposit in Peru, or Diamond Fields’ discovery of the Voisey’s Bay nickel deposit in Labrador). Lenders are at the low end of the riskheward curve; a further distinction may be made between corporate and project lending. Typically, project lending is riskier than corporate lending, which is reflected in slightly higher fees and loan margins. Tenor (i.e. loan life) varies for both corporate and project loans. The outside limit for term loans for mining projects is about twelve years; however, in the current market an acceptable tenor is more typically six to eight years.

PROJECT FINANCING DEFINED Project financing is the financing of an independent capital project that a sponsoring company (or companies, as there may be more than one) has segregated from its assets and general-purpose obligations. Economic prospects of the project, combined with commitments from the sponsor and third parties, provide the required support for extensive borrowings with limited recourse to the parent company. During the construction (pre-completion) phase, the sponsors must guarantee a project loan; however, in the operating (post-completion) phase, lenders look to the cash flows and earnings of the project to service the loan, and to the project’s assets as collateral for the loan. The key characteristics of project financing are further summarized as follows: 0

0 0

0 0

the venture is established as a separate business and legal entity, relying heavily on debt leverage (typically 50% or more) for its capital needs; borrowing is linked directly to project assets and cash flow potential; commitments by third parties (including suppliers, customers, government agencies) and sponsors form key elements of credit support; sponsor’s guarantees to lenders usually do not cover all risks involved; project debt may be differentiated from parent (sponsoring) company’s direct obligations.

PROS AND CONS OF PROJECT FINANCE There are several benefits for borrowers considering project finance as an option in financing their project; these are as follows:

0

A source of finance: Project lending is attractive to banks seeking higher returns than may be available from corporate lending. Therefore the universe of potential lenders for a project finance deal may include financial institutions that are not interested in participating in a conventional corporate loan facility. The sponsor retains control of project: Within the constraints imposed under a typical project finance deal, the project sponsor retains control of the project. In many instances, the only other source of funding would be an equity issue (leading to share dilution) or the sale of an interest in the project to another party. The participation of Export Credit Agencies (ECA’s) or multilateral agencies may be the only way to secure financing for projects in riskier “sub-investment grade” countries. ECA’s offer political risk insurance (PRI), which may pave the way for commercial banks that might not otherwise be interested in participating in the project. Even if no PRI is purchased, ECA’s - especially the likes of the U.S. Export-Import Bank - provide a “halo effect”. Governments of less stable countries tend to think twice before acting against a foreign investment that could provoke a highly visible quarrel in an international context with a large, influential organization. The reduced risk exposure to sponsors with a non-recourse loan: Once the project has passed the Completion Test, the loan becomes non-recourse and the sponsor is effectively insulated from the project’s creditors. Should the project fail to service the project debt,

372

0

0

0

the creditors are restricted to the project assets in seeking a remedy. Furthermore, the sponsoring company’s debt capacity is freed up, providing capacity for the sponsor to pursue other business opportunities. Well-suited for joint owners: often there are two or more sponsors participating in a project through a joint venture ownership vehicle. A project financing type loan structure may provide the most attractive platform for raising finance for an investment. Risk reduction: A sponsor may be reluctant to absorb all the risks in an investment situation. It may be more prudent - and offer a better fit with a company’s business strategy - to allocate resources over several different investments than to devote 100%of the firm’s resources to a single project. Leverage: Debt finance is almost always a materially cheaper way for a project owner to raise funds than by issuing shares.

Project sponsors considering project finance as an option for raising funds should also be aware of potential disadvantages and costs, including the following: 0

0

0

0

Deal costs: project financing transactions typically require the services of financial advisors and consultants for technical and marketing aspects, insurance and legal counsel. The work involved is considerably more extensive - and therefore more expensive - than is required for corporate loans. In addition to the advisor and legal fees, up-front fees and loan margins charged by lenders are typically more expensive than for corporate loans. Lengthy, detail-oriented process: Project finance deals are typically highly structured and the process may seem rather cumbersome to the uninitiated, who may not appreciate the commitment of both time and resources required to launch a project financing. Restricted access to the project’s cash flow: Pay-out of dividends and other forms of cash distribution to the sponsors is at least partially restricted until the project loan has been fully repaid; if the project performs exceptionally well, typically half of any surpluses may be directed to pre-paying the project loan. Full recourse to sponsors prior to “Completion”: Until the project attains Completion, a senior project loan is fully guaranteed by the sponsor. Should the project fail to meet performance expectations, the sponsors could be called on to fully repay the loan.

RISK FACTORS IN PROJECT FINANCING Credit ratings for corporate bonds (assigned by credit agencies such as Standard & Poor’s and Moody’s) are used by bond purchasers to gauge risk and to set yield, thereby fixing bond price. Commercial banks generally rely on their internal credit rating procedures to gauge a client’s creditworthiness, but they also monitor external credit ratings. Banks will accept minimal conditions in corporate loan agreements with borrowers that enjoy strong credit ratings because of the relatively low probability of default on such loans. By contrast, loan agreements in project financing are characterised by extensive deal structure designed to protect the lenders’ interests over a variety of possible outcomes. Project loans are generally considered riskier than corporate loans since only a sole asset supports the loan in the post-Completion phase. Projects in developing countries are further exposed to country or political risk, which is the potential for government intervention that could negatively impact the venture and thereby impair the loan. Consequently, non-recourse project loans are typically priced at a premium to corporate loans. Banks consider several areas of risk in examining a mining project loan: 0 0

0 0 0

Sponsor quality Country/political risk Deposivreserve risk Construction risk Performance risk

373

0 0 0

Market risk Price risk Environmental risk

Sponsor quality ranks as the primary issue to lenders; banks must have confidence in a sponsor’s financial strength as well as its capabilities as a mine developer and operator. Few lenders will progress very far in their analysis until this key condition has been satisfied. A strong sponsor must have the financial resources, the management skills and technical know-how required to meet the challenges of launching a new project and to solve the myriad problems that will arise during the construction and start-up phase. Country or political risk issues also attract close attention from lenders. A recent and painful reminder of is Argentina’s default on debt obligations, resulting in substantial losses to many banks. Banks and credit agencies independently rate individual countries and assign risk ratings just as they do for corporations; these ratings are used both to price loans and to set limits on credit exposure that banks will accept. Internal bank policy-makers also place term limits to loans on a country-by-country basis. The net result is the implementation of guidelines and constraints that may restrict a bank’s ability to conduct business in a given country at a given point in time. Each new loan consumes a portion of precious country limit, leaving less room for the next one. Country risk may be mitigated by the participation of Export Credit Agencies (ECA’s) and through the sponsors taking out political risk insurance. In Argentina, the Alumbrera coppergold project financing included ECA coverage, while the Cerro Vanguardia gold project did not. It is noteworthy that both these projects have continued to service their respective project loans without incident (as at March 2002) despite Argentina’s financial crisis. Some form of political risk protection may be essential in order to make a project financing palatable to the commercial banks. This area of financing can be complex and there are several organizations that specialize in structuring these deals. Multilateral development banks, such as the European Bank for Reconstruction and Development (EBRD), may provide the ingredients that pave the way for commercial banks. ECAs may act as primary lenders, provide guarantees or insure loans; examples of ECA’s are Export Development Canada (EDC) and the U.S. Export Import Bank (US Exim). The World Bank’s Multilateral Investment Guarantee Agency (MIGA) is another institution that can provide the necessary structure to encourage the commercial banks to participate in project loan deals in riskier countries. It is also possible to purchase political risk insurance from commercial providers such as Lloyd’s of London and American International Group Inc. (AIG). Depositheserve risk refers to the risks associated with the ore deposit including grade and tonnage, and the uniformity and attitude of the ore body. Confidence level is generally a function of the quantity of work carried out in developing resources and reserves, including drilling, sampling, assaying and metallurgical testing (from bench-scale to pilot plant testing). The amount of work required to develop a sufficient “comfort” level in the statements of resources/reserves will be largely dependent on the geological nature of the deposit and its genesis, e.g. contrast porphyry copper deposits with high-grade narrow gold quartz vein structures. As a rule, lenders restrict the basis of their lending to “2P’ (i.e. Proven and Probable) reserves. The sponsoring company’s track record is certainly relevant, particularly with the type of deposit and mineral commodity involved for the project at hand. Financial institutions rely heavily on the advice of independent engineering firms that are competent in the technical aspects of geology, mining and metallurgy. In order to protect the interests of the lenders, a reputable consulting firm is hired as the “Independent Engineer” to audit the reserve work, to opine on the feasibility study and construction plan, to monitor construction progress and to sign off on the completion test as laid out in the loan agreement. Construction risk is the risk of capital cost over-run occurring after a project financing deal has closed and the project loan is committed or drawn. Causes of over-runs include delay due to bad weather, underestimating logistical challenges, change or redesign orders, construction

374

deficiencies, off-spec equipment, strikes and other forms of labour disruption, shortages of skilled labour and various start-up or “teething” difficulties. Project sponsors are expected to fully cover capital cost over-runs, which is why lenders insist on sponsors having the financial resources to back up their initial investment. Since many projects are not completed on time/on budget, lenders try to anticipate and quantify the magnitude of potential variances from plan, where the plan is defined by the feasibility study. Prevention is the first line of defence against a significant cost over-run. Lenders look for evidence of a sound project concept backed by appropriately detailed engineering work followed by competent execution during construction. The reputation and ability of the engineering and construction firms involved in the project are therefore of importance to the lenders. Adequate contingency also provides further assurance that the project will stay within budget. Performance risk is the risk of a production shortfall, excessive operating costs or offspec product that may be rejected by off-takers. The presence of any of these conditions will impact the operating margin of the project, and therefore impair the project’s ability to service debt. Operating deficiencies may have a variety of causes including, for example, below-plan ore grade (due to inadequate sampling, insufficient drilling of the deposit or excessive ore dilution), process design flaws (particularly with new or unproven technology on a commercial scale), premature equipment failure, labour strife or environmental non-compliance. Lenders mitigate performance risk by screening out all but the more robust projects with high quality feasibility work performed by reputable sponsors and engineering firms, and having this work reviewed by trusted independent technical consultants. It is a standard prerequisite that the project maintains a comprehensive insurance program assignable to the project lenders. Marker risk is particularly relevant for operations producing semi- or unfinished products that subsequently require further processing by a third party to extract the final product value. An example is copper concentrate that must be smelted and refined to produce finished copper metal meeting LME specifications. The concentrate may also contain significant quantities of gold and silver recovered through smelting and refining. There is an active global market for copper concentrates with a supply/demand cycle that affects a producer’s ability to place its product for custom treatment and to maximize the net smelter return. Project financing deals typically require that at least 80% of production during the loan life be secured under long term purchase and sale contracts with approved smeltinghefining companies. Price risk refers to the uncertainty surrounding future prices (particularly over the loan life) for a project’s products - conventionally, metals such as copper, zinc, nickel and gold. In fact, many mineral commodities may be project-financed, including a broad range of industrial minerals - even diamonds. As anyone who has tested sensitivity of the forecast economics of a mining project knows, metal price is usually the most sensitive variable in the analysis. Industry observers know that metal prices exhibit a high degree of both volatility and cyclicality. Metal markets are truly global - metal consumers are less interested in brand names than in standard specifications of product content and impurities and in reliability of delivery. Transportation costs for the metals tend to be of limited importance since these costs normally form a small component of total production costs and metal price. Lenders are attracted to projects that are forecast to be low cost producers since these projects may be expected to cover cash operating costs and service senior debt in times of low prices. Price risk may be partially mitigated - at least for some commodities - through hedging. There is a variety of hedging instruments available including, but not limited to, forward sales, purchase of put options and various other derivative instruments. Hedging can be a conservative management tool provided hedge commitments are covered with forecast production over matching time periods. Environmenfal risk is the risk of a project failing to operate within acceptable standards of environmental compliance. Most lenders will not provide financing for a new mine that is expected to operate outside of international standards for environmental compliance - regardless of the mine’s location in the world. In a worst-case scenario, the principle of “deep pockets” could result in the project lenders being held liable for damages and restoration costs if the borrower lacks the funds to pay. Furthermore, most lenders are concerned about their corporate

375

image and can ill-afford the bad publicity arising from an environmental incident. Reputable mining companies are motivated to improve their environmental record in response to pressure from shareholders and other stakeholders. An essential responsibility of the Independent Engineer acting on behalf of the lending group is to comment on the project’s present and future expected compliance with environmental laws of the host country and international standards - usually as covered by World Bank guidelines.

PROJECT LIFE CYCLE Figure 2 is a conceptual illustration of the spending and funding requirements for a new mining project, from exploration and discovery, through feasibility and construction, and finally through operation, assuming that “green lights” are encountered at each step of the way. In the early exploration and reserve development phases, prior to the feasibility, mining projects are as a rule too risky for lenders, and funding is restricted to equity sources.

Project

Go Ahead Decision

FUNDING:

I EQUITY1

DEBT + EQUITY

Figure 2 - Project Life Cycle A promising project may support the sponsor’s decision to conduct a feasibility study, often undertaken with the assistance of a reputable engineering or consulting firm. A favourable feasibility study will provide the company’s Board of Directors with sufficient confidence to support a go-ahead decision and form the basis for initiating discussions with potential financial advisors and lenders. The feasibility study would include a life-of-mine plan based on reasonably detailed engineering work, and the project economics would be demonstrated in an accompanying financial forecasting model.

376

LOAN STRUCTURE FOR PROJECT FINANCE Project loans are tailored to the specifics of each project. Determination of the loan amount begins with a review of the capital budget and takes into consideration the project economics as forecast in the financial model. In addition to construction and EPCM costs, the capital budget should include owner’s costs, initial working capital (including inventories and “first fill”), financing and legal fees, and interest during construction. Ideally, sponsors will have a loan agreement in place before the project enters the construction phase, when heavy spending commences. Sponsors are normally required to contribute their own funds (equity) to the project at each draw down of the loan in line with a pre-agreed debvequity ratio. Virtually all projects have a ramp-up period, i s . the phase between commissioning and the operation attaining design capacity. The first loan repayment typically commences within a few months of the project passing the Completion Test, as defined in the loan agreement. The loan amortization schedule will depend on specific circumstances, but equal, semi-annual payments of principal (together with interest due) are common. The maximum tenor (i.e. the period from the signing of the loan documents to the final scheduled loan payment) for a mining project finance loan is about twelve years, but six to eight years is more typical in the present market. A typical scenario might be a three-year construction and ramp-up period for a mine with a ten year operating life (as projected from the Feasibility Study, based on Proven and Probable reserves), with the loan amortizing in ten equal semi-annual payments over the first five years of operation. A standard rule of thumb in project lending is to restrict the loan life to a maximum twothirds of the project life on the basis of proven and probable ore reserves, thereby leaving a 50% “reserve tail” as a cushion against deviations from plan [see Figure 31. The reserve tail provides a measure of assurance that the project will be able to service debt beyond the loan life in the event that the operation fails to deliver projected cash flows due to either operating deficiencies (e.g. excessive operating costs, production shortfalls) or external circumstances (e.g. metal prices). It is noteworthy that there are few examples of project financing for underground mines since deeper deposits tend to be less well defined - in terms of Proven and Probable reserves - than nearsurface deposits (amenable to open pit methods) at the start-up phase when project financing is being arranged.

PROJECT LOAN LIFE c 2/3 PROJECT LIFE

Basis :Proven & Probable Reserves LOAN LIFE

RESERVE TAIL

4

~

PROJECT LIFE

Figure 3 - Typical Restriction on Loan Term The key terms and conditions of a project financing negotiated between the sponsor and lead group of arranging banks are summarised in the Term Sheet, which forms the basis of the

377

formal loan agreement - the final legal documents binding the parties over the duration of the loan. The loan agreement provides specific remedies for a broad range of situations that could arise should the project deviate from the path charted by the feasibility study. To the uninitiated, the process of negotiating the term sheet and agreeing the details of the loan agreement may appear daunting due to the complexity and level of detail. Financial covenants in the loan agreement provide triggers for lender protection. A typical feature in a loan agreement is a cash sweep mechanism, whereby surplus cash (beyond that required to fund ongoing operating costs and sustaining capital expenditures and to service senior debt) is shared equally between the sponsor (as “restricted payments” in the form of dividends or payments against subordinated debt) and the lenders (as prepayments against senior project debt).

Escrow/Collateral Accounts Borrowers are normally required to maintain segregated accounts with the Administrative Agent (one of the lead banks appointed to that role) to collect revenues and to hold cash balances sufficient to cover, for example, the next 90 days of the mine’s cash operating costs and the next senior debt payment including principal and interest. For projects located in developing or higher risk countries, borrowers must usually maintain offshore US dollar accounts, thereby reducing exposure to cross-border currency risk and insulating sales proceeds from potential interference by governments in which the project is located. The Completion Test The purpose of the Completion Test is to demonstrate that a project has attained sustainable commercial production in line with the parameters laid out in the feasibility study. Terms and conditions of the completion test are negotiated, with banks favouring tighter, restrictive terms, while sponsors seek the opposite. While the Independent Engineer (IE) may or may not assist the lending group in the early negotiations, the IE will always play a key role in monitoring the completion test and in signing off on the agreed test components as they are met. The trial period over which a project must demonstrate commercial viability is typically 90 days. Key operating parameters include mine tonnage (ore and waste), mill throughput, metal recoveries, metal production quantities, product quality and operating costs. The completion test is a significant milestone in the life of a project for two reasons: (1) most importantly, for the transfer of risk, and, (2) for loan pricing. When a project passes the completion test, the sponsor’s guarantee falls away and the debt obligations transfer from the sponsor to the project on a stand-alone basis. For lenders, the worst-case scenario would occur post-Completion, in which the project fails to meet scheduled debt and interest payments and the sponsor has the right to abandon the project, leaving the lenders as the new owners of the project. Without the sponsor’s guarantee, pricing of the project loan is no longer linked directly to the sponsors, and instead reflects the project specifics (as forecast when the loan agreement was signed), and the loan spread usually increases.

ELEMENTS OF A SUCCESSFUL PROJECT FINANCING In conclusion, one can identify five main elements needed for a successful project financing; these are as follows: 1. The involvement of strong sponsors with established credentials in building and operating

major mine projects, and that are financially sound, i.e. have moderate levels of debt, the ability to generate cash from existing operations, good access to capital markets with existing lines of credit and the capability to raise funds in the equity market. 2. Country or political risk may be an issue for projects located in countries perceived as risky (i.e. sub-investment grade) by the international financial community. The participation of one or more export credit agencies in the financing andlor the provision of

378

political risk insurance to safeguard the loan facilities may be necessary to provide sufficient comfort to commercial lenders.

3. The project must be robust, i.e. show a high probability of becoming a low cost producer underpinned by a substantial base of proven and probable ore reserves. Forecast economic performance must indicate that the project is likely to provide good debt coverage while offering sufficiently attractive rates of return to the sponsors under a range of sensitivities. The Feasibility Study, which essentially defines the project for both sponsors and lenders, should be a highly credible piece of work undertaken by a well-qualified, reputable firm. 4.

Key licences and long-term contracts are normally required to support the project at least for the duration of the project loan, if not beyond. Essential licences typically include property and right-to-mine leases, and, in some cases, water supply licences (e.g. important in arid regions of Chile). Key contracts may include supply of electrical power, natural gas, coal or other fuel, maintenance-and-repair contracts (MARC) for the mining fleet, essential process input materials, land and/or ocean transportation, port facilities, insurance and off-take contracts (e.g. long term frame contracts for smelting and refining of concentrates).

5 . The financing structure should strike a reasonable balance of riskheward among the sponsors, the lenders and other stakeholders. A typical range for the level of debt is 50% to 65% of total funding requirements, although the actual debt capacity will depend on the merits of the project and anticipated cash flows.

379

Mineral Processing PlantKircuit Simulators: An Overview J. Herbst,' R.K. Rajamani,' A. Mular,' and B. Flintof

ABSTRACT Over the past 40 years or so simulation has emerged as a powerful and seemingly ubiquitous tool for the design, analysis/optimization and control of mineral processing unit operations, circuits and plants. This chapter examines the evolution of mathematical modeling and simulation in mineral processing, introducing the various approaches and applications. Particular emphasis is placed on population balance modeling looking at the use of both steady state and dynamic simulation. Also included is an introduction to the more recent developments in high fidelity simulation, and in particular the application of discrete element modeling to the optimization of comminution device internals. INTRODUCTION Process simulation is well established as a very important tool for mineral processing engineers. The principal application areas include the design, analysis/optimizationand control of processing systems. These areas are not mutually exclusive, and it is quite common for the same mathematical models to be used in two or more of these applications, and for a combination of models to be used for a particular application. There are other uses of process simulation, for example, operator and technical training, or research and technology development. These have been reported in the literature, but will not be discussed in this chapter. The appropriate dictionary definition (source: Encarta World Dictionary) of the term simulation is: sim-u.la*tion [simmya lriysh'n I COMPUTING STATISTICS construction of mathematical model: the construction of a mathematical model to reproduce the characteristics of a phenomenon, system, or process, often using a computer, in order to infer information or solve problems. Mathematical models are the fundamental building blocks of process simulation, and it is common to hear the two words modeling and simulation used as synonyms, since the latter is inevitably the outcome of the former. There are generally three kinds of models that can underpin a process simulation. Empirical: often a set of algebraic equations developed by regression, multivariate statistics, or neural networks. These are the so-called black box models, which are "trained" on sets of input and output data. (e.g. Bond's equation is a semi-empirical model) Phenomenological: often a set of algebraic and differential equations arising out of application of some engineering principles, physics and chemistry, but requiring plant calibration. (e.g. Population Balance Models) Fundamental: often a set or algebraic and differential equations based on fundamental laws of physics and chemistry, requiring minimal calibration. (e.g. Discrete Element Methods, Computational Fluid Dynamics) Figure 1 illustrates model hierarchy, highlights the tradeoffs in accuracy and computation effort, and attempts to convey that moving up the hierarchy tends to increase the modelhimulation detail. As one might expect, there has been uneven development in the models, as this activity has been driven by the economic and technical needs in the industry. For example, comminution 'Metso Minerals -Minerals Processing Business Line, York, PA. 'University of Utah, Department of Metallurgical Engineering, Salt Lake City, LIT. 'University of British Columbia, Mining Depr, Vancouver, BC.

383

related models exist at all three levels, but empirical models are still the dominant means to characterize classification size as in hydrocyclones. The consequence is that many of the simulation-based problem solving techniques and tools use a combination of two or more types of models. (Lynch and Narayanan, 1986; McKee and Wiseman, 1991) have tackled the general topic of simulation applications in mineral processing, with a view to documenting industrial practice and speculating on future developments.

.

Micro Scale Simulation

(e.g. Population Balance Models) Empirical (e.g. Regmssion Equations)

Figure 1 The mathematical modeling hierarchy This paper provides an application-oriented review of simulation in mineral processing, beginning with a brief chronologically oriented introduction to the subject. It offers an overview of the current applications of the three simulation approaches highlighted in Figure 1, with a particular emphasis on the phenomenological and fundamental approaches.

A CHRONOLOGICAL INTRODUCTION TO SIMULATION IN MINERAL PROCESSING Comminution and in particular, grinding mills and circuits, have received the greatest attention in modeling and simulation, and for that reason ball milling has been chosen to briefly illustrate the history of development. Since empirical and population balance methods are pretty well integrated in our working culture, less time is spent on these topics. The Period up to the Early 1960’s Mathematical models have been around for as long as engineers, scientists, economists, etc. have been thinking about problems in a quantitative framework. However, prior to the introduction of low cost digital computation and high level programming languages, the need for analytical solutions limited the complexity of system models to a few algebraic or even fewer ordinary differential equations. A particularly good illustration of the kind of model that evolved in this era is Bond’s (1952) Work Index Model, given by Equation (1). This model derives from a blend of theory, i.e. the energy requirements to create new particles, and practice, i.e. an enormous experimental database supporting the fundamental premises of model structure.

where: W = specific energy (kWWt) required to comminute the feed. P = power draw (kW) T = tonnage processed (tph) Wi = work index (kWh/t) Ps0= 80% passing size of the product, in microns

384

Fso= 80% passing size of the feed, in microns t = short ton (1 ton = 2000 lbs) Over the years Bond’s equation has been used to design many rod and ball mill circuits, and it is still the basis of much of the ball mill design work. As the scale of equipment and breadth of ores treated has widened, the database around the standard Bond test has increased yielding correction factors/methods that account for limitations in the original equation (Rowland, 1982, Napier-Munn et al., 1996). In addition to its application in design, the Bond model has also been used in process analysis of comminution devices/circuits, such as illustrated in Figure 2. By making the appropriate measurements, one can rearrange and solve Equation (1) for the operating work index. Comparing this estimate to the laboratory value for the same ore provides a measure of efficiency (Austin et al., 1984, McKen ef al., 2001).

P Figure 2 A (Standard) Closed Circuit Ball Mill The Period of the 1960’s through the 1980’s Although the Bond model has considerable value in ball mill design, it does not have sufficient detail to be useful in analysis and optimization work. In response to these limitations, the late 1950’s and early 1960’s saw research activities begin to focus more on phenomenological models of comminution devices. The general approach is based on what are known as population balance methods/models or PBM’s. Many universities, including Queensland, Witwatersrand, Utah, Pennsylvania State, British Columbia and Alberta to name just a few, began to publish on the structure and application of these models for the analysis and optimization of comminution circuits. The work of Lynch et al., 1967, is a representative example of one of the early efforts to demonstrate the power and benefit of this approach. In this particular example, simulation studies pointed toward better efficiency with a simple circuit reorganization. Plant experiments (before and after) validated the predictions. The benefit was taken as a finer grind (from 70% to 77% -200 mesh) at essentially the same tonnage - a very significant change! The PBM simulation approach was investigated extensively in the 1970’s and early 1980’s, particularly with respect to the development of better models, and in the creation of methodologies concerning calibration (e.g. sampling equipment and rules, experimental procedures, mass balance methods, parameter estimation techniques, etc.). Borrowing from the jargon of the information technology (IT) world, this was a time when researchers and practitioners alike were focused on “content.” An important outcome of the research activity was the incorporation of modeling and simulation in university curricula. This led to an increasing number of plant process engineers who were knowledgeable in the use of the technology, and keen to apply their new skills. Symbolic of the changing mindset of this period is the fact that the SME volume on Plant Design first published in 1978 contained a single paper (ref. Mular and Herbst, 1980) introducing the framework for applying simulation in design and optimization. In contrast, the companion publication on the Design and Installation of Comminution Circuits published several years later

385

had five papers on this subject, four illustrating specific applications of simulation technology. Moreover, Comminution - Theory and Practice, published in 1992, had 15 papers focused primarily on modeling and simulation, and numerous others where modeling was employed in process analysis or control applications. Up to the mid 1980's, simulation software had been developed on a one-off basis by various researchers and practitioners, but as acceptance grew there was a clear need for commercial (robust and supported) software and standardized methods. Again borrowing from the IT jargon, this was the period where researchers and practitioners alike were focused on delivery. In Canada, the response to this challenge was the Simulated Processing of Ore and Coal (SPOC) project , initiated in 1980. Sponsored by CANMET (Laguitton, 1984), this was one of the first collaborative efforts by industry, academia and government to pull together a complete package of well-documented computer programs that would facilitate the use of simulation in industrial applications. Similar efforts were underway in other countries, and in 1984 King and Ford published on MODSIM, a general purpose steady state mineral processing simulator, which was commercialized in 1985. In 1986, the JKMRC introduced the first commercial version of JKSimMet (a general purpose steady state simulator for comminution circuit^)^, and BRGM launched the first version of USIMPAC. All four packages have undergone extensive development over the past 15 years or so, and while they all may arguably have a somewhat different technical focus these days, they still represent the de fucro industry standards. The advantages of PBM over the empirical approach in process analysis can be illustrated by the following example. The simulated circuit is shown in Figure 3. Equation (2) is presented only to show the relative complexity in the PBM approach5, when compared to the Bond model of Equation 1. Figure 4 presents the predictions of a modified Bond model (one that includes a size distribution expression to characterize the product stream), in this case presented as a family of curves labeled with the respective (operating) work index values. Figure 4 also includes the predictions of a general purpose simulator, in this case JKSimMet, as well as the experimental data points - one of which corresponds to the calibration data set, and the rest show measurements where the only input variable that was significantly different was the fresh ore feed rate.

Figure 3 A rod mill - ball mill circuit o = (I-CC)(I-(BSe +I-S)VBCC)-lk(l-C~(BS~ +I-SR)(I-CR(BSR + I - S ~ ) r l r ~ f (2)

where: o = cyclone overflow product size vector f = rod mill feed product size vector B = ore dependent lower triangular breakagetappearancematrix S = device dependent diagonal selection or rate of breakage matrix C = device dependent diagonal classification matrix vR,vB= stages of breakage

The JK has other commercial simulation products aimed at flotation circuits, minerals sands plants, etc. In fact, some of the terms in Equation (2) are nonlinear (e.g. C,, SR,SB),and the solution is not algebraic, but iterative based on what simulationists would term the sequential modular approach.

386

05 c

JKSimMet

80

65

Calibration Point

60 250

275

300

325

350

375

Tonnage (tph)

Figure 4 Simulated and experimental results Since the ore work index was essentially constant over the period of the testing, it is clear that the phenomenological models provide a much superior prediction of circuit performance than does the empirical model. In other words, the additional work required to gather calibration data yields benefits in terms of increased confidence and detail in predictions.

The Period of the 1990’s to the Present Although PBM dramatically improved the fidelity of the predictions in ball mill (circuit) performance, a number of important variables had to be accommodated through empirical correlations that were not rigorously derived, and not necessarily generally applicable. For example, ball sizing, ore hardness, and mill speed were incorporated based on expected changes in the rate of breakage function. Other variables such as slurry rheology and liner profile, both known to have an impact on mill performance, were essentially ignored in almost all mill models, In the very late 1980’s and early 1990’s, the application of Discrete Element Methods (DEM) to milling simulation was investigated at the University of Utah (Mishra and Rajmani, 1992,). Over the past decade, facilitated by the exponential increase in computer processing speed (Moore’s Law), there has been increased development activity in this area, and quite a number of industrial applications have been recently reported. As the name implies, DEM involves tracking the motion of a large number (typically c lo6) of discrete elements (e.g. balls) in response to the applied forces within a device (e.g. a rotating ball mill). Essentially this involves the solution of a very large set of differential equations describing each element’s (or particle’s) translational and rotational motion in 2D or 3D space. Mishra and Rajamani, 1992, provide a good introduction to the underlying mathematics. Recent development activities have focused on increasing computational speed to reduce simulation times and solve larger problems. The focus has been on parallelization of the codes and on better algorithms to “anticipate” particle interaction and simplify solutions. Despite these efforts, a reasonably large (5 x lo5elements) and complex (3D with non spherical shapes) problem may still require on the order of 5 days of simulation time to solve. Work is also underway to develop better pre-processing (problem set-up) and post-processing (data presentation and interpretation) tools. However, three dimensional simulation of plant scale mills (Rajamani and Mishra, 2001) too can be brought down to less then 24 hours of computing by maintaining both the feed and discharge conical ends with corresponding end lifters but using a shorter length in the cylindrical section. Figure 5 presents the mechanical analog to describe collisions between balls, or between balls and the mill lining as the balls and mill move. Each collision is resolved into an impact (normal) and a shear (tangential) component.

387

Figure 5 The mechanical analog (spring + dashpot) for collisions The energy losses on each impact can be tracked and used to develop energy spectra, such as is illustrated in Figure 6. In addition, animations of the model output can provide some helpful insights as well.

*

6 ,25 mm

Ball Load=35%

P)

r5-4

HC

--ee* 0

4.

0 0.01

0.1

1

10

100

1000

Single Collision Energy (Jncg)

Figure 6 Impact energy dissipation spectra for a ball mill (with a specific liner profile and speed) Figure 7 illustrates the translational velocity mapping of the media in the mill, and indicates the degree of cataracting and ball on liner collisions. For the interested reader, Cleary, 1998, has reported on a fairly broad ranging study of ball mills using DEM simulation. The shear and impact energy dissipation spectra are particularly useful in the estimation of liner wear (e.g. Qiu ef al., 2001), a phenomena that can be incorporated in the DEM computations to study the effects on such factors as power and the energy spectra over time. One of the more exciting recent developments is the prediction of comminution results based on the energy spectra. Bwalya et al., 2000 and Buchholtz er al., 2000, have described methods of incorporating breakage into DEM simulation.

Figure 7 Snapshot from a ball mill charge motion animation (with a specific liner profile and speed)

388

Nordell et al., 2001, have also reported on developments in this area, including discrete grain breakage, as well as ways to unify PBM and DEM. This later development is of particular interest to industry, as it builds on the proven strengths of both approaches to extrapolate into new applications of simulation technology. Figure 8 illustrates the possibilities by showing that it’s now possible to economically optimize the design of mill internals by trading off wear (liner cost + downtime) against throughput /grind. 4

50 Relative Throughput

4c Relative Wear, % 3(

2

a

-

1

1

D3

I2 Release N2 Numberof lifters

N3

%

0 NuderN of lifters

D 2 Release angle, deg

Net Present Value = f(throughput, grind, liner costs, maintenance costs,downtime costs, etc.)

Figure 8 Optimized design of mill lining systems with PBM/DEM Summary The history and principal application areas of the three major simulation approaches have been described, using ball mill circuits as the topical thread. Not surprisingly, in current practice no one approach provides all of the answers, and some combination of tools is frequently required. Moreover, it is unlikely that we’ll see a convergence in the foreseeable future, instead it is probable that the higher fidelity methods will be used to selectively enhance the predictive capability of the lower fidelity approaches. To conclude, the current best practice in ball mill design underscores the need for combinations of tools, specifically: 0 Bond’s model is still used for basic design and equipment selection 0 PBM’s are used to calculate material balances in the circuit and conduct sensitivity analyses to changes in selected operating and design variables 0 HFS methods are used to explore liner profiles, ball sizing, speed, etc. As is probably evident from the descriptions, the current design methodology is to start with the simplest simulations and work toward the most complex. Research work continues to improve the understanding and applicability of all the tools.

STATISTICAL (EMPIRCIAL) MODELING In the next three sections the authors have chosen to illustrate how various simulation approaches have been employed in the principal application areas for mineral processing, using “typical” case studies. The intent is to demonstrate the extent to which simulation technology has been assimilated into today’s work practices. This first section deals with the application of empirical methods. Design As one might expect, equipment design has a long history with empirical models/methods, and a good example was presented in the chronological introduction. The design of crushing and screening equipment is also largely based on empirical methods, and a screen example is presented for illustration.

389

The basic equation for determining the screen area required for a given application is:

Area =

T ABCDEFG,HI

(3)

where: Area = screen area in m2 T = screen undersize in feed (tph finer than hT) A = unit area capacity of reference material (tph/m2) B = oversize correction factor C = half-size factor D = deck location factor E = wet screening correction factor F = solids density correction factor G,= near-size correction factor H = relative open area correction I = a collection of additional manufacturer corrections (e.g. aperture geometry “slottedness”, grizzly bar, multi-flow deck, particle shape, surface moisture, etc. factors)

pi =1-0.f

Figure 9 Empirical screen modeling There are empirically deduced formulae for all of the correction factors (see Karra, 1979, and King, 1990). Figure 9a illustrates the application of this methodology to calculate the required screen area. It is clear from the symbology that there are some nuances (e.g. the calculation of the effective through fall aperture, hT), that the user must know about. Using the computed area (or substituting the appropriate value for a standard screen), the product sizing from this separation can be computed using a standard partition curve approach. In this case (Figure 9b) the expression for the corrected cut size (dsh) is expressed in terms of the through fall aperture and the correction factors. The sharpness of separation is taken to be a constant (5.9). The solid curves in the figure illustrate the estimated product sizing. Although this calculation appears rather trivial, lending itself to a spreadsheet implementation, the problem arises for closed circuit calculations, where the screen feed is dependent upon another unit (e.g. a crusher and/or another screen). Under these circumstances process design simulation software is employed, and at least one of these programs is described in subsequent chapters.

AnalysidOptimization The example selected to illustrate the application of empirical modeling to optimization was first reported by Bazin et al., 1994. It has subsequently gained broader acceptance in the industry, and a number of authors have reported on industrial applications (e.g. Sosa-Blanco et al., 1999 and

390

Edwards and Vien, 1999). The specific example presented here is based on the work of Edwards and Vien, 1999. Industrial grinding operations produce products for downstream operations such as flotation and leaching. The feed size distribution to the separation stage can be predicted from PBM modeling, or empirically as follows. The product size distribution can be characterized by one of a number of common empirical distribution models - e.g. the Rosin-Rammler model as shown in Figure 10a. Assuming self-similar distribution curves, it is possible to estimate the product size for finer or coarser grinds by assuming the distribution modulus (a)is constant, and varying the size modulus (p). Perhaps the most interesting finding by Bazin ef al., 1994, is that there is a good and often rather stable correlation between the cumulative fraction of metal (or mineral) passing a given size, and the corresponding solids fraction passing the same size. This is illustrated in Figure lob, where the raw data points are modeled empirically using a constrained quadratic (to ensure the curve passes through the point [lOO,lOO]). Both of the cumulative curves in Figure 10 can be discretized to give the mass fraction of solids and corresponding metal (or mineral) grade in a particular size fraction. 100

100

P

6 8 75 n

I-

cn

L E

Qmtrainad Qladatic

G

w

0

10

10

a)

100 Passing Size (microns)

1M30

0

b)

25 50 75 Cum. Solids %Pass.

100

Figure 10 (a) Fitting the particle size model; (b) Fitting the metal distribution model 100

90

c.

75

$

a 1

10

100

Passing Size (Microns)

70

40

1WJJ

b)

60 80 %Passing 150 Mesh

100

Figure 11(a) Metal recovery on size; (b) Estimated size - recovery relationship Experimental studies have shown that under essentially similar conditions, the recovery for a certain size fraction in leaching or rougher flotation processes is essentially constant, and independent of the feed size distribution. Thus, an experimental recovery on size curve, such as illustrated in Figure 1l a can be used to compute the overall recovery from the discretized data. By varying the feed size (p), one can use the empirical models to compute the corresponding overall recovery, and effectively simulate the size - recovery relationship for the ore under study, as shown in Figure 11b. This is particularly useful in optimizing the economic balance between throughput, grind and recovery. In the case of flotation, if the minerals are used instead of the metal, then it is also possible to infer concentrate grade.

391

Control Although one might not normally think of these applications as simulation, empirical models for characterizing process dynamics have been used rather extensively in the field of process control. For example, the simple first order plus time delay empirical model of a step response can provide: (a) in the single input - single output case, the process dynamics required for tuning a PID or Dahlin regulatory control loop (Vien et al., 2000); (b) in the multiple input - multiple output case, the empirical process model is required for implementing model predictive (i.e. simulation-based) control (e.g. Wen et al., 1991). In the same vein, empirical time series models can form the basis of control loop diagnostics (e.g. Perry et al., 2000) In the advanced control of mineral processing operations, it is more common that multivariate statistics or neural networks are used for process modeling. These can be used to simulate performance, off-line, say as part of an optimization strategy. For example, Twidle et al., 1986, describe an algorithm for optimizing reagent and level set points in a flotation circuit. In this case two multi-linear regression models, one predicting concentrate grade and the other metal recovery simulate the flotation process. Every half hour these models are “fit” to the last 24h of properly filtered (e.g. outliers, missing measurements, etc.) production data. This ensures they are adapted to the current process behavior. Using a grid search technique on the manipulated (independent) variables, it is possible to select a combination of set points that maximize an economic objective function. Marklund and Oja, 1996, have also reported on the use of Partial Least Squares to construct off-line process models, which were helpful in developing an understanding of AG mill optimization issues, and in demonstrating the utility of charge volume measurement. Neural networks (NN) have been widely discussed as candidates for advanced process control, either as a soft sensor, or as a process model for use in predictive control and/or optimization. Although one of the authors (BF) is aware of a number of such applications, there are few publications on industrial case studies. Simulation based exploratory work (e.g. Duarte et al., 2001) concludes that a properly designed NN control strategy is as effective as a number of other multivariable control strategies that have been successfully applied in industry. In a more conventional mineral processing application, the NN model is likely to be used in conjunction with an expert system. A common software structure is shown in Figure 12. In this illustration the NN could be used as a soft sensor, e.g. to predict the output from a critical sensor when that unit has failed. If the NN model accurately represents the process, then the optimizer can simulate the process response to changes in controlled variables to discover set points that should enhance process performance, which it would then “suggest” to the expert system. Using process data and a suitable scheme to decide when the NN model must be retrained, it is possible to adapt the model to changes in process performance characteristics, ensuring the predictive accuracy is maintained. Set points

tI Sensor Inputs

I

tI

I

Expert System

I *Estimated Variables

I Suggestedr Set points

Training Database

.

-+

I

v

Training Alg.

Figure 12 Advanced expert control using a neural network process model

392

In both of the illustrations above, the development of the empirical model (e.g. picking the right independent variables) can be quite challenging (e.g. Qin and Rajagopal, 1993), and when used in an adaptive scheme, these models can exhibit large prediction error if the process is not persistently excited.

POPULATION BALANCE MODELING Population Balance Models have a -40 year history in the mineral processing industry, so it is not surprising that they have seen application in essentially every area of interest. Once again, the relative simplicity, installed base, and economic importance of grinding has fueled the broadest development in this area, although there have been significant developments in flotation. Consequently, the examples in this section relate to some aspect of grinding or flotation. Design PBM’s emerged as a design tool for AG/SAG milling in the early 1990’s. Up to that point, pilot plant data and various customized versions of the Bond empirical approach had been applied. The overall development history is interesting, since PBM’s were initially used almost exclusively for analysis/optimization work. However, as the databases grew and model predictive capability was validated, PBM’s gradually became a de fact0 standard tool in the process design of AG/SAG mills. (More recently, refined empirical methods have been explored, e.g. Kosick er. al., 2001, but PBM remains the tool of choice.) Since the application of PBM’s in AG/SAG milling will be explored elsewhere in this volume, the authors have elected to illustrate this simulation application with a simple open circuit ball mill example. Consider the problem of grinding two or more materials of different hardness in the same mill. Although this is quite a common situation in milling circuits treating blended ore, it assumes great importance in iron ore or cement applications, where product chemistry is of particular concern. Herbst and Rajamani, 1982, have published on a methodology to use PBM for scale-up from laboratory batch to continuous milling. This technique is of special interest in this case, since it allows one to quantitatively consider the concurrent grinding of two materials, something empirical methods cannot accommodate. Very briefly, Equation (4) is the well-known phenomenological model for batch grinding. i-1

I d(Hp’) - -siHpi + C biisjHp

(4)

dt j=1 where: H = solids mass holdup in the mill pi = mass fraction of solids in size class “i” si = selection function or rate of breakage for size class “i” b, = fraction of material reporting to the i’th size class from a breakage event in the j’th size class On the basis of prior work (Herbst and Fuerstenau, 1980), it has been shown that for ball milling, one can estimate a specific selection function S: as:

sEI

=#I

where: P = power draw of the mill

393

1

.oo

0.01 0

a)

.a

1

0.0 1

2

3

4

5

6

GrindingTime, minutes

0.0

0.5

1.o

1.5

specltic Enegy Inpn, kwM

b)

2.0

I

Figure 13 S1 and SIEfor a ball mill operating under different conditions (Herbst and Fuerstenau, 1980) Loolung at the top size disappearance kinetics, Figure 13 illustrates the advantage of this transformation, since the SIEare seen to be independent of mill operating conditions, and can be assumed independent of mill size. On the basis of this methodology, batch grinding tests were performed on the two different materials of interest. The media size and slurry 9% solids values were similar to what was expected in the operating plant. From this data the S y and bij values were extracted for both materials. Using a multi-component ball mill model (in MinOOcad - Herbst and Pate, 2001), knowing the feed size distributions and mass flow of each component, it is possible to predict the performance of the continuous mill from the batch results. (A continuous mill transport model was assumed, based on prior plant test work.) Finally, the continuous mill performance under essentially the same operating conditions was measured, to validate the results. Figure 14 summarizes the data for this example: (a) Figure 14a illustrates the results of model calibration on the batch tests; (b) Figure 14b shows model predictions and actual performance for the full-scale mill. Clearly the predictive accuracy is within the normal expectations for design error, demonstrating the applicability of PBM's in process design. 1oc

100

.-

2 d g

90

-

m E

01)

80

Comp. No. 1

B

Comp. No. 2

8

'O

80 7a

60

0

0

60

50

'Pred No. 1 Meas No. 1 Pred No. 2 Maas No. 2

4c

50

10

a)

Full Scale Continuous

go

100 Size (microns)

1000

10

b)

100 Size, rnlcrons

1000

Figure 14 Batch scale calibration and plant scale prediction for concurrent grinding AnalysidOptimization PBM's have a very long and rich history in process analysis/optimization. Applications range through circuit re-engineering, ball size optimization, water addition optimization, tonnage - size relationships, grate/pebble-port design, cyclone orifice optimization, and SO on. It would be particularly easy to choose an example from any of these topics. However, an exciting new area of study is mine-to-mill optimization, and this effectively utilizes PBM in a new way. Mine-to-Mill exists on a number of levels ranging from enterprise simulation (to be discussed elsewhere in this volume) to exploring for the best (lowest cost) combination of comminution

394

processes (blasting - crushing - grinding). The example presented here focuses on this later topic, and is drawn from the work of Herbst and Blust, 2000. In this particular case, the plant is required to treat four distinct ore types (characterized by hardness), and the question was what blending strategy would yield the best performance, assuming the crushing and grinding circuits were controlled to maximize production, regardless of the feed. Figure 15 presents a schematic of the flowsheet from the mine face through to the grinding circuit product. The simulation program (Herbst and Pate, 2001) used in this study employs a multi-component approach and energy-based comminution models, where the latter make it possible to look at tradeoffs between finer fragmentation in blasting to changes in gyratory crusher settings and SAG mill operation. Using a case study approach, numerous alternatives were investigated. Table 2 presents a summary of the key findings. The results show best blending strategy would to be mix all ore types in the proportion they exist in the mine, and to run with a more open gyratory crusher setting (OSS). Evidently, altering the run of mine size by changing powder factor (PF) has little effect. From an operating perspective the important points of the study are to maximize blending prior to the crushing process, and to ensure there is sufficient coarse material in the SAG feed to provide maximum rates of breakage.

Figure 15 Mine-to-mill optimization flowsheet

395

Table 2 Summary of simulation results (from Herbst and Blust, 2000) Ore Feedrate

Total Energy

MTPH

KWhlMT

Option 1: Four ores crushdground separately 285 (PF = 0.177 kg/MT and OSS = 200 mm) Option 2: Four ores blended at crusher (PF = 0.177 kgA4T) OSS = 150 m m OSS = 175 mm OSS = 200 mm OSS = 225 mm

Option 3: Four ores blended at crusher (OSS = 200 mm) PF = .200 kg/MT PF= .177 kglMT PF = .lo0 kgA4T

18.7

287 312 326 334

18.6 17.1 16.4 16.0

326 326 326

16.5 16.4 16.3

Control Figure 12 presented the most common software architecture for the advanced control of mineral processing circuits. Figure 16 shows the most common industrial implementation of this structure, where the model-based piece of the controller is based on a PBM or phenomenological formulation. Since things change, these models are continuously adapted to the process using an advanced statistical technique known as the extended Kalman filter. The combination of the model and the filter provides better estimates of measured process variables, and it permits the calculation of unmeasured variables (i.e. the soft sensor), such as rates of breakage or rates of flotation, which provide some quantitative insight into the current state of the process. Moreover, since the models are based on the laws of physics and chemistry, they hold some significant advantages over their empirical counterparts, namely: (a) the PBM model is generally applicable over a much broader range of operating conditions; (b) the PBM model can accommodate non linearities (e.g. a SAG mill volume overload) without having to be “trained” on data including this condition; and (c) estimation errors are easier to detect as the PBM model parameters (e.g. selection function) have a physical significance. Set points

Sensor Inputs

Expert System (with fuzzy logic) Suggested Set points

Estimated Variables

Optimizer

PBM Model Kalman Filter

Figure 16 Model-based fuzzy expert control (MBEC)

396

The application discussed here is more fully described by Johansson er al., 1998. In this case the MBEC control strategy was commissioned on a rougher-scavenger flotation circuit including four parallel banks of cells, as well as a regrind and a conditioning step. The process was quite highly instrumented with all principal solids, slurry and reagent flows measured, as well as cell levels, air flows and the composition (on-stream XRF) of selected process streams. Johansson ef al. classified the PBM models as proprietary, but from the description the approach was to employ a rather comprehensive mechanistic description of the flotation process in each cell. (It is likely that the work of Bascur, 1982, provided a basis for PBM development.) The authors treated the ore as being composed of 8 different species of floatable and non-floatable valuable mineral, gangue, and assemblies. Given the complexity of the circuit (47 manipulated variables) and the number of species, this would have created a very large and complex model. Nevertheless, the application is reported to have successfully achieved the goal of maintaining the bulk rougher concentrate grade while maximizing the recovery of the valuable mineral species (e.g. a 0.5% Copper recovery improvement was mentioned in the paper). In addition to providing filtered and soft sensor estimates of process variables, the adapted PBM model was employed by the optimizer to explore the effects of various set point changes on overall performance. In this instance, all of the internal control loops (e.g. level control loops) are incorporated, and the entire process model is integrated to produce the steady state estimates of ultimate (disturbance free) circuit performance, given a combination of set points. A multivariable search algorithm was used to solve the for the best combination of set points, and these were passed to the fuzzy expert system for sanity checks and implementation. The benefits of using models that are faithful to the physics and chemistry of the process are underscored in this example.

HIGH FIDELITY SIMULATION High Fidelity Simulation (HFS) is the phrase used to describe the integrated collection of first principles modeling techniques that include Discrete Element Modeling (DEM), Computational Fluid Dynamics (CFD), Discrete Grain Breakage (DGB) and Finite Element Methods (FEM). DEM is the base methodology for HFS, and the one that has seen the greatest use to date. HFS, and in particular DEM has some advantages over PBM in terms of user acceptance, i.e.: (a) it is based largely on first principles requiring little or no calibration; and, (b) it has benefited from the wide scale embrace of modeling and simulation technologies fueled by the use of PBM. Although there are few publications on the applications of HFS in mineral processing, the authors are aware of applications in sampling, chutes, conveyors, gyratory and cone crushing, AGISAG, ball and stirred mills, screens, and even flotation froths. In an earlier section it was mentioned that DEM has thus far been used to solve problems that are large enough to provide meaningful results (lo5 to lo6 elements), yet small enough to be solved in a reasonable amount of time 10' to lo2 hours. The examples below help to underscore this observation, and have been selected to illustrate how DEM has truly extended simulation technology. Not surprisingly, there have been no control applications of HFS reported, although it has been used in instrument development (e.g. Dupont and Wen, 2001 ), in addition to providing a better understanding of mill charge phenomena. Design Even though the 2D discrete element algorithm simulates a thin slice of the SAG mill, it has far exceeded expectation. The projected power graft of plant scale mills has been extremely accurate (Dennis and Rajamani, 2001). As shown in figure 16A; the simulation of the two dimensional slice of the mill shows cascading and cataracting charge. These simulations suggested that besides increasing the space between rows of lifters, manipulation of the lifter face angle offered the best means of adjusting the trajectory of the charge. Using a hat-type lifter design of 48 rows (down from 72) and a face angle of 30°, the packing was relieved and throughput increased by approximately 30% for the hard ore types, achieving the designer's predicted average throughput rates. In addition, mill availability was enhanced, due to extended lifter life in the absence of liner

397

cracking and breakage due to high-energy ball impacts. Consequently, grinding media consumption has also been reduced by about 20% relative to original practice (Sherman and Rajamani, 1999). The power of DEM is very much illustrated by the Alumbrera SAG or Collahuasi (Villouta, 2001) mill examples, as a consequence, one is tempted to do full scale high fidelity simulations with 3D DEM, which is described in this paper.

Figure 16A. Two dimensional DEM simulator of 40ft SAG mill at 76 96 critical speed and 23% filling. One of the more interesting issues in AG/SAG milling these days is the efficiency of removal of material from the mill, i.e. the efficiency of the pulp lifters. On occasion this has proven to be a bottleneck, as the material that passes through the grates and ports cannot be effectively removed from the mill. Hart et al., 2001, have described such a situation for the Cadia mill, where radial pulp lifters had been installed to facilitate bi-directional mill operation. However, inefficient removal of pebble-sized material was observed to increase pulp lifter wear (recycle) and thought to inhibit mill performance. 3D DEM was used to look at the potential of moving from radial to curved lifters, and initial feedback seems to indicate that DEM predictions related to trends in wear and throughout have been validated. To illustrate the approach, a hypothetical example is presented to look simply at the solids removal capability of the two different pulp lifter designs, as illustrated in Figure 17. In this case the simulation was designed to classify particles as: (a) those that leave the mill - Production; (b) those that recycle to the pulp lifter (called backwash) - backflow; and, (c) those that reenter the mill through the gratedports - Recycle.

398

Figure 17 Two different pulp lifter geometries for DEM evaluation The results for the two geometries in Figure 17 are summarized in Table 1. The differences are quite significant, and it is fairly clear that DEM could be used to look at many different geometrical variants, mill speed effects, particle size effects, etc. Table 1 Particle removal efficiency for radial and curved pulp lifters (DEM Simulation Predictions)

(c) Recycle Just to reiterate, this contrived example is simply to illustrate the power of DEM to look at a class of problems that have heretofore been restricted to a long and sometimes costly experimental evaluation. AnalysidOptimization The example chosen to illustrate this application of HFS could just as easily qualify as an example of design, highlighting the multidimensionality of most industrial problems. Nevertheless, we are conditioned to categorize problems and this is considered an optimization analysis/opportunity for the purposes of this paper. For continuity with the design example above, the example chosen here is a preliminary investigation of the differences between unidirectional and bi-directional AG/SAG milling. In this particular case the DEM simulation includes liner wear models to explore the effect of wear on performance. Although there are a number of choices, the initial lifters were selected to have “standard” geometries, and the same mass, which translates to some initial lifter height differences. Figure 18 (from Qiu et aE., 2001)shows the initial lifter profiles, and the worn profiles at various stages of the life of the lining. In the case of bi-directional milling, the direction of rotation was changed at each stage.

399

ln

8

0.3 x C

9

0.3

x 0

0.2

E

c

I-

0.2

L

.--gI

0.1

2 = -I

0

e

-0.1

-0.1

0

a)

0.2

0.1

0.3

0.4

0.5

Lateral Distance

-".-0.1

0.6

o

0.1

0.2

0.3

0.4

0.5

n RI

Lateral Distance

b)

Figure 18 Simulated wear in a) unidirectional and b) bi-directional milling In this preliminary analysis several performance measures were computed corresponding to each stage of the liner life. Figure 19 shows two of these. In both cases, all measures are relative to the unidirectional mill with a new liner. Figure 19a provides the relative measure of cumulative impact energy dissipation (i.e. the sum over collisions of all energies). Since impact collisions are thought to play a controlling roll in breakage, this study indicates that bi-directional milling should provide better throughput at the same grind, or better grinds at the same throughput. The second chart in Figure 19b is a (crude) measure of comminution efficiency, here taken as the ratio of total impact energy dissipation to mill power draw, showing that bi-directional milling provides benefits, until near the end of the liner life. 1.08

a)

1.10

Black=Wir. Grey =lhdir.

1

2

3 4 Stage of Life

Black = Bidir. Grey = Uxlir.

5

1

2

3 4 Stage of Life

5

Figure 19 Performance comparisons for unidirectional and bi-directional milling

CONCLUSIONS The intent of this paper was to provide an overview of the extent to which modeling and simulation have become an integral part of our work practices in equipmentkircuit: (a) design; (b) analysis/optimization;and, (c) control. Simulation terminology has entered our jargon, and model parameters have become an effective means to communicate quantitative information. Succinctly, the impact of modeling and simulation is quite apparent in the way we do business, as operators are increasingly loohng for more scientific (compelling) approaches to support assertions made with respect to design or operational change. Some companies have even discussed the merit of requiring that Capex/Opex requests be accompanied by supporting simulation. Empirical simulation methods are still common, particularly in equipment design, but more and more advanced statistical methods are used in analysis/optimizationand in control, to develop simulations of processes that are not amenable to mathematical or heuristic approaches. PBM has a very long history of evolution in mineral processing applications, and has arguably been the catalyst for the development of a set of methodologies for the design and execution of field tests, and the analysis of experimental data, which inevitably lead to process improvement. HFS is a relative newcomer, but is making fast inroads as computer speeds increase and faster and more efficient codes are developed, although acceptance is driven mainly by it's unique capabilities. Although we choose to distinguish between problems relating to design, analysis/optimization and control, almost all such problems have multiple dimensions. Moreover the simulation

400

techniques we apply are generally a judicious combination of different methods to get the best answer, with the highest confidence, and at the lowest cost. On this note, there is no indication of an imminent convergence of these simulation technologies. Rather, one can predict with some confidence that there will be a rapid growth of HFS applications, and consequent improvements in both empirical and PBM methods.

REFERENCES Austin, L.G., Klimpel, R.R., Luckie, P.T., 1984, Process Engineering of Size Reduction: Ball Milling, SME, pg. 50 Bascur, O., 1982, Modeling and Computer Control of A Flotation Cell, PhD Thesis, University of Utah Bazin, C., Proulx, M., Chapleau, C., Lavoie, G., 2000, Evaluation of the Quebec Cartier Mining Pelletizing Plant Grinding Mill Efficiency as a Function of Mill Speed, Proc. 32ndAGM CMP, CIM, Ottawa, pp. 87 - 106 Bazin, C., Grant, R., Cooper, M., Tessier, R., 1994, A Method to Predict Metallurgical Performances as a Function of Fineness of Grind, Minerals Engng., Vol. 7, No. 10, pp. 1243 1251 Bond, F., 1952, The Third Theory of Comminution, Trans AIME, Mining Engineering, May, pp. 484 - 494 Buchholtz, V., Freund, J., Poschel, T., 2000, Molecular Dynamics of Ball Mills, Bulk Solids Handling, Vol. 20, No. 2, pp. 159 - 171 Broussaud, A., Guyot, O., McKay, J., Hope, R., 2001, Advanced Control of a SAG and FAG Mills With Comprehensive or Limited Instrumentation, in Proc. Intl. Conf. AG/SAG Grinding Technology, ed. Barratt, Allan and Mular, Pacific Advertising Printing and Graphics, Vancouver, Vol. 11, pp. 358-372. Bwalya, M., Moys, M., Hinde, A., 2000, The Use of the Discrete Element Method and Fracture Mechanics to Improve Grinding Rate Prediction, Minerals Engng., Vol. 14, No. 6, pp. 565 573 Cleary, P., 1998, Predicting Charge Motion, Power Draw, Segregation and Wear in Ball Mills Using Discrete Element Methods, Minerals Engng., Vol. 11, No. 1 1 , pp. 1061 - 1080 Dennis, M. and Rajamani, R.K., 2001, Evolution of the Perfect Simulator, Roc. Intl. Conf. AG and SAG Grinding Technology, Pacific Advertising Graphics and Printing, Vancouver, pp. IV.24 - IV.33 Duarte, M., Suarez, A., Bassi, D., 2001, Control of Grinding Plants Using Predictive Multivariable Neural Control, Powder Technology, Vol. 115, pp. 193 - 206 Dupont J., Vien A., 2001, Continuous SAG Volumetric Charge Measurement. Proc. 33* Ann. Can. Min. Proc. Mtg., ed. Smith, CIM, pp. 51-68 Edwards, R., Vien, A., 1999, Application of a Model-Based Size-Recovery Methodolgy, in Process Control and Optimization in Minerals, Metals and Materials Processing, ed. Hodouin, Bazin and Desbiens, MetSoc CIM, pp. 147 - 159 Ford, M.A., King, R.P., 1984, The Simulation of Ore-Dressing Plants, Int. J. Min. Proc., Vol. 12, pp. 285 - 304 Hart , S., Valery, W., Clements B., Reed, M. Song, M., Dunne, R., 2001, Optimisation of the Cadia Hill SAG Mill Circuit, in Proc. Intl Conf. AG/SAG Grinding Technology, ed. Barratt, Allan and Mular, Pacific Adverting Graphics and Printing, Vancouver, pp. 617 - 631 Herbst, J., Fuerstenau, D., 1973, Use of Specific Power Information for Ball Mill Simulation, SME Tranbsactions, Herbst, J., Rajamani, R., 1982, Developing a Simulator for Ball Mill Scale-Up - A Case Study, in Design and Installation of Comminution Circuits, ed. Mular and Jergensen, SME, pp. 325-342 Herbst, J., Blust, S., 2000, Video Sampling for Mine-to-Mill Performance Evaluation: Model Calibration and Simulation, in Control 2000, ed. Herbst, SME, pp. 157-166 Herbst, J., Pate, W., 2001, Dynamic Modeling and Simulation of SAG/AG Circuits with MinOOcad: Off Line and On Line Applications, in Proc. Intl. Conf. AG/SAG Grinding

40 1

Technology, ed. Barratt, Allan and Mular, Pacific Advertising Printing and Graphics, Vancouver, Vol. IV, pp. 58-70 Johansson, B., Bergmark, B., Guyot, O., BouchC, C., Broussaud, A., 1998, Model-Based Control of Aitik Bulk Flotation, SME Preprint 98-173 Karra, V.K., 1979, Development of a Model for Predicting the Screening Performance of a Vibrating Screen, CIM Bull., April, pp. 167 - 171 King. R.P., 1990, Simulation - The Modern Cost-Effective Way to Solve Crusher Circuit Processing Problems, Int. J. Min. Proc., Vol. 29, pp. 240 - 265 Kosick, G., Dobby, G., Bennett C., 2001, CEET - Comminution Economic Evaluation Tool for Comminution Circuit Design and Production Planning, SME Preprint 01-152, Denver Laguitton, D., 1974, On-Line Software Demonstration for Coal and Mineral Processors, Proc. Ann. Mtg. Can. Min. Proc., Ottawa, pp. 151-168 Lynch, A.J., Whiten, W.J., Draper N., 1967, Developing the Optimum Performance of a MultiStage Grinding Circuit, Trans IMM, Section C, pp. 169 - 182 Lynch. A.J., Narayanan, S., 1986, Simulation - The Design Tool for the Future, in Mineral Processing at the Crossroads: Problems and Prospects, ed. Will and Barley, Martinus Nijhoff Publishers, pp. 89 - 116 Marklund, U., Oja, J., 1996, Optimisation of an Autogenous Grinding Circuit Through Mill Filling Measurement and Multivariate Statistical Analysis, in Proc. Intl. Conf. AG/SAG Grinding Technology, ed. Mular, Barratt and Knight, Pacific Adverting Graphics and Printing, Vancouver, Vol. I1 pp. 617 - 631 McKen, A., Raabe, H., Mosher, J., 2001, Application of Operating Work Indices to Evaluate Individual Sections in Autogenous - Semi Autogenous/Ball Mill Circuits, in Proc. Intl. Conf. AG/SAG Grinding Technology, ed. Barratt, Allan and Mular, Pacific Advertising Printing and Graphics, Vancouver, Vol. 111, pp. 151-164 McKee, D.J., Wiseman, D.M., 1991, General Plant Simulators for Metallurgical Plant Performance, in Evaluation and Optimization of Metallurgical Performance, ed. Malhotra, Klimpel and Mular. SME, pp. 219 - 229 Mishra, B.K., Rajamani R.K., 1992, Analysis of Media Motion in Industrial Ball Mills, in Comminution - Theory and Practice, ed. S.K.Kawatra, SME, pp.427 - 440 Mular, A., Herbst, J., 1980, Digital Simulation: An Aid for Mineral Processing Plant Design, in Mineral Processing Plant Design - 2"dEdition, ed. Mular and Bhappu, SME, pp.306 - 338 Napier-Munn, T., Morrel, S., Morrison, R., Kojovic, T., 1996, Mineral Comminution Circuits: Their Optimisation and Control, JKMRC, pp. 210 - 212 Nordell, L., Potatpov, A., Herbst, J., 2001, Comminution Simulation Using Discrete Element Method (DEM) Approach - From Single Particle Breakage to Full-Scale SAG Mill Operation, in Proc. Intl. Conf. AG/SAG Grinding Technology, Pacific Adverting Graphics and Printing, Vancouver, pp. 235 - 251 Rajamani, R.K. and Mishra, B.K., 2001, Three Dimensional Simulation of Charge Motion in Plants Size SAG Mills, Proc. Intl. Conf. AG and SAG Grinding Technology, Pacific Advertising Graphics and Printing, Vancouver, pp. IV.48 - IV.57 Perry R., Supomo A., Mular M., Neale A., 2000, Monitoring Control Loop Health at P.T. Freeport, in Control 2000, ed. Herbst, SME, pp. 71-81 Qui, X., Potapov, A., Song, M., Nordell, L., 2001, Prediction of Wear of Mill Lifters Using Discrete Element Methods, in Proc. Intl. Conf. AG/SAG Grinding Technology, ed. Barratt, Allan and Mular, Pacific Advertising Printing and Graphics, Vancouver, Vol. IV, pp. 260 27 1 Rowland, C., 1982, Selection of Rod Mills, Ball Mills, Pebble Mills and Regrind Mills, in Design and Installation of Comminution Circuits, ed. Mular and Jergensen, SME, pp. 393 - 438 Qin, J., Rajagopal B., 1993, Combining Statistics and Expert Systems with Neural Networks for Empirical Process Modeling, Proc. ISA Mtg., Paper 93-41 1, pp. 171 1 - 1720

402

Sosa-Blanco, C., Hodouin D., Bazin, C., Lara-Valenzueala, C., Salazar, J., 1999, Integrated Simulation of Grinding and Flotation - Application to a Lead-Silver Ore, Minerals Engng., Vol. 12, NO. 8, pp. 949 - 967 Twidle, T., Engelbrecht P., Koel, J., 1986, Optimizing Control of Lead Flotation at Black Mountain, Proc. 15” IMPC, Vol. 3, pp. 189 - 198 Vien, A., Fragomeni, D., Larsen, C.R., Fisher, D.G., 1991, MOCCA: A Grinding Circuit Control Application, SME Ann. Mtg., Denver, Colorado, February Vien A., Edwards R., Perry. R., Flintoff B., 2000, Making Regulatory Control a Priority, in Control 2000, ed. Herbst, SME, pp. 59-70 Villouta, R.M., 2001, Collahuasi: After Two Years of Operation”, Proc. Intl. Conf. AG and SAG Grinding Technology, Pacific Advertising Graphics and Printing, Vancouver, pp. 1.31 - 1.42

403

BRUNO: Metso Minerals' Crushing Plant Simulator Dean M. Kaja'

ABSTRACT Manual crushing plant calculations often required hours to produce limited information about one possible solution to a crushing problem. With the new tools like the BRUNO Crushing Plant Simulator, it is possible to quickly run through many alternatives to optimize the crushmg plant. In addition, more mformation is available about each of the alternatives, so that a better understanding of the advantages of each alternative can be obtained.

INTRODUCTION Since 1994, Metso Minerals, formerly known as Nordberg has used a simulation program called Bruno to facilitate the equipment selection process. This program was developed as an object oriented flow sheet simulator, where the flow sheet components are selected from a menu bar and connected with lines that define the relationshps between the components. By graphically building the circuit, it is mathematically defined. Parameters for each component can be entered in data screens that open by clickmg on the component. The circuit is solved by clicking on the solve button and each of the flow paths can be investigated for detailed gradation information. Bruno is used by Metso sales and application engineers to provide our customers with the best estimates possible of what equipment is required to produce a given size product at the required tonnage rate. It is also used to help troubleshoot existing operating crushing plants, to compare actual production with the theoretical projection and thus to indicate areas where crusher or screen efficiencies may be low, and thus where to concentrate further field analysis of the plant to improve the plant performance. The name Bruno comes from the founder of the Nordberg Manufacturing Company, Bruno Nordberg. Bruno was born in Finland and immigrated to the United States in 1879. He settled in Milwaukee the following year working as a draftsman and designer of steam engines. In 1886 he founded the Bruno V. Nordberg Company which eventually became the Nordberg Manufacturing Company and today is known as Metso Minerals, a division of Metso Oy of Helsinki Finland.

THE BRUNO SIMULATION PROGRAM The program is based in part upon a previous program called SIMU, which was a DOS based mass balance program that kept track of the tonnage rate of each size fraction in the various circuit flows. It combined flows that were jointed, divided fractions of a given stream into separate flows when processed by a screen, and transformed flows that went through a crusher. The program gave percentage loading values for crushers based the calculated flow through the crusher compared to the book tonnage for the selected setting, corrected for material density. The program adjusted the gradations of the flows from the screens based on screen efficiency calculated by a screen model.

' Metso Minerals, Milwaukee, Wisconsin 404

The major advance of Bruno over SIMU was the use of a graphical environment to define the circuit components and their relationships. A "pallet" is provided which provides the most common circuit components. By clickmg on the icon of a crusher, the crusher can be positioned in the circuit and defined using the data window. The same can be done for a screen, stockpile, feeder or other circuit component. The pallet contains, feed material, plant feeders and grizzley feeders, primary gyratory crushers, jaw crushers , cone crushers, impact crushers, Mycrusher (where you can enter the characteristics of a non-Metso crusher from the product bulletin or enter the data for a particular crusher application using field sample data), screens for material classification, silos and stockpiles. The starting point for a circuit definition is to use the template that comes up when you open the program. The template has a feed material object already located on the desktop, which you can open and define. The specific gravity and the identifying name of the feed material can be entered. If the feed gradation is known, it can be entered using the MyFraction selection, by entering the screen mesh sizes known along with the corresponding cumulative percent passing for each mesh size. Otherwise, one can select a representative gradation based on the choices given, which are based on field application experience gained over years of field sampling. The plant description and info lines can be filled in to define the specific plant and which alternative is being represented. The template includes the name of the person running the simulation and the date. Next one selects either a classification machine such as a grizzley feeder or a screen or selects a crusher such as a primary gyratory or jaw crusher. By clickmg on the icon in the pallet and then clicking on the location on the desk-top, you place the object. Double clicking on the object opens the data screen to define the model, size, setting and capacity factor for the feeder, screen or crusher. After placing the crushers and screens desired onto the desktop, they are connected using the connect mode. By clicking onto the output note of a first object and the input node of a second object, a line connecting the two is created along with the mathematical model definition for the particular flow path. After defining and connecting the all of the equipment, the circuit simulation can be run by clicking on the calculator icon in the toolbar. After viewing the result, corrections and adjustments are made to optimize the circuit to achieve the desired result.

EXAMPLE The following example will demonstrate the simulator's capability. The example is a gold ore crushing circuit with a grizzly feeder, a jaw crusher, an intermediate stockpile, a screen and a cone crusher. See Figure 1. The plant has a required production rate of 300 tph (330 stph) with a product size of 100% passing 32 mm (1.25"). Using an availability factor of 75% the peak tonnage rate for the circuit should be 400 tph (440 stph). The blasted ore is represented by the feed material triangle, which indicates that the top size is 900 mm (36"). At this point, we have used a "typical" gradation curve for 900 mm ore. See Figure 2.

405

If a specific gradation curve is known, it can be specified in the gradation definition using MyFraction and providing a name for the gradation or ore type. The percent passing each selected mesh size is entered to form the definition of the gradation curve. See Figure 3. The jaw crusher can be selected using the jaw crusher data screen. The crusher model and size as well as the crusher setting and capacity factor can be entered. See Figure 4. In this case, we chose a C140 jaw crusher set at 130 mm (5.1 in.) and a capacity factor of 1.0 which means that we have one crusher and the material is considered to be of average crushability. The Jaw Crusher Data Display shows the limits built into the specific size crusher selected. See Figure 5. After the jaw crusher, a surge pile and feeder were inserted to de-couple the primary and secondary sections of the plant. This allows flexibility in operating hours and allows for maintenance to be done on one section of the plant without disrupting the operation of the other section. In this case, the primary section could be tied to the mining schedule, while the secondary section of the plant could be tied to the down-stream requirements. It is often beneficial to oversize the primary section to allow it to keep the intermediate stockpile filled for continuous running of the secondary section. This avoids disruptions in the down-stream plant operations. The screen can be defined using the Screen Data Screen. See Figure 6. In the example we chose a double deck calc screen, meaning the computer will calculate the area required to perform the screening at the selected mesh sizes and efficiencies. The top deck was selected as 75 mm steel plate with an efficiency of 91% and the bottom deck was selected as 32 mm wire mesh with an efficiency of 95%. And the cone crusher can be defined using the Cone Crusher Data Screen. See Figure 7. In this case we selected an HP500 crusher at a setting of 32 mm (1.25 in.). Once the icons are all defined, the flow paths are directed using the connect mode. Finally, the calculate button is pressed to calculate the flow balance of the selected circuit. If the tonnage for the plant feed material is entered as 0 tph, then the program will calculate the maximum tonnage rate that can occur without exceeding 100% loading on any crusher in the circuit. See Figure 8. In this case, we Variations in the circuit design can be quickly compared to optimize the equipment selections. The crusher settings can be varied to see the effect on flow rates and gradations in the circuit and in the final product of the circuit. It is useful for determining recirculating load tonnage in closed circuit applications. The value for the tonnage at each flow path is displayed along with the percentage of book capacity for each of the crushers. The information about screens will show the % efficiency for each deck unless the Calc Screen selection is made, in which case, the screen model will force the desired screen efficiency, and calculate the required area to accomplish this efficiency, The flow paths can be queried to check the gradation. It can be done by either specifying a mesh size to find out the percentage passing or by specifying a percent passing and finding out what mesh size would correspond to that percentage passing. See figure 9. In h s case we wanted to know the 80% passing size in flow to the product stockpile. The result was a P80 of 25.2 mm (.99 in.).

406

The program is designed to provide customer-ready reports in an easy manner. The flow diagram is one view. The flow diagram that you just created is printed along with the plant name and the information about the author of the flow diagram and date. Each flow path tonnage is listed along with the percentage loading for each of the crushers. It also shows the screen mesh sizes and efficiencies. A second view is to use the text report format. Here detailed information is provided for each piece of equipment. This information includes the settings and percentage loading calculated. Selected flow path gradations can also be displayed with user defined screen mesh sizes. See figures 10,ll and 12.

Finally, a graph of the flow path gradations can be made using the graph report function. See Figure 13. Here the selected flow paths are graphed on a semi-log graph showing the cumulative percentage passing versus the screen mesh size.

SUMMARY The Bruno plant simulator is a very useful tool for calculating the capacity and gradation of a crushing and screening circuit. It provides the opportunity to vary circuit components to optimize the performance of a circuit and to optimize the equipment selection, resulting in a more cost effective crushing solution. '

REFERENCES Nordberg 1886-1986, The First 100 Years; Internal Nordberg Publication

407

0 P CD

Figure 1 Two stage jaw-cone gold ore crushing circuit

Figure 2 Feed material data screen

P

0

Figure 3 MyFraction definition screen

Figure 4 Jaw crusher selection screen

Figure 5 Jaw crusher data screen

zw

Figure 6 Screen data screen

P P

Figure 7 Cone crusher data screen

Figure 8 Results of circuit calculation

Figure 9 Flow path query screen

Figure 10 Report of circuit data

P W

Figure 11 Report of circuit data

P W

Figure 12 Report of circuit data, showing flow gradations

0 1->2

02/0->3

03->4

+ll->7

A7->6

x6/2-> 10

100 90 80

P 0 N

20 10 0

0 . 0 0 1 B E 1 3 c I ~ ~ B c 1 ~ 5 ~ 3 1 ~ . 0 6 2 0.25 9 0 . 1 20.5 5 1 2 4 8 Screen hole size [mm]

16

Figure 13 Graph of circuit flow path gradations

32

64

128 256 512 1024 2048

PlantDesigner@:A Crushing and Screening Modeling Tool Per Hedvall', Martin Nordin'

ABSTRACT PlantDesigner" is a PC-based program for flow sheet design in crushing and screening applications. Using a flexible and friendly graphical user interface, and Sandvik AB's extensive unit operation model library, a user can quickly configure, test and document a variety of potential flow sheets. This paper describes the structure of the Planmesigner@program in detail. Then drawing on recent industrial experience, three PlantDesigner" predictions are compared with the actual results subsequently obtained in their respective real-life applications.

PREAMBLE The Sandvik Rock Processing simulation program, PlantDesignefi for crushing and screening plants enables the user to construct process flow sheets with simultaneous flow calculations quickly. The program today runs on PC in Windows environment with a graphical interface where one can easily modify flow sheets to test new designs, conduct sensitivity analyses, and perform process diagnostics. INTRODUCTION Successful application engineering for crushing and screening plants is based on knowing the operating objectives (and the degree of flexibility) required of the crushing and screening plant, having a good understanding of the characteristics of the material to be processed and having the freedom to choose the right machines. Simulation is an extremely important tool to demonstrate how the crushing and screening plant will work in order to produce the required products and to calculate the load for each of the chosen machines. The final step is to show the material flow, presented on a flow sheet for the calculated crushing and screening plant. Since 1979, the Sandvik team has been obtaining and developing precise tools in a computer environment for quick and simple evaluation of alternative solutions to meet crushing and screening plant inquiries, for flow sheets, layouts and quotations. In the future we intend to continue designing, developing and using new tools to provide excellent application engineering. In accordance with this philosophy we have created our PlantDesignerB program, intended for the simultaneous generation of a flow sheet drawing and the calculation of the material flow in the simulated plant. The PlantDesignerB software is a PC program in Windows 98ME and/or Windows 2000KP for the design of flow sheets, simulation of processes and calculation of mass balances for crushing and screening plants. We have developed the PlantDesignerB simulation program both to generate flow sheets and to be able to simulate and calculate the performance of crushing and screening plants. 0

The flow sheet is created by selecting icons from a database. The icons are dragged and dropped into the drawing as the proposed process is designed. The characteristics of each of the units in the process are selected in dialog boxes, which appear when the icon is double-clicked in the flow sheet.

'Per Hedvall is General Manager of Process Technologyat Sandvik Rock Processing AB. Martin Nordin is an IT Specialist and Principal PlantDesigner Program Developer at Sandvik Rock Processing. 42 1

Flow calculations can be made as soon as all of the connections between the symbols have been made in the flow sheet. Our PlantDesignerB simulation program is a program in which a better input gives a much better answer. A very simple array of building blocks allows the user to build models rapidly. Boundless and unlimited hierarchical composition of the model to make even complex systems easy to build and understand. Full connectivity and interactivity with other programs and platforms through CopyPaste, import/export, text (ASCII) files, or DLLs and DDE (dynamic data exchange) and OLE. To select the right machines, the right equipment or the right crushing & screening plant, the following parameters need to be taken into consideration: The material fed to the plant The desired capacity and products The machines and equipment available The appropriate technology and engineering Specificationsand requirements Investment and running costs One’s own requirements One’s own experience One’s own innovation and creativity

Figure 1: How PlantDesignera interacts with starting and final properties.

422

THE PROGRAM The targets for our PlantDesigner@program are applications of Crushing and Screening Plants at our application centers in North America, Appleton WI, in Europe at Clichy, France and Svedala, Sweden. Moreover, our most skilled sales personnel also use the program.

PlantDesigner uses Imagine That Znc’s program, Extendmas a simulation platform and engine and we have added on our modules for our machines and equipment. USING THE RIGHT MODELING TOOL

A distinction should be made between what results our program will provide versus what results we could obtain from a spreadsheet. With PlantDesigner, the user creates a block diagram of a process where each block describes one part of the process. In a spreadsheet, the user lays out numbers in a two-dimensional tabular format; in Extend, the user lays out a process in a two-dimensional or threedimensional drawing environment. PlantDesigner allows the user to create models of real- world crushing plant processes that are too complex to be easily represented in a spreadsheet. This is done very quickly, because PlantDesigner comes with all the blocks needed to create most simulations. These blocks act like macros, so the user builds models without having to create algorithms or type equations. All the many blocks used in the simulation can be assembled into a simple model, thus describing a very simple process.

THE ModLm PROGRAMMING LANGUAGE This programming language is a structured subset of C with object-oriented extensions. The user types in the script, which describes the parameters, initializations, and behavior of the block. Extend then checks, compiles, and saves the block when the block is closed. If there are any syntax errors in the script, the user is alerted to correct them, A script can be as simple as a single equation, or as sophisticated as an entire program. Like C programs, a ModL script starts with type declarations and constant definitions. Because you declare these at the beginning of the script, before any message handlers or user-defined functions, they are considered static or permanent. They are therefore valid throughout the block’s script unless overridden by a local variable declaration.

423

After the type declarations, there are procedure and function definitions, and many message handlers. The procedures, functions, and message handlers are just definitions: they need to be called in order to be executed. Message handlers interpret messages that come from the simulation or from the user. Extend runs the message handler whenever one of the messages is passed to the block.

424

Creating or Modifying Blocks: Blocks are created using two windows: a dialog window and a structure window. The dialog window is where a block’s dialog box is created. The structure window is where a blocks icon and on-line help is created, and where the behavior of the block is programmed.

To access an existing block’s dialog and structure windows, the block is selected and the “Open Block Structure” command in the define menu is used (or ALT-double-click the block). You may also choose Open Library Window at the top of a library menu, or ALT-double-click the block in the library window. To build a new block, select the Build New Block command in the Define menu. When editing a blocks dialog window, clicking once on a dialog item selects it for re-sizing or moving. Double-clicking on the item opens it for editing its type and its script variable name. Variable names can be used in the script to input and output values from and to the dialog box. Pictures for a blocks icon can be pasted into the icon pane within the Structure window,

425

When the program starts two different windows appear. One window for makingldrawing the flow sheet and one window with library of the machines. The flow sheet is created by dragging and dropping the machineslequipment from the library onto the flow sheet.

Drag & drop

Figure 4 demonstrates the graphical method of establishing the flow sheet, as icons are simply connected by attaching the line to the input or output of one unit, and attaching this to the output or input of another unit.

Input

output Figure 4. Inputs and output.

426

To change the properties for the machinedequipment you double click on the icons in the flow sheet and the machine dialog opens.

Machine characteristics set in dialog

Figure 5: The dialog box for changing the properties of a Hydrocone crusher All dialog items have dialog names (for labels, the dialog name is optional). These dialog names are variable names and message names used by the script to interact with the dialog, both to read the data and set the data visible to the user. Some dialog items have titles or text that appear in the dialog.

427

Editable text and parameters can be made “Display only” in the block dialog editor. This allows the user to view the values, but not change them. Because deleting dialog items can misalign the data in existing models, dialog items can be hidden if the developer wants to “remove” them. This maintains data and variable names to interact with the script, but hides them in the dialog. The dialog items are: Buttons A click from the user sends the message name to the script. The title can be changed by setting its name to a string. The OK button closes and updates the block’s data from the dialog box. Its message can be aborted with the ABORT statement to prevent closing, if the data is not entered properly. Cancel just closes the dialog box without updating the block’s data. Its message cannot be aborted.

Feed rate:

Buttons

Feed malefiai: Bulkdensity (lbsn13)

~oisture content(%)

Abrasion index

Feed dislribueon

comments

428

CheckBoxes A click by the user toggles the appearance and the value between TRUE and FALSE. It can be used as a message name and variable and has the value TRUE or FALSE (1 or 0). It can be assigned a value and will update its appearance.

Capacity Crusher Molsture content C7 Prrnrlrrminrn mnshire cnntenl

Data Table A two dimensional table of parameters (similar to a spreadsheet). The titles of the columns are separated by semicolons. Titles default to numbers.

Capackf Crusher Moisture content R Predeterminedmolsfure content NeWmOlStUrecontent

Data Table

429

Editable Text A textedit item that can be assigned a string value. This can be read or written to as a string by the script. Editable text items that are two lines or greater create a scroll bar on the item. It can be made “Display only” in the block dialog editor.

Feed rate

STPH

Feed material Bulkdensity (IbSrn3)

1P.fnrrlnrf-w

Moisture content (%)

Abrasion index

Feed distribution.

Parameters A textedit item that can be used as a Real variable and can be assigned a value. A click by the user also sends the message name to the script message handler. It can be made “Display only” in the block dialog editor.

Feed rate. Feed material: Bulkdensit$ (IbSrn3)

Workindex

Moisture

Parameters

content (%)

Feed distribution.

comments

430

Popup Menus A popup menu that allows selection of an item from a list. This is a good replacement for radio buttons where space is at a premium. The list is dynamic and can be changed with a function call.

Feed rate

STPH

Feed matenal. Bulkdensity (ibM3)

Warkindeu

Moisture content (96)

Abrasion index

Popup Menus Cornrnents

Radio Buttons A click by the user affects the appearance and the value of all the settings in its radio group. It can be used as a message name and variable and has the value TRUE or FALSE (1 or 0). It can be assigned a value and will update its group's appearance.

Deck 2

Separation size Screening surface opening. Particle shape: Open area ERlciency:

Radio Buttons

431

Slider A control that resembles a slider on a stereo. You can drag the knob to change a value, or your block can move the knob to show a value.

Feed rate

I STPH Pariicie size dlstrlb

Feed material Bulk densily

(lbsm3)

Work index

Sleve (In) Moisture contenl(%)

Passlng through (96)

Abrasion Index

Feed distribution’

comments

Static Text (Labels) Used as titles, but can also be assigned strings or numbers. The value is always the string entered in the dialog box, so even if you assign a different string to it, its value on the right side of an assignment statement will never change. (i.e.: If stat is “The value is “, the equation: stat = stat+” $”+5; will show up as: “The value is $5” in the dialog box.

432

When the flow is finalized, all icons are connected with flow lines. We can make the material flow calculation for the plant, by pressing the Ctrl and R keys at the same time.

The calculation report appears on the screen. Here you can identify the machinedequipment in the plant, capacities in each stream of flow, loads and setting of the machinedequipment.

Figure 6: Calculation report

433

On the top of the report you see the customer’s name or the project name and the properties for the feed material used in the calculation. There are several other functions in PlantDesignerB to view the results such as particle size distribution diagrams and gradation tables that you can monitor directly on the screen.

Cap: 100 MTPH Passing Sieve through (mm) P90 P50 P20

70 9 2

Topsize: 600 mm

Figure 7: Gradation table

PLANTDESIGNERG3 SIMULATION VS. REALITY The following examples provide quantitative comparisons between simulation and actual plant operation. Example no. 1: Primary station. Feed material Location: Impact Work Index Wi Bulk Density

Blasted Diorite Asia 27 1.63 tJm3

434

Flow sheet:

Reciprocating Feeder

Primary Screen

Jaw Crusher JM1312 CSS 125 m m

Figure 8. Primary station flow sheet Capacities.

250 270

Average PD estimate

Figure 9: Jaw Crusher capacities and the predicted capacity by PlantDesigner

435

Particle size distribution.

P 90 P

70

Figure 10: Jaw Crusher product curves and the predicted particle size distribution by PlantDesigner Remarks. The variation in capacities from ajaw crusher depends to a very high degree on the changes in the size distribution in the feed. The product gradations are very close to those calculated by Planmesigner@'.

436

Example no. 2: Tertiary station. Feed material Impact Work Index Wi Bulk Density Location: Flow sheet:

Crushed Iron ore 8 2.8 t/m3 Europe

Hflrocone Crusher H-4800-F ASR+

Figure 11: Hydrocone crusher flow sheet

Capacities.

356 323 407 362 350

Test no. 1 Test no. 2 Test no 3 Average PD estimate

____

Figure 12: Hydrocone crusher capacities

437

Particle size distribution.

100

70

60 50 Feed PD Test 1

40

Test 2

30

Test 3

20 10 0 1

10

Size (mm)

Figure 13: Product curves from Hydrocone crusher

Remarks. The average capacity (362 MTPH) from the tests above is very close to the PlantDesigner@estimate of 350 MTPH. The PlantDesigner@calculated product distribution lies well within the family of product curves achieved by the tests.

438

Example no. 3: Secondary crushing station Feed material Impact Work Index Wi Bulk Density Location: Flow sheet:

Primary Crushed Copper Ore 50 - 200 mm 14 1,6 t/m3 America

Hy drocone H6800 ASR

Screen Hole size 45 mm Rubber element

Figure 14: Flow sheet

Capacity:

438 450

Sample PD estimate Figure 15: Hydrocone crusher capacity

439

Particle size distribution: 100

90

80 70

30

20 10 '

0 10

20

30

40

50

80

70

80

Square Hole slze (mm)

Figure 16: The product size distribution from the Hydrocone crusher 6800 Remarks. This single test with copper ore shows that the throughput was less than the estimated from [email protected] the product was a little finer than calculated. Comparing both capacity and product at - 45 mm shows PlantDesigner@: 450 * 0.60 = 270 MTPH finished amount Test: 438 * 0.64 = 281 MTPH finished

440

90 100

WHAT’S NEXT? The Figure below presents the development and design highway for the short term.

Figure 17: The ongoing timetable for PlantDesigner@fall 2002.

CONCLUSIONS All process simulation programs try to capture the real material flow in the plant. To obtain such good results with PlantDesigner@ we have more than 12 000 tests of different material with our machines/equipment in the PlantDesigner@database. With all that incorporated, the tool can easily predict and calculate new applicatiordprocesses with high accuracy. Not surprisingly, PlantDesigner@ is an integral part of the Sandvik Rock Processing work culture as an engineering tool, customer communication channel, and knowledge management facility. ACKNOWLEDGEMENTS The authors would like to thank both the Management of Sandvik Rock Processing, Svedala, Sweden, in their support of the developments of PlantDesigner@,and Imagine That Inc., of San Jose, CA for their permission to publish this paper on PlantDesigner@. Special thanks to Fernando Romero, application and automation engineer, Sandvik Rock Processing, Appleton, WI, USA, Kristian Eriksson and Tony Dumpleton, applications engineers, Sandvik Rock Processing, Svedala, Sweden for their technical information and editorial assistance.

44 1

JKSimMet :A Simulator for Analysis, Optimisation and Design of Comminution Circuits R.D Morrison' and J.M. Richardson2

ABSTRACT JKSimMet was conceived as a user-friendly face for a suite of FORTRAN programs which had been developed at the JKMRC to analyse and model comminution and classification processes. In the early 1980's a simulate-only prototype was developed as part of the AMIRA p9 project. The data analysis capabilities were added by JKTech and the first commercial release was in 1986. JKSimMet development has since progressed to Version 5 and some 300 sites world wide. Combined with JK Breakage testing, JKSimMet is now an industry accepted tool for analysis, optimisation and design of comminution circuits. Many new JK models have been developed over the last 15 years and made available via JKSimMet. Over the past decade, comminution modelling has been extended to blasting leading to Mine to Mill optimisation for the complete comminution chain with often substantial improvements in productivity. BACKGROUND Definition of Simulation. Simulation can best be described as artificial experience. This artificial experience can be used to evaluate options at low cost and low risk. Actually, simulation is a key part of human existence, where experience and imagination are used to consider the consequences of possible courses of action. Simulation is employed in a wide variety of enterprises, from military preparations to maximization of business opportunities, but in all cases, the objective is to minimize risk. Regardless of the application, the successful use of simulation is an iterative process, involving these steps: 0 Experience (i.e., measure) 0 Analyze (develop model parameters) 0 Act (make those changes which appear to have the most promise) 0 Measure again History of Development of JKSimMet. The Julius Krutschnitt Mineral Research Centre (JKMRC), was founded in the 1970's as a direct result of research into the application of simulation in mineral processing technology. As early as 1965, Draper and Lynch (Draper, N. and Lynch, A. J., 1965) had demonstrated that using simulation techniques to rearrange the grinding capacity in a particular existing milling circuit could result in significant throughput improvement at a given grind size. This work formed the basis for what was to come. Since its founding as an arm of the University of Queensland, and with the support of the Australian Mineral Industry Research Association (AMIRA), the JKMRC has focused on developing models of unit operations that are of direct use to practitioners of mineral processing. By the mid- to late-1970's the JKMRC had developed a significant collection of FORTRAN subroutines, each subroutine designed to model a different mineral processing unit operation. The subroutines and their associated non-linear

' Deputy Director and Technical Director, Julius Krutschnitt Mineral Research Centre President, Contract Support Services, Inc.

442

parameter estimation and mass balancing routines, known as the GSIM package, ran on the minicomputers of that era and were made available to industry. The models and analysis techniques were well documented by Lynch (1977). For various reasons, probably not the least of which were minimal availability of computer skills and lack of integration, industry did not embrace the GSIM package and it saw only limited use outside of JKMRC research projects. In the early-l980’s, as the pace of computer hardware development began to rapidly accelerate, a project was begun to integrate the individual GSIM routines into a single software package with a “userfriendly” graphical front end, designed to facilitate the analysis and simulation of entire flowsheets. The project ultimately resulted in the commercial release of version 1 of JKSimMet, which allowed users to specify and simulate comminution circuits via graphical construction and included parameter estimation, or “model-fitting” capabilities to permit adjustment of the models to match existing plant data. By the end of the 1980’s, mass balancing capability was added, along with several additional unit operations models. Integration of the basic tools of data analysis, model-fitting and simulation with a large library of unit operations models, under the framework of an easy-to-use graphical user interface, led to wide industrial acceptance of the technology. The base of industrial users of JKSimMet climbed rapidly throughout the 1990’s. By the end of the 199O’s, the original DOS operating system-based JKSimMet, born on PC-AT class microcomputers, had evolved into the forerunner of the current Windows operating system-based version 5 , which is designed to take advantage of the speed and power of modern Pentium class machines. The functionality of JKSimMet has remained essentially the same, but its applicability has widened considerably and the productivity of end-users has grown exponentially.

TASK SYNERGY AND SIMULATION SOFTWARE INTEGRATION Fundamental Tasks. Beyond the evolution of the microcomputer as a universal tool for engineers, the key to the wide acceptance of a package such as JKSimMet was recognition of the need to integrate certain fundamental tasks in one piece of software. A list of tasks that have considerable “synergy” with one another includes: 0 Flowsheet drawing 0 Data Analysis 0 Model-Fitting 0 Simulation 0 Reporting These tasks are integral to the iterative nature of the simulation process. It is worth noting that the requirement to be a skilled user of FORTRAN - or any other computer language - is no longer an issue. JKSimMet is a user-friendly package for mineral processing engineers; data is entered and problems are defined and resolved by the user through the execution of tasks in a manner analogous to how engineers think and function on a daily basis. Flowsheet Drawing Engineers think of mineral processes in terms of flowsheets, due to both training and natural inclination. Thus, a necessary part of any integrated software package is a graphical interface in which the user describes the process to the software by drawing a flowsheet. Icons, representing individual unit operations, are chosen by the user and arranged in order of occurrence in the process and connected by graphical representations of flowstreams. An example of how this works in JKSimMet is shown in Figure 1.

443

Figure 1 Defining a flowsheet with JKSimMet Data Analysis Generally speaking, the task of measuring the process must take place outside of the simulation software package. However, once the data describing the current state of the process have been collected and entered via the graphical flowsheet interface, they must be sorted and analyzed for validity. Do the data accurately represent the true state of the process or are there inconsistencies? The function of data analysis (also known variously as reconciliation or mass balancing) is to determine if a given set of data is “good” or “bad” is complex and is best handled by the integrated software package. User friendly mass balancing forms the basis of a sister product called JKMBal as well as being almost certainly the most-used module of JKSimMet. The balancing module can be applied to many separation processes - and even to metallurgical accounting. Once the user has defined a flowsheet by drawing it and entering the individual flowstream data, JKSimMet permits the user to set up and run a mass balance to analyze the data. Figure 2 demonstrates this process for a copper flotation circuit analysis. Proceeding stream by stream, the user indicates the quality of the data by entering standard deviation estimates, used by JKSimMet to set weights which determine how much the software is allowed to “adjust” each data point in order to achieve a balance. In addition to total stream flows of solids and liquids, JKSimMet allows the user to specify the number and type of measurable components to be included for each flowstream. Allowable components include percent solids, chemical or mineralogical assays or percent retained on size fractions.

444

Figure 2 Defining and running data analysis (mass balance) with JKSimMet Model-Fitting If mass balance analysis shows the data to be good, the next task in the iterative procedure of simulation is to establish parameters for the various models which match the observed performance of the process embodied by the data. Model-fitting is the step whereby the user “tunes” or adjusts the individual unit operations models until they are accurately predicting the actual measured data. As with mass-balancing, the user describes the quality of the data to JKSimMet via standard deviation estimates. The user also tells the software which variables (model parameters) can be adjusted to match the measured data. Figure 3 illustrates this process.

445

Figure 3 Model fitting with JKSimMet Simulation Once accurate parameters and an accurate “base case” model of the existing process have been established by the model-fitting step, the user is ready to actually simulate the flowsheet. In this step, model parameters determined by the fitting step are held constant and various process options are tested by making changes to the model design variables (also known as operating variables) in order to compare the outcomes against the existing process. Model design variables include, but are not limited to, items such as process solids feed rates, water addition rates, solids and pulp specific gravities, ore breakage characteristics, equipment dimensions and flowsheet arrangement of unit operations. Figure 4 shows a flowsheet in which these values are being set and the main control window by which the user executes a simulation.

446

Figure 4 Setting up and running a simulation with JKSimMet

Reporting Just as data must be input to the software by the user, in order to perform simulations, so the outcome of the data analysis, model-fitting and simulation steps must be communicated back to the user by the software. Good reporting capability is essential to an integrated simulation system. JKSimMet V5 uses a configurable reporter with options for summarised or detailed data which can be printed to paper, the Windows clipboard or to files which can be read by other applications for ease of data transfer. Simulation results can also be displayed as size distribution graphs in a variety of sizing formats, including weight percent retained, cumulative percent retained and cumulative percent passing. Figure 5 shows the control window for report construction, along with typical report output and a typical size distribution graph from the same simulation.

447

Figure 5 Report generation with JKSimMet How JKSimMet Integrates Fundamental Tasks - The JKSimMet Structure. JKSimMet User Interface The current version 5 of JKSimMet incorporates all of the essential tasks listed above, into a single piece of software. JKSimMet is designed for use with MS Windows-, meaning it is designed to function within the framework of the Windows operating system, including all versions of MS Windows from MS Windows 95 through the current MS Windows XP.

Database Management The graphical user interface of JKSimMet communicates with a Microsoft Access database that stores and organizes the user’s data by projects and flowsheets within a project. Because of Windows compatibility, JKSimmet can communicate easily with other software such as spreadsheet programs and corporate online databases for easy input of data, and word processors and spreadsheet programs for customized output of data. Once the flowsheet is drawn and the data is inside JKSimMet, the internal database management system allows for seamless integration of the tasks of data analysis, model-fitting, simulation and reporting. Results from one task are readily available to all other tasks and the user can switch between them at will and in any order. Library of Unit Models JKSimMet contains an extensive library of unit models for use in all three major functions of data analysis, model-fitting and simulation. These are the building blocks of any flowsheet and they can be used in any combination and quantity. The user needs only to specify which unit models are required for the current flowsheet case, connect them with flowstreams and to specify the necessary design variables which define the current or starting operating condition for each unit. Figure 6 shows a portion of the unit model library.

448

Figure 6 Part of the unit model library in JKSimMet

Flowstreurn Data Model Flowstreams are also models in JKSimMet. Streams are modeled by specifying total flows of solids and liquids, as well as solid phase characteristics, such as specific gravity. Because JKSimMet is a software package for modeling comminution and classification processes, the solid phase is subdivided into individual particle sizes and each particle size is further characterized by its “assay”, that is to say, the weight percent retained on, cumulative percent retained on or cumulative percent passing that size. Figure 7 demonstrates the nature of the flowstream data, showing total flows for one stream and the particle size distribution data for another.

Figure 7 Flowstream data

449

Design Variables vs. Model Parameters The integration of tasks within a single software package works well because JKSimMet makes a strong and necessary distinction between design (operating) variables and model parameters. Users are taught to understand the difference between these two classes of variable and also the important differences between the three primary functions of JKSimMet: data analysis, modelfitting and simulation. Design variables are any variable which an engineer can specify, measure (within practical limits) or control; examples of design variables include flowrates, equipment dimensions (mill internal diameter, ball charge as 9% of mill volume, cyclone vortex diameter, number of operating cyclones), water addition rates, percent solids. Model parameters are variables specified by the developer of a mathematical model of a unit operation and usually relate design variables to unit performance (output). Because they are conceptual and somewhat abstract, model parameters are not easily specified or measured by engineers; they must be determined by model-fitting a given set of performance (survey) data using the model-fit function of JKSimMet. Adding New Models. JKSimMet Version 5 has an optional “Software developer’s Kit” (SDK) which facilitates the development of new mathematical models in a user friendly environment. Because the user interface of all JKSimMet models is contained in a data base the main component of the SDK is a data base editor. The other component is a FORTRAN template for the general model structure. Hence the developer only need learn about modelling - not about all of the intricacies of data base design or user interface development. USING JKSIMMET FOR DATA ANALYSIS Overall Approach and General Objectives. The data analysis process does not follow a rigid path but it is possible to provide some guidelines for a strong methodology which is reasonably similar for both mass balancing and model fitting. However, before starting, it is worth considering briefly the objectives of mass balancing and model fitting. If the data adjustments (for mass balancing) and residuals (for model fitting) are drawn from the same population as the errors in data measurement, then a very simple way of looking at the data is to divide each adjustment or residual by our estimate of its standard deviation. On average, these should have similar values and the expected value sum of their values squared should simply be the number of values considered, less the number of constraints active. In other words, adding up a series of numbers with an average value of one (or one squared) is likely to give a sum which is similar to the number of values. Dividing this sum by the number of points measured (less the number of constraints) yields a single number - the pooled standard error - which provides a measure of the accuracy of the complete balance or model fit. Further, an examination of how the weighted sum of squares (i.e. the sum of the adjustments or residuals divided by each standard deviation and squared) varies with small changes in flow rates or model parameters, results in an estimate of how well each of these is defined without resorting to a Monte Car10 analysis. Intuitively, if a small change in a parameter or flow rate causes a large change in the sum of squares, then that parameter is well defined by the data.

450

Now the overall objective of the process is easily and well-defined: to achieve a pooled standard error of around value equal to one (1) with well-defined (less than 10% relative error) flow rates and parameters. In practice, these objectives are a little too ambitious for most mineral processing data. A pooled standard error with value of two (2) or less is good data for grinding and classification and value of three (3) or less is good data for secondary and tertiary crushing and screening. For primary crushing and SAG mill feed, low numbers of coarse particles in a sample dominate the errors and each case should be considered separately (see Napier-Munn et al, 1996 for a general approach and recommended sampling methods as well as a detailed description of the process models and data analysis techniques). If the data are poor, the measurement should be rectified and repeated. NB-Neither mass balancing or model fitting will “fix” bad data. Either process can help to obscure errors and lead to poor decision making. The data analysis function of JKSimMet provides a tool which can help to pinpoint problems in data collection - but it cannot “fix” them. If there is sufficient redundant data, it may be possible to omit the problem data. It is also worth noting that a good balance fit may be obtained through chance and the balance may still be nonsense - fortunately, the model fitting step will usually identify these data sets as the models will probably be a poor fit. Mass balancing can have application beyond data analysis . Richardson (Richardson, 1991) describes how mass balancing can be applied to metallurgical accounting and the reconciliation of mine and plant inventories, once the quality of the data has been established. Model Fitting Methodology. Figures 8 and 9 outline a methodology that is easy to implement using JKSimMet. Once Mass Balancing (as shown in Figure 8) is completed, figure 9 outlines the methodology for Model-fitting that is best employed using JKSimMet.

45 1

h lnwt (Draw) Flow Sheet

0

Input Measured Stream Assays or Sizes and Standard Deviation (SD) Estimates Input Measured Stream Flow Rates and SD Estimates (at least one should be well-defined or set SD to value equal 100 if none are well-defined) Input Measured Water Flows and SD Estimates

I

~~

Run a Mass Balance on the Complete Flow Sheet

I

0 0

0

Select First Node Junction and Run a Balance CheckPSE If PSE less than 2, add the next node in the flowsheet and rerun balance, continuing on adding nodes until PSE greater than 2 When PSE greater than 2, examine weighted residuals for the node at which this happens, to identify measurement or tying problems

If measurement problems are found, then SDs can be increased for the problem data points to reduce their effect, BUT repeating the survey is recommended, once measurement problems are identified.

v Continue on to Model-Fitting

+ Figure 8 JKSimMet data analysis methodology

452

4

Transfer Unmeasured Flow Rates and SDs from Mass Balance (but NOT adjusted data values)

Select Initial Model Parameters from Experience or the Supplementary Parameters Manual (parameter database supplied with JKSimMet)

Run Preliminary Simulations to Check for Reasonable Results (finite circulating loads, underflow solids not too high, etc.) Not Reasonable (try new initial parameter guesses)

Reasonable

+I Run Model Fit and Check Pooled Standard Error (PSE) PSE greater than 2

PSE 2 or less

Model Fit Each Process Unit

v

Parameter Estimates

Go on to Simulation and Optimization 4

Model Fit Entire Process Flowsheet

PSE 2 or less Run Model Fit and Check Pooled Standard Error (PSE) PSE greater than 2

+I

Examine Weighted Errors in Detail - Look for Measurement and Other Data Problems I

I

Figure 9 JKSimMet model-fitting methodology

453

Data Analysis Application Example. For the more sophisticated and powerful JKSimMet models, a single representative set of data may often suffice. However, for secondary parameter dependencies, JKSimMet can simultaneously model fit several data sets to derive a wide range model. This capability is very useful. The database structure of JKSimMet version 5 allows for separate experimental, mass balanced, model fitted and simulated data types and allows the user to track and manage all of them. The “model fitted” data type provides a natural base case for optimisation by simulation. A series of eight crusher surveys were run on a secondary crusher at a range of conditions including variations on feed rate, feed sizing and crusher gap. Each data set was fitted individually to produce four parameters 0 K1 all particles smaller than K1 fall through the crusher 0 K2 all particles larger than K2 are broken 0 t,, particles which are selected for breakage experience sufficient breakage for t,, % to pass one tenth of the original passing size. 0 A multiplier which relates crusher net power draw to model power. A detailed description of the crusher model is provided in Andersen and NapierMunn (1991). Each of these model fits also produces a pooled standard error and a weighted sum of squares as shown in the table below.

Of the eight data sets, only Test 3 shows a poor fit. Figure 10 shows why. There is tramp coarse material in the measured product. This might indicate crusher overload and the crusher opening the gap or it may represent a problem in sample handling. In any case, this data is not suitable for model building. The model parameters are also unlike those derived from the other test data sets. Figure 10 uses the Graph property associated with each piece of equipment.

454

and Cal

100

90 80 70

,

60

,

%so, 40

30 20 10

’ ’ *

- ,

0 4

*

*

, - I

I

.

* I

1 I .

10

1.

100

i

1000

Size {mm) Figure 10 Quick graph of feed and product for test 3 Figure 11 shows two of the other data sets using the built using the Configurable Graph tool which can combine sizing and other data from any of the equipment used within a flowsheet. The symbols on this graph represent experimental points while the solid line are the results of the “best-fit” model fit. This provides a compact and easy to read format for comparing fitted and measured data or simulated data with a base case.

Model Fitting Results for Tests 1 and 2

.-$!

40

a

20 10

0

1

X A

v

I 10

1:Feed 1 Combiner, Exp 2:Cone 1 Roduct, Ex 3:Feed 2 Combiner, Ep!

-A,

Fun

- .

100

Screen Aperture (rnm) 1:Feed 1 Cohiner, Ft 2:Cone 1 Roduct, Fit 3:

Figure 11 Comparing measured and fitted data using the configurable graph tool.

455

A simple linear model form can link all of the acceptable models together. This facility is most useful for building wide range circuit models. K1= A0 + Al*CSS +A2*TPH + A3*F80 K2= BO + Bl*CSS +B2*TPH + B3*F80 T10= CO + Cl*CSS +C2*TPH + C3*F80 The two obvious approaches to deriving these combined models are: 0 Carry out multiple linear regression on the parameters tabulated above, or 0 Add parameters to a linked model fit in an analogous approach to stepwise multiple regression. JKSimMet allows the user to link together the same parameter in instances of the same model type within a flow sheet. All of the required data is added to the objective function defined by the Data List. While the first method is quicker, the second encourages the user to seek simpler models. The estimate of parameter standard deviation produced by the fit is helpful for deciding whether a term is significant - if the parameter plus or minus the standard deviation includes zero, the parameter can usually be omitted.

The above table shows the results of a combined fit for all tests except number three with dependencies added for throughput and closed side setting. The PSE is 1.83 for a combined SSQ of 340 - of the same order of fit as each test individually. The standard deviation estimates for each parameter suggest that some might be omitted. After the user has achieved a satisfactory model, simulation for optimisation can begin.

USING JKSIMMET FOR PROCESS OPTIMIZATION Methodology. JKSimMet lends itself well to a very straightforward and rigorous methodology for process optimization. The following figure outlines that methodology, which can be applied to almost all comminution and classification circuit optimization exercises. The simplest application of simulation is to match the comminution grind size to the classifier product size. This usually achieves a 5 to 10% improvement in throughput at a similar grind size. The more complex the circuit, the larger the potential gains which can be achieved by balancing the stages to each classifier. A more interesting application of optimisation is to adapt an existing circuit to an ore body which grows harder at depth. The ore at the Kidston Mine started at a Work Index of less than 10 kwh/t and over a few years more than doubled. Kidston staff used JKSimMet to consider the effects of pre-crushing and two stage recycle crushing.

456

Throughput as maintained keeping the operation viable for substantially more than the planned mine life. For a detailed description, refer to Needham and Folland (1994) A similar problem at the ASARCO mine - a 50% increase in hardness was overcome by partial pre-crushing of SAG mill feed. (McGhee et al, 2001)

Select Criteria for Optimisation

w Survey Circuit

Mass Balance

I JKSimMet Good Data

(Simulate)

Poor Data

Adjust Simulation Model to Maximise Optimisation Criterion Simulate)

Make the Changes I

b I

Figure 12 JKSimMet process optimization methodology Mine to Mill Optimisation. Perhaps the most profitable of all optimisation possibilities has been the matching of run of mine sizing to milling circuit characteristics. This approach uses JKSimBlast to estimate the product size distribution from blasting. Two successful examples are referenced - Kanchibotla et a1 1998 for work at KCGM and Kojovic etaL1998 for maximum iron ore lump product at Marandoo. The first aimed to maximise throughput and the second to minimise to production of -6mm iron ore “fines”.

USING JKSIMMET FOR PROCESS DESIGN Methodology. While the development of model parameters is a bit more problematic and hypothetical, given that the plant being designed does not yet exist, it is still possible to develop and

457

adhere to a rigorous methodology for hocess Design, using JKSimMet. The following figure outlines that methodology, which can be applied to many comminution and classification circuit design exercises.

Establish Design Criteria

1 TesVMeasure Ore Characteristic

Select a plausible circuit and equipment

4

I JKSimMet

Draw model parameters from test work and/or data base

imulate chosen circuit and check constraints

Yes

+I

IEvaluate for best capital and operating costs*

Yes

I

No

1 Figure 13 JKSimMet process design methodology Design Examples. Design is perhaps the most challenging application of all. Broad based application of JKSimMet to design became practical after a strong data base of industrial data and model parameters was available. As this database has grown so have design opportunities. A very large number of AG/SAG, ball mill circuits with or without recycle crushers have been designed using JKSimMet. The more challenging design cases have involved opportunities to take advantage of particular ore characteristics. For example, the Red Dome project in North Queensland required closed circuit SAG operation for soft flotation ore, SAGhall for harder free milling gold ore and a recycle crusher for still harder deeper ore. JKSimMet models were used to investigate all of these ore feeds. The same milling circuit was used for the entire life of mine. The largest SAG mill in the world used simulation (as well as conventional techniques) as a key part of its design process (Dunne, Morrell et al, 2001) Two of the most innovative circuits used closed circuit autogenous milling. The JKSimMet models in each case suggested promise for these ores. However, very careful pilot testing is recommended as many of these circuits have failed - without benefit of JKSimMet simulation!

458

Two which have succeeded very well are at Olympic Dam (Muller, 1998) and the Cannington Mine (Willey,1997). In each case, run of mine ore is processed into flotation circuit feed in a single stage of grinding at high energy efficiency. The single mill concept provides high capital and operating cost efficiency. However, it is certainly not applicable to many ore types.

POTENTIAL SIMULATION PITFALLS AND HOW TO AVOID THEM As with any complex human endeavor, simulation has certain pitfalls that need to be avoided. In particular: 0 Even good quality simulators can generate plausible “nonsense” when the basis is poor quality data 0 Attempting to use a simulator outside of the range of validity for a particular model will often result in such plausible “nonsense” 0 As the time-worn adage pertaining to all software says: “Garbage In = Garbage Out” 0 Attempting to optimize a design simulation in fine detail wastes valuable time and lends a level of accuracy to results that is not warranted by the accuracy of the data. 0 Attempting to optimize either an existing process or a new design without knowledge of real-world process constraints wastes valuable time and leads to solutions that often cannot be implemented. 0 Allowing users without sound knowledge of mineral processing fundamentals in general and the simulated process in particular will also waste time and lead to infeasible solutions. To avoid these pitfalls: Take full advantage of software training and documentation where offered Take pains to ensure data is of the highest quality. A fundamental knowledge of sampling theory and statistics will help to make it so Know the model and software limitations. A good simulation package such as JKSimMet will have a section describing “known restrictions” for each unit model Keep the level of detail for a given simulation exercise commensurate with the level of accuracy of the data and the objectives of the exercise. Tune the actual operation in detail, as suggested by the simulation, not the reverse Plan simulation test campaigns in advance. Know the real-world process constraints and understand which process design variables can be adjusted and which cannot. Make changes according to a plan rather than randomly. Simulation software is best used by engineers and professionals possessing a sound knowledge of mineral processing fundamentals, as well as their own flowsheets. The software can produce plausible looking results that are not achievable in the real world, but often this cannot be discerned without engineering knowledge and experience.

SUMMARY AND CONCLUSIONS Steady-state process simulation is mature technology for: 0 Data analysis 0 Plant optimization 0 Plant design (in conjunction with traditional methods) Integrated simulation software packages take advantage of the synergies of certain tasks, and incorporate these tasks into a seamless structure. Integrated packages allow

459

the user to enter data, analyze it and perform model-fitting, simulation and reporting of results, within the same software package and during the same session if necessary. In the future, the JKMRC plans to integrate liberation, flotation and gravity models within JKSimMet, extending the applicability beyond comminution and classification to all sections of the concentrator. Eventually, integration will extend to blasting models, allowing users to link mining practice with concentrator performance. ACKNOWLEDGEMENT JKSimMet and JKMBal would not exist without the long term support for applied industry research provided by the Australian industry through AMIRA (Australian Mineral Industry Research Association (predominantly via the P9 Project) and for the last 16 years of industry support world wide via JKTech. REFERENCES Andersen J. and Napier-Munn T.J. 1988. Power Prediction for Cone Crushers. Third Mill Operators Conf, Cobar, NSW (AusIMM) Dunne, R., Morrell, S., Lane, G., Valery, W. and Hart,S. (2001) “Design of 40 foot SAG mill installed at the Cadia Gold Copper Mine”. SAG 2001 ed. Mular and Barrett, Vancouver Kanchibotla, S, Morrell, S, Valery, S and O’Loughlin, P 1998 “Exploring the effect of blast design on throughput at KCGM’. Mine to Mill conference, AusIMM Brisbane Australia October pp 153-158. Kojovic, T, Kanchibotla, S, Poetschka, N and Chapman, J (1998) “The effect of blast design on the 1ump:fines ratio at Marandoo Iron Ore Operations”. Mine to Mill conference, AusIMM Brisbane Australia October pp 149-152. Lynch A.J. 1977. “Mineral Crushing and Grinding Circuits: Their Simulation, Optimisation, Design and Control”. Elsevier, 340 pp McKee D.J. and Wiseman D.M.,1991. General plant simulators for metallurgical plant performance. “Evaluation and Optimization of Metallurgical Performance” edited by Malhoutra, Klimpel and Mular, published SME .pp219-229. Richardson J.M., 1991. Reconciliation of mine site, mineral processing plant, concentrate and tailings inventories. “Evaluation and Optimization of Metallurgical Performance” edited by Malhoutra, Klimpel and Mular, published SME, pp 55-60. Muller, H, 1998. “Olympic Dam expansion- a study in retrofitting”. Mineral Processing and Hydrometallurgy Plant Design. AMF Perth Napier-Munn T.J., Morrell S., Morrison R.D. and Kojovic T., 1996, “Mineral Comminution Circuits: Their Operation and Optimisation”, JKMRC, University of Queensland, Brisbane, pp. 413. Needham T M & Folland G.V. 1994. “Grinding Circuit Expansion at Kidston Gold Mine”. Presented at SME Annual Meeting, Albuquerque, New Mexico. February 14 -17. McGhee, S, Mosher, J. Richardson, M, David, D and Morrison RD. 2001. “SAG Feed Pre-Crushing at ASARCO’s Ray Concentrator; Development, Implementation and Evaluation”. SAG 2001 ed. Mular and Barrett, Vancouver Willey,G, 1997. “The Cannington project case study. crushing and grinding in the mining industries,” IIR Conference, May, Perth.

460

JKSimFloat as a Practical Tool for Flotation Process Design and Optimization MC Harris', KC Runge', WJ Whiten' and RD Morrison'

ABSTRACT A practical model of flotation has been developed, incorporating new ideas in the defmition of the sub-processes of flotation, in terms of hydrodynamics, mineral floatability, froth recovery and entrainment. A methodology for circuit analysis using the new model has been developed and validated in a number of full-scale applications. This paper presents an overview of the model and its application, and discusses the development of a computer simulation package, JKSimFloat V5, which will allow the features of the new model to be captured by practicing metallurgists, for use in process design, process optimization and equipment selection. INTRODUCTION Froth flotation is undoubtedly the most important and versatile process for mineral concentration, yet despite a vast amount of research and development over many years, the design and operation of full scale flotation circuits is still substantially based on experience and heuristics. This often results in inefficient plant designs (usually over-designed, for safety), excessively long commissioning times, and initial operating performance levels below achievable targets. Over time, with appropriate process modifications, performance of new plants typically evolves towards optimum levels. However, the capital implications of this process development path are frequently extremely high, especially as many plants probably never reach their performance potential during their lifetime of operation. Appropriate use of modeling and simulation appears to be an obvious answer to solving many of these problems. However, describing the flotation process on the basis of fundamental relationships is extremely difficult. The number of variables affecting the kinetics of a floating particle have been considered by Arbiter and Harris (1962), who reported that as many as 100 variables contribute to the flotation of a particle in a flotation cell. Despite 'these difficulties, many attempts have been made to describe the overall kinetics of the process using a variety of sub-processes models. Some of these models lump all the sub-processes together while others try to describe each sub-process individually. In addition, several attempts have been made to develop relatively fundamental models of the flotation process. However, the difficulty of measurement of many of the physical process variables used in these models, and the unrealistic assumptions used in their derivation, render the application of these types of models on industrial plants impossible at this time. The general consensus is that the most useful flotation models are semi-empirical in nature. The development of an effective model of the flotation process involves making realistic simplifications which will reduce, to within reasonable limits the amount of experimental work required to derive the model parameters, but will not cause gross inaccuracies in prediction (Lynch et al. 1981).

'

Department of Chemical Engineering, University of Cape Town,Private Bag, Rondebosch 7701, South Africa. Email: [email protected] Julius Kruttschnitt Mineral Research Centre, Isles Rd, Indooroopilly, Queensland 4068, Australia. Email: [email protected]

461

This paper presents a practical methodology that has been developed for flotation circuit modeling and simulation, as part of a collaborative research project involving the Julius Kruttschnitt Mineral Research Centre (JKMRC) at the University of Queensland, Australia, the Mineral Processing Research Unit at the University of Cape Town, South Africa (since 1996), and the Mineral Processing Group at McGill University, Canada (since 2001). The project is being conducted as part of the Australian Minerals Industry Research Association (AMIRA) P9 Research Project, titled “The Optimization of Mineral Processes by Modeling and Simulation”. This is a multi-national project currently sponsored by 39 companies. Since 1992, the project has deployed considerable resources towards the development of models of the flotation process for application in simulation studies of industrial operations. The modus operandi of the project involves postgraduate students and research staff conducting a combination of diagnostic, development and research studies on operating plants, with the support and involvement of plant personnel ranging from operating staff to senior management. The broad sponsor support base of the project has presented the opportunity to develop, test and refine a wide variety of measuring techniques and analysis procedures on full-scale and pilot-scale operations on four continents, treating a wide range of different ore types. The outcomes of this work have been: 0

0

0

A set of measuring devices and procedures for the diagnostic performance evaluation of flotation cells and circuits A methodology for modeling the performance of flotation circuits for optimization studies A methodology for modeling the performance of flotation pilot-plants for design studies.

The logical extension of these outcomes is the provision of a simulator that will allow industrial personnel to efficiently access and use the information provided by these technologies. This will take the form of JKSimFloat V5, which is being developed in three phases, with each phase associated with a particular type and level of functionality. The following sections will present and discuss, in turn: 0

0 0

0

The models used in JKSimFloat V5 The measurement and analysis techniques used to calibrate the models The methodology used to apply these techniques and models to the analysis and simulation of flotation circuits The philosophy and structure of the new simulator, its functionality and use in the current development phase and its development path into the future.

TKE MODEL The processes of flotation circuit design and optimization begin with the development of a suitable kinetic model. As a first requirement, the model must be able to describe the recovery response of the various components of the feed into the concentrate or tailings of a flotation cell or bank of cells. This is relatively easy, and there are many, often very simple models that easily meet this criterion. However, to be useful for anything beyond simple diagnostic studies, the model must also be able to predict a response based on a change to the process. Clearly, as the number of process changes to which a model can respond increases, its potential usefulness for predictive simulation increases. However, the amount of information required to apply the model also increases, and this information can often be progressively more difficult or costly to acquire, Hence, the ideal kinetic model is one that is adaptable to different levels of refinement for a range of applications. This is an important consideration in the discussion that follows. The recovery of particles in a flotation process can be considered to occur via two primary mechanisms: 0 0

Trueflotation - selective recovery of hydrophobic particles by attachment to air bubbles Entrainment - nonselective recovery of both hydrophobic and hydrophilic particles in the water recovered to the concentrate.

462

Each of these mechanisms will be discussed in turn in the sections that follow. True Flotation The selective recovery of particles by attachment to bubbles in flotation is generally considered to be a first order rate process, provided that there is sufficient gas addition, and the concentration of hydrophobic solids isn’t so high as to cause bubble overloading. Work conducted by Gorain et al. (1997; 1998) studied the rate constant achieved in a variety of industrial flotation cells of different types and sizes operated at a range of different air rates, impeller speeds and froth depths. They concluded that the difference in the flotation rate constant (k) achieved in these different cells was due to a change in either the feed ore floatability (P), the bubble surface area flux (Sb) generated in the cell or the loss in recovery across the froth phase (Rf), i.e.

Based on these findings, the performance of a flotation unit can be considered to arise from the interaction of a stream property, the particle floatability (P), with parameters that characterize the operating conditions of the pulp and froth zones of the unit (Sb and Rf). This provides a very useful basis for linking a change in the flotation rate to its cause, a key requirement for the development of a predictive model. Floatability is the propensity of a mineral particle in a flotation pulp to be collected by true flotation, and reflects the interaction of the mineral properties with the pulp chemical environment. It arises from a range of particle properties such as size, density, shape, mineralogical composition, degree of surface oxidation, type and extent of reagent coverage, and degree of aggregation of the particles in the system. Hence, it is not surprising that most flotation feed streams cannot be described by a single floatability value, but rather by a spectrum of floatabilities, as has been noted by many previous researchers (Imaizumi and Inoue 1965; Harris and Chakravarti 1970; Huber-Panu et al. 1976; Dowling et al. 1985). The interaction of a mineral-bubble aggregate in the froth zone is affected by the conditions under which the froth zone is operated, such as froth height and gas rate, and the pulp chemical environment, particularly with respect to frother concentration. However, particle properties such as size, density and hydrophobicity also affect the froth zone performance, so the froth recovery factor, Rf, must also be represented as a distributed property. On this basis, equation 1 can be expressed as: k i = P i .Sb.Rf, In equation 2, ki can be considered to be representative of a set of particles that present a similar flotation performance under fixed operating conditions and fured chemical environment. Entrainment Many researchers have shown that the recovery of nonattached particles by entrainment occurs in proportion to the recovery of water over a wide range of operating conditions. The degree of entrainment arises from an interaction of the froth zone conditions and particle properties such as size and density, and must therefore also be considered a distributed property. Johnson (1972) has shown that this can be well described by a classification function to account for the effect of the particle properties: R(Entrainment), = Cfi R,

(EQ 3)

Where: Cf, R,

= the classification function for size i, = the recovery of

water to the concentrate

463

The degree of entrainment increases as particle size decreases, owing to the decrease in particle drainage rate with decreasing size. Hence the classification function, Cfi, in equation 3 ranges from a value of one for very fine particles (i.e. no particle drainage) to zero for coarse particles (i.e. particle drainage rate is so high that no particles are recovered to the concentrate). Combined Model Equations 2 and 3 can be combined, with the addition of an appropriate flow model, into an expression for overall recovery, by both true flotation and entrainment, for a grouped class of particles with a similar set of properties. For a perfectly mixed cell, which represents the flow conditions prevalent in most conventional flotation units, the following expression can be obtained

Ri =

Pi S, T R ~( l, - R , ) + C f i R,,, ( l + P i S, T R f i ) ( l - R w ) + C f R, i

(EQ 4)

Where: T

Sb Pi

= the residence time (s) = bubble surface area flux (m2/s/m2)

= floatability of particle class i (-) Cfi = classification function of particle class i (-) Rfi = froth recovery of particle class i (-) R, = water recovery (-)

The overall recovery for a unit feed divided into n classes of similar particles is then given by: R=xmiRi

(EQ 5 )

i=l

Where: mi

= mass fraction of particle class i in the feed (-)

Equation 4 represents the general expression of the flotation model developed in the P9 Project for use in the simulation, optimization and design methodology presented in this paper. It can readily be seen that if entrainment is not considered (ie. Cfi = 0), equation 4 takes on the familiar form: Ri =

ki T ( l + k , T)

However, this often-used form of the rate equation provides no mechanistic link between recovery and the flotation conditions and properties that affect it. The rate constant, k, whether distributed or constant, represents a purely empirical parameter. The advantage of equation 4 over equation 6 is that it provides a useful framework for the diagnostic analysis of the performance of a flotation process. Measurement of responses related to particle properties (Pi, Cfi, and Hi) in relation to unit operating characteristics (Sb, T and R,,,) can aid in the identification of cause and extent of circuit performance problems, and auditing the effect of circuit changes. In addition, when conducted over a range of unit and circuit operating conditions, these measurements can be used to develop appropriate sub-process models within a consistent, mechanistic framework that can be used for predictive modeling and design studies.

464

MEASUREMENT AND ANALYSIS TECHNIQUES Floatability In flotation, floatability is the property that is exploited to effect a separation of a mineral from an ore. It therefore follows that quantifying the floatability distribution of the feed to a flotation process is critical to understanding the behavior of the system. This is true whether one is interested in operating and controlling a process, improving a process, adapting a process to new requirements, or designing a new process. The problem that arises with respect to the flotation process is that floatability is not amenable to any direct measurement procedure. As a result, many flotation plants are both designed and operated based on the use of lumped kinetic parameters, such as provided by equation 6, with heuristic scale-up procedures related to the origin of the data (e.g. laboratory batch tests, locked cycle tests, pilot-plant campaigns or plant surveys). There are two parallel approaches that can be used to address this problem: 0 0

The use of a Distributed Property Floatability Component Model The use of an Empirical Floatability Component Model.

Each of these methods will be discussed in turn, with their associated advantages and disadvantages. Distributed Property Floatability Component Model (DPFC-Model). This approach involves dividing the feed into classes based on some measurable physical property or set of properties of the particles in the system. Each class is then assumed to float with a discrete floatability. The recovery of each class in a flotation system is used to detennine the rate constant or floatability of each class. Properties used to divide the feed include size (Thomlinson and Fleming 1965; Thome et al. 1976; Kawatra et al. 1982), mineral association or liberation (King 1976; Niemi et al. 1997; Savassi et al. 1998; Gorain et al. 2000) and chemical surface coverage (Niemi et al. 1997). The use of size, alone, as a floatability distribution parameter in this type of model will almost always be inadequate for use in modeling and simulation studies; for data of this type, the Empirical Floatability Component Model discussed below is generally more appropriate. However, the models distributed by mineral association (by size) and chemical surface coverage (by size by mineral composition) offer exciting possibilities, and are likely to be extremely important in the future. Both models offer a potential framework for the predictive simulation of changing grind size and liberation, and, rather more speculatively, simulation of chemical conditioning. There is no doubt that linking these processes in a robust simulation environment would represent a major advance in the design, optimization and control of industrial flotation plants. These models also have the advantage that they intrinsically contain size information and therefore can easily incorporate entrainment and froth recovery models. However, it is important to recognize that both these models remain somewhat problematic at the present time: 0

0

0

0

Mineralogical analysis is very costly and time consuming and there is no guarantee that the distribution of the floatability can be totally accounted for by size and mineralogical effects. For example, in systems where particles are severely aggregated or partially oxidized or in conditions of collector starvation, variations in the floatability of totally liberated particles in narrow size intervals has been observed (Imaizumi and Inoue 1963; Chander and Polat 1994). The error associated with the parameters in these types of models, can be relatively large (Chander and Polat 1995). During the fractionation of a sample into size and liberation classes, error may be introduced at various stages of sampling, weighing and analysis. Analysis of chemical surface coverage adds significantly to both cost and time, and further compounds the problem of analysis errors. The amount of test data available to even explore the relationship between measured particle properties and comminution andor reagent conditioning is currently somewhat limited. The large amount of high quality data required to develop and validate predictive

465

models of these units appropriate for use in a simulator is unlikely to be available for some time. Nonetheless, the increasingly widespread use of mineralogical analysis systems such as the PhilipsIJKMRC MLA and the CSIRO QEM*SEM by mineral processing operations and research organizations suggests that the amount and quality of the available information should increase rapidly. A number of projects are currently being conducted within the P9 Project on models distributed by mineral association and chemical surface coverage, with encouraging results. Empirical Floatability Component Model (EFC-Model). This approach involves apportioning the mass of particles in each stream of a flotation circuit into a number of floatability components each with a mean flotation rate constant (Imaizumi. and Inoue 1965). A floatability component is defined as a set of particles which present a similar flotation rate under fixed operating conditions and fixed chemical environment (Lynch et al. 1981). This methodology assumes that the floatability values (P) are conserved throughout the circuit as long as no regrinding, oxidation or reagent addition occurs. The principle of the method, and the parameters involved are presented in Fig. 1. Class PI, ml

Class PP, m;,

Class

......

PI,m

Class P”, m,

......

Class 0 Non-floating mineral mo P o = 0)

Gangue

Need to determine: 1. Number of Classes, n 2. Floatability of each class, P 3. Proportion of feed in each class, m

Ore

Figure 1 The division of a mineral (or minerals) in an ore into floatability fractions The key to the application of this model for circuit simulation is the determination of the floatability distribution of the feed into a circuit, as this property can then be used as an input property to a simulator. In the previous section, the floatability distribution of the feed to a circuit was assigned on the basis of a measurable property in relation to its response. The lack of a measurable property in the EFC-Model can be overcome by consideration of the response of the ore when treated in a flotation circuit. The mineral content of the various streams within a circuit represents the outcome of a fractionation process of the mineral originally in the feed to the circuit, based on its original floatability distribution under the physical and chemical conditions that exist within the circuit. This implies that if the circuit is operating under constant, steady state conditions, and the floatability of the ore is not changing with time, then the floatability distribution of all the streams within a flotation circuit can be directly related to the “parent” floatability distribution of the mineral in the feed ore. This principle can be used to regress a feed

466

floatability distribution that is consistent with the measured mineral content of the streams in the circuit. As can be seen in Fig. 1, a relatively large number of parameters can be required by the EFCModel: the number of classes, n; the floatability of each class, Pi, and the mass fraction of each class in the feed, mi. Even a simple, 3-class distribution requires 4 parameters (assuming a nonfloating class), with the number of parameters increasing by 2n-2 for increasing n. This means that the amount of data available from a survey of circuit streams is usually insufficient to resolve the floatability distribution with any reasonable degree of confidence: 0

0

The cells in different parts of the circuit are likely to be operated under different conditions (e.g. froth height and gas rate), which means that any estimate of floatability would be significantly affected by the accuracy of models used to account for these effects. The floatability distribution of even a single mineral in a circuit would be expected to require that the feed be distributed into 3 or more particle classes to provide a robust description for use in design simulation. The relatively large number of parameters required to describe such a distribution requires a large amount of high quality data.

This problem can be resolved by conducting simple benchmark tests on streams within the circuit, using a batch flotation cell. Such tests are best performed “hot”, on samples of slurry taken directly from the circuit, with no adjustment of the pulp chemistry. In this way, each stream provides a recovery-time profile, obtained under the same conditions of air rate, impeller speed and froth depth (which should be shallow to avoid froth effects). In this way the amount of data available from the circuit is substantially increased, and it becomes relatively easy to fit a set of floatability parameters to the data. Standard sum-of-error-squared minimization techniques, such as available in most standard spreadsheet software will generally suffice (Runge et al. 1997; Harris 1998). Batch and survey data can also be combined into a linearized model to facilitate this process (Alexander and Morrison 1998). The performance equation applicable to the analysis of the batch test data is: R; = z m ; , j ( ~ - e x p ( - ~cbalch i , j t))

(EQ 7)

i=l

Where: RS

= recovery of mineralj

in a batch test performed on stream s;

mtj

= proportion of the ith component in mineralj

of stream s;

= ore floatability of component i of mineralj, Pi,j Cbltrh =batch test cell operating constant.

The proportion of each component in each stream of the process is calculated from an overall circuit simulation, which is itself a function of the ore floatability of each component, the proportion of each component in the feed to the process and the cell operating variables (sb, Rf and 7). In equation 7, the batch test cell operating constant (Cb,,,) is a fitted parameter that combines all the hydrodynamic operating characteristics of the batch cell into a single number. It is treated as a constant for all batch tests performed using the same operating conditions (i.e. air rate, impeller speed and froth depth). In data sets where the operating variables (Sband Rf) of the industrial cells are not measured and therefore unknown, it is possible to use the rates achieved in the batch flotation test as a proportional benchmark of the performance of industrial cells by a scale-up number (Ccea):

k;,:

= C , , kpj

467

Where: ky.,

= the batch rate

constant of component i of mineralj

Experimental data are used to estimate the batch flotation rate constant of each mineral specific component, the proportion of each mineral component in the feed to the process and the scale-up number of each process unit. A batch cell operating constant is not required and the batch test recovery is simply calculated using equation 9. R;

= r m : , j ( l - exp(- kej t))

(EQ 9 )

I=I

Simulations using cell scale-up numbers have a larger number of derived parameters than those performed using measured cell operating variables. This reduces the confidence that can be associated with any predictions made using this type of model. The approach is nonetheless usehl as it is relatively easy to implement, and can provide information for screening ‘‘fxst pass” alternatives both quickly and cheaply. One of the main advantages of the use of stream kinetic batch tests in the application of the EFC-Model is that the assumption of constant floatability in the circuit can be tested directly, independent of the use of any model, using the technique of nodal analysis (Runge et al. 1997). A node represents any point in the circuit where: One input stream is split into 2 or more output streams (e.g. a flotation cell, or bank) To or more streams combine to form one output stream (e.g. a sump). Mass must be conserved around a node, so the batch test recovery-time data of streams entering a node (for example) can be combined, based on their relative flows, to produce the predicted recovery-time curve that would be obtained if these streams were combined and then subjected to a batch flotation test. This hypothetical combined stream recovery- time data set is given by:

2 -

R CornbinedSuem.lrninulcr Mineral j

R Sucm

;,:;F :

s=l

1 rninutcs

Mincralj

2

(EQ 10) Fs’le”s, Mineral]

$=I

Where: Suamr FMine,al j

R Svcarns.irninutcs Mineral j

= flowrate of mineral j

in stream s

= cumulative batch test recovery

of mineral j in stream s after t minutes

The recovery-time curve obtained using equation 10 can then be compared to the experimentally measured curve conducted on the combined stream. If the two curves cannot be separated on a statistical basis, it can be concluded that the floatability of the particles is conserved around the node. This procedure can be repeated around all the nodes in the circuit, each providing a further check of the constant floatability assumption. Occasionally, there is a need to reduce the pulp density in the batch cell (with plant water from same stream) when very high-grade circuit concentrate streams are treated, owing to bubble overloading effects. This possibility should be explored when it appears that floatability is not conserved around a “high-grade’’ node, especially when it does appear to be conserved around “low-grade” nodes in the circuit. In circuits where floatability is changed, whether by the addition of reagent, or in a regrind step, the nodal analysis procedure can be performed around nodes before and after the change, provided that the respective sections are kept independent (i.e. they don’t share streams by

468

recycle), which is usually the case. More importantly, by conducting tests on the streams around the node in which the floatability is changed, the extent of the change can be audited by comparing the predicted recovery-time curve of the combined input streams to the predicted recovery-time curve of the combined output streams. This can be an extremely valuable diagnostic tool, as it provides a direct, quantifiable llnk to the effect of a change (whether grind size, or reagent addition, or both) on the intended target property, floatability (Runge et al. 2001). The effect of such changes when audited by standard survey data, on the basis of recovery and grade, is often masked and confounded by changes in the operating characteristics of the circuit. This technique has been tested on many full-scale and pilot-scale plants over the last 5 years, for a wide range of both base and precious metal sulfide ores, and floatability has been found to be conserved in the vast majority of cases. It represents an easy and highly effective technique for analyzing flotation circuit behavior. To summarize, the EFC-Model represents a highly flexible method of deriving a floatability distribution for use during circuit simulation. It has been applied to a wide range of ores, treated in both pilot-scale and full-scale equipment over a wide range of sizes and operating conditions. It can be applied to virtually any set of standard circuit survey data, but its use in combination with batch test data is strongly recommended, as this substantially improves confidence levels for prediction, particularly for design simulation studies. Circuit analyses with respect to gas dispersion, froth recovery and entrainment are also recommended where possible or justified. The main problem with this approach is that grade prediction is often poor, especially when unsized sample information is used, as accounting for entrainment can be problematic. This can be remedied to some extent by assaying a few key streams in the circuit by size, and using this information to calibrate a gangue entrainment response for the circuit. However, any components in the feed recovered by true flotation that cannot be resolved from the assay data will also cause significant errors in grade prediction. It is important to note that an error in any one of the component recovery models will give rise to an error in the grade prediction, even if all the other component recovery models are perfect. More elaborate analysis of the circuit or the batch test samples can be used to improve the derived model, but this type of information will typically add considerable time and cost to the exercise. However, it may well be justified under some circumstances. In fact, where a DPFC-Model is used in conjunction with mineral association by size information, it is recommended that the EFC-Model also be used to analyze the circuit. This adds very little to the overall effort required in such a study, and the results of the EFC-Model could well significantly simplify the analysis, interpretation and model fitting of the much more complex, and relatively unproven DPFC-Model approach. Entrainment Measurement of entrainment involves sampling the pulp in a flotation cell in a quiescent region below the pulp-froth interface, in a manner as to exclude the collection of any bubbles (and their associated attached particles). This sample can be considered representative of the solids entering the froth by entrainment. This sample is sized, and compared to a sized sample of the concentrate from the cell. If there is a substantially liberated component of the ore that is known to be nonfloating, this can be used to determine the entrainment by size relationship of this component. These values can then be used for the prediction of the entrainment behavior of the other particles, based on the model presented below, as long as their density and shape characteristics are not substantially different from the calibration component. If there is no non-floating, liberated component present in the ore, a non-floating mineral tracer can be introduced into the cell feed, and this can be used to obtain the necessary calibration. As mentioned previously, entrainment can be well approximated using an appropriate classification function, Cf. Work by Savassi et al. (1998) has shown that the following 2parameter classification function, ENT, is able to describe the entrainment response very well over a wide range of cell operating conditions, and this equation is used in JKSimFloat V5 to predict entrainment when sized data is available:

469

ENT, =

[

)")

(dpi exp 2 . 2 9 2 ~E20

+exp

[-

(dpi)&J) 2.292 x E20

(EQ 11)

Where:

E20 = particle size, dpi, at which ENTi = 0.2 6 = froth drainage parameter ENT essentially represents an entrainability curve for the ore, a property that is highly comparable to floatability. The two parameters of the function, E20 and 6 are to a large extent the analogues of the properties P and Rf used to model the response of particles recovered by true flotation. However, unlike Rf, the effect of cell conditions on the entrainability curve in a particular circuit is typically relatively minor. On this basis, the prediction of entrainment by simulation is simple: measure entrainment on a circuit as discussed above, fit equation 11 to the data, and these (average) parameters can then be used to represent the feed entrainability for subsequent circuit simulation. Note that constant entrainability does not imply constant entrainment. The mass recovered by entrainment in a particular cell in a circuit will depend on the particle size distribution and the water recovery in that cell.

Froth Recovery Froth recovery is a term that can be defined in a variety of ways. In this paper the term only applies to particles transferred into the froth by true flotation. It is the fraction of particles that entered the froth attached to bubbles that are recovered to the concentrate. This means that there are two ways that such a particle can be recovered. It can be recovered to the concentrate while still attached to a bubble, or it can detach in the froth, and then be recovered by entrainment. The direct measurement of froth recovery on industrial cells can be problematic, and the development of a robust measure of this important performance characteristic is currently being explored in a number of studies within the P9 Project. The simplest method is to measure the mean flotation rate as a function of froth depth in selected cells in the circuit where the mineral content varies significantly (e.g. in a rougher cell, a cleaner cell and a scavenger cell). This information can be resolved by an extrapolation procedure to determine Rf in these cells (Vera et al. 1999), and these measured values can be used to calibrate the froth recovery model used during simulation. The main drawback of this approach is that it is often difficult, sometimes even impossible, for use in the analysis of industrial cells, but it can be relatively easy to perform in pilot-plants for use in design simulation studies. The alternative approach is to simply assign froth recovery values to cells for use during simulation based on values calibrated from base-case survey data. This is often quite adequate for many types of study. The predictive simulation of froth recovery in JKSimFloat V5 uses the following model: R , =(l-Pd)+PdENT, R w

(EQ 12)

P, = 1- exp(- p FRT)

(EQ 13)

470

Where: Hf

= froth height (m)

Ax

= cell cross sectional area (m2)

FIOUl

in the froth (-) in the cell (m3/s/mZ) - concentrate volumetric flow rate (m3/s) = gas hold-up

E Air

= superficial gas velocity

J, Q;QI~;~W~C

Froth recovery in a cell is considered to be a function of the time particles spend in the froth phase, which can be approximated by using either the Froth Retention Time (FRT) of the air, equation 14, or the FRT of the slurry, equation 15. Equation 14 is more useful for prediction and simulation, as it is not dependent on the prediction of water recovery (required to calculate the concentrate flow rate). However, equation 15 is generally more accurate for the diagnostic analysis of cell performance in a circuit where the concentrate flow of the cells has been measured directly. The model postulates that particles are either recovered to the concentrate while still attached to a bubble, or by entrainment after dropping off a bubble. Thus froth recovery is a function of the probability of detachment (Pd), the entrainability (ENT) and the water recovered (Rw)in the cell (equation 12). The probability of detachment is exponentially related to the froth residence time (approximated by FRT) and a probability of detachment rate constant (p) (equation 13) (Gorain et al. 1998; Mathe et al. 2000). Gas Dispersion

The effect of gas dispersion on the flotation rate is accounted for by the bubble surface area flux, Sb, in the cell (Gorain, 1997). Sb is calculated as follows:

Where: J,

= the

superficial gas velocity

d,

= the

Sauter mean bubble diameter

Both J, and db are measurable using the UCT bubble size analyzer (Tucker et al. 1994) and a superficial gas velocity probe (Gorain et al. 1996). The Mineral Processing Group at McGill University is currently developing more robust tools for gas dispersion measurement on industrial cells within the P9 project. Sb can also be predicted using a correlation developed by Gorain et al. (1999). This was developed using a large number of data sets collected from different base metal flotation plants: Sb =123 J:" N:M A:'' P80a4*

(EQ 16)

Where:

N, = impeller tip speed A, = impeller aspect ratio (impeller widtldimpeller height) P80 = cell feed 80% passing size

471

Water Recovery The measurement of water recovery is relatively simple. The prediction of water recovery during simulation is rather more problematic. A number of options are available, but none of these approaches can be used with any confidence when a cell is simulated in operation under conditions very different to those used in the model calibration base-case. For example, if it is intended to simulate circuits that employ recleaning, it is important to calibrate a recleaner water recovery model on data collected on a cell performing at this duty. In fact considerable caution should be exercised whenever a simulation indicates significant departure from measured cell duties relative to the base-case calibration data. There is currently a considerable effort underway in the P9 Project to develop a more mechanistic approach to this key performance characteristic. Three options are currently available for predicting water recovery in JKSimFloat V5: Fixed Concentrate Percent Solids. A fixed value of percent solids in the concentrate is assigned to a cell, and water recovery is calculated such that the concentrate percent solids equals that specified by the user during simulation. Solids dependent model. The concentrate water flow (Qw)is related to the solids concentrate flow (Qs). The model has two empirical parameters, a and b, which are obtained by calibration to the base-case data:

Q, =aQ:

(EQ 17)

Water rate function. Water recovery is assumed to follow fust order kinetics and is cell residence time dependent. The water recovery rate equation for the perfectly mixed flotation cell is shown as equation 18. The water rate constant is once again obtained by calibration to the base-case data. It is important to recognize that this approach has no basis in reality whatsoever, as the amount of water in the concentrate is, at best a very weak function of residence time. However it can be a mathematically convenient way of accounting for water flow during simulation under some circumstances.

Residence Time Residence time is calculated during simulation based on the effective volume of a flotation unit (i.e. excluding mechanism volume, with a correction for gas hold-up). It can be based on either the unit volumetric feed rate or the volumetric tails rate, with the latter generally considered most appropriate. Despite this, measurement of residence time of cells or banks of cells using liquid tracers can provide valuable information, especially when the flow characteristics in a unit could be expected to depart from a fully mixed flow regime. Perhaps more valuable, however, is that this technique can provide a much more accurate measure of flow rate of large volume streams in fullscale plants than those provided by commonly used plant instrumentation. Errors in measured stream flows can severely compromise model calibration, and any subsequent simulation studies. THE METHODOLOGY Flotation Plant Design The application of the models and techniques presented in this paper to flotation plant design is best considered following a description of typical current practice: 1.

Mineralogical examination in conjunction with a range of grinding tests 0 Establish economic liberation target, as the basis for selecting primary grind size Establish possible strategies for regrind in the flotation circuit

472

2. A range of laboratory scale batch tests and locked cycle tests Reagent suite evaluation Mineral recovery and selectivity Flotation kinetics 3. A circuit design based on scale-up of the laboratory kmetics, and recovery-grade data 4. Preliminary economic evaluation of ore body 5 . Pilot-plant tests of the circuit design Evaluate circuit performance based on recovery and grade Mineralogical examination of some circuit streams (us. tails and recycle streams) Refinement of the design, reagent suite, grind strategy 6. Economic evaluation 7. Full scale plant design. The first problem with this strategy occurs in step 3, where the specification of a plausible plant design for economic evaluation and pilot testing is based largely on rule-of-thumb scale-up of laboratory data. This procedure is highly dependent on the experience of the design engineer, especially in terms of knowledge of plants operating on similar ores, where these exist. That this procedure works reasonably most of the time is more related to the relatively forgiving nature of the flotation process, and a tendency to build in healthy safety factors into circuit capacity estimates. The second problem occurs in step 5. Current practice typically attempts, as far as possible, to run pilot plants in the configuration of the intended full-scale design. This practice is strongly indicative that the principles of the flotation process are poorly understood. If performance can be linked to the properties of the ore, rather than being linked to a lumped combination of both ore and circuit, this “mirror” approach is no longer required. This can have important implications, as pilot plants have very different operating characteristics to larger plants. Pilot plants become progressively more difficult to operate at steady state in complex configuration with multiple recycle streams. Not infrequently, the dynamics of these systems can be such that they never reach steady state. This can severely compromise the information derived from the tests. Complex pilot plant circuits are also very labor-intensive, both in terms of operation and sampling. The large number of samples typically limits the extent to which the majority are analyzed; detailed analysis, whether by size, or mineralogy is usually limited to key process streams such as final concentrate and tailings, and a few recycle streams. Thus the performance of the test circuit is usually primarily considered in terms of recovery and grade. Once again, this limits the meaningful interpretation of the information in terms of full-scale plant design. The models and procedures presented in the first part of this paper offer the opportunity to significantly improve the process described above. It is important to note that the most important aspect of the techniques that have been described are always related to circuits. A simulation model, however modest its foundation, requires some process of multiple fractionation to resolve any meaningful information. Firstly, the data derived in steps 1 and 2 above provide a reasonable basis for a preliminary economic evaluation, as is typically the case, but this information should not be considered a valid basis for specifying a design beyond the requirements of the evaluation. Where the design appears to represent a key component of the preliminary evaluation, an attempt should be made to conduct some multiple-stage tests in the laboratory. This can be used with the EFC-Model to determine a more meaningful set of kinetic parameters than can be obtained from single stage batch or locked cycle tests. Subsequent pilot testing should then be used to develop a basis for the plant design. A simple configuration should be used initially, such as rougher-scavenger-cleaner,possibly with a cleaner scavenger, with no more than one recycle stream. This circuit can be subjected to a nodal analysis, and application of the EFC-Model, to provide an estimate of the floatability distribution of the minerals in the ore. In addition, a preliminary entrainment analysis can be conducted around one of the units, so as to obtain an initial estimate of the entrainability function (ENT). This simple set of information can readily be analyzed in a spreadsheet to provide a basis for specification of a pilot circuit for more detailed evaluation. The information can also be used to reassess the preliminary economic evaluation.

473

In line with the comments above, the second phase of pilot testing should not attempt to mirror a full-scale design. It should be set-up to provide information on the response o f units expected to perform a similar duty, in terms of mineral content and particle size, in the full-scale operation, in as simple a configuration as possible. This can frequently best be achieved by running stages in open circuit where feasible. The level of subsequent analysis of the circuit then relates to the level of confidence required to proceed with the final design. The analysis can take the form of EFC-Model based simulation, combined with simple stream assay data, all the way through to a DPFC-Model, with all the units analyzed for entrainment and froth recovery. Obviously, the cost difference between these two extremes will be substantial. The modeling should not be based on the operation of a single circuit configuration, however. The circuit should be changed in some way, usually best achieved by redirecting one or more recycle streams, and reanalyzed in the same manner as the first circuit configuration. This provides a much stronger basis for the calibration of the simulation models. Finally, a third configuration should be tested for comparison to model prediction. This need only constitute a simple survey based on stream assay data. As is the case with the level of sample analysis, this can be extended as warranted by the level of confidence required to proceed with the design. The pilot plant can also be used very effectively to audit the performance of units in the circuit such as regrind mills, and other retreatment equipment where relevant, based on the nodal analysis procedure. This can also be used to refine the reagent suite, or the use of staged reagent addition. Not all the information derived from a pilot plant should be considered with the same level of confidence in terms of scale-up for full-scale plant design. The ore characteristics, as defined by the component model properties are likely to be very robust in most cases; certainly far more so than is the case with the procedures used at the moment. However, performance of the pilot-scale units with respect to froth phase performance can be somewhat less reliable. Poor scale-up of volume and mass flows can significantly compromise any full-scale design predictions. It is in this area that experience and access to information on full-scale operations treating similar ores can be very usehl. Flotation Plant Optimization The application of the proposed methodology to the simulation of existing plants appears at first glance to be somewhat easier than design simulation: 0 0 0

0

The chemical and physical condition of the ore is provided by the plant. Scale-up is no longer a key issue. A large volume of historical operating data is usually available. Plant personnel typically provide a large body of operating experience.

However, there are also a number problems related to the analysis of full-scale plants relative to pilot plants: 0

0 0 0

The ore is frequently subject to a high degree of variability. Access to reliable sample points can be limited, and representative sampling difficult. Often difficult to vary unit operating conditions to calibrate unit models. Plant management are frequently reluctant to implement circuit changes for model validation.

With these issues in mind, development of a simulation model of an operating plant should really only be attempted if the following requirements can be met:

0

The plant operation can be maintained in a sufficiently stable condition to obtain at least two base-case surveys, with access to the majority of major streams for sampling and kinetic batch test analysis. At least one survey can be conducted with a significant change to the circuit.

474

0

Representative units form key stages in the process can be analyzed with respect to performance over a reasonable range.

The level of sample analysis should reflect the objective of the simulation exercise, but analysis by size is highly recommended in most cases. In addition, a mineralogical analysis of a few key streams can be very useful. Where the above conditions cannot be met, or in any case where a sophisticated simulation model is required, it is recommended that a pilot plant be used to run in parallel with the plant, operated on feed diverted from the plant. This scenario offers the best possible combination of the strengths of both types of operation. It can readily be seen that none of the problems listed above with respect to the analysis of full-scale plants represents a problem on pilot-scale plants. This offers the opportunity to both calibrate and verify a simulation model with minimal disturbance to the full-scale plant. The full-scale operation provides information for calibration of the mass and volume flows in the simulator. This strategy for simulation model development is a key component of the approach currently used in the P9 Project in plant modeling studies, based on the use of a custom designed pilot plant, the Wemco@ FCTR (Floatability Characterisation Test Rig) developed in partnership with Eimco. To date, this technology has been applied on a number of Lonmin platinum operations in South Africa, and Western Mining nickel operations in Western Australia, and has been extremely successful in both cases (Rahal et al. 2000).

SIMULATOR DEVELOPMENT The work presented in this paper represents an integrated methodology for gathering, interpreting, generalizing, and applying information to the separation of minerals in a flotation process. Many of the tools and procedures that have been described are readily accessible, and easily applied by practicing metallurgists. However, access to the full potential of this type of technology requires a high level of integration of a very complex body of information. This in turn requires either a considerable amount of expertise by the user, or an interface that can process, organize and present the information with a high degree of automation: A simulator. In 1993 the JKMRC completed the development of the first version of a flotation simulator, JKSlmFloat in a small AMIRA-funded project. JKSimFloat V2 followed at the end of 1995. Both versions were MS-DOS based, and allowed for the simulation of flotation circuits with a few classical flotation models, and a column flotation model. The new models discussed in this paper are not compatible with the structure of this program, and this lead to the development of Versions 3 and 4 of JKSimFloat, both as internal prototypes for use by the research team. These prototype versions were useful for proof-of-concept testing of the simulation methodology in a number of research case studies. However, problems with the underlying structure of these DOS-based versions in relation to the new models severely limited their use by the research team and their potential for commercial release. Consequently, a plan to develop version 5 in a Windows-based environment was initiated in 1998. However, the formal specification of an appropriate structure and interface for JKSimFloat V5 was found to present a formidable task. The simulator was required both as a vehicle for delivery of the current technology to industry, and as tool for the ongoing development of the technology by the research team; satisfying the requirements of both types of user within the same software environment presented a considerable challenge. However, this was considered to be of critical to the ongoing success of the work, which had arisen primarily from the continuum of involvement of researchers and industry in the development of the technology. The outcome of a three-year software design process is a specification for a Windows-based steady state computer simulation package, JKSimFloat V5,that will be implemented as a suite of modular products delivered over three phases of development. The functionality that will be associated with each JKSimFloat V5 development phase is discussed in the sections that follow. Phase 1 JKSimFloat V5 Phase 1 will provide a system that will allow a user to determine the metallurgical performance of a flotation circuit based on defined flotation models and flotation circuit stream

475

data using a simulation engine. Data entry and data viewing will be performed via a flowsheet based graphical interface. The simulation engine is being designed to interact with a highly flexible stream structure that readily accommodates the range of requirements of the models currently in use (DPFC-Model and EFC-Model), while offering virtually unlimited capability for developing and testing new models in the future. It is important to note that the JKSimFloat Phase 1 software module will not have the capability of calibrating the unit models used in the flowsheet. Appropriate sets of model parameters for a particular operation will have to be determined offline by experts. What the software will provide, however, is a vehicle than can be used to deliver the outcome of a circuit modeling study to a specific user in the form of a calibrated simulation model of a process. The software would be preconfigured, with hnctionality customized to user requirements, to a level justified by the data used to generate the models. It is envisaged that these preconfigured systems will have the capability of assisting: 0

0

0

0

Metallurgists in assessing the impact of plant throughput, unit operating conditions, unit volume, circuit layout, stream destination, new unit addition and change in plant feed sizing or liberation on metallurgical performance Plant Operators to understand the effect of day to day changes in circuit operation (e.g. unit operating conditions, plant throughput, feed properties) on metallurgical performance Researchers to test and evaluate new flotation models Academics with the training of both undergraduates and postgraduates in the principles of flotation Consultants to develop and demonstrate process optimization and design for client operations.

Phase 2 The Phase 2 development of JKSimFloat V5 will involve the incorporation of a mass balancing algorithm, a liberation data importing function and a range of viewing and charting facilities into the existing Phase 1 system. The mass balancing algorithm will be able to balance both one dimensional type plant data (e.g. size, assay or mineral class data) and two dimensional type plant data (size by assay, size by mineral or size by liberation class data). The graphical user interfaces developed in Phase 1 will be used (with some minor modifications) to enable mass balance data entry and the results of the mass balance data to be viewed and analyzed within the system. This mass balancing module has the potential to be used by researchers, consultants and plant metallurgical staff to assess the quality of plant data, estimate unknown plant data and adjust plant data for use in subsequent analysis. To complement the mass balancing algorithm, a database import function will enable data from mineral liberation analysis systems such as the PhilipsIJKMRC MLA or CSIRO QEMSCAN to be imported into the streams of the simulator. This data can then be used in liberation-based simulation models or balanced using the mass balancing algorithm. The flowsheet-based graphical interface will be used, in combination with a range of charting and overview functions, to assist researchers, consultants and plant metallurgists to view and interpret this type of data.

Phase 3 The phase 3 development of JKSimFloat V5 will involve incorporation of model fitting, model assessment and optimization modules into the system. The model fitting algorithm utilizes numerical techniques to determine the value of parameters required for the flotation models. The graphical user interfaces developed in JKSimFloat Phase 1 will be used (with some minor modifications) to enable model fitting data entry and the results of the model fitting to be viewed and analyzed within the system. To complement the model-fitting algorithm, sensitivity analysis techniques (e.g. Monte Carlo analysis) will also be incorporated into the system. The aim of these techniques is to provide model developers with tools to mathematically assess the quality of the developed model. The

476

function of the optimization module will be to determine optimum operatingkircuit conditions to meet specified economic criteria. Development Schedule JKSimFloat V5 will be the product that results from the completion of Phase 3. It is anticipated that this process will take some 3 years to complete, although a number of aspects of the Phase 2 and 3 plans still require more research before a definite schedule can be defined. However, Phase 1 development is at a very advanced stage, and is scheduled for completion in March 2003. At this stage, the implementation of a number of customized simulators will begin, on behalf of the group of companies within the P9 Project who have supported the software development. CONCLUSIONS The methodology presented in this paper has proved to be a powerful, yet practical approach to the modeling and simulation of industrial flotation plants. Many of the techniques that form part of this methodology have already been accepted and integrated into the practice of the companies who partnered the research in which they were developed. JKSimFloat V5 represents the vehicle for the transfer of the overall methodology as an integrated technology for flotation circuit design, characterization and optimization by simulation. As importantly, it will provide researchers with a powerful platform for developing and testing new models of the flotation process, and linking these models to other key unit processes such as comminution and reagent conditioning. It is anticipated that the use of a common platform for research, development and application studies will greatly facilitate technology transfer in the future. ACKNOWLEDGEMENTS Most of the work presented in this paper is based on the results of research conducted within the P9 Project, since 1992. The authors would like to thank: 0 The companies whose funding and support have made, and continues to make, this work possible 0 The Australian Minerals Industry Research Association (AMIRA) who manage and broker the Project 0 The staff and students at the JKMRC, UCT and McGill who have contributed to this substantial body of knowledge. The authors would also like to thank the P9 sponsors of the JKSimFloat development project, which is also supported by a generous commercialization grant from the Collaborative Industry Venture (CIV) Program of the Queensland Department of State Development. The JKSimFloat V5 functional specification was prepared by a JKMRC user group comprising Dr Stephen Gay, Dr Ricardo Pascual, Kym Runge, Dr Andrew Schroder and Prof. Bill Whiten, chaired by Prof. J-P Franzidis, with technical review by Martin Harris (UCT) and Dr Rob Morrison. REFERENCES 'Alexander, D.A. and R.D. Morrison. 1998. Rapid estimation of floatability components in industrial flotation plants. Minerals Engineering. 11:2 133-143. Arbiter, N., and C.C. Harris. 1962. Flotation kinetics. In Froth flotation, 5dhAnniversary Volume. ed. D.W. Fuerstenau , New York: AIME. Chander, S. and M. Polat. 1995. Flotation kinetics: Methods for estimating distribution of rate constants. Proceedings 19'hInternational Mineral Processing Congress. 105-1 1 1 . Dowling, E.C., R.R. Klimpel, and F.F. Aplan. 1985. Model discrimination in the flotation of a porphyry copper ore. Minerals and Metallurgical Processing. May 87- 101. Gorain, B.K., J.P. Franzidis, and E.V. Manlapig. 1996. Studies on impeller type, impeller speed and air flow rate in an industrial scale flotation cell. Part 3: Effect on superficial gas velocity. Minerals Engineering. 9:6 615-635.

477

Gorain, B.K., J.P. Franzidis, and E.V. Manlapig, 1997. Studies on impeller type, impeller speed and air flow rate in an industrial scale flotation cell. Part 4: Effect of bubble surface area flux on flotation kinetics. Minerals Engineering. 10:4 367-379. Gorain, B.K., M.C. Harris, J. P. Franzidis, and E.V. Manlapig. 1998. The effect of froth residence time on the kinetics of flotation. Minerals Engineering. 11 627-638. Gorain, B.K., J.P. Franzidis, and E.V. Manlapig. 1999. The empirical prediction of bubble surface area flux in mechanical flotation cells from design and operating data. Minerals Engineering. 12 309-322. Gorain, B.K., J.P. Franzidis, K. Ward, N.W. Johnson, and E.V. Manlapig. 2000. Modelling of the Mount Isa Mine copper rougher-scavenger flotation circuit using size by liberation data. Minerals and Metallurgical Processing. 17:3 173-180. Harris, C.C. and A. Chakravarti. 1970. Semi-batch froth flotation kinetics: species distribution analysis. Trans AIME. 247 162-172. Harris, M.C. 1998. The use of flotation plant data to simulate flotation circuits. Proceedings SAIMM Mineral Processing Design School, SAIMM , Johannesburg. Huber-Panu, I., E. Ene-Danalache, and D.G. Cojocariu. 1976. Mathematical models of batch and continuous flotation. Flotation: A.M. Gaudin Memorial Volume,. ed: M.C. Fuerstenau. New York : AIME, 2 675-724. Imaizumi, T., and T Inoue. 1965. Kinetic considerations of froth flotation. Proceedings Sixth International Mineral Processing Congress. 58 1-593. Kawatra, S.K., P.J. Suardini, and W.J. Whiten. 1982. The computer simulation of an iron ore flotation circuit. Proceedings 14Ih International Mineral Processing Congress. 10.1-10.19. King, R.P. 1976. The use of simulation in the design and modification of flotation plants. Flotation: A.M. Gaudin Memorial Volume, ed: M.C. Fuerstenau. AIME, 2 937 - 961. Lynch, A.J., N.W.Johnson, E.V. Manlapig, and C.C. Thorne. 1981. Mineral and coal flotation circuits: their simulation and control. Elsevier Scientific Publishing Company, Amsterdam. Johnson, N.W. 1972. The flotation behaviour of some chalcopyrite ores. PhD Thesis, University of Queensland. Mathe, Z.T., M.C. Harris, C.T. O’Connor. Modelling the influence of the froth phase on recovery in batch and continuous flotation cells. Proceedings XXI International Mineral Processing Congress. B8a 33-39. Niemi, A.J., R. Ylinen, and H. Hyotyniemi. 1997. On characterisation of pulp and froth in cells of flotation plant., International Journal of Mineral Processing. 51 5 1-65. Rahal, K.R., E.V. Manlapig, J.P. Franzidis. 2000. Flotation plant modelling and simulation using the Floatability Characterisation Test Rig (FCTR). Proceedings International Congress on Mineral Processing and Extractive Metallurgy (MINPREX 2000). AusIMM. 339-344. Runge, K.C., M.C. Harris, J.A. Frew, and E.V. Manlapig. 1997. Floatability of streams around the Cominco Red Dog lead cleaning circuit. Proceedings 6Ih Annual Mill Operators Conference. AusIMM, Publication series 3/97. 157-163. Runge, K.C., M.E. Dunglison, E.V. Manlapig and J.P. Franzidis. 2001. Floatability component modelling - a powerful tool for flotation circuit diagnosis. Proceedings 41h International Symposium on Fundamentals of Minerals Processing. CIM. 93-107. Savassi, O.N., D.J. Alexander, J.P. Franzidis and E.V.Manlapig. 1998. An empirical model for entrainment in industrial flotation plants. Minerals Engineering. 11:3 243-256. Thorne, G.C., E.V. Manlapig, J.S. Hall, and A.J. Lynch. 1976. Modelling of industrial sulphide flotation circuits, Flotation: A.M. Gaudin Memorial Volume, ed. M.C. Fuerstenau. AIME, 2 937 - 961. Todinson, H.S., and M.G. Fleming, 1965. Flotation rate studies. Proceedings 6Ih International Mineral Processing Congress. 563-579. Tucker, J.P., D.A. Deglon, J.P. Franzidis, M.C. Harris, and C.T. O’Connor. 1994. An evaluation of a direct method of bubble size measurement in a laboratory batch flotation cell. Minerals Engineering. 7516 667-680. Vera, M.A., J.P. Franzidis, and E.V. Manlapig. 1999. The simultaneous determination of collection zone rate constant and froth zone recovery in a mechanically agitated flotation cell. Minerals Engineering. 12:lO 1163-1176.

470

USIM PAC 3: Design and Optimization of Mineral Processing Plants from Crushing to Refining S. Brochot, J. Villeneuve,J. C. Guillaneau, M.V. Durance and F. Bourgeois'

ABSTRACT Simulation is now widely used in process industries. USIM PAC has been developed over the past 16 years for the Mineral Processing sector. Recently, it has gained capabilities in the field of contaminated soils and waste treatment. In its latest version, USIM PAC 3.0 handles the design and optimization of an entire processing plant with a comprehensive set of mathematical models for unit operations that span from crushing to refining. Moreover, novel concepts have permitted addition of environmental impact assessment criteria to the traditional technical and economic objectives. A variety of objective functions and tools, such as the simulation supervisor facilitate the global multi-criterion optimization. Features and capabilities of USIM PAC 3.0 are reviewed through a case study in the field of gold ore processing. INTRODUCTION Since 1986, BRGM has been developing a powerful process simulation software package, USIMPAC (Broussaud 1988; Durance et al. 1993; 1994; Guillaneau et al. 1997). It is a userfriendly steady-state simulator that allows mineral processing engineers and scientists to model plant operations with available experimental data and determine optimal plant configuration that meets production targets. The simulator can also assist plant designers with sizing unit operations required to achieve given circuit objectives. The software package contains functions that can manipulate experimental data, calculate coherent material balances, sizes and settings of unit operations, physical properties of the processed materials, simulate plant operation and display results in tables and graphs. Widely used in industrial plant design and optimize, with more than one hundred fifty licenses sold in thirty countries, this software has been continuously improved, through successive versions, to make it more accurate and easier to use. These last years have seen significant developments in mineral processing technologies, particularly in hydrometallurgy, bio-hydrometallurgy (CCzac et al. 1999; Brochot et al. 2000) and mineral liberation. In addition, it is now necessary to take into account the environmental impact at each stage of a mining project, including water and power consumption, waste treatment and disposal (Sandvik et al. 1999; Guillaneau et al. 1999). This new version of the simulator, USIM PAC 3.0, incorporates these modern developments. Indeed, its structure and tools allow the user to take into account, at the same time, a wide range of technological, economic and environmental aspects (Brochot et al. 2002). The main features of USIM PAC 3.0 will be presented through the description of the design and optimization methodologies. The significance of these features will be illustrated by an example of design of a gold ore treatment plant.

' BRGM, OrlCans, France.

479

THE SIMULATION BASED APPROACH Process modeling and simulation are used at all stages in the life of a mineral processing plant: from process development to site rehabilitation, including feasibility studies, engineering design, plant commissioning, plant operation and upgrading right through to decommissioning. From the beginning, the simulation-based approach gives an idea of the behavior and performance of the future plant. This idea will be more and more precise owing to the capitalization of knowledge acquired through laboratory tests, pilot plant campaigns and plant operation. There is a continuous exchange between reality and the virtual plant constituted by its steady-state simulator. A simulator combines the following elements (see Figure 1): 0

0

0

0

A flowsheet that describes the process in terms of successive unit operations and material streams. This flowsheet encapsulates the experience of the engineers responsible for the plant design or optimization. It can reflect various scenarios so they can be compared against given criteria. It takes into account numerous plant features such as reagents distribution, water recycling or wastes treatment. A phase model that describes the materials handled by the plant (raw material, products, reagents, water, wastes) so that unit operations and plant performance, product and reagents quality (grades and undesirable element level), waste characterization (e.g. longterm behavior) for environmental impact can be evaluated. The phase description is critical for analyzing and optimizing the process. This statement reinforces the vital importance of field data and sampling protocols. A mathematical model for each unit operation. This model formalizes the current scientific knowledge about the unit operation, and its level of complexity depends on the data available and the targeted objectives (i.e. flowsheeting, unit operation sizing or optimization). The model parameters - dimensions, settings and calibration factors - are calculated or validated from field data. A set of algorithms for data reconciliation, model calibration, unit operation sizing, full material balance calculation, power consumption and capital cost calculation. These algorithms are interfaced to a set of data representation tools. As a result, the plant simulator constitutes a highly efficient communication vector between the different actors who play a part in the plant life. Physical properties Input Flowrates Sizes, Grades, etc. ,

MODEL

Output@)

Parameters

Feed stream description

Plant operation Plant capital cost

Equipment configuration Model parameters Equipment configuratior

Steady- state simulator

Plant operation Objective

Figure 1 Functions of the models and the steady-state simulator The simulation-based approach may be used for three main purposes (Durance et al. 1994). The frst one is the preliminary plant design. The simulator is used to calculate an overall plant

480

balance with quantitative information on all streams (mass flowrates, grades, size distributions). This yields selection of the key unit operations and a first estimate of investment costs. At this stage, the simulator can be used to eliminate possible processing routes. The second one is the advanced plant design. Based on detailed laboratory and pilot plant test data, the simulator is used to predict the plant operation at a level sufficiently advanced for flowsheet optimization and precise unit operation sizing. Used in advanced plant design, the simulator is an invaluable aid at the plant commissioning stage or for operator training. The third classical application is plant optimization (including audit, retrofit and upgrade), in which actual plant operation data is used to build a simulation that mirrors the plant behavior. This approach is classically used to improve the performance of operating plants.

Preliminary plant design Figure 2 shows the methodology used to produce a preliminary plant design. The first step consists in assessing the plant requirements in terms of flowsheet and stream descriptions based upon feed characteristics and main performance objectives. A preliminary material balance is established by direct simulation, which yields an ideal description of all the streams. The next step uses reverse simulation to back-calculate the dimensions of the main pieces of equipment. The final step simulates the future plant operation and calculates the capital investment. This process allows the process engineer to compare several flowsheets according to their technical performances and their financial implications.

Conventional flowsheet hypothesis

Laboratory experimentation: -measurement of Bond index

. mesurement of flotation kinetics

USIM PAC simulation of the plant using simple models without equipment dimensionino

USIM PAC

dimensioning the equipment for the nlant

USIM PAC technical results

printout of a full report with flowsheet, graphs and tahlrx

USIM PAC simulation of the full operation of the plant. Comparisons between several possible flowsheets USIM PAC economical resutk

calculation of the approximate capital cost of the main equipment and the overall cost of the plant

Figure 2 Methodology for preliminary plant design

481

Advanced plant design Figure 3 illustrates the method for designing a new plant from a pilot plant test. The first step uses material balance techniques to reconcile all experimental data coming from sampling campaign during pilot plant test. The second step consists in building a simulation of the pilot plant by calibrating each unit operation model using these coherent data. After multiplication of all streams by the scale factor, the next step uses reverse simulation to back-calculate the dimensions of the main pieces of equipment in industrial conditions. As previously, the final step simulates the future plant operations in various configurations and calculates the capital investment.

I

Objective Industrial plant design

Collection of data in the pilot plant

USIM PAC establishment of detailed coherent material balances describing the operation of the pilot plant

USIM PAC calibration of the simulation models for the equiment used in a Dilot olant

USIM PAC dimensioning of the simulation models for the equipment used for the industrial plant

USIM PAC technical results printout of a full report with flowsheet, graphs and tables

USIM PAC simulation of various configurations (tlowsheetr, settinas. etc.) for the industrial olant

USIM PAC economical results calculation of the approximate capital cost of the main equipment and the overall cost of the plant

Figure 3 Methodology to design an industrial plant from a pilot plant campaign Plant optimization and upgrade Figure 4 proposes a simplified box diagram of the methodology used for optimizing the flowsheet of an existing plant (pre-control optimization or plant upgrading due to new production objectives or operating constraints). The first step uses material balance techniques to reconcile all available plant operation data. The second step consists in building a simulation of the existing processing plant by calibrating each unit operation model using the coherent plant data. The final step is about using the simulator to test different processing scenarios and analyze the simulation results in technical (characteristics of the products, power drawn by the main equipment), environmental (tailings stability, waste minimization, water recycling) and economic (estimation of the capital cost investment and reactive consumption) terms.

482

Collection of data on the operation of the existing plant

USIM PAC establishment of detailed coherent material balances describing the oDeration of the dant

USIM PAC calibration of the simulation models for the main units of equipment in the plant

USIM PAC simulation of the operation of the plant at its future capacity. Identification of bottlenecks

USIM PAC dimensioning of additional eauiDment

Figure 4 Methodology for optimization of an existing plant Data processing Processing and analysis of field data require various algorithms that are available in USIM PAC:

0

0

0

0

0

Data reconciliation by statistical material balances (Le Guirriec, Brochot, and Bergounioux 1995) gives a set of coherent stream data from a set of experimental incoherent data coming from different measurement sources (on-line sensors, pilot plant tests, laboratory tests, etc.). Each data set is characterized by its own confidence level. The only rules used in this algorithm are the material conservation laws. Direct simulation uses a sequential modular iterative algorithm. It calculates all stream data from feed data and unit operation parameters. Direct simulation can be used only on a selected sub-flowsheet or on a single unit operation. Optimization algorithm seeks unit operation parameters that yield output streams that match the objective streams as closely as possible. This algorithm is used for unit sizing, model calibration or physical property calculation, the difference appearing in the set of desired parameters: dimensions, adjustment parameters or physical properties. The objective stream can be an entire stream data set or some stream data parameters such as the d80, a component grade or a recovery. It can also be an output unit operation parameter such as the power consumption. Objective Driven Simulation (ODS) is a mix between direct simulation and optimization (Villeneuve et al. 1992). At each simulation iteration, it calculates, for each unit operation, the parameters that meet a specified target. This algorithm is generally used to improve the plant calibration previously done unit by unit using optimization. Global optimization uses the same algorithm as the unit operation optimization, however it applies to the entire flowsheet. It finds parameters of specified unit operations that meet a plant performance objective. For example, it can be used on a flotation circuit to evaluate the number and volume of cells per bank that gives the best compromise between grade and recovery such that profit is maximized. Capital cost estimation calculates the investment cost corresponding to each unit operation with a cost model and the overall plant construction cost with a hierarchical

483

0

ratio system (Mular 1982). The list and level of the ratios is entirely configurable by the user. The supervisor can be used either as a sensitivity analysis tool or for visual Optimization (Guillaneau et al. 1995). It calculates user defined plant parameters (global or local performances, constrained parameters, etc.), named sensors, when some input parameters (feed data or unit operation parameters), named actuators, are varying. It is then possible to draw sensor variations as a function of actuators and observe the sensitivity of a plant to given changes. In the case of a multi-criterion optimization, it is easier to choose the best configuration by examination of such a graph rather than by converting the target into an objective function.

Each algorithm has many options for translating simulation objectives into mathematical problems. However, these algorithms can be easily used with a set of default options chosen for their ability to fit most situations. This is one reason why USIMPAC can be used by process engineers as well as by researchers.

PRELIMINARY DESIGN OF A GOLD ORE TREATMENT PLANT The objective of this case study is the design of a grinding/classification/leaching/adsorptionplant capable of treating 100 t/h of a gold ore with 95 % recovery. Characteristicsof the feed are: 0 0 0

Size distribution: 0x8 mm Gold content: 7 ppm. Specific gravity: 3.

Plant specifications are: 0 0 0 0 0 0

Primary grinding with a rod mill in open circuit. Secondary grinding with a ball mill in close circuit. Classificationby hydrocyclone with a circulating load from 150 % to 250 %. Leaching tank series with d80 = 75 pm for the leaching feed. CIP tank series with 50 ppm of gold in the recycled carbon. Dewatering of the barren pulp in a thickener with water recycling for solid percent regulation.

Laboratory tests give: 0 0 0 0 0

Work Index: 14 kWh/st. Maximum recovery of gold by cyanidation: 98 %. Leaching rate constant: 0.3 h-' (assuming a first order kinetic). Adsorption rate constant: 700, time constant: 0.3 (assuming the kn equilibrium model). Maximum solid percent after clarification: 70 %.

The following sections details the succession of steps used in preliminary plant design (see Figure 2): plant modeling with flowsheet drawing, phase model description and choice of mathematical models for each unit operation, stream data input, direct simulation and unit sizing algorithms, results display using graphs and sheets. Flowsheet drawing The first element of the plant model is the flowsheet that describes the different unit operations and material streams. The &lowsheet Drawing>)tool (see Figure 5 ) is dedicated to drawing the

404

flowsheet. It is entirely graphic and icon driven with over 100 unit operation icons available through a toolbox. Streams and units are referenced by a unique number and can be given a specific label, which the software will conveniently use through all its user interfaces. Colors and text fonts can be adjusted to increase readability.

Figure 5 Flowsheet drawing tool Phase model In USIMPAC the material flowing into the streams is described in terms of physical phases. There are three types of phases corresponding to the three material states: solid, liquid and gas. A fourth type named “ore” is defined to identify the raw material from other solids like carbon. Each phase is uniquely described by its flowrate and some description criteria among “particle size”, “composition” (mineralogical or chemical composition), “particle type”, “floating ability” sub-populations and “user defined sub-populations”. Two criteria can be crossed to have a finer description such as mineralogical composition by size class. A set of predefined phases offers different ways for material description according to the available data and the level of complexity imposed by the objectives and the unit operation models (see Figure 6). The ability of USIMPAC and its mathematical models to work with a wide range of predefined or user-defined material phase descriptions is a very strong feature of the software. Communication mechanisms between the phase model and the unit operation mathematical models have been largely presented in previous papers (Durance et al. 1993; Brochot et al. 1995). For the case study at hand, the ore is described in terms of a size distribution for grinding and classification and a global gold content for leaching. The predefined phase named “Gold ore” fits this description (see Figure 7). The other phases present in the process are the water used for wet treatment and leaching and the carbon used for the CIP stage. These phases are described by their gold content.

485

Figure 6 Different phase models

Figure 7 Phase model used in the example Equipment description Various mathematical models can be associated to each unit operation drawn on the flowsheet. Mathematical models calculate the output streams data from the input stream data and model parameters (see Figure 1). These parameters can be equipment sizes, operating conditions, physical properties, model adjustment parameters or simply performance criteria. Depending upon the simulation objective and the data available, different mathematical models can be used for the same equipment. In USIM PAC, mathematical models are divided into four levels: 0

Level 0 models enable the user to specify directly the performance of the units. For example, the performance of a classification unit can be modeled by a partition curve for which the user specifies the bypass, the imperfection and the d50 (the cut-size). Such models are flowsheeting models that do not take into account any sizing parameters. During the simulation, the performance of the unit will be independent of its dimensions and the flowrate of the ore feeding it.

406

0

0

Level I models take dimensional parameters into account. They require little (sometimes no) experimental data. A typical example is a ball mill model, which uses only the Bond Work Index as its single experimental parameter. If no data is available, it is even possible to estimate the Work Index. Obviously the precision of such models is limited, but they are simple to use. Models of higher level are more accurate but they require the estimation of some of their parameters. This estimation can be carried out either on the basis of experimental data obtained from the continuous operation of the unit (level 2 models) or from such data supplemented by information obtained from specific tests, generally carried out in the laboratory (level 3 models).

Over 120 mathematical unit operation models are available covering a wide range from crushing to refining, from ore dressing to waste management. These include comminution (SAG, Pebble/Rod/Ball mils, SAM, etc.), classification (Hydrocyclones, Screens, Rake/Spiral classifiers, etc.), concentration (ConventionaYColumn floatation, Gravitymagnetic separation, etc.), hydrometallurgy (Leaching, Bioleaching, CIP, CIL, Precipitation, Cementation, Solvent extraction, Electrowinning, etc.), solidliquid separation (Filtration, Sedimentation, etc.), waste treatment (Collection, Sorting, Incineration, composting, etc.).

Performance models used for the preliminary material balance At this stage of the study, only level 0 mathematical models are used for describing the performance of each piece of equipment. The global objectives issued from the plant specifications are translated into local objectives unit by unit (see Table 1 and Figure 8). The size reduction from the plant feed d8O of 4,800 pm to the leaching feed d80 of 75 pm is a three-stage process: it uses a rod mill product with a d80 of 1,OOO pm, a ball mill product with a d80 of 120 pm and a cyclone overflow with a d80 of 75 pm. The Mill (OA) model generates a Rosin-Rammler size distribution for the product with the same slope as the feed and the desired d80. The Hydrocyclone (OB) model calculates OF and UF size distribution using a RosinRammler partition curve. The default values for the short circuit and the imperfection are used. The partition coefficient d50 is calculated such that the d80 specification of the OF is satisfied.

Figure 8 CIP - Carbon-In-Pulp(0) mathematical model parameters The specified gold 95 % recovery is also divided into two local objectives: 95 % recovery during leaching and 95 % recovery during CIP. The water recycling will permit achieving the global 95 % recovery.

487

Table 1 Level 0 models Units #1- Feeder #2 - Rod mill regulator

Models and main parameters Values Mixer (0) Density Regulator (0) Percent solids at regulator output (%) 70 Mill (OA) #3 - Rod mill d80 at the mill discharge (mm) 1 #4 - Hydrocyclone regulator Density Regulator (0) Percent solids at regulator output (%) 40 #5 - Hydrocyclone Hydrocyclone (OB) Short circuit of fines (%) 25 d80 of output fine stream (mm) 0.075 Corrected partition curve imperfection 0.3 #6 - Leaching Leaching (0) Leached percentage per component of ore and solid 95 phases (%) - Gold CIP - Carbon-In-Pulp (0) #7 - CIP 95 Adsorbed percentage per component of the liquid phase (%) - Gold Solid/Liquid Separator (0) #8 - Thickener Percent solids of the slurry stream (mass %) 70 Liquid Split (0) #9 - Splitter Maximum flowrate of liquid to the specified output (th) 180 Density Regulator (0) #10 - Ball mill regulator Percent solids at regulator output (%) 55 #11 -Ball mill Mill (OA) d80 at the mill discharge (mm) 0.12

Figure 9 Stream data entry

488

Feed data The input interface for the stream data follows the phase model structure. It gives, stream by stream, the list of phases and solid-liquidpairs (see Figure 9). Available data for each phase are the mass flowrate, the volumetric flowrate and the density if they are known, the component grade and the size distribution for the ore phase. Depending on the study, stream density may be required and descriptions such as composition by size classes or floating ability by component can be necessary. Available data for each solid-liquid pair are the pulp mass flowrate, the pulp volumetric flowrate and density if they are known and the solid percent. For a given solid-liquid pair, only two values among both phase flowrates, pulp flowrate and solid percent are necessary. The other two are calculated. The solid-liquid pairs are configured at the phase model stage. Preliminary material balance calculation The ideal description of all the streams is calculated by the direct simulation algorithm from the feed description using the chosen performance models. This preliminary material balance predicts a first estimate of 0 0 0 0

The circulating load in the grinding circuit, The recycling level of water and the fresh water consumption, The d80 and the gold content for each stream, and The overall gold recovery.

This ideal material balance will be used as the new objective, called “target”, during the equipment sizing stage. Full stream description can be displayed using the stream data entry interface (see Figure 9), the stream overview sheet or more synthetically with graphics and global results. Graphics There are seven distinct forms of graphical representations: size distribution curves, size partition curves, density distribution curves, density separation curves, split curves, stream and component bar graphs. These graphs are entirely configurable. Some predefined graphs can be drawn directly from the flowsheet popup menu. It is possible to draw the size distributions of all solid components directly from a stream submenu. Furthermore, size partition and split curves can be drawn directly from a unit operation submenu. Figure 10 gives an example of a graph showing the size distribution curves for the hydrocyclone feed, overflow and underflow streams. Figure 11 gives the gold partial flowrate in each phase and each stream as a bar graph. It clearly indicates the amount of gold in the grinding circulating load or in the recycled water as well as the phase transfer between ore and water and then between water and carbon. Global results The global result sheet is a results summary. It displays a set of calculated parameters -given in columns - for each stream -given in lines. These parameters are user-defined from a large list of possible parameters. It is then possible to draw a stream bar graph from the selected column of a global sheet. Figure 12 shows a global results sheet that displays the d80 of the ore size distribution, the partial mass flowrate of gold for the combined phases and the water “recovery” relative to the water loss in tailings, giving a measure of the water recycling level.

489

Figure 10 Size distribution curves of the cyclone feed, overflow and underflow

Figure 11 Stream bar graph of the gold partial flowrates

490

Figure 12 Global results sheet Higher level models used for equipment sizing The mathematical models associated with the rod mill, the ball mill, the hydrocyclone, the leaching tank and the CIP circuit are turned to level 1 or 2 models that takes equipment sizes into account (see Table 2). The Rod Mill (1) and Ball Mill (1) models are based on the energy-based theories of grinding and in particular on Bond’s law and the Allis Chalmers methods for dimensioning grinding mills (Rowland and Kjos, 1978). The choice was made to use one rod mill and one ball mill with predefined conditions such as mill shape, grinding media loading and speed. Only the diameter will be calculated to fit the target size distribution. The Hydrocyclone (2) model is based on the empirical equations established as a result of experimental work (Plitt, 1976). This model accounts for the roping effect. The number and dimensions of the hydrocyclones are calculated. The leaching (1A) model uses a first order kinetic equation for the gold transfer in solution using a maximum recovery and a constant rate (McLaughlin and Agar 1991). The leaching tank icon represents a series of 6 tanks for which the volume is determined to achieve the 95 % gold recovery. The CIP (1) model is derived from the kn model (Fleming, Nicol, and Nicol 1980). The CIP icon represents a series of 6 tanks with a counter-current carbon flow. The tank volume is fixed to an arbitrary volume and the resident time of carbon is adjusted to achieve the 95 % gold recovery. It is then possible to calculate the volume of carbon in the tank and hence the tank volume from the ideal carbon concentration in pulp.

49 1

Table 2 Level 1and 2 models Units Models and main parameters Values Rod Mill (1) #3 - Rod mill Number of mills in parallel 1 ???? Inside mill diameter (m) 2.72 Length/diameter ratio 2 Percent volumetric loading of rods 35 Fraction of critical speed 0.7 Rod specific gravity 7.8 Work index per component (kWh/st) 14 #5 - Hydrocyclone Hydrocyclone (2) ???? Number of hydrocyclones in parallel 2 ???? Cyclone diameter: D (m) 0.5833 ???? Distance between the underflow and overflow nozzles / D 1.8402 ???? Diameter of the feed nozzle / D 0.3843 ???? Diameter of the overflow nozzle / D 03356 ???? Diameter of the underflow nozzle / D 0.2146 #6 - Leaching Leaching (1A) ???? Tank volume (m3) 744 Number of tanks in series 6 Maximum recovery per component of ore and solid phases (%) 98 - Gold Rate constant per component (l/h) - Gold 0.3 CIP - Carbon-In-Pulp (1) #7 - CIP Number of tanks in series 6 Tank volume (m3) 500 Rate constant per component of the liquid (
492

Figure 13 Unit of equipment sizing CONCLUSIONS An example was used to highlight the ability of the USIM PAC simulation software to model and simulate a wide range of process types and technologies. The simulator offers various powerful tools in response to the increasing demand for a multi-criterion and global approach by plant designers. It takes into account a wide spectrum of design criteria, including: 0

0

0

Economic criteria such as capital cost, reagent and power consumption, production quality in terms of valuable mineral grade or undesirable elements level; Technical aspects with the evaluation of various configurations and processing technologies, a complete and detailed description of all material streams and their behavior during process; Environmental factors such as water consumption and recycling, pollutant production or waste treatment.

USIM PAC is an extremely flexible simulator. It can be used equally by process engineers for plant design or optimization, by researchers for process development, as well as by academics for teaching process engineering students. The latest version, USIM PAC 3.0, represents a significant milestone towards integrating different industries through a global approach. It is currently possible to simulate treatment from the mine to the metallurgical plant. Current studies on a global approach in urban waste management (Sandvik et al. 1999) or metal life cycle (Reuter 1998) already use steady state process simulation techniques. These studies are indicators of the evolution of the mineral processing field, which will integrate the industrial chain starting from the raw material production right through to ultimate waste of products.

ACKNOWLEDGMENTS The authors acknowledge the financial support from BRGM Research Division for this ongoing project. This paper is the BRGM publication n"2099. REFERENCES Brochot, S . , M.-V. Durance, J.-C. Guillaneau, and J. Villeneuve. 2002. USIM PAC 3.0: New Features for a Global Approach in Mineral Processing Design. Proceedings APCOM 2002 Conference. (to be published). Brochot, S . , M.-V. Durance, S . Foucher, J.-C. Guillaneau, D. Morin, and J. Villeneuve. 2000. Process simulation to enhance complex flowsheet development: examples in biotechnology. SME-Control2000 Conference.

493

Brochot, S., M.-V. Durance, G. Fourniguet, J.-C. Guillaneau, and J. Villeneuve. 1995. Modelling of the Minerals Diversity: a Challenge for Ore Processing Simulation. Proceedings EUROSIM’95 Conference. 861-866. Broussaud, A. 1988. Advanced Computer Methods for Mineral Processing: Their Functions and Potential impact on Engineering Practices. Proceedings XVIth International Mineral Processing Congress. 17-44. CCzac, P., X. M. Truong-Meyer, X. Joulia, S. Brochot, and D. Morin. 1999. A New Modelling Approach of Bioleaching Process. CDROM of the ECCE2, Second European Congress of Chemical Engineering. Durance, M.-V., J.-C. Guillaneau, J. Villeneuve, G. Foumiguet, and S. Brochot. 1993. Computer Simulation of Mineral and Hydrometallurgical Processes: USIM PAC 2.0. a Single Software from Design to Optimization. Proceedings International Symposium on Modelling. Simulation and Control of Hydrometallurgical Processes. 109-121. Durance, M.-V., J.-C. Guillaneau, J. Villeneuve, S. Brochot, and G. Fourniguet. 1994. USIM PAC 2 for Windows: advanced simulation of mineral processes. Proceedings 5th International Mineral Processing Symposium. 539-547. Fleming, C.A., M.J. Nicol, and D.I. Nicol. 1980. The optimization of a carbon in pulp adsorption circuit based on the kinetics of extraction of aurocyanide by activated carbon. Ion Exchange and Solvent Extraction in Mineral Processing Meeting. Mintek Randburg. South Africa. Feb. Guillaneau, J.-C., J. Villeneuve, S. Brochot, M.-V. Durance, and G. Fourniguet. 1995. The Supervisor of Simulation: a step further to meet the Process Engineer Needs. Proceedings XIXth International Mineral Processing Congress. Guillaneau, J.-C., J. Villeneuve, M.-V. Durance, S. Brochot, G. Foumiguet, and H. Durand. 1997. A range of Software for Process Analysis. SME Annual Meeting. Preprint # 97-202. Guillaneau, J.-C., S. Brochot, M.-V. Durance, J. Villeneuve, G. Foumiguet, H. Varine, K. Sandvik, and M. Reuter. 1999. From mineral processing to waste treatment: an open-mind process simulator. CIM 1999. Le Guirriec, E., S. Brochot, and M. Bergounioux. 1995. An Augmented Lagrangian Method for Problems Arising in Mineral Processing. Proceedings 17th IFIP TC7 Conference on System Modelling and Optimization. Vol. 1,65-68. McLaughlin, J., and G.E. Agar. 1991. Development and Application of a First Order Rate Equation for Modelling the Dissolution of Gold in Cyanide Solution. Minerals Engineering. 4: 12:1305-1314. Mular, A.L. 1982. Mining and Mineral Processing Equipment Costs and Preliminary Capital Cost Estimations. Canadian Institute of Mining and Metallurgy. Plitt, L.R. 1976. A mathematical model of the hydrocyclone classifier. CIM Bulletin. Dec. 1976. Reuter, M.A. 1998. The Simulation of Industrial Ecosystems. Minerals Engineering. 11:10:891-918. Rowland, C.A., and D.M. Kjos. 1978. Rod and Ball Mills, chapter 12. A.L. Mular and R.B. Bhappu ed., Mineral Processing Plant Design. SME. Pp. 239-278. Sandvik, K.L., J. Villeneuve, M.-V. Durance, and H. VCdrine. 1999. Development of a Mineral Processing Program as a tool for optimal decision in Waste Treatment. Proceedings REWAS’99, Global Symposium on Recycling, Waste Treatment and Clean Technology. Vol. 1, 55-64. Villeneuve, J., J.-C. Guillaneau, R. Pilotte, and A. Broussaud. 1992. Objective Driven Simulation: a new Approach to Improving the Efficiency and Usefulness of Steady-state Simulators of Mineral Processing Plants. SME Annual Meeting. Preprint # 92- 168.

494

Emergence of HFS as a Design Tool in Mineral Processing John A. Herbsr’ and Lawrence K. Nordell’

ABSTRACT High fidelity simulation (HFS) involves the use of detailed multiphysics models to describe device behavior. In this paper the use of discrete element methods @EM) for granular mechanics, computational fluid dynamics (CFD) for slurry flow, discrete grain breakage (DGB) for particle breakage and population balance models (PBM) for size distribution accounting are explored for comminution equipment design and performance optimization. Applications to SAG milling, ball milling, and crushing and screening are described. Insights into scale-up relationships are highlighted. INTRODUCTION Selecting appropriate pieces of equipment, accurately s u i g equipment for a given task and optimizing its performance have been amongst the most important tasks of equipment designers and mineral processors for hundreds of years. Not surprisingly each of these activities has had largely empirical roots. Once a concept was developed a prototype was built and tested. If the device was successful commercially then gradually a data base was built up for different ores, different sizes of the equipment and operating variable levels. Beginning about 100 years ago mathematical models found their way into the design and optimization process. These early models were found to be especially valuable in the equipment sizing aspect of design. As time has gone on models have become increasingly more important in all aspects of design including concept development and testing, equipment selection and sizing as well as optimization. “Theoretical models” of the late 1800’s and early 1900’s were generally single parameter models supported by relatively little confirming data. By the mid 1900’s efforts like that of F.C. Bond which were based almost exclusively on empirical observations, dominated the design scene (Bond, 1952; Rowland 62 Kjos, 1978). A few decades later another attempt at theoretical modeling (in this case based on conservation equations) was made which allowed property distributions of particles (multiparameter) to be included in mineral processing operations performance (Hulburt and Katz, 1964; Randolph and Larson, 197 1). These population balance models are phenomenological in nature, that is, they get their form from theory but model constants are derived from experimental measurements. This type of model has been successfully used for both design (Austin, 1973; Herbst and Fuerstenau, 1980) and optimization (Martinovic et al, 1990; Samskog et al, 1990). In spite of their practical success these models failed to provide significant insights into design based on the physics of the process or the equipment. To achieve such insights models were needed which separate machine contributions fiom material property contributions to rate processes occuning in equipment. Beginning in the 1980’s microscale models began to emerge which considered material characterizationand internal mechanics as separate steps (Cho, 1987; Herbst and Lo, 1992; SchUnert, 1995).

’ Metso Minerals Optimization Services, Kealakekua, Hawaii

’Conveyor Dynamics, Inc., Bellingham, Washington

495

Today, we are getting closer and closer to “first principles” models of mineral processing operations. This is occurring through the use of high fidelity simulation (HFS) models which predict particle mechanics, slurry flow and ore particle breakage in a fundamental, physics based manner (Mishra and Rajamani, 1992; Cleary, 1998; Nordell et al, 2001). The potential of HFS models for design is just beginning to be realized. This paper explores a wide range of possibilities.

HFS TOOLS High fidelity simulation tools of value for mineral processing design include discrete element modeling (DEM), computational fluid dynamics (CFD), discrete grain breakage (DGB) and population balance modeling (PBM). DEM simulations focus on discrete “particles” by solving Newton’s Second Law of motion applied to a particle of mass m, moving with velocity v, when it is acted upon by a collection of forcesJ, including gravitational forces and particle-particle, particle-fluid and particle boundary interactive forces, i.e.,

If particle motion is confined to two directions the simulation is referred to as 2D-DEM; if full three directional movement is allowed the simulation is referred to as 3D-DEM. For mineral processing design applications the “particles” are generally ore particles, grinding media pieces or bubbles. Constitutive equations can be provided for interactive forces, energy dissipation, wear and breakage. At the present time the most advanced code allows the behavior of up to 1,000,000 “particles” to be solved in 3D simulations with a moderate-to-high powered, multi-processor PC in a reasonable time h e . This is enough detail to allow a realistic simulation of an entire large 3D-SAG mill such as the one shown in Figure 1. For the first time in history the mineral processing equipment designer has an opportunity to see what is going on inside of equipment and has a chance to improve designs to achieve better performance without building the equipment and modifying it in a trial and error fashion.

Figure 1. Photo of 11.4 x 8.5 m SAG mill and snapshot of associated particlehall motion DEM simulation.

496

CFD simulations focus on continuous flow behavior of fluids and slurries modeled as pseudofluids by solving the full Navier Stokes Equation, i.e.,

Dv DZ

(A)

v i m + - f;:

p--=-VP+?p 2

at any point in the continuous phase x,y,z. The last term is a fluid-particle interaction term which accounts for losses resulting fiom mutual interactions (Tsuji et al, 1995). CFD codes differ significantly depending on how particle-fluid interactions and fiee surface conditions are modeled. Today a realistic full 3D mill simulation of particle and slurry motion is achievable as shown in Figure 2. The addition of slurry basically doubles the computation time for a DEM simulation of "particles" alone.

Figure 2. Snapshot of CFD/DEM simulation of partirle/ball and slurry motion in 11.4 x 8.5 m SAG mill.

DGB simulations focus on discrete particles in the same way that DEM does except in this case each physical particle is made up of discrete grains into which strain energy can be storedreleased and cracks can propagate along their boundaries governed by the energy conservation equation which governs crack extension force, G, i.e.,

where t( is the stored strain energy around the crack, a is the crack length and t is the crack width.

497

Computer codes for breakage differ in allowable grain shapes and grain numbers. With efficient code, a realistic simulation of 3D breakage events such as that shown in Figure 3 can be completed in 24 to 48 hours. The computation time explodes exponentially as the number of progeny created increases.

Figure 3. Snapshot of DGB/DEM simulntion of n ball breaking particles.

PBM simulations focus on particles of a given type and how their numbers change as a result of discrete and continuousevents. The governing number conservation equation is:

where I+I is number of particles per unit volume with properties in a defined range, D and B are the death rate and birth rate for particles in the defined range while xl and vl are the spatial and internal property variables and the rate of change of these variables, respectively. This tool, although non-stochastic, allows one to track the average behavior of any number of particles. The challenge then becomes coupling of DEM/CFD/DGB and population balance models. This coupling is the key to turning otherwise impossibly large (in particle numbers) scale-up problems into mathematically tractable problems with acceptable time horizons for solution.

USE OF HFS TOOLS TO ACCURATELY PREDICT THE EFFECT OF SCALE In this section four examples of the application of HFS tools for scale-up design will be presented. The first example will show that DEM simulations of large and small ball mills coupled with microscale selection function values allow the accurate prediction of full scale performance fiom batch tests. The second will show that DEM simulations can be coupled to population balance models through single particle impact breakage tests to predict SAG mill performance. The third will show that coupled breakage and DEM motion provides necessary tools for crusher scale-up. Finally the fourth example shows that DEM simulations alone provide a first principles look at screen efficiency changes with changes in screen surface.

498

Example 1 - Ball Milling In order to test the accuracy of HFS scale-up calculations for ball mills, a batch mill grinding test was conducted by inputting a fixed amount of energy into a -1Omesh iron ore concentrate in a 10 x 11.5 inch energy monitored mill (see Figure 4, left). This test yielded the ground product size distribution also shown in Figure 4 (right).

100 S h . mkmra

10

Figure 4. Energy monitored batch grinding mill and ground product size distribution for iron ore concentrate.

In addition, a DEM simulation of the laboratory batch mill was performed for the same conditions of ball size, ball load, particle load, water addition and speed as used in the batch test.

Subsequently, microscale breakage rate parameters were estimated f?om the experimental batch data and the DEM energy spectra (Herbst and Lo, 1992; Herbst, 2002). The result is shown in Figure 5. r

g

1oO0.00100.00 10.00 1.000.10 -

-'

Y

Y

s

0.00 i 10.0

100.0

1oO0.0

1oooo.o

Size, microns

Figure 5. Snapshot of DEM simulation of energy monitored batch mill also shown is microscale selection function.

Following this step, a DEM simulation of charge motion in a 13.5 x 26.0 ft ball mill was performed and the impact energy dissipation for this mill was calculated as shown in Figure 6.

499

Figure 6. Snapshot of DEM simulation of 13.5 x 26 ft ball mill also shown is a plot of impact energy delivered to balls and particles versus size.

The predicted impact energy dissipation was combined with the estimated microscale parameters, to calculate a macroscale selection function which were in turn used to perform the PBM simulations for the large mill shown in Figure 7 (Herbst and Pate, 2000; Herbst, 2002).

'10

Figure 7. Comparison for measured and predicted product size distributions for 13.5 x 26.0 ft ball mill.

As shown in the figure the product size distribution is accurately predicted by this procedure. Further the power draw and feedrate predicted of 1999 kW and 106 MTPH are extremely close to the measured values of 2004 kW and 106 MTPH respectively. These results provide important validation for HFS ball mill scale-up procedures.

Example 2 - SAG Milling In order to test HFS predictions for SAG mills, impact breakage tests were conducted at various energy inputs. This test yielded the product size distributionsshown in Figure 8 (right)

500

Tost EZ Tost E3 1 0.01

1

Figure 8. Characterization of breakage behavior for SAG milling through impact tests.

The mill performance to be predicted is 11.4 x 8.5 m with a 20% ball charge and 12% ore charge. In this case the calculation of full scale selection functions involved simulating the motion in the full mill and determining the impact energy spectra as shown in Figure 9.

XI

g f

15

P

i

‘O

B f

s

0 0.001

0.01

0.1

1

10

im

ioa,

II

m

Figure 9. Snapshot of DEM simulation of ball and particle motion in a 0.5 m wide slice in the center of a 11.4 x 8.5 m SAG mill also shown is plot of impact energy spectra derived for different ball sizes (B) and particle sizes (P) from simulations.

As shown in Figure 10, in the full mill, power per unit of axial length and calculated selection function vary in a systematic way down the length. It is significant that the average power draw is calculated to be 16.9 M W while plant data yields values between 16.4 and 17.5 MW.

50 1

Figure 10. Calculated power draw and selection function as they depend on position in mill (slice 1 = feed cone, slice 17 = exit cone.

As linerllifters in the mill wear the average selection function changes and the associated throughput changes as shown in Figure 1 1. Here we see that the predicted throughput follows closely the experimental data extracted at the plant providing additional model validation.

End

0

200

100

400

300

500

Llner Life (Days)

Figure 11. Prediction of throughput over the wear life of a liner set

Of course with the help of these tools the effect of mill size on power draw and throughput can be readily evaluated. Table 1 shows predicted power and throughput for the 11.4 x 8.5 m mill and a 9.75 x 7.27 m mill operated at the same hction of critical speed and percent filling. Also shown are modified Rowland and Kjos estimates of power (Herbst et al, 1993). The empirical estimates are seen to be about right, but the DEM results which include the influence of liners/lifters and mill discharge design are found to be more accurate. Table 1. Power draw and throughput predictions for SAG mills

Diameter

Length

RPM

Ballhad

Rockbad

1 1.40 9.75

850

9.54 12.1

20 20

12 12

Power,

Power, Modified R&K, MW 19.3 13.1 ~~

7.27

502

19.3 12.1

-

Example 3 Primary Crushing Primary crusher design provides an opportunityto increase breakage efficiency by controlling the way forces are applied to particles and to reduce wear by controlling particle flow along breaking surfaces. It is now possible to model concave and mantle profiles and their effect on breakage and flow as illustrated in Figure 12.

Figure 12. Model building for primary crusher.

In addition the influence of crusher size on performance can also be evaluated through 2D and 3D simulationsas shown in Figure 13 and Table 2.

-0.271 Kwm/

1

10 0.1

Sla/W.x. Slt.

Figure 13. Predicted product size from small and large crushers.

503

Table 2. Predicted throughput and power of small and large crusher.

MTPWm Powerim

Small Crusher

Large Crusher

1.24 289

1.42 386

With proper hgmentation and wear characterization,it will be possible to predict throughput over the life of crusher liners increasing the power of these models considerably over that of their empirical counterparts.

Example 4 - Screening Screen models of the past have been based on simple probability models and empirical models involving design variables (Matthews et al, 1985). DEM provides an opportunity to calculate screen performance as a function of design and operating variables from “first principles”. Figure 14 shows snapshots of particle motion down the surface of a vibrating screen.

Chute

Screen

Figure 14. Snapshots of vibrating screen simulation - full view and zoom.

From this type of simulation screen efficiency can be calculated for various lengths of screen

as shown in Figure 15.

100

E

(II

8

80

v)

al

60

E al

B

40

I-

20

.Y

0

s

0

0

5

10 Size, mm

Figure 15. Screen efficiency curves for different lengths of screen.

504

15

20

With this model it is possible to look at the effect of chute design, screen dimensions and angle, feedrate, particle shape, screen media type,hole shape screen loading, screen open area and vibration modes. It is expected that soon details of wear will also be accurately predicted. These developments represent significantadvances over the currently used empirical design equations.

A STRUCTURE FOR FORMALIZING HFS DESIGN PROCEDURES HFS design procedures are more complicatedthan current procedures because the number of tools

that need to be managed is large (and variable) and the computationaleffort is significantly greater. In any case, an overall methodologv for using these tools is emerging as follows: 1.

2. 3. 4. 5.

Perform materials characterizationtests. Choose appropriate characterizationmethods, e.g., breakage impact tests, batch tests, pilot tests or full scale tests; while wear resistance can be found by fundamental wear tests, abrasion index tests or full scale tests. Run experiments. Perform high fidelity simulation of characterizationtest(s). Run HFS for test conditions. Calculate microscaleparameters. Determine the microscale parameters which make the model output match the test output. SpeciQ design requirements. Search for best design Set design parameters - Perform DEM/CFD/DGB/PBM simulation - Checkperformance - Iterate

-

Step 5 is generally the most time consuming step. Obviously one wishes to limit the number of iterations. Our experience has been that classical design relationships provide a good place to start. A good starting point can result in convergence in one of two iterations. Virtually all of the initial work that has been done with HFS is for comminution devices. Extension to other mineral processing operations is also very promising. Model building and validation should be explored for flotation, gravity concentration, magnetic separation etc. The key to making HFS work in these complex systems is having high speed computers and efficient computer code. For the actual design process, computationally efficient search algorithms will be required.

CONCLUSIONS During the last year or two a new set of equipment design tools has begun to emerge. These High Fidelity Simulation tools are based on highly multi-physics models for particle mechanics (DEM), fluid motion (CFD), particle breakage (DGB) and overall particle accounting (PBM). Rigorous validation exerciseshave shown that these models are beginning to provide realistic simulations of equipment performance for crushers, tumbling mills and size separators. In this paper, the ability of the models to accurately predict the effects of scale was demonstrated. In addition, a kunework for formalizing HFS design procedures was offered. Finally some of the challenges to making these procedures routine were presented. ACKNOWLEDGEMENTS The authors wish to thank Dr. Xiangjun Qiu, Dr.Alexander Potapov, Dr. Ming Song and Dr. William Pate for performing the high fidelity simulationspresented here.

505

REFERENCES Austin, L.G., 1973, “Understanding Ball Mill Sizing’’, Ind. Eng. Chem. Process Design and Development, v. 12, No 2, p. 121. Bond, F.C., 1952, “The Third Theory of Comminution” Mining Engineering, May, p 484. Cho, K., 1987, “Breakage Mechanisms in Size Reduction” PhD Thesis, University of Utah. C l e w , P.W., 1998, “Predicting Charge Motion, Power Draw, Segregation Wear and Particle Breakage in Ball Mills Using Discrete Element Methods” Minerals Engineering 1 1, p. 1061. Herbst, J.A. and Fuerstenau, D.W., 1980, “Scale-up Procedure for Conthous Grinding Mill Design Using Population Balance Models”, International Journal of Mineral Processing, v. 7, No l , p . 1. Herbst, J.A. and Lo, Y.C., 1992, “Microscale Comminution Studies for Ball Mill Modeling”, Comminution Theory and Practice, Ed. S.K. Kawatra, SME ,Littleton, CO, p. 137. Herbst, J.A. and Pate, W.T., 2000, “Dynamic Flowsheet Simulation - A New Tool For Mine Through Mill Optimization” IMPC, Rome. Herbst, J.A., 2002, “A Microscale Look at Tumbling Mill Scale-up using High Fidelity Simulations”, Proceedings of the X European Comminution Symposium, Heidelberg, Germany. Hulburt, H.M., and Katz, S., 1964, “Some problems in particle technology: a statistical mechanical formulation”, Chem. Eng. Sci., v. 19, p. 574. Martinovic, T.I., Herbst, J.A., and Lo, Y.C., 1990, “Optimization of Regrinding Ball Mills at Carol Pellet Plant of the Iron Ore Company of Canada” Control ’90, SME, Littleton, CO, p. 223. Matthews, C.W., Colman, K.G., and Stavenger, P.L., 1985, “Screening”, SME Mineral Processing Handbook, p. 3E- I . Mishra, B.K. and Rajamani, R.K., 1992, “The Discrete Method for the Simulation of Ball Mills”, Applied Math Modeling, 16, p. 598. Nordell, L., Potapov, A., and Herbst, J., 2001, “Comminution Simulation Using Discrete Element Method Approach - From Single Particle Breakage to Full Scale SAG Mill Operation”, SAG 2001, Vancouver, v. IV,p. 235. Potapov, A. and Campbell, C., 1996, “A Three Dimensional Simulation of Brittle Solid Fracture”, International Journal of Modem Physics, 7 (5), p. 7 17. Randolph, A. and Larson, M., 197 1, Theory of Particulate Processes, Academic Press. Rowland, C.A. and Kjos, D.M., 1978, “Rod and Ball Mills”, Mineral Processing Plant Design, Ed. A. Mular, SME, Littleton, CO, p. 239. Samskog, P, and Bjorkman, J., Herbst, J.A. and Lo, Y.C., 1990, “Optimization of the Grinding Circuit at the Kiruna Plant of LKAB using Modeling and Simulation Techniques”, Control ’90, SME, Littleton, CO, p. 279. Schhert, K., 1995, “Comminution from Theory to Practice”, Proceedings of XIX W C , v.1, p. 7. Tsuji, Y., Kawaguchi, T. and Tanaka, T., 1995, “Discrete Particle Simulation of Two Dimensional Fluidized Bed”, Powder Technology, 77. 506

Reducing Maintenance Costs Using Process and Equipment Event Management Osvaldo A. Bascur and J. Patrick Kennedy

ABSTRACT Mining and metallurgical plants use predictive maintenance policies based on statistical analysis and special techniques, like vibration analysis, and oil and lube analysis for critical equipment. Both process equipment and control strategies require maintenance from local personnel at the mine. Workers get insight from the overall process effectiveness control index based on key variables such as losses due to plant availability, performance efficiency and rate of quality. The availability of real time data and plant equipment and process history enhances the discovery of opportunities for optimization. The Word Wide Web leverages existing computers and networks with these data to provide highly profitable opportunities. This paper describes a plant integration paradigm that can be used to simplify the implementation and integration to ERP (Enterprise Resource Management) systems. Creating an environment where upgrade and change are allowed is becoming an accepted methodology to accelerate and simplify the integration of applications such as production monitoring, equipment performance management, yield and inventory management, and equipment downtime management. This paper introduces the application infrastructure and gives several examples of where it is used, including gross error detection, yield and inventory data unification, process and equipment event tracking, and asset downtime analysis. Three case studies in large metallurgical plants are presented. Keywords: Integrated Mine, Concentrator, Smelter and Refineries Management, Performance Monitoring, Client Server Computer Architecture, Process Control Monitoring, Asset Management, Reliability Maintenance, Continuous Improvement and Innovation, Plant Coordination Work Flow.

WHERE ARE THE OPPORTUNITIES? Increasing competition in the global market place has forced companies to seek new ways to achieve costeffective production. Previous projects emphasized production rather than asset availability and reliability. Cutting maintenance budgets was often seen as a fast easy way to reduce production costs, but it is a shortterm strategy. An integrated availability and reliability system enables organizations to coordinate and analyze large volumes of data. Existing maintenance and process control, cost, and reliability data can be integrated to point the way to maximum productivity with real, quantifiable saving. In the U.S. alone, process industries spend $70 billion annually on maintenance. This amounts to more than 9% of the cost of goods sold in those plants. Preventive maintenance combined with reliability analysis provides large opportunities for simultaneous cost reduction and productivity improvements. Until now, predictive maintenance has been a good concept that could not really be implemented because of low investment and elementary tools. Future systems must emphasize repairing root causes of failures rather than just the results. This provides greater equipment availability while simultaneously improving product quality. The relationship between losses and equipment effectiveness parallels production quality and equipment availability. The Overall Production Effectiveness (OPE) is a measurement of equipment and process productivity. OPE is part of a total productive manufacturing (TPM) methodology to improve equipment OSI Software, Inc. Houston, TX and San Leandro, CA.

507

productivity. The elements of TPM strategy include cross functional teams to improve equipment uptime, rigorous preventive maintenance programs, improved maintenance-operations management efficiency, better process and equipment training, and providing information to help develop better next generation equipment. There are developments in both extending both process capabilities as well as equipment availability. Equipment problems need to be identified quickly; only then can they be solved. Most companies already have reams of data about their equipment coming from: Maintenance Management systems Cost Systems Predictive Systems Production Systems Manufacturer specifications and reliability data What is needed is not necessarily more data (the problem is there is too much of it), but an environment that simplifies integration of the data with tools available to understand and analyze it. Asset optimization seeks improved operating practices through the use of process analysis and diagnostic monitoring to notify operations and maintenance system of quality deviations and to permit further improvements. Asset optimization involves the manipulation of real time process and equipment status to improve performance, equipment availability and overall process effectiveness. The strategies behind predictive maintenance and asset optimization have been around for more than a decade. The major drawbacks have been traditional management of the information collected and the limits of the typical plant organization. Usually, independent functions and islands of automation have precluded their implementation. The next section reviews the advances in technology that enable a simplified environment to close the loop. The closing of the loop is at the industrial desktop within an application framework. This graphical, adaptable user environment promotes continuous improvement, provides tools and facilities to help the user analyze and make discoveries about the plant and business processes, and most importantly, helps the user to implement the findings. In a nutshell, it promotes both continuous improvement and innovation. (Bascur and Kennedy, 1995,1998).

EQUIPMENT DEGRADATION AND FAILURE PATTERNS As the performance of the belt conveyors, crushers, mills, pumps, compressors, thickeners, and other mineral processing equipment deteriorates over time, equipment efficiency decreases, power consumption goes up, throughput is reduced and operating costs rise. Process and equipment performance event monitoring using desktop computers and remote access via the Internet offers an effective way to offset that toll. Condition based monitoring has been decoupled in six different patterns (Moubray, 1997). Studies revealed that 68% of all failures were attributed to pattern F, while only 5% of the failures were attributed to pattern B. Pattern F is defined as: at the start, high rate of failure, and then consistent operation with very slow increase of conditional based failure. Pattern B is defined as: at the start, stable, followed by continued stability, then a wear-out period. Moubray feels this study is substantial enough to translate across industries. Djuric, 2002, has shown that a new failure paradigm can be used for conditioned based equipment monitoring. Based on the asset condition curve (or equipment degradation in time), he defines several stages. A failure condition starts early in the curve. Between this early start and the old definition (equipment broken) a time interval between Potential Failure and Functionally failed (P-F interval) is defined. The new definition of failure is based on equipment not performing the intended function.

508

Customers

support

Mine

Mill Port Others Shiff Reports, Produdion Reports, Recovery, Grades, Costing Plant Availability, Plant Reliability Event Management, Mine Status Port Status Business Status

DCS Data Samples, Assays, Flows, Weights, Event Management, Stoppages. Maintenance Orders Consumables,Shiff Reports, Mine Status Port status Business Status

Event Management DCS Data Samples, Assays, Flows, Weights, Consumables, Adjustments, Metal Balances, Shifl Reports, Mine Status Port Status . Business Status ,

Work Flow Server

PI-Host Server SQL Server

WEB Server

Figure 1 Empowering people to access their information according to their roles and responsibilities. This change in paradigm calls on a more general approach to use the process variables and equipment condition based data. There are many ways to transform the same data into information. Typical maintenance data call for lubrication analysis, vibration analysis, infrared thermography, motor circuit analysis, operator rounds, visual inspections, and mechanical inspections. The process and equipment operating condition events become the most important set of information available for process analysis and diagnosis. Real time process and equipment information can be used to facilitate the transformation of data into actionable information. This information may lead to improved maintenance cycles, fault detection and rectification, and process improvements identified through performance trends. The ability to use the process plant topology to store equipment connectivity information and the use of equipment templates to store the equipment attributes, and the links to detailed asset information from the manufacturer can connect to real time process information to continuously monitor the asset under process conditions. Figure 1 shows the different functions in a plant with the different types of information requirements. Users have the flexibility to adapt their desktop according to their roles and responsibilities. The transformation of information into wealth means that more members of the firm must be given opportunities to know more and do more. A structured performance monitoring approach allows for the accurate comparison of the actual versus design conditions, and improved decision-making based on derived performance and economic data.

The major requirements for such a change in culture are:

-

-

A universal data server to collect and archive the information from laboratory systems, maintenance systems and control systems. A common application framework environment to access the real time data according to context.

509

-

A common view to access the integrated information by operation, maintenance, engineering and management.

The common view is divided amongst powerful desktops on the LAN for process troubleshooting, diagnosis, configuration and maintenance. An Internet real time application for viewing and remote access should also be provided.

AN APPLICATION FRAMEWORK AND THE INDUSTRIAL DESKTOP Process systems and the procedures they represent are often a major problem when a company tries to become more responsive. The cost to change these systems is great in terms of both money and effort. Over the past two years, the philosophy for the design and construction of new systems and upgrades to existing systems has changed dramatically. A term that is becoming more familiar in the computer software industry is “Three Tier Architecture”. The three-tier architecture is a logical, not physical, computer architecture that was a key software development, without which large client/server systems would not be possible. The client server architecture may access data and methods from dissimilar sources in order to provide these applications, but it is the three-tier architecture that addresses how one can maintain clientherver systems and permit them to operate reliably in a distributed environment. In industry, we are mostly concerned about real time systems. Most processes and equipment systems have measurements, which are stored in process historians. Fault detection and root cause analysis can only be performed if this information is available at their original resolution. Most companies have realized that they need to roll out a modem integrated desktop for the office workers including word processing, spreadsheet and communication software. The industrial desktop (Kennedy, 1996). because it is proximate to a forbidden, proprietary place, is often neglected even though it supports the people who operate, monitor and improve industrial facilities. Its users have a wide range of formal and job related skills not found in the office; they work in a harsh and sometimes dangerous environment, they work on rotating shifts, and they have the same basic requirements as the office workers plus many plant related needs. The industrial desktop has a dramatic effect on productivity, cost and yields of the plants, many of which are operated 7 days a week, 24 hours a day. It is clear that large mega projects for productivity improvement are short sighted because they are based on an individual FWQ and not the aggregation of many software standards, which are the bulk of the technology. Instead, the only enduring solution to challenge constant change lies in the deployment of a real time information system that can adapt and change as fast as the organization it supports. This is a radical departure from the time-honored practice of developing new applications from scratch to meet new business requirements. It requires competitively focused companies to deploy software systems of sufficient flexibility and scalability that can be quickly modified in response to new opportunities and easily allow, building, integrating, and discarding applications as required. The PI Application Framework (PI-AF) is a major step towards the consolidation of the PI system as a plant infrastructure technology for operational related information. The application framework isolates the client applications from the data, allowing the continuous improvement of the underlying technologies and embedded know how. The application framework integrates the process data, the “Application Models” (plant, physical, organizational, etc.), and the scheduling of calculations and procedures, providing a scalable and evolving infrastructure for the implementation of the client applications. Plant models are time referenced, including specific business rules as defined from the “Cases”. This infrastructure for the operational data management must go from a strategic decision a “Package that solves a set of identified needs” toward an infrastructure where the needs are implemented, while allowing the addition of new applications, plug-ins and modules.

510

The PI Application Framework is built upon Microsoft.NET. Not only does this provide the means to build distributed applications that communicate as Web Services, it also provides standard tools for "wrapping" existing code until rewritten, support for all languages, integrated memory management, garbage collection, security and debugging in a multi-server, multi-client market.

PI-AF supports software where the user is expected to graphically construct a model of something (their plant, their organization, etc.) with standard "objects" that represent connected elements in the model. These elements are stored in the PI-ModuleDB - a standard package from OSIsoft for structuring modules for viewing plant data. Once these models are created, the user can pick any number of compiled plug-ins to apply to that model, and choose output templates for reporting. PI-AF allows third party developers (who may or may not be exposed to .NET) to extend this tool with their own plug-ins. PI-AF allows users to abstract the PI database into a maintainable structure that is meaningful to them in a reusable manner. It allows users and vendors to automate calculations, reports, etc. Visualization is as important as the server based rules and data methods. PI-ICE provides a package for creating real time Browser based displays. Combined with PI-ProcessBook and PI-DataLink, this allows all users access via graphics, spreadsheets or a Web Browser - the only three kinds of software that can be used today with no training. PI-ICE is also based on the new Web technology - SVG displays driven by SOAP objects. Figure 6 shows a web browser with the real time performance status of a mineral processing complex.

Operational System Overvil

Port Conc Receivable &Storage Ship Loading Port Operations and Delays

Mine

Block Advanced Tomes Broken Ore Hoisted

Tomes TNcked Ivhe Operations Delays POWK& Water Consumablm Usage

Operations Mgt System

Figure 2 Universal data server and application framework schematic Figure 2 show a typical mineral processing plant block diagram receiving material from the mine, followed by the concentrator, sending the concentrates via a pipeline to a filtration plant and port facilities for shipping via trucks or ships. Further processing of the concentrates will be done in the enterprise smelters or sold to other independent smelters. The bubbles in Figure 2 show the typical type of information requirements for plant operation decision-making. At the bottom of Figure 2, shown as cylinders, are all the databases for storing information such as a laboratory systems, a mine system, a DCS system and the ERP systems. On top of these systems, a universal data server integrates and collects process, laboratory,

51 1

inventories data and events. The PI-AF schematic foundation and additional applications can coexist within the plant information system.

Framework to rethink how to visualize and analyze operational information

Transactions Spatial and Functional Structures Eouiument Locations Geography Functions Accounts Hierarchies

PI ACE

Temporal Orgauization Fixed or Variant

PI SDK

PI Module DB

Others

PI SDK

Sigmafine PI AF

Other Apps

)I-SDK,pI-fi

..

RLINK

Events Lots campaigns

PI Archive

RDBMS Web Server

Process

Figure 3 Real time information management infrastructure. Figure 3 shows the universal data server. The PI-AF allows data access and data transfer between applications that are shared between the different plants or provided by others. PI-SDK and PI AF-SDK exposes the data, organization models and common calculations and procedures to client applications, Plug-In modules and external applications. Additionally, OLE DB also exposes the data and data organization as a “SQL linked server”, providing another standard way for “plug-in” to external applications. A strategic decision involves the identified set of needs in addition to the framework to allow the system evolution, as new technologies arise, for the support of business and operational needs. The increasing availability of “Web-Services” (and the increasing availability of communication bandwidth) will completely change the type and scope of applications, allowing reuse between sites, support from third parties, even those that are remotely located. From a strategic perspective, the deployment of an infrastructure for managing the operational data is a completely different scenario than the one based on applications based on a core data management system (i.e. a RDB). This infrastructure allows for new implementations as the needs arise and for the implementation of “Organization Models”, which can have access to Plug-Ins, Modules and Applications. To build the plant connectivity the basic concepts are the typical block and the process flow diagrams guidelines used in economic process evaluations, process design and process analysis. The PI-AFcaptures the plant topology and the links to the plant data according to the objectives of the application.

512

Block flow diagrams (BFD) The block diagram is introduced early in the education of process engineers (Turton et al, 1998). In early courses in material and energy balances, often the initial step is to convert a word problem into a simple visual block flow diagram. This diagram is a series of blocks connected with input and output flow streams. It includes the operating conditions (flow, temperature and pressure) and other important information such as conversion and recovery. It does not provide details regarding what is involved with the blocks, but concentrates on the main flow of streams through the process. According to Turton, the block flow diagram can take one of two forms. First, block flow diagram maybe drawn for a single process. Alternatively, a block flow diagram may be drawn for a complete plant complex involving many different metallurgical or chemical processes. We differentiate between these two types of diagrams by calling the first a block flow process diagram and a second a block flow plant diagram. Figure 2 shows a block flow diagram for the mineral processing plant including the mine, mill, pipeline and port facilities.

Table 1 Conventions and format recommended for laying out a Block Flow Process Diagram 1. Operations shown by blocks 2. Major flow lines shown with arrows giving direction of flow 3. Flow goes from left to right whenever possible 4. Light stream (gases) toward top with heavy stream (liquids and solids) toward bottom 5. Critical information unique to process supplied 6. If lines cross, then the horizontal line is continuous and the vertical line is broken. 7. Simplified material balance provided Process flow diagram (PFD) The process flow diagram (PFD) represents a quantum step up from the block flow diagram in terms of the amount of information that it contains. The PFD contains the bulk of the chemical engineering data necessary for the design of a metallurgical or chemical process. For plant information management the objective is to capture the necessary information for performance monitoring of the process (production quality, equipment reliability and environmental monitoring). The process information can be used to predict quality variables or to define quality of asset alerts when non-conformance to target plans. For all of the diagrams discussed in this paper, there are no universally accepted standards. The PFD from one company will probably contain slightly different information than the PFD for the same process from another company. Having made this point, it is fair to say that most PFDs convey very similar information. A typical commercial PFD will contain the following information: 1. All the major pieces of equipment in the process will be represented on the diagram along with a description of the equipment. Each piece of equipment is assigned a unique equipment number and a descriptive name. 2. All process flow streams is shown and identified by a number. A description of the process conditions and chemical compositions of each stream will be included. These data are displayed either directly on the PFD or included in an accompanying flow summary table.

The same principles used to draw Block Flow Diagram (BFD) or Process Flow Diagram (PFD) can be used to develop powerful information linked to real time and historical data. The elements making up the Process Flow Diagram are created from templates whose additional properties and attributes can have data references to external databases. For example, the templates attributes can store the equipment numbers, and descriptions. Figure 4 and 5 shows the modeling tools and a view of the PI-Module DB.

513

Figure 4. The application framework using PI Modeler Add-in. Let’s take the block diagram in Figure 2 and define a process flow diagram for the mineral processing plant using the application framework to implement a metallurgical accounting subsystem using the Sigmafine plug in. Figure 4 shows the model of the process flow diagram represented graphically by the tools in Process Book. The display is composed of several parts. The top left allows connection to different servers. The bottom left displays the templates and elements that have been defined in the PI AF that are available for creation of a connectivity model representing the process flow diagram. The center portion is the connectivity model graphical representation. The top right displays the connectivity model in a tree structure. The bottom right shows the properties of the module the user has selected in the tree view. The properties window content changes based on which module the user selects.

Figure 5 PI Module Database Editor showing the properties for the Port Site Module.

514

The elements in the graphical view represent the plant connectivity and are necessary to define the mass balance constraints required to solve the data reconciliation problem. All inventories, flows, and stream compositions are connected using the process flow sheet tools. All rules for data collection and the plug in manage are aggregated internally. A set of procedures are configured to close the mass balance and to rectify any gross errors found in the process. Figure 5 presents the PI Module DB editor with the properties associated with the port site module. These properties can be defined individually or by the application framework plug in.

BOTTOM LINE, YIELDS AND INVENTORY MANAGEMENT One of the biggest challenges to process plant management is the accumulation of accurate information on process operations. This information is necessary for any analysis and decision-making within the plant and enterprise. Therefore, there is a requirement for meaningful, accurate and consistent data. Material balances calculated from data measured at various locations around process units, tanks inventories, stockpiles, silos, bins, and assays are useful for many purposes, such as yield accounting, online control, and process optimization (catalyst selections, reagent schemes, liner replacements, water management, utilities management). To achieve material balances, gross errors or anomalies in the production data must first classified, detected, and the source of the data examined. Often, it is possible to calculate material balances by several independent procedures when excess measurement information, i.e., redundant data, is available. Clearly, if the data were collected without measurement errors (a theoretical condition never found in practice) all material balances calculated from redundant data would be in agreement. The real situation is that errors exist in practical measurements, so that the results of material balances determined from available optional procedures differ. Consequently, best-fit computational procedures to adjust the material balances by taking measurement errors into account can be implemented (Bascur and Soudek, 2001a, 2001b). In this case, the application framework is used to develop a process flow diagram connecting flow meters, tank inventories, stockpiles and composition analysis for the defined streams. A plug-in for data reconciliation and gross error is used to perform the calculations. Figure 4 shows the process flow diagram a mineral processing plant. Once the process topology is defined, the Sigmafine plug in can be used to reconcile the data from inventories, flows, and compositions. Figure 6 shows the results in a WEB environment for access by management, personnel and external resources. A thin client called PI Internet Configurable Environment (PI-ICE) enables any one with a browser and a security password to access the information in real time.

In addition to the unified yields and inventories, the total variable costs associated with the processing of the ore for a certain block can also be included. The real time information will access the associated consumables during the period of time when this ore type was processed. Analysis of the metallurgical performance can be performed linking the graddrecovery with the grindinghlast strategies. At the same time equipment downtime and equipment availability can be incorporated for the assets. Real time based costing emerges as a reporting exercise when the proper application framework for real time information management is used. This integrated approach enables collaboration between operations, engineering, accounting and management to drive the organization’s bottom line according to their business strategy. At the same, personnel can look for opportunities using alternative processing methods and strategies (grinding efficiency, reagents, blasting methods) to adapt to the changes in ore type to produce the least cost concentrates.

515

Figure 6 Web browser showing real time performance indicators such as yields, inventories, asset availability.

PROCESS AND EQUIPMENT EVENT MONITORING The PI-AF extends the use of the traditional event management subsystem. The most common structure to build the process and equipment event monitoring includes a main module with the key performance indictors such as ore production, overall equipment availability, and specific power consumption. Submodules are defined for all equipment in the area (the grinding section). For each grinding mill, triggered events are defined (calculated digital sets) for operating conditions (running good, in trouble, overloaded, not running) and equipment conditions (operating, unavailable), operator comments, manual data entry, alarm status and process values. These data and events are captured based on a defined time interval or on a main event such the production of a lot or the movement of certain feed material. As such, all consumables and production records can be reported together with the process operations, laboratory data and any additional information included in the asset modules. Maintenance logs are classified into: 1. - Working at standard rate 2. - Waiting or working at less than standard rate Trigger calculated or measured digital values can be defined to capture in real time information such as: manpower downtime, materials downtime, logistic downtime, environmental downtime andor administrative downtime. Operator logs are classified into: 1 Operable 1.1 Operating

1.1.1 1.1.2

Full Capacity (no downtimes) Rate Delayed (operating less than planned capacity)

1.2 Standby

516

1.2.1 1.2.2 2

Administrative Downtime (equipment not scheduled) Logistic Downtime (no materials or utilities)

Inoperable 2.1 Predictive Maintenance (Scheduled) PM Downtime (equipment down for inspection) 2.1.1 2.1.2 Overhaul Downtime (scheduled repairs) 2.2 Corrective Maintenance (not scheduled) 2.2.1 Chargeable downtime (part wore out) 2.2.2 Unchargeable downtime (part broke)

These real time events are automatically captured by the system. The following calculations are made by the system based on the time intervals of these events; asset availability is defined as the ratio of operable time/all time, asset utilization is the ratio of outputkapacity, failure ratio is chargeableNnchargeable and maintenance delays is the ration of predictive/corrective. Transformations of data such as these add valuable information for asset management decisions. With both events and historical data available, sophisticated joined queries can be resolved rapidly for analysis of critical equipment. Human response is seldom based solely on the current value; we need to classify and look for trends and changes in the patterns. Many of the new events are simple alerts, but adding the batch event and transfer record into a structured environment and combining them with history affords unique views.

Figure 7 An Event based query for an asset showing operating states in a Gantt Chart ActiveX Component The batch event defines a set of coordinated data over a period of time, rather than data at an instance of time. Batch events are a result of rules that are used to define triggers marking the start and stop times (e.g. an outage or production run) and then cause the resulting events to be stored. For example equipment startup can be considered a batch event. Another kind event is the transfer record that includes a source and sink as well as the times. This event is used to define a high level view of material movements that are essential for reconciliation of the flow data. Examples of the data added to the historian are operations within a batch process (charging, cooling, heating, flooding), detected conditions for equipment (cavitations, fouling, high vibration, high amperage, down, running, maintenance, troubleshooting), product quality (on spec, out spec, cycling), movements (receipts, shipments, transfers), exception (process/quality alarms, collections of operator comments, collection of manually entered unmeasured data) and calculations (performance indicators). Modem systems must have much wider scope (e.g. financial and

517

structural information) to provide these views. They also need a sophisticated computational layer to prepare the presentation. Figure 7 shows the results of a query for the unit displayed in the tree view on the left. The tree view contains all the events such as operating conditions, manually entered data, operator comments and the associated sensor data connected to the unit. The operating events for the time interval are shown in a Gantt chart. Availability, downtime and cycle time analysis is performed to check that the standard procedures are consistent with the economic targets. A statistical control chart (SPC) to plot the quality variable to check that the process is under the statistical process control limits or within the normal bounds of the process is shown on the left hand side of the figure. SPC events can be automated to generate process or equipment alerts to notify maintenance or metallurgical operations. Events add information needed for diagnostic actions, but special tools are required to analyze the combination of these events and real time data. F o r m or query systems based on relational databases are simply not up to the task. Tools must allow the non-specialists to get all events and data by product, order numbers, from any unit in the plant without programming or writing SQL queries. A good example of combining financial and plant data is marginal economics where operations can see the margin on each order in real time before it is shipped. The actions, rules and calculations are themselves important data and must be versioned, agreed to and shared by the entire organization.

EQUIPMENT DOWNTIME EXAMPLE Downtime incidents are often not correctly documented by operators. Many times the most painful downtimes are those that are of short duration but occur frequently. By capturing these events, together with a simple selection of the downtime code, valuable statistics can be obtained to take root cause action at the operating or equipment level. Often, operators become used to clearing a system fault quickly and no longer consider downtime associated with 10 or 20 quick fixes to be more than one incident during a shift. These system faults can also be symptoms of other problems. Operators may fail to notice other problems that occur upstream or downstream as a result of starting or stopping a production line. This leads to downtime that is erroneously attributed to a given piece of equipment. In this case, the process flow topology can detect the weak links from an availability perspective and identify the global impact in determining the part replacement, methodology or operation training.

Figure 8 Real time graphical interface and spreadsheet reporting. This example uses the PI-AF for downtime analysis to monitor process events and equipment conditions. The conditions are based on rules to indicate if there has been a violation in the current design practice. It enables the operator to track performance, assign a process event code for the violation by equipment, and

518

enter comments and reason codes by equipment. It supports the process topology structure of the concentrator and incorporates changes made to the data structure without rewriting the calculation rules for downtime. The steps used to develop such an example are: 1. Define the process structure model. 2. Define the element templates to be used in the process, 3. Define the attributes for the type of element (grinding mill power). 4. Define the categories for grouping and querying capability to visualize the model structure. 5 . Define the elements to fill in the specific attributes (tag for grinding mill power, maintenance operator, unit type, unit number, location code in maintenance data base, typical power rating, etc.). 6. Define the analysis plug in (how to calculate the downtime).

The plug in requires the following steps: collect the inputs for the analysis, run the analysis and publish the analysis. In addition, the PI-AF requires a time rule to schedule the downtime calculation. Figure 8 shows a Process Book display for a unit in the plant and the addition of a reason code and operator comments. The spreadsheet shows a simple shift report with all events and reason codes. The structured environment of the PI-AF significantly reduces the programming requirements for the plug in. All the basic templates are available as objects to be reused. The applications support changes to the elements without programming.

CONNECTIVITYTO ERP SYSTEMS The SAP-certified RLINK gateway to W3’s PP-PI, QM and PM modules reduces enterprise integration costs. The result is a standard R/3 configuration that enables process engineers and management to leverage production information. RLINK supports Microsoft and SAP standards to provide organizations with all the information needed to make profitable business decisions: Asset Efficiency-operating to capacity and maximizing equipment uptime with timely, conditionbased maintenance using operating hours, number of starts, and alarms (e.g., temperature, pressure, vibration) to trigger maintenance Increase Profits-plant control systems don’t have access to data required to optimize profits like material costs, energy costs, market demand and prices for finished products Enhance the level of coordination and collaboration between manufacturing, maintenance and logistics business functions Analyze tradeoffs to satisfy business objectives of reducing operational costs and inventory, improving delivery reliability and response time, and service to the customer Cycle Time-reduce time from product order placement to customer delivery Available-To-Promise-provide reliable delivery date to customer (i.e., when order is taken) based on real-time view of finished goods inventory, production plan, raw materials, etc. Reduce magnitude and complexity of production management reporting Overcome problems associated with manual entry: data entry is slow and error prone, manual calculations are often required, limited volume of data that can be handled

R/3 must be notified when problems occur in manufacturing (Stengl, B. and R. Ematinger, 2001). Without a real time connection, R/3 may only recognize a problem when it’s too late (i.e., if product is not shipping). Reacting to problems is inefficient, so critical plant events must be escalated to R/3. As companies move to an E-Business environment integrating suppliers and customers over the web, greater demands will be placed on internal coordination between manufacturing and sales. RLINK enables companies to react to unplanned manufacturing events. Therefore, companies will be required to provide a more timely and accurate view of manufacturing to the ERP system to compete in the E-Business world.

519

PM Module. The R/3 PM (Plant Maintenance) module tracks maintenance history and schedules maintenance activities. PM supports measurement points and counters for each piece of equipment that will be maintained based on actual data from the plant. That equipment is also monitored in PI and RLINK to calculate the runtime (measurement counter) or determine alarm conditions (measurement point) that should be sent to W3. The PI-Alarm subsystem detects that an alarm (e.g., high temperature) has occurred and RLINK passes the alarm to SAP. This creates a measurement document and an optional maintenance notification. The measurement document and the notification number are returned to RLINK and sent to PI for analysis with production data. The information collected in the substation operator’s weekly inspection is a critical component of the maintenance decision-making process. Abnormal values or conditions found during the weekly inspection are used to create a maintenance notification in SAP for follow up via a corrective maintenance work order. Counter readings on transformer tap changers and circuit breakers are used to trigger preventive maintenance activities in SAP. This enables a company to move from a calendar based maintenance program (such as maintain every 4 years) on circuit breakers and tap changers to an operations based maintenance trigger (such as maintain every 10,OOO operations for transformer tap changes). PM notifications are also generated based on rate of change conditions for the tap changers (i.e., too many or not enough in a given period). For circuit breakers, if the quantity of gas added is more than 5 Ibs and less than 10 Ibs per month on average over a 6 month rolling period, then an outage is scheduled to repair the seals. Transformer temperature readings on top oil and hot spot are used (along with other information) to assess the transformers ability to carry rated load.

RELIABILITY ANALYSIS Reliability analysis applications provide more sophisticated statistical and graphical tools for investigating reliability issues. Examples are: standard analytical techniques like Weibull, Lognormal, and Growth Model (Potts, 1996). In many organizations reliability analysis is based only on equipment history and very often is only done on a per equipment basis. For example, a separate analysis might be done on pump A, B and C. If pump A failed 20 times over the observation period, you would consider this to be a sample of 21 cases (assuming that the pump was working at the end of the study). The reliability application doesn’t replace commercial statistical packages by providing all possible analytical tools. But by providing basic statistical analysis capabilities within the application you’ll save time formatting and transferring data to those packages. At the same time, the reliability application provides some functions not available in commercial packages, like Weibull and Growth analysis.

EXTENDED PROCESS CONTROL PERFORMANCE Entech Control reports that 80% of distributed control system loops actually increase process variability as compared with manual control systems (Bialkowski, 1996). The main reasons for this variability are20 8 due to design causes, 30%due to control tuning and another 30%is related to equipment performance. To maintain process controls on line, we need to look at their performance and continuously improve their behavior (Bascur, 1990). We provided some guidelines to achieve high results by providing on-line audits to verify that the business needs still hold, training of the operators, developing a good set of documentation, and a implementing a program for management of change of these advanced controls. The key performance indices of the plant (Mine, Concentrator, Hydrometallurgical plant, Smelter, Refinery, Tailings, Port facility) can be used as global indicators of a set of controllers to achieve the business objectives. The basic idea is to look at the plant as divided up into individual unit operations. Each unit operation will have a key measurement of some type. For a crushing circuit that measurement might be the TPH at a certain grind. In a grinding circuit, it might be the k W t o n s , throughput, or grind. In a flotation circuit, it might be the total metal recovery, tailing losses. Ideally, these key measurements would be on-line instruments, but in many cases, those measurements come from inferential calculations or manual data entry. In either case, a Process Control Monitor will display the key variables and the key

520

variable statistics. Each process unit would also have several control loops that are important in the control of that unit. Theoretically, if the control loops are running well, then they key strategic variables mentioned above should be controlled as well. A process control monitor displays the key variable as a dependent variable and the control loops as independent variables. The key strategic performance variable (i.e. kWh/tons) is plotted as a run chart, The

Maintenance Notification Integration Measurement

Document Details

mco

SAP MM Plan

FLOW

DCS

FLOW

SAP PP Plan

Figure 9 Mine-mill real time information management integration and maintenance notification. unit manager, process operator, or instrument technician can scan this display to view the control performance for yesterday, last week, or last month. The quality statistics can also be compared with similar processes at other locations. This simple application ensures both control availability and production efficiency and maintains rate of quality. These control actuators need to be monitored all the time for any control strategy to work well. Other traditional problems in mineral processing are final control measurements such as density gauges, on line particle size analyses and on line chemical analyzers. The calibration and maintenance of these measurements are facilitated using these tools. REAL TIME INFORMATION MANAGEMENT INFRASTRUCTURE CASE STUDIES There are many examples where the requirements for process, laboratory, maintenance and business systems have driven the decision for data integration to empower the work force. The PI-AF adds value to the current openness to access the data at the original resolution for transformation into actionable information. Many simple projects are generated with high return on investment rather than a few isolated traditional projects. A few selected cases are described in the following paragraphs. These were selected because of their objective to reduce maintenance costs and increase overall process effectiveness. In all

521

cases, the customers integrated to their current maintenance system with the PI System (i.e., SAP PM, Advantis, Maximo, MIMS).

Situation 1: Large South American Mine/MilI/Pipeline/Port Critical Issues: Decreasing Ore Grades Operating costs 30%due to maintenance costs Compliance with Production and External Constraints Management concerns with existing systems Capabilities needed: Integration to all existing DCS, MMIs, and PLCs and business systems Engineers should have data for fast process troubleshooting and root cause detection Remote view of key performance indices Access to process data by all personnel (operations, maintenance, engineers, planning, and managers) Easy access to historical information using MS Office Capabilities Provided: Integration of MineMillPort real time data with S A P PM module PI data is available at all locations PI linked to SAP PM using RLINK gateway Integration of operational data from MilVConcentrators DCS, PLCs, Analyzers, Lab Systems Results: RLINK PM has been operating since July 2001. They have installed RLINK PM to integrate these systems to pass equipment maintenance parameters between the new business system, the mine and the processing plants. This integration provides the ability to generate maintenance notifications from any of these systems manually or automatically. Currently, 180 pieces of equipment from the mine and 500 pieces of equipment from the ore processing plant are configured in RLINK. Notifications are created in PM based on asset hours, failure codes, and GPS positions. In additional, RLINK provides valuable maintenance parameters for equipment failure assessment and maintenance planning. This was critical since maintenance is normally 30%of the annual cost in an asset intensive industry like mining. Any percentage savings has significant economic implications.

Situation 2: Large steel producer in North America This company wanted to change their maintenance culture from Equipment repair to Asset Management without increasing maintenance costs. Using the PI system, they implemented conditioned based equipment monitoring. They changed their failure paradigm to an interval of potential failure and functionally failed (equipment not performing has intended function). They have changed from: 70% Total Maintenance hours and 30%Proactive maintenance to 20% Reactive Maintenance and 80% Proactive Maintenance. Average equipment availability has gone from 78% available and 22% unavailable to 91% available and 9% unavailable. Figure 10 shows their graphical presentation of how the modgun nozzle to the tap hole face fit has deteriorated into an alert. They have gone from scattered knowledge to business process and practices with consistent organized way to capture and use information and knowledge.

---+Actionable Knowledge: Easy access to knowledge repository ----+Consistent Action: the right work at the right time

Inconsistent actions Maintenance work

In Blast Furnace #4, they have extended the furnace campaign from 8 years to 15 years, resulting in a savings of $ 1 MM per year, o r $ 7 MM for 7 years. For Blast Furnace #3 they have extended the campaign from 8 years to 20 years, resulting in a savings of $ 1 MM per year, which results in a savings of $ 12 MM for 12 years. Their projected savings are $ 19 MM just for this example.

522

The PI System has enabled them to develop many small projects within their iron and steel industrial complex such a caster breakout program, energy management, and others.

Figure 10 Real time functional failure tracking using statistical bounds. Situation 3 Large Copper/Gold Metallurgical operation in South America Critical Issues: Green field site Reduce Production Costs High Variability of ore characteristics Remote site Capabilities needed: Integration of Mine System, Concentrator DCSRLCs, Maintenance and Production Systems Fast track implementation of performance monitoring and analysis Real time information access of mine, concentrator, pipeline at the desktop for all functions Capabilities Provided: Total hours and tonnages are automatically sent to maintenance system. Automatic generation of work orders for predictive maintenance is based on asset conditions. Critical asset lubrication planning and control using real time information Integration with Mining System providing critical ore data for each truck dumping ore at the stockpile. Information is used for real time metallurgical planning based on ore characteristics. Simplified metallurgical statistics providing production, chemical analysis, size analysis, yields and inventories Equipment and Process Control Event State Management Results: Increased nominal production rates by more than 20% Increased equipment availability Enabled Mine/Mill optimization Cost management. Tracking grinding mill relining and ball consumption No need for specialized programming to generate real time graphical displays or reports Implemented a real time plant wide water management Implemented a real time reagent inventory and purchasing system. Automatic purchase order and shipments

523

Implemented safety, tire and security systems Their main objectives were the integration between personnel functions for collaboration and the integration of their mining, DCS, production and maintenance systems for real time decision-making and debottlenecking of the mine/mill operations. Figure 11 shows a trend to capture the events for detailed troubleshooting and analysis. They use their system to identity and to eliminate constraints within the entire system: equipment availability, process, systems and inventory management. An event management system was implemented to identify the downtime causes, and equipment and process constraints. They have defined the following variables: 1.- uncontrolled variables: rock type and weather factors, which affect energy availability 2.- controlled variables: stock pile management, water availability, and inadequate operation of the equipment Their continuous improvement objectives include: 1. - Minimize bottlenecks within the integrated mill 2. - Generate shift, weekly and monthly variable costs/production control and variance tracking 3. - Adjust predictive maintenance and the relation with steevtreatment for increases overall equipment availability.

Figure 11 Event tracking for root cause analysis and elimination

CONCLUSIONS There is a critical need to integrate legacy systems into real time information management infrastructures. This environment should enable users to transform process data into actionable information. A methodology based on adding the process structure (plant topology) and knowledge of the measurement system and its strategic locations will minimize the global error based on satisfying the material balance constraints. Process topology is key to defining the operational management database for implementation of variable cost management; yield accounting, dynamic process and equipment performance monitoring, downtime analysis and asset monitoring. Figure 12 shows the integration of real-time information with metallurgical laboratory, maintenance systems, mining system and financial systems (Bascur, 1990, 1991). The tools have drastically evolved facilitating the integration of systems, data collection, and easy access by users and applications.

524

The technologies are available today to rethink plant operations and to increase the performance of current production systems. Data unification simplifies the integration of information from the process, laboratory, receipts and manual data entry. It generates high quality performance information from process data. The synergy of combining process data with transactional data provides a deeper understanding of the data for continuous improvement and innovation. It simplifies and speeds up the accounting process and detection of gross errors. It simplifies the identification of process, equipment, quality problems and opportunities. These data can be used for plant optimization, environmental reporting, costing, gross error detection, accounting, instrument management and assessment of the global measurement strategy.

Materials

Production Track Key Performance Indicators based on real time data Aggregate production, material consumption Generate Operational Reports Inventory and Composition Tracking of tanks and stockpiles Yields, Recoveries Energy Consumptions

Coordination

'

I

JI

Maintenance Delay Logs, Runtimes Equipment Event Management Maintenance Planning based on Equipment Condition Equipment Alerts based on history for work orders

1

"-I Preparation.

Quality Certificates of Analysis Batch and Composition Tracking Statistical Process Control, NonLinear SPC, Multivariate Statistics.

Figure 12 Integration of metallurgicallaboratory, maintenance system, mining systems, and financial systems with real time information infrastructure. Linking process and asset information is the key to extending equipment availability and reducing operating costs. This application environment enables users to identify the best application and to continuously improve the application without disruption of the real time information infrastructure. The unified information can be exposed into web centric environments. The Web is quickly becoming a key driver of data cleanliness (or dirtiness) as it gains ground as a way for knowledge personnel, managers, research groups, service providers and customers to input and access business information. The key to re-engineering is linking people, business processes, strategies and the best enabling technologies. It is important to recognize that the cleaning data is a process. As such, several groups (instrumentation, maintenance, process engineering, accounting and managers) collaborate in the data unification process. This team effort should be rewarded. The successful application should be judged on how it provides added value to the overall information system, such as new ways of storing data and events, classifying, aggregating, and combining and visualizing existing data and information.

525

The major ingredients for a successful implementation are: A solid real time information management infrastructure that provides historical database access and historical database structure access. Implementing total quality management guidelines and people empowerment for continuous improvement and innovation. A desktop environment to unify the access of enterprise information. REFERENCES Bascur, O.A., 1990, “ Human factors and aspects in process control use,” Plant Operators Svmposium Proceedings, SME, Littleton, CO. Bascur, O.A., 1991, Integrated GrindingFlotation Controls and Management, Proceedings of Comer 91, Volume 11, Ottawa, Canada, pp.411-427. Bascur, 0. A., 1993, “Bridging the Gap Between Plant Management and Process Control,” Emerging Computer Techniques for the Minerals Industry, B.J. Scheiner et.al. Eds. SME, Littleton, CO, pp. 7381. Bascur, O.A., (1999). “The Industrial Desktop - Real Time Business and Process Analysis to Increase Productivity in Industrial Plants,”IPMM99 Proceedings, Prof. John Meech, Editor, UBC, IEEE Conference, BC, Canada. Bascur, O.A., and Herbst, J.A, 1985, Dynamic Simulators for Training Personnel in the Control of GrindingFlotation Systems, Automation for Mineral Resource Development, IFAC Proceedings, July. Bascur, O.A. and J.P. Kennedy, 1995, “Measuring, Managing and Maximizing Performance in Industrial Plants,” XIX IMPC Proceedings, SME, Littleton, CO. Bascur, O.A. and J. P. Kennedy, 1996, “Industrial desktop - information technologies in metallurgical plants,” Mining Engineering, September. Bascur, O.A. and J.P. Kennedy, 1998, “Overall Process Effectiveness in Industrial Complexes,” Latin American Perspectives: Exploration, Mining and Processing, Bascur, O.A., Ed., SME, Littleton, CO. Bascur, O.A. and J.P Kennedy, 2002,” Web Enabled Industrial Desktop: Increasing the Overall Process Effectiveness in Metallurgical Plants,” Preprint 02- 134,2002 SME Annual Meeting, Phoenix, AZ. Bascur, O.A, Soudek, A., 2001a, “Improving Metallurgical Performance: Data Unification and Measurement Management Proceedings VI SHMT, Rio de Janeiro, pp. 748-755. Bascur, O.A, Soudek, A., 2001b,”Integration of Mineral Processing Operations via Data Unification and Gross Error Detection”, International Autogenous and Semiautogenous Grinding Technology 2001, Editors Barratt, D., Allan, M. and Mular, A., IV-34-47. Bialkowski, W., 1996, “Auditing and reducing process variability in your plant,” Fisher Rosemount System Users Group ’96, November, Houston. Djuric, V, 2002, “How PI Played a Key Role in Achieving Maximum Equipment Reliability at Dofasco,” OSIsoft Users Conference, Monterrey, CA. Fuenzalida, R.E., 1998,” Economic Operations Management in Copper Concentrators,” Latin American Perspectives: Exploration. Mining and Processing, Bascur, O.A., Ed., SME, Littleton, CO, pp311. Kennedy, J.P., 1996,”Building the industrial desktop”, Chemical Engineering, January. Llao, M, 2001, “ PI Us0 y Expectativas en la Industria Minera”, OSIsoft Technical Users Seminar, Buenos Aires, Argentina. Luyt, C., 2002, “BHP-Billiton Escondida RLINK Plant Maintenance Implementation,” OSIsoft Users Conference, Monterrey, CA. Mah, R.S.H., 1990, Chemical Process Structures and Information Flows, Boston: Butterworks. Martin, J., 1995, The Great Transition, Using the Seven Disciplines of Enterprise Engineering to Align People, Technology, and Strategy, AMACON, New York. Moubray, J., 1997, Reliability Centered Maintenance, 2“dEdition, New York, New York. Industrial Press Inc. Potts, D.R, 1996,”Reliability and Availability Analysis,” OSI Software Plantsuite Seminar, Microsoft Advanced Technical Center, Houston, TX, February.

”.

526

Rojas, J.H., and H.M. Valenzuela, 1998, " Strategic Plan for Automation and Process Management at the Chuquicamata Mine," Latin American Persoectives: Exoloration. Mining and Processing, Bascur, O.A., Ed., SME, Littleton, CO, pp281-292. Senge, P.M., 1990, The Fifth Discipline, The Art and Practice of the Learning Organization, Currency Doubleday, New York. Stengl, B and R. Ematinger, 2001, SAP R/3 Plant Maintenance, Addison Wesley, New York. Turton, R., Bailie, R., Whiting, W., Shaeiwitz, J., 1998, Analysis, Synthesis, and Design of Chemical Processes, Prentice Hall, New Jersey, 7, 34.

527

Enterprise Dynamic Simulation Models David W. Ginsberg

ABSTRACT Enterprise dynamic simulation models have been developed and used to support decision-making by mine operators and service organizations. For operators of mine facilities, dynamic simulation models have addressed the effective deployment of capital, risk management, market opportunities and operating strategy. Service organizations typically use them to assess strategic issues, such as how profitably corporate assets and market opportunities match over time. Enterprise dynamic simulation models take a holistic approach to providing strategic direction. Optimizing profit in one facility may be sub-optimal for the enterprise as a whole. This paper presents key aspects of this approach with case studies, and discusses web-based model deployment. INTRODUCTION Simulation is considered by many to be a vital problem-solving tool for making complex technical and business decisions. For example, engineers designing a comminution circuit operating under stated conditions can use simulation to build and then model the associated material flows (Napier-Munn 1996). Results can be analyzed and compared with alternative designs and assumptions. At the other end of the spectrum, Peter Senge’s Fifh Discipline (1990) explores simulation and “systems thinking” in terms of the learning organization. One of Senge’s key messages is the importance of obtaining a clear perspective without being drowned in details. An enterprise and its associated suppliers, partners and customers form a complex system made up of numerous sub-systems. This paper explores the use of simulation as a decision-making tool that permits the objective assessment of alternative strategies and simplifies masses of detail to define clear strategies. Enterprise simulation models can help companies increase profitability by optimizing their operations, and can support and guide capital investment and allocation decisions. If their organizations are to outperform the competition, executives concerned with company market valuation must consider total company asset effectiveness as they relate to markets. This requires a holistic view that is often impossible to achieve without sophisticated analytic quantitative tools and techniques. No one person can process simultaneously the technical and financial variables that describe the highly complex supply chains from mines to markets, because they are too numerous, and many of them fluctuate significantly over time. Many fluctuations in system performance are caused by random events that can be quantified but not deterministically predicted, necessitating the inclusion of system dynamics in the analysis and management of risk. The traditional steady-state financial analysis of grass-roots projects, for example, does not fully address supply-chain management issues that connect operations with markets. These issues may present serious risks (such as transportation delays, insufficient storage capacities, or supply shortages) that could have a significant impact on profitability. The definition of an enterprise simulation model is totally dependent on the specific issues that must be addressed and can be quantified. Business simulation games (Dixon, 2001) can illustrate general strategies, but every company’s management faces a unique set of circumstances. For example, a company may have an enterprise model to assess strategies for global asset management, and the impact of adding new facilities or mines; but a large mine or metallurgical

528

facility may be considered an enterprise in its own right, and the strategic issues that must be addressed during its development (such as capital deployment and risk mitigation) may be distinct from issues of operations-wide optimization (such as advanced strategies and new market opportunities). Although enterprise simulation modeling is an important part of what we term “strategic quantitative analysis”, it is considered an “art” to define the system to be modeled, just as it is “science” to correctly characterize the system with equations, statistics and logic. Business rules and system definition are obtained by integrating all relevant aspects of the supply chain. Hardearned experience determines the questions to be addressed using these techniques, the scenarios to be tested, the appropriate data and level of detail, and the toolset for the job. While we always advocate a “simpler is better” approach, this does not mean that it can be a simplistic one. Many corporations have seen their optimization techniques for decision support yield inadequate or misleading results.

TOOLS Commercially available simulation software was once categorized as either a simulator or a simulation language. The former allows users to piece together flowsheets from a library of predefined models, and is chiefly used for process simulation. Simulation languages, like any programming language, rely on the coder for meaningful model definition and accuracy. They are more costly to apply but also have more flexibility for precisely modeling a system. Today, the lines between these categories are blurring. Steady-state optimization calculations can be coded (using C, for example), but spreadsheets are often a more convenient alternative. The techniques most often used for analyzing systems in the mining industry are well summarized in the SME Mining Engineering Handbook (Hartman 1996). For enterprise simulation, tool selection is a lower priority than setting and quantifying business objectives. Toolsets and algorithms are selected to meet the task at hand. Meaningful models are often made up of stochastic (Pegden 1995), logical and empirical components that are simulated dynamically over time, and may even include optimization algorithms such as linear programming. DATA SOURCES While enterprise models play a role in predicting overall performance and setting strategy, they do not eliminate the need for the numerous detailed technical and financial models used to simulate and assess each operation. These models have different purposes and levels of detail. For example, a large-scale mine’s ore-to-product supply chain may include: 1. Mine operations, whose size, principal ore characteristics, and development are planned using 3D mine planning software. As more information becomes known about the orebody, the 3D model is updated to better reflect reality and influence operating practices. This model can be applied from start-up through to the end of the mine’s life. 2. Process operations, where metals are liberated through a range of comminution and chemical processes. The process and its configuration are planned and optimized using a range of mass, energy and chemistry simulation models. A dynamic simulator can assess alternative process equipment configurations, processes and process control strategies, as ore characteristics and throughput change over time. 3. Material handling and storage systems, which are sized using a range of mechanical and electrical analysis tools. In the case of a pipeline, for example, the physics of the process must be reflected in the model. 4. Forecast, scheduling and network models, which are used to schedule transport (shipping, rail, trucks) as a function of contract needs and mine planning 5. Financial models developed on a spreadsheet, which address cash flow and profitability, and are used to assess strategic direction.

529

In addition to these models, data and statistics are collected from a range of online systems (controls, trending, middleware, ERP) to assess quality, throughput and maintenance. Smelting and refining operations must often handle concentrate feed from multiple mining operations. This requires monitoring and planning tools for concentrate receiving and product shipping, in which various modes of transport and logistics, and differing product qualities, must be reconciled with process needs and client contracts. To arrive at an appropriate system definition, enterprise simulation models assess only the essential characteristics of each area’s detailed analysis. Even if it were possible to incorporate all the above-mentioned models and data, the execution time would be prohibitively long, and strategic cause-and-effect results would be lost in the details. Good strategy requires clarity.

CASE STUDY A multi-billion-dollar grassroots mine operation in South America comprises multiple operations with complex interactions and variations in performance. The operation’s organization is complex, with multiple owners who have their own experts and ways of doing business. The design and construction consultants comprise another complication, as does the newly hired operating team. A mine-to-market enterprise simulation model was commissioned principally to support capital deployment decisions, quantify projected overall system performance (quantity and quality), and identify areas of risk. The model includes appropriate aspects of the mine plan, process facilities, major storage facilities, and transportation and logistics systems (Figure 1). It also includes long-term market contracts, production forecasting and complex scheduling algorithms.

Markets

Demand

Mine

Production Forecast & Shiament Schedulina

Shipping to Markets

Port Operations

Process Facility _Supply Multiple Concentrates

Mine-site Storage

TransporVLogistics

Figure 1 Mine-to-market enterprise simulation model

530

Port-site Storage

The model addresses:

0

0

Predetermined mine design sequences, expected grades and statistical variability obtained from dedicated mine planning models. In addition, conditional simulation is used to consider alternative mine plans. Empirical relationships modeled from metallurgical testwork, coupled with historical operational statistics and logic. Scenarios with alternative storage capacities and configurations, to minimize capital costs while obtaining the desired performance. Transport and logistics models, incorporating project-specific operating conditions such as detailed weather models, which can have a major impact on shipping. Commercial arrangements and contracts, based on marketing and sales data.

Timing has a significant impact on cash flow and capital deployment in every factor of operations. Ore types, quantities and grades vary over the course of the mine’s life. The process facilities produce corresponding variations of concentrates at varying rates. Supply is also affected by equipment availability, production schedules, operating strategies, shipping conditions, etc. On the demand side, concentrate production forecasts and shipping schedules must match actual production over time. Blending may be required to meet commercial contracts. Matching system supply and market demand, while maximizing profitability, is a very complex challenge. The right concentrates and strategies must be in place for blending to ensure commercial arrangements are met without excessive waste. The required blends must be available for export when ships arrive, to keep port operating charges to a minimum. Berthing capacity and scheduling must avoid ships queuing to dock. Ideally, concentrate should arrive at the port on a just-in-time basis for shipment, but the storage facilities must be large enough to deal with delays in both concentrate production and the arrival of ships at the berth. These multi-million-dollar facilities must be carefully sized, to avoid unnecessary expenditure without jeopardizing operations. A spreadsheet can predict only steady-state profitability, but a dynamic enterprise simulation model can additionally assess the likelihood of cash constraints. Mine planning and scheduling models’ predictions of annual shipments don’t allow for the inevitable up- and downstream bottlenecks, but an enterprise model can simulate the interactions among supply-chain components to assess their impact. It can apply variations in costs, transport and clients (and hence product mix) to determine the most robust supply-chain configuration. It can assist in intelligent trade-offs on which facility should service which client; whether a low-cost or more flexible shipping scheme is best; and how much cash flow can be tied up in inventory to maximize profitability. ORGANIZATIONAL ALIGNMENT AND INTERFACE MANAGEMENT When traditional analysis suggests the need to expand facilities, a good systems strategy can help keep costs down. In Leveraging Engineering Services Through Decision Support for the Mining Industry (Ginsberg 1999), case studies are described in which dynamic system simulation and financial models for operations and enterprises are combined to develop and test strategies, and then objectively compare alternative expansion and operations scenarios. Figure 2 illustrates a scenario analysis to support equipment selection for a large metallurgical facility. Each scenario is defined by equipment selection, management strategies, and markets addressed. Scenario D produces the largest production throughput, yet has the worst associated NPV.

531

Annual Production Scenarios

Corresponding NPV

5

.-E X

9

Scenario D

a, 0)

m G

a,

2 Scenario D

Figure 2 Scenario analysis How does this tie into other existing or planned operations, and what market-driven scenarios produce the best return? For an enterprise, optimizing key performance indicators (typical indicators illustrated in Figure 3) from one facility may be sub-optimal for the enterprise as a whole. If the organization has more than one facility that can produce a certain product, a choice must be made among them to determine which one should service a particular contract at any point in time.

Operating margin Fixed costs Capital expenditures Cash flow

Figure 3 Key performance indicators

532

In making such decisions, it’s vital to remember that enterprises are made up of people with individual talents, perspectives and experience. Implementation will fail unless there is buy-in from key constituencies. An enterprise model provides an excellent way to align organizations to a common purpose, by illustrating how everyone fits into the winning strategy. It provides an objective and holistic analysis that includes everyone’s input. This alignment is particularly relevant for: 0 0

0 0

owners, consultants and engineers at the capital investment phase integrating design and operating considerations among engineering disciplines, operating groups, and sales and marketing matching business objectives with technical operations transitioning from a capital project to operations.

When reviewing operations to optimize business performance, the biggest opportunities present themselves at the interfaces between different areas of responsibility. This is because organizations are traditionally designed by function and influenced by corporate history. Enterprise modeling helps focus on the most important business needs, and allows design by process. This pushes accountability and decision-making to the point of action, enabling all parts of the organization to understand how they fit into the big picture. Enterprise modeling also helps identify areas suitable for outsourcing, and those that are critical to business success.

SERVICE ORGANIZATIONS Much has been written about analytic techniques for assessing process operations, but little on the use of enterprise simulation models for service organizations. In models built to support strategic initiatives related to such organizations’ business efficiency, risk, and growth, the variables are seldom deterministic and must be addressed using statistical probability. An analytic framework is required to integrate these variables with business processes, and to deal effectively with their complexity. Consider an enterprise simulation model developed for an engineering services organization, which generates revenue by bidding on, winning and executing projects of varying types and sizes. The model’s principal purpose is to assess the organization’s current ability to support its stated objectives and markets. In particular, the model can be used to compare alternative strategies for achieving corporate targets and assessing risk. Should management pursue growth, or should they downsize? Is the current employee pool structured efficiently to match market realities? Unlike an accounting tool, this modeling approach makes use of all known information to simulate what the future may hold rather than simply analyzing the past. As with process operations, some of the greatest benefits of enterprise simulation modeling result from the alignment of purpose between, for example, Operations (“How busy are we?”) and Finance (“What’s our margin?”). The model can help staff understand and articulate the idiosyncrasies of the business, gain a better appreciation of the importance of planning and strategy, and avoid decision-making based on personal or narrow organizational agendas. This service organization model (Figure 4) categorizes employees by line of business, type of expertise, salaryhilling information and numerous other parameters. Revenues are calculated over time, based on charge-out rates and the size and nature of contracts. Direct and indirect expenses are accrued and paid using applicable corporate processes. The user can specify business opportunities (identified through empirical evidence) and associated activities for each area and project type, and can simulate them under varying business-specific conditions.

533

Opportunity Marketsand Services

+

Project Execution

*

Revenues in

market demand resource supply * commercial terms

workload profile

Key Performance Indicators

-

-i Fixed costs out

full-time contract other

Figure 4 Service organization model This model permits business scenarios to be simulated over a reasonable planning horizon. Typical outputs include the number of projects completed by type, margin, and opportunity costs (arising, for example, from lack of available resources). Regional, group, and staff-category utilization and profitability can be assessed over time, for the development of hiring policies, outsourcing decisions and other strategies. Dynamic plots of cash flow and outstanding receivables can provide useful insights into the health and risk of the business. New scenarios can be run to simulate planned growth, possible decline, and changes precipitated by acquisitions. Staff strength and overhead structure can be manipulated to optimize profits and reduce risk for any scenario. Even a mechanism to factor in the effects of litigation costs can be applied, to compare business risk and reward.

MANAGEMENT AND DEPLOYMENT OF MODELS Like all forms of intellectual capital, enterprise models should be maintained and refined beyond their initial use. A model developed for a particular capital project can be applied to provide benefits across the organization. This is an important future trend for enterprise simulation models. AMEC has explored and tested sophisticated web-based systems that address this very issue, with encouraging results. The tested system can collect the results from stand-alone systems and applications and feed selected data into an enterprise model that is globally accessible for decision support. This is particularly desirable for mining organizations, which typically have far-flung operations and a limited supply of experts to assess complex scenarios. Data can be synchronized and shared, and models transitioned from capital investment to operations. Best practices can be disseminated for use on similar projects. Intellectual assets can be maintained, protected and applied across the organization. Proprietary enterprise models remain resident in the corporation, with no risk that key data and resources will be lost when key people resign or files are misplaced. Web-based deployment systems have proven to be viable and promise future commercial benefits, but they pose numerous challenges that are currently being addressed. In the mining industry, experience indicates that there remains a large gap between simplistic statements about “knowledge management” and commercially secure web-based systems that are suitable for deploying enterprise simulation models.

534

SUMMARY This paper has described how enterprise dynamic simulation models can increase profitability and guide key decisions for operating companies and service organizations. The analytic techniques and systems approach for assessing future scenarios are the same for both types of organizations, yet the details of the models can be very different. Experience is very important for successful implementation. It is essential to set and quantify business objectives properly, so that the model is properly articulated to produce meaningful and comprehensible results. This can be achieved only when no constituent subsystem is considered more important than the whole. Experience suggests that enterprise dynamic simulation models save millions of dollars when used to support capital deployment decisions; but even greater benefits can result from their ability to align the members of a project team, an operation or a service organization to a common purpose. An enterprise model can show how everyone - whatever their talents, perspectives and experience - fits into the winning strategy. It helps people focus on their most important business needs, and supports organizational design by process rather than management through history. Finally, enterprise models are important intellectual assets for an organization, and their maintenance and deployment must be considered in that light. Careful consideration must be given to the processes and technology used in their implementation. Their deployment on the web has yielded very encouraging results, and should prove to be an excellent method for increasing their value to an organization. ACKNOWLEDGEMENTS The author wishes to thank the clients and staff of AMEC who, in one way or another, have contributed to the body of work that forms the basis of this paper. In particular, thanks to Dr. Bear Baker for his helpful suggestions and input, and Andrew Solkin for his editorial support. REFERENCES T.J. Napier-Munn, S. Morell, R.D. Morrison, T. Kojovic. 1996. Mineral comminution circuits: their operation and optimisation, Julius Kruttschnitt Mineral Research Centre, Australia. P.M. Senge. 1990. The Fifth Discipline, Doubleday. H.L. Hartman (senior editor). 1996. SME Mining Engineering Handbook, second edition. D.W. Ginsberg. 1999. Leveraging engineering services through decision support for the mining industry, Engineering Management Journal, Vol. 11, No. 4. C.D. Pegden, R.E. Shannon, R.P. Sadowski. 1995. Introduction to simulation using SIMAN, second edition, McGraw-Hill. G. Dixon. 2001. Game plays by rules of new economy, Globe and Mail, February 22 2001.

535

Factors which Influence the Selection of Comminution Circuits Derek Barratt' and Mark Sherman2

ABSTRACT During the last twenty years, most new or expanded grinding projects have selected some form of Semi-Autogenous Grinding technology. However the project's design team must rigorously evaluate the basic reasons for that selection before rejecting alternatives such as: multi-stage crushing followed by either single-stage ball milling or rod milling/ball milling, the use of high pressure grinding rolls, dry grinding or, in some special cases, a combination of ore washinghcrubbing and grinding. Factors that should be examined are lithology, alteration, mineralogy, geotechnical parameters, ore hardness, standard comminution parameters, pilot-scale test results, mining rate, and production schedule by ore type, etc. The variability and interaction of all of these factors determines their overall effect on metallurgical performance and project economics. INTRODUCTION Summary of Factors The high capital and operating costs associated with comminution circuits, and their impact on overall project economics, demands the selection of the most cost-efficient circuit. The most costeffective option, however, itself depends upon the type of project being considered. If the project is a new mining development, then most, if not all of the factors identified in this chapter, will have to be analyzed. If the project investigates the treatment of a new ore body adjacent to an existing operation, or is simply the expansion of the existing operation, the factors that need to be analyzed will be influenced by the design constraints of the existing operation. Not all of the factors may be pertinent. The factors that have been identified in this chapter are those that will have to be considered for the most complex of projects, that of developing a new operation. The most important step in the development is the analysis and understanding of ore characteristics. The broad range of ore types encountered requires that each characteristic be treated individually. Thus hardness of the ore; i.e., its resistance to both impact and abrasion grinding forces, its abrasiveness, friability, moisture content, grade and mineralization including gangue mineralization, liberation size, chemistry, and other characteristics should be the first to be analyzed. The next factors to be analyzed should be the determination of plant size, throughput rate, location, climate, accessibility, and availability of water. The selection of samples for comminution test work, be it for initial bench-scale or largerscale testing such as pilot plant test work, is very critical. Its importance in ensuring the success of the project cannot be over emphasized.

' DJB Consultants, Inc., North Vancouver, B.C., Canada

* Sherman and Associates, Pty Ltd., Canberra ACT, Australia

539

Although the metallurgical flowsheet dictates the liberation criteria, in many cases it is the comminution engineer rather than the metallurgist who dictates the number of grinding stages. The advantages and disadvantages of the various comminution circuits are many, but the final analysis involves practical considerations such as the availability of good quality grinding media, trained people to operate a sophisticated plant, etc., as well as economics. Finally, it is essential to have a proper understanding of current practices in similar and related industries to understand, and learn from, why others did what they did, the mistakes that they made, and the constraints that prevailed (Barratt and Sochocky 1982).

Outline of Various Comminution Circuits under Consideration The wide variety of possible circuits provides significant challenges for the comminution engineer. To keep this analysis manageable, only the predominant circuits, each with its own advantages and disadvantages, are reviewed. These are: Crusher, Rod Mill, Ball Mill Crusher, Rod Mill, Pebble Mill Crusher, Single-Stage Ball Mill Crusher, Multi-Stage Ball Mill Single-Stage Autogenous Autogenous, Ball Mill Autogenous, Ball Mill, Crusher Autogenous, Pebble Mill Autogenous, Pebble Mill, Crusher Single-Stage Semi-Autogenous Semi-Autogenous, Ball Mill Semi-Autogenous, Ball Mill, Crusher Pre-Crushing, Semi-Autogenous, Ball Mill, Crusher. Relatively simple preliminary tests such as the Autogenous Media Competency and Bond’s Low Energy Impact Crushing Work Index tests allow for the elimination of a number of the above circuits. This significantly reduces both the amount of time and expense required for the selection of the most cost-effective comminution circuit.

Project Development Sequence and its Effect on the Choice of Factors It is a pre-requisite for a successful plant design that a comminution engineer be specifically assigned at the beginning to work with the metallurgical staff. The gathering and evaluation of design criteria and data can be done by the owner’s staff or by an independent consulting firm, but in either case, the starting point must be the geologists’ report. This will delineate the ore body and provide a guide to initial sample acquisition. The quantity of the ore samples at this point is usually limited thereby restricting studies to bench-scale in order to define the starting parameters. The bench-scale results, together with the comminution engineer’s experience, will allow the proposal of preliminary ore treatment flow sheets. At this point, a preliminary technical and economic viability study should be prepared, considering factors relative to plant location, ore type, product tonnage and quality, even if many factors have to be assumed. These factors should be considered only in as much as they affect comminution circuit design; e.g., the fineness of grind or the need for multi-stage grinding with intermediate concentration steps between grinding stages. Based on this information, the comminution engineer should conduct the appropriate comminution tests. For the purpose of an initial, or pre-feasibility study, a conventional CrushindSemi-Autogenous/Ball Mill circuit can be assumed. This is a practical first step that facilitates the evaluation process when considering alternative schemes.

540

Samples should also be sent as soon as possible to an accredited laboratory for autogenous media competency, Bond work index, and abrasion index testing that allows for the commencement of a more detailed full-scale feasibility study, assuming that the initial technical and economic evaluations have returned favourable results. For the detailed feasibility study, many decisions will have to be made by the comminution engineer who, by this stage, should have results of the autogenous media competency, Bond work index, and abrasion tests. This study will require the thorough investigation of all of the factors. Depending upon the location and size of the project, and type of deposit, the development will require the acquisition of a larger representative sequence of samples, more thorough bench-scale tests, and perhaps a pilot-scale investigation. The decision will have to be reached as to which comminution circuits should be tested and to what level of effort. Each circuit tested will have to be evaluated taking into account economic, metallurgical, and practical considerations with regard to equipment sizing. The final selection should be the consensus of the best available knowledge.

IDENTIFICATION OF FACTORS Geological Interpretation of Drill Core and Bulk Samples Sampling methods for the feasibility studies depend upon the geological characteristics of the deposit. Initial sample acquisition methods will vary and should be adapted to each particular deposit, but at this stage of the project development, it is likely that the samples used for development of the design parameters will be of diamond drill core, and possibly trench or surface grab samples (Ashley 2002). Factors of interest to the comminution engineer from the geologists’ report would be the identification of the mineral constituents, their relative quantity and degree of dissemination, the number of distinguishable ore zones, and the main characteristics of each. The following factors should be evaluated to provide the initial background information for further decision making. Mineralogical Analysis Examination of the mineral specimens can determine the identity, character, grain size range and middling associations of the ore minerals and constituents of the gangue rock. This information will provide liberation size analyses for primary, intermediate or regrind applications of grinding, and the appropriate concentration steps. Occasionally, the minerals present in the ore are in such complex associations that simple optical or X-ray techniques cannot identify them. In these cases, electron probe analysis; e.g., QEM*SEM or QemSCAN, can be used to determine the mineral characteristics and the occurrence of the element(s) in question. Chemical Analysis Whilst more relevant to the design of the beneficiation flowsheet, a complete chemical analysis is always valuable to the comminution engineer as it provides an early indication of potential problems. For example, if there is evidence of alkalis or sulphates, sliming may be a problem. Sulphates may also produce an acidic slurry that results in higher media wear rates or predicates the need for an ore washing stage in the flowsheet. Physical Properties The first visual observations can immediately indicate the physical properties of the ore; e.g., the hardness, blockiness, friability, amount of primary fines, and clay content. These properties are indicative of potential problems to be encountered in crushing, screening, and grinding the ore and will influence flowsheet design. Examples are the avoidance of slime generation when processing tin, tungsten, lead, chromite, or tantalum ores; the release of molybdenite at a coarse grind from fracture planes in a blocky

541

granodiorite; and the elimination of fine crushing problems by using semi-autogenous grinding when processing a clayey ore. Whilst the results of the Bond impact and grindability tests, abrasion tests, ore specific gravity, and autogenous media competency tests for such samples are a tool for obtaining the final design parameters, the results also enable the metallurgical staff to chart a course through to a better understanding of the ore characteristics. Analysis of the information will allow for the rationalization of circuit options. It should be emphasized that the aforementioned tests should be performed on each of the ore types that will be presented as the major components of the mill feed. Individual testing of each major component of the mill feed will indicate the variation in mill performance that should be expected. This is a critically important management issue that needs to be analyzed, and understood by the project owners due to its impact on project cash flows.

Review of Circuit Feed Parameters It is essential that the comminution engineer have a clear understanding of the overall project. The engineer should be familiar with the mining plans, schedules, mining methods, rates, and equipment sizes, as these will affect the choice of processing equipment sizes in particular the crushers and primary mill. Other parameters that will be affected include the operating hours for the crushing circuit, the need for ore storage, manpower requirements, and the location of the crushing and/or grinding plant. The primary crusher is in most cases a surface gyratory or underground jaw crusher. The size of the crusher will be dictated not only by the mining rate and the top size of the feed but also by the product size requirements; e.g., for pre-crushing and screening ahead of SAG milling. Another important parameter required by the comminution engineer in the early stage of the project is the throughput requirement. The economics of the project will dictate the required final product quantity and quality, which in turn will allow for the calculation of throughput rates based on the predicted head grade and the engineer's experience in designing circuits of the type being proposed. Review of Sampling Requirements It cannot be stressed often enough that any process design for any given ore body is only as reliable as the sample upon which it is based. The type of deposit and mineralization define the sampling requirements; i.e., there is no definitive pre-determined amount of sample that ensures good results for all ore bodies. The relative costs and time taken for obtaining samples and testing them are significant factors in the choice of circuit to be investigated. As such, the project team would be well advised to invest the time and money required to obtain the appropriate samples, based on the characteristics of the deposit being investigated. Preliminary samples are usually obtained as drilling progresses to delineate the ore body. The cores are split in half, with one half being used for metallurgical testing. Sometimes, surface grab samples and/or trenching is also used to generate material for preliminary test work. Generally, approximately 23 kg (50 Ib) of sample is required for testing Bond's rod mill (Wi,,) and ball mill (Wi,,) work indices, whilst lumps capable of generating 7 cm (3-in.) rocks are necessary for testing the low energy impact crushing work index. Over the last 20 years, some semi-autogenous grinding circuits have been designed on the basis of the three Bond work indices alone, Wic, WiRM,and WiBM,with reference to inherent rock strength; e.g., point load index (PLI) and/or unconfined compressive strength (UCS). In these instances, whole core intervals either PQ (85 mm dia.) or HQ (64 mm dia.) have been sampled to identify each ore/mineral lithology and alteration type and to assess ore variability. Fracture frequency and rock quality determination (RQD) are important parameters when considering the potential for autogenous grinding (Barratt 1986, 1992). Within the last 5 years, the Minnovex (formerly Starkey) SAG Mill Power Index (SPI) has matured into an indicator of ore variability that can be compared to the results of established

542

power-based models and pilot plant testwork. Sample requirements for the test are also based on whole core (Amelunxen 2001), and it is becoming common practice to twin samples with those used for Bond work indices. In this manner, variability of ore hardness and grindability can be mapped out for a resource prior to selecting bulk samples for pilot plant testwork, by sourcing from large diameter drill core or ROM excavation. If an economic ore deposit has been established and preliminary bench-scale tests have shown promising results, then further sampling will be needed. Also, if preliminary autogenous media competency tests have indicated that the ore has autogenous or semi-autogenous potential, samples must be obtained for further testing. For either an open pit or underground operation, preliminary mine development work traditionally consists of driving adits or shafts and driftskrosscuts into the deposit. These are a good source of samples for larger-scale testing; e.g., for autogenous and semi-autogenous grinding test work, typically 25 to 50 tonnes of representative sample is required for processing in one or two tests for each ore type through a 1.8 m (6 ft) diameter pilot mill. The samples should be runof-mine prepared in the normal manner by drilling and blasting to about 20 cm (8-in.) top size. In some cases, pilot plant testwork has been conducted on 150 mm dia. core drilled from surface. Identification of Contiguous Properties Each piece of comminution equipment has its own inherent characteristics; e.g., rod mills produce fewer fines than ball mills, closed-circuiting reduces over grinding, autogenous mills treat ore along grain boundaries and can liberate values at coarser sizes. These basic tenets, together with properties of the ore, influence the comminution engineer’s thinking when selecting equipment that must produce the product for the subsequent concentration process. If an ore has a tendency to slime, the circuit may need to incorporate high efficiency classification in closed-circuit with the grinding step. The high efficiency classification circuit may call for high-speed screens or two-stage classification. Removal of troublesome clay constituents may require dry grinding and pre-concentration before the ore can be wetted. Conversely, the beneficiation flowsheet might have to accommodate a mesh of grind that a piece of comminution equipment naturally grinds to; e.g., autogenous mills or rod mills. This consideration can be illustrated by a flowsheet for taconite iron ore. Comminution circuits for a taconite ore would traditionally be designed as two-stage circuits incorporating either rod mill-ball mill or autogenous grinding followed by ball milling. An alternative flowsheet used single-stage autogenous grinding in closed-circuit with magnetic separators and cyclones, Whilst the power efficiency of this circuit was not as favourable as the common two-stage plant, the simplicity of the flowsheet and the reduced capital costs made economic sense (McDermott 1972). Thus, it is apparent that the comminution aspects of the flowsheet are closely inter-related with the beneficiation processes and close co-operation in designing these is indispensable. Feed and Product Specificationsfor each Comminution Step Feed and product specifications are defined in the earliest stages of flowsheet development. Size analysis of the plant feed or mine product is the first to be determined since the proportion of fines can influence the necessity for screening after primary crushing. Open pit mines can adjust the top size to the primary crusher requirements. More recently, the “mine-to-mill” movement has been emphasizing with some success the importance of optimal blasting practice and its anticipated advantages with respect to optimizing SAG mill feed rates through the creation of fines. In the authors’ experience, such benefits can sometimes be equally achieved by choke feeding the primary crusher (also Dance 2001). Also in this regard, care should be exercised in optimizing feed top size in the case of autogenous grinding, and in recognizing the potential impact of “mine-to-mill’’ on the creation of critical size, the effect of the additional critical size on mill throughput rates, and how to deal with it (Sherman 2001). In underground mines, especially if the primary crusher is underground, the mine product can be much finer as dictated by the handling capabilities of the hauling or hoisting equipment.

543

Most comminution equipment is designed to work in a specific size range. However, as usual, there are grey areas that make life for a comminution engineer both interesting and challenging. Which powder factor to use and/or how fine to crush before grinding are some of the challenges. Powder factors will be driven by mining phenomenon, such as their effect on pit floor and wall surfaces. However, there is usually some capacity for manipulating this parameter. How fine to crush is one of the parameters under the comminution engineer’s control. Recent developments in high efficiency crushing circuits and the installation of larger crushing units make it more acceptable to do away with the rod mill and proceed directly to ball mills, which do not have the same product and equipment size limitations. This approach can sometimes result in an economic alternative to SAG milling. It can also be influenced by the potential application of high pressure grinding rolls (HPGR) as a final stage of crushing ball mill feed (Parker 2001). If rod milling is used, it is preferable to prepare the feed with closed-circuit crushing in order to avoid problems with segregated oversize; i.e., forcing rods apart (Sochocky 1972, Rowland 1970, Flavel 1981). If crushing is done underground and the mine product size is too fine for primary autogenous milling as practised in North America, then the more even lengtwdiameter ratio type of mill, as used in South African and Scandinavian practice, may be the answer.

The Importance of the Work Index and Abrasion Index The Bond Work Index is a very useful and convenient tool for a comminution engineer. Application of test results to the Bond Work Index formula provides the initial indicator of the specific power consumption (kWh/t) required to grind any particular ore. Additionally, because the method requires only a small amount of sample and costs relatively little, the specific power consumption for grinding can be ascertained quickly and accurately for a number of different ore types, by utilizing drill core samples, early in the development of the project. Work index can be reported as a dimensionless number in both short and metric ton formats without reference to “power consumption;” i.e., Wi (metric) or Wi (s.t.). The work index should be determined for different applications of comminution; e.g., crushing, rod milling, and ball milling. It is a false assumption to expect the work index to be uniform for all incremental stages of size reduction. Abrasion tests serve as an indication of metal wear that can be expected in crushing and grinding. Once determined, they can be the first indicator that autogenous grinding should be investigated, as steel wear can be a significant cost impost on an operation. Once an overall composite of the ore deposit is established, Bond grindability work should be run to obtain the overall specific power consumption. Mathematical averaging of the Bond Work Indices for a series of ore types normally does not produce the correct average specific power consumption. Differential grinding causes each ore type to be ground to a different size distribution, some finer and some much coarser than the average (Rowland 1980). If pilot testing is needed due to the complexity of the flowsheet and the size of the deposit, Bond grindabilities should be used during the test work to check the grinding data produced by the pilot mill. Frequently, these mills as AG/SAG mills are oversized with respect to the pilot beneficiation circuit and are run in separate campaigns in order to achieve mill load stability. In smaller mills as rod or ball mills, it is often difficult to measure the power consumed for grinding because of the high proportion of power that is absorbed by the drive inefficiencies and fixed load characteristics. The major requirement for a pilot mill in grinding testwork is to establish a reliable relationship between grinding power and measured power. This is done by: 0 0

Taring the mill, measuring the stable no-load power after a test; and By calibrating the gross power against the net power for the subject pilot mill.

In this manner, reliable estimates of specific power consumption can be obtained.

544

Determination of Specific Power Consumptions Specific power consumption for each stage of crushing and grinding can be calculated from the relevant Bond work indices in the case of: Primary and Secondary Crushing, Wic (low energy impact). Tertiary Crushing, Wic (twin pendulum testing). Rod Milling, WiRM to a specified mesh dictated by process. Ball Milling, WiBM to a specified mesh dictated by process. Single-Stage Ball Milling, WiBM and WiRM dictated by process. Pilot plant testwork is more commonly used to determine specific power consumptions for primary and secondary grinding stages in the case of: Single-Stage Autogenous. Autogenous, Ball Mill. Autogenous, Ball Mill, Pebble Crusher. Autogenous, Pebble Mill. Autogenous, Pebble Mill, Crusher (excess pebbles). Single-Stage Semi-Autogenous. Semi-Autogenous, Ball Mill. Semi-Autogenous, Ball Mill, Pebble Crusher. Pebble Milling. Pre-Crushing a SAG Mill Feed. For any option, design criteria can be prepared following an analysis of ore variability. This can be done by testing composites of mill feed that have been selected according to the mine production plan and making a decision with respect to the risk or frequency of harder and softer ore blends. The SPI test can assist in this decision making process in respect of AG/SAG grinding, as can the Bond work index for rod milling in orebodies for which it is a clear indicator of the development of critical size; i.e., when Wic < WiRM > WiBM. In sizing mills for AG/SAG grinding, emphasis is placed on having enough primary mill power (as determined by mill size, mill speed, ball charge volume, total mill charge volume, ore specific gravity, pulp density, pebble crushing, transfer size, and motor design, etc.) to process the desired range (frequency) of ore hardnesses (i.e., feed rates) so that average monthly, quarterly, or annual production targets can be met during the payback period and beyond (Barratt 2001). In this respect, secondary grinding power is usually determined from Bond work index testwork, combined in some cases by scale-up from pilot plant testwork and screen analysis of primary mill circuit product. Expected variations in the feed size analysis (new feed to secondary grinding) and acceptable variations in fineness of grind to the subsequent process are taken into account in order to assess the contingency that should be applied (Barratt 1989). Emphasis is placed on developing the maximum required power at the fastest acceptable mill speed and the highest acceptable ball (or pebble) charge volume when sizing the secondary mills. For traditional grinding circuits, where less feed rate variation is usually expected, more emphasis is placed on having enough mill power to process the most significantly (i.e., greater than 10% of the mineable ore reserve) hardest ore blend at the required fineness of grind. In those cases where power-based models with Bond work indices (Wit, WiRM, and WiBM) as input are used to determine specific power consumptions for AG/SAG circuits, or only limited pilot plant testwork is possible on one or perhaps two composites, the following test procedures are available to supplement the use of Bond work index tests, autogenous media competency tests, and geotechnical parameters; e.g., PLI, UCS, and fracture frequency (Hanks 2002). JK Tech. The drop weight test creates energy-based parameters across a range of size classes. These parameters are used to create a model of a comminution circuit that can be used to simulate

545

different process conditions; e.g., variation of feed size distribution, feed rates or classification size, etc., for a specific breakage environment; e.g., mill size, ball charge volume, ball top size, mill speed, and pulp density. Mill power draw can also be estimated. In this respect, the model is useful, but the results are specific to the sample that was tested and are subject to some limitations: 0

0

0

Cost. Compared to the spectrum of comminution test data that can be obtained from the autogenous media competency test and Bond work index testing, this test procedure is expensive. Breakage. Breakage rates have to be determined for a particular environment; e.g., ball charge volume, ball top size, and mill speed, which can only be done with pilot plant testing or by reference to sampling a relevant industrial milling circuit. For example, the difference between breakage by impact and abrasion and changes in the relative balance of these mechanisms between one environment and another, such as in a high aspect mill at low mill speed vs a low aspect mill at high mill speed, could only be determined by circuit surveys and sampling of the mill charge in each case (Powell 2002). Identifying a relevant circuit can be difficult as issues such as volumetric flows, feed size and pulp discharge assembly design can have a significant impact on the model’s output. In the absence of measured breakage rate data, the effect of changes in the grinding environment can be estimated in the model through use of the “variable rates” data base and approximations; e.g., by ratioing a change in mill speed. At present, the effects of changes or differences in shell linedlifter design are yet to be fully incorporated into the variable rates data base and continue to be researched. Use of approximations should be reflected by increased contingencies. Circulating load: Although the model has been updated since its inception, JK Tech states that its use for mill diameters in excess of 10 m is subject to continuing research into the perceived restriction of the mill charge on the magnitude of the circulating load and exit rate of pebbles (JKMRC 2001).

The JK Tech test procedure and model are used to estimate mill power draw and simulate process streams under defined operating conditions. They are not used to specifically estimate specific power consumptions for design criteria, but they can be used to test variability in ore characteristics. MacPherson. The MacPherson autogenous work index test has been in use for nearly thirty years and the results can be applied to estimate the specific power consumption for primary grinding. Limitations with respect to feed size (minus 32 mm), feed size distribution (proportion of minus 14 mesh), and lack of impact forces in the test has prompted the use of correction factors (MacPherson 1977). Its cost is comparable with that of the JK Tech test procedure, and the test can be used to test variability in ore characteristics. MacPherson’s approach to estimating the specific power consumption for the overall comminution circuit by using the Bond work index for ball milling, WiBM,without any contingency carries with it some risk, especially in situations where WiBMis significantly less than WiRMand possibly Wic. Thus, specific power consumptions for many conventional and traditional circuits can be determined by Bond grindability procedures, by various laboratory mills which have been calibrated against existing producers and, if need be, by pilot plant testing.

Circuit Selection In general, power is the most critical item in any grinding circuit evaluation due to the fact that the comminution circuit is probably the major power consumer in any mineral processing plant. While cost analysis may show that a circuit, which is less efficient in the use of power, can be more economical overall, the possibility of future power cost escalation requires careful examination so as to keep power consumption to a minimum (McDermott 1972, Barratt 1979, Bassarear 1981).

546

This becomes very important; especially if one considers that specific power consumptions can be up to 25% higher for power-efficient semi-autogenous or autogenous milling circuits for more competent and harder ores compared to traditional crushing and grinding circuits. Specific power consumptions for pebble milling can generally be as efficient as ball milling if one considers the power that is used to grind the pebbles as well. However, in real terms, the maximum power drawn by a given mill used for pebble milling could be of the order of 50% of the power used by the same mill if it were used as a ball mill, and hence the tonnage processed would be half (Kjos 1979). Rod mill-ball mill circuits offer the highest specific power efficiency and offer 95% or greater availability. This is marginally lower than a single-stage ball mill or multi-stage ball milling circuits, which offer approximately 97% availability, because of down time for rod charging. Primary ball milling circuits, which accept crusher product as feed, have an inherently lower specific power efficiency compared to normal secondary ball milling installations and a factor of 10%-20% has to be added to the standard power calculations (Rowland 1980). Grate discharge ball mills are capable of consuming 15% more power than overflow ball mills for the same specific power efficiency, but maintenance costs and availability would be less favourable. The advent of larger crushing units has prompted some projects to re-examine the necessity for rod mills in favour of primary ball milling circuits. The availability of SAGBall milling circuits ranges between 88% at commissioning to 94% at maturity, while for autogenous/pebble milling it could be equal or up to two-percentage points lower, depending upon circuit complexity and age of the plant with respect to mill drive type; e.g., gearless vs pinion or redwedpinion balanced against a usually higher mill liner availability. The Bond twin pendulum test has been developed to establish the size analysis of the crusher product and specific power consumption so that crushing and screening flowsheets can be designed to minimize capital and operating costs and maximize power efficiency (Flavel 1981). More sophisticated circuit controls, including the use of Expert Systems, have made mill operations more efficient and should be fully exploited. In summary the unit power cost, which is determined by the mode of power generation and location of the comminution plant, is probably the most significant factor that can influence the selection of equipment and plant design in order to keep overall power costs to a minimum and thereby achieve the most economical capital and operating cost. For projects forced to absorb high power costs, it is highly unlikely that potential savings in liner and media costs associated with autogenous grinding can offset the costs associated with potentially higher specific power consumptions compared to traditional crushing and grinding.

Effect of Circuit Selection on Metallurgical Efficiency With the standard traditional circuit, the principal questions are how much to grind per stage and whether to close circuit a mill or to operate in open circuit. Feed variation to the mill, although undesirable, could be tolerated at least from the perspective of the comminution circuit design. Autogenous grinding experience has highlighted the complications associated with changes in feed size and characteristics. If the ore breaks easily along the grain boundaries, the value minerals can be liberated at a coarser grind and at higher efficiency than with media grinding. With other ores, a lot of slimes could be generated. Another parameter that can often be overlooked in autogenous plant design is that for a given product specification, the mill may have 50% - 100% variation in feed rate (depending upon ore sizing, hardness, and other characteristics) that will have to be accommodated by the downstream metallurgical processes. The need for blending, for autogenous feed homogeneity, and close process performance monitoring is obvious. Generally, with both semi-autogenous and autogenous grinding, there have been significant improvements in minimizing the learning curve between start-up and attaining design throughput to a few weeks or months for a well-designed and engineered plant. Co-ordination and training of operating crews with commissioning teams is necessary to fully realize this benefit, particularly with respect to process control. Inevitably, some minor adjustments to the process flowsheet may have to be made. On the plus side, if an ore is wet and sticky and a real problem for crushing and

547

screening exists (or is envisaged), a SAG mill may be the answer and can override all other considerations. The cost benefit of time and relative production losses or gains for such a choice must be calculated (Bassarear 1981). Rod mills have been chosen in many cases because of their inherent feature of preferentially grinding down the top size and generating fewer fines than primary ball mills. This can affect magnetic recovery as in taconite flowsheets where cobber wet drum magnets follow the first grinding step. It is a metallurgical axiom that valuable minerals should be recovered from the stream as soon as they are liberated. Ideally, multiple grinding steps, each followed by a beneficiation step, would constitute a more metallurgically-efficient flowsheet than one-step grinding followed by one-step beneficiation. However, such a flowsheet might not be cost-efficient from the operating andor capital cost aspect, therefore some metallurgical or grinding power efficiency may have to be sacrificed to reach the optimum balance. Flowsheets for processing tin, iron, tungsten, chromite, lead, platinum, and tantalum ores are typical in this regard. When compared to autogenous grinding, systems using traditional grinding and steel media to mill complex sulphide ores can impair the flotation recoveries of copper, lead and zinc. It should be emphasized that this effect is not always observed with these types of ores; nonetheless, the physicochemical mechanisms that underline this behaviour continue to be studied. These investigators suggest that factors that also influence the results are ore type and iron content, the nature of the gangue minerals, the types of collectors and modifiers used, the degree of oxidation of the sulphides, and the oxygen potential of the pulp. The general consensus is that if such ores are competent enough to grind autogenously, then such a circuit should be investigated (Rey 1960, Thornton 1973, Iwasaki 1981).

General Cost Considerations for Various Circuits The objective of decreasing capital and operating costs has resulted in an increase in comminution equipment sizes to larger and larger units. Constraints in equipment size could be structural in terms of cast heads and trunnions, for example, rather than process. More attention is being paid to the benefits of large fabricated mills mounted on riding rings with peripheral discharge of pulp. Debate continues on the efficiency of larger diameter ball mills as there are indications that capacity can be limited to below the mill’s full potential if maximum power efficiency at the target grind size is the objective. Experience indicates that there could be a material flow or transport problem through the mill because flowrate is related to the available cross sectional area, which increases in proportion to the square of the mill diameter, and pulp density. In contrast, the power draw increases as the mill diameter to the power of approximately 2.5. Potential mill capacity is also a function of the mill diameter to the power 2.5. The larger diameter ball mills have experienced varying degrees of inherent inefficiencies involving ball charge level, mill speed, feed sizing, media size, liner configuration, length: diameter ratio, material residence time, pulp density, and mixing, all of which require examination before selection (Morrell2001). Other considerations are: 0

0

0

Plants with multi-unit single-stage primary mills could be more capital intensive than plants with two-stage grinding. This is because primary mills require expensive feeding and ore storage systems, generally with more lines in fine crushing and screening plants. However, single-stage comminution circuits cost less to operate as labour and maintenance costs are lower; Since pebble mill media invariably has a lower density compared to steel media, pebble mills are larger and more expensive for the same power as ball mills, but unit operating costs are usually less (MacPherson 1980); Semi-autogenous mills draw more power than autogenous mills of the same size and therefore are less expensive to install for the same throughput. They are more tolerant to feed variation and have proven to be a popular choice. However, they must be kept full of

548

ore otherwise liner and ball breakage will result. Surges in throughput are the result of the feed becoming finer and/or softer, therefore downstream equipment must be sized to accommodate these surges.

Water Supply, Source Availability and Cost versus Process Requirements Water is the normal media for physical beneficiation of the ore. However, there are areas in the world where water is scarce or unavailable. Some of these are located near the coast and seawater can be used. With others, dry pre-concentration is sometimes possible. Dry primary grinding, with drying of process air a necessity, might be followed by dry magnetic cobbing, or dry gravity concentration. The pre-concentrate is then shipped to an area where water is available for further processing. Dry AG/SAG mills, ball mills, and rod mills have been used; e.g., refractory gold ores ahead of roasting. They require more power and close temperature and moisture control and require sophisticated dust control equipment. Compared to a wet grinding environment, consumption of grinding media and liners is less. Many dry grinding installations also exist for material that cannot be wetted; e.g., cement clinker, talc, and goethitic direct shipping iron ores which are ground to pelletizing fineness and could never be filtered. These ores would be processed dry even if water were plentiful. Fine Grinding Concepts The amount of grinding necessary for an intermediate concentrate or a final concentrate to generate additional surface area can only be developed when the flowsheet is established and an upgraded product is available for testing. Whereas, previously, this may have seemed an easy assignment due to the fact that only a ball mill would have been used, contemporary options include tower mills, sand mills, and bead mills, etc. Care should be exercised when applying the Bond grindability calculations to such applications, as the Bond test is ideal for reduction ratios for approx. 6 : 1. Special tests need to be performed and some interpolations made, often using pilot test mills. In many instances, jar mills or Abbey mills are used and a timehurface curve is established. This kind of test mill can be calibrated with known ores and a specific power consumption can be determined. Whether to run the regrind mill in open- or closed-circuit depends on the product characteristics required and on current practice for similar applications. Plant Layout Considerations Familiarization with the geographic location, topography, climate, precipitation, accessibility, and the physical characteristics of the mine is paramount when designing a comminution plant. Whilst the location of the ore body cannot be changed, the physical layout of the process plant and the equipment selection can be adapted to take advantage of the conditions. For instance, under arctic conditions a compact plant design in one building offers opportunity to conserve energy and provide comfortable working conditions. Primary crushing, located close to the mine which may be at high altitude and subject to avalanche, could be separated from the fine crushing and grinding plant by underground conveyor to a safer point at a lower altitude or in underground caverns; e.g., Chilean copper projects. Roughness of terrain or inaccessibility may preclude the use of larger equipment; e.g., limitations of splitting crusher frames and mill shells into smaller components to pass road bed loadings or rail tunnel dimensions. Pre-concentration or crushing prior to shipment to a concentrator or re-processing plant is common in the iron ore industry, in which crushed ore or concentrate is railed a few hundred miles to a finishing plant located near tide water. Dry grinding may have to be used to avoid chemical attack of the grinding media or to circumvent water quality/quantity problems. Considerations that may apply to a large plant may not hold true for a small operation and in each case this becomes a question of economics. A small plant is a relative term, and it still has to

549

be large enough to be economic and this may vary with many factors, but comminution circuit design considerations may also be affected. For large plants, many of the service functions such as roads, railroads, warehousing, etc., are part of the basic design without which a large plant cannot be contemplated; i.e., there is such a thing as economics of scale. For smaller plants, factors such as accessibility or location may be a much more critical cost consideration. Supply of critical items necessary for daily operation, such as flotation reagents and grinding media, may be decisive factors in circuit selection especially if the economics of scale have been researched. More often than not, supply firms will set up special operations to service a major mine operation while small plants do not have that kind of leverage. The size and number of unit operations required for a smaller plant can influence the capital cost and operating cost significantly when one considers, for example, the space and labour requirements for a five shifts per week versus a ten shifts per week crushing plant producing the same feed tonnage. With large plants, equipment is manufactured to cope with larger incremental increases in throughput and there is often less flexibility in the number of shifts operated per week. One consideration, which is often overriding, is the effect of future plans for plant expansion on the selection of crushing and grinding equipment. For larger plants, processing lines are usually added whereas with smaller plants, an increase in ball charge volume and mill speed can achieve significant increases in throughput.

SPECIFIC SITUATIONS This section contains brief descriptions of selected grinding plants and the factors which influenced circuit selection.

Uranium Comminution practices in the North American uranium mining industry are presently confined to Canada, following the change to in situ leaching in the USA. The early producers in Canada processed ores from the Beaverlodge area of Saskatchewan and the Bancroft and Elliot Lake areas of Ontario. Their milling circuits were traditional crushing and grinding in the usual multi-stage fashion. Although the Elliot Lake conglomerates were capable of generating pebbles, SAG and pebble milling was not incorporated until later during plant expansions; e.g., the Quirke mill. Consequently, all the material was crushed and ground to the liberation size of the uranium mineral in the matrix, typically 50% minus 200mesh. The Madawaska mill at Bancroft incorporated pebble milling as part of its reactivation in 1978 to process a pegmatite ore. Grinding of uranium ores moved to semi-autogenous grinding in an attempt to reduce the release of radon gas into the atmosphere during the crushing process. By grinding in a wet environment with as coarse a feed as possible, radon is dissolved in water. The operation at Beaverlodge pioneered this development in 1964 with the installation of a double compartment mill fitted with a central peripheral grate discharge operating in closed-circuit with a static screen. Other operations that have taken advantage of this concept are Rabbit Lake, which had a SAG mill - ball mill circuit, McArthur River and Key Lake in Canada, and Bluewater, Panna Maria, Church Rock, and Bear Creek in the USA, all of which employed single-stage semi-autogenous grinding in closed-circuit with cyclones or screens. The Cigar Lake operation in Saskatchewan uses a waterflush crusher followed by a ball mill, both located underground (Edwards 2002). The requirements of government and industrial health monitoring agencies influence the design of uranium comminution circuits. Iron Comminution practices in the iron industry are some of the more varied. Iron ores can range in character from blocky magnetites like the Minnesota taconites, to specular hematite, similar to the Labrador and Quebec ores, and to weathered ores of goethite or limonite, which are lateritic in their physical behaviour, similar to the ores from Venezuela or Western Australia. Each particular ore type is processed using certain established methods of comminution. But even here the

550

practices can vary significantly as can be illustrated by the plants in Minnesota treating magnetic taconites. These ores average between 25% and 35% Fe and 60% to 70% of the total is magnetic ore. The ore is very hard and abrasive and must be ground to between 80% and 95% minus 44 microns for liberation. There are presently seven pelletizing operations: National Steel, Hibbing Taconite, Minntac, Eveleth, Ispat (Minorca), North Shore (Reserve), and Tilden. Hibtac evaluated various flowsheets with the goal of obtaining the lowest possible capital cost, and the provision that additional capital could be expended if the operating cost savings returned the additional capital within three years. Consequently, the Hibtac plant has single-stage autogenous milling in closed-circuit with cyclones. In subsequent operation, pebble crushing is now under investigation to cope with variations in ore type, and screens precede the cyclones. Both Empire (presently shut down) and Tilden utilize two-stage circuits with primary autogenous and pebble milling, and crushing of surplus pebbles. High pressure grinding rolls (HPGR) were used at Empire to fine crush crushed pebbles for recirculation. At Tilden, a ball mill is used to grind crushed pebbles. HPGR is used to re-crush autogenous mill oversize at SNIM, Mauritania (Patzelt 2001). The majority of HPGR in the iron ore industry are used in pellet feed preparation with moisture contents in the range 8.5% to 12%. Eighteen HPGR have been commissioned in this role since 1996 (Klymowsky 2002). Copper, Nickel, and Molybdenum Crushing and grinding practices used in the processing of copper, nickel, and molybdenum ores have been influenced by the following factors: Ore type. Whether the ore is a low grade porphyry or higher grade massive sulphide. Region. Whether plant design allows for the processing of wet sticky ores from open pits in high rainfall areas, frozen ore during prolonged winters or, in a more salubrious climate, multistage crushing and screening; i.e., taking advantage of the operating maxim that it is sometimes “cheaper to crush than to grind”. Throughput. Whether a particular concept for equipment selection and plant design at variable levels of throughput can demonstrate savings in capital and operating costs, minimize production losses and problems at start-up, and show a more attractive rate of return on investment compared to alternatives. Ore Characteristics. Whether the ore is hard and abrasive or soft and clayey. Technology. Whether there are limitations placed on equipment sizing and the number of units employed by the current state of technological development; e.g., mill length: diameter ratios and slurry transport rates in large diameter ball mills. Power Cost. Whether the overall specific power consumption for one concept is more or less in comparison to alternatives and in the last analysis contributes to a higher overall operating cost. Sampling. Whether the costs incurred for development, collection and testing of bulk samples of significant ore types within the ore body can demonstrate a satisfactory return on investment compared to cheaper alternatives, such as drill core samples, after consideration of other capital costs and risk. Until 1967, traditional crushing and grinding circuits were used to treat both massive sulphide and porphyry type ores with the number of stages being dependent upon operator preference and cost analysis. Commissioning of autogenous and semi-autogenous circuits in North America began in 1967 and since then other plants have come on stream in British Columbia, Labrador, Ontario, Quebec, Arizona, Colorado, Idaho, New Mexico, Nevada, Utah, and in other parts of the world, Argentina, Armenia, Australia, Chile, China, CIS, Finland, Indonesia, Mongolia, Morocco, Papua-New Guinea, Peru, the Philippines, South Africa, Sweden, and Turkey (Jones 2001). For the purposes of this paper, discussion of molybdenum ores will be included under the porphyry ores even though they are usually associated with granitic stocks; and copper-nickel sulphide ores are similarly associated with mafic rocks. Traditional practice varied based on operator preference and to some extent the differences can be assigned regional boundaries. In the porphyry copper provinces of the southwestern United

551

States and northern Mexico, single-stage ball milling found acceptance over rod milling-ball milling over a span of 40 years (Barratt 1982). Rod mill - ball mill plants were the normal choice for processing massive sulphide ores and, until 1971, were also chosen for large tonnage porphyry copper and molybdenum plants in Canada. As ore grades declined and operating costs escalated, attention was turned to the advantages of eliminating the secondary and tertiary crushing and screening plants through the use of autogenous and semi-autogenous grinding mills. Capital and operating cost savings have been significant enough to encourage many operators to take the plunge. Other factors that influenced the decision to use autogenous or semiautogenous grinding are highlighted for selected operations, at the risk of oversimplification and or omission: 0

0

0

Palabora, South Africa, processes an igneous complex containing magnetite, copper sulphides (chalcopyrite, bornite, cubanite, valleriite, and chalcocite), and some heavy minerals (uranothorianite, baddleyite) in carbonatite and foskorite members. Ultramafic rocks contain apatite, phlogopite, and vermiculite. There are two grinding divisions: single-stage autogenous and traditional rod mill - ball mill. Autogenous mill feed is ideally > 25% magnetite, > 40% plus 150 mm, < 25% minus 25 mm, and < 4% dolerite. Dolerite occurs in dyke swarms and much of it is sorted to waste in the pit. A prime reason for this sorting is the high rock strength and homogeneity of dolerite which builds up in the mill charge and results in lower mill throughput. Comparative rock strengths are (by rock type): carbonatite 113-159 m a , foskorite 65-99 MPa, and dolerite 360 m a . The autogenous mills are operated in closed-circuit with 2.5 mm aperture screens and 685 mm dia. cyclones to produce a coarse overflow to copper flotation (80% minus 300 microns) under conditions that minimize overgrinding. Heavy minerals are progressively recovered from copper flotation tailings. The carbonatite and foskorite mineral structure, inherent rock strengths, and magnetite content all promote the amenability of these ores to autogenous grinding. Rated throughput of the autogenous mills is currently 30,000 tpd and pebble crushing is required to deal with any accumulation of critical size (van Heerden 1996). Utah Mines Ltd, Vancouver Island, B.C., was located in an area of high rainfall averaging 80 to 100 inches per year. Due to the high clay content of the ore, operation of a fine crushing and screening plant would have been difficult. The original six Koppers autogenous mills were converted from single-stage autogenous to primary semi-autogenous operation and five secondary ball mills were added. The prime reasons for this conversion were to grind ore blends that were different from those piloted and to realize the design throughput of 33,000 stpd. The progressive addition of ball mills, and forced-air cooling of the SAG mill motors to accommodate a ball charge of 8% v/v (which was nominated in the mill specifications), enabled mill throughput to be increased to 38,000 stpd in the first instance, and then to the ultimate 55,000 stpd by 1988. Operations, which began in 1971, ceased in 1995 (Brown 1974, 1985). The Island Copper grinding circuits were unique in that the transfer size (new feed to the ball mills) was a proportion of primary cyclone underflow instead of a conventional screen undersize. Highland Valley Copper, B.C. (formerly Lornex) conducted pilot plant tests on 20,000 tons of ore over a period of nine months. Alternative circuits investigated were single-stage ball milling, single-stage autogenous, autogenous/semi-autogenous and secondary ball milling, and the traditional rod mill - ball mill combination. Bond work index (Wi,,) ranged from 13 to 22. Semi-autogenous grinding followed by ball milling offered the best results with regard to grinding media consumption, the circuit’s ability to cope with fault gouge mud seams, and ore variability.

552

0

Upon commissioning in 1972, steel balls were added until a 4% - 8% ball charge by volume was reached and steel consumption was 0.25 kg/t, which was twice that predicted in the pilot plant. Following expansions in 1981 (‘‘(2” Line) and 1989 with addition of the Highmont autogenous mills (ABC configuration) in “D’ and “E’ Lines, processing of Valley Copper ore since 1986, and implementation of a “mine to mill”/primary crushing optimization program, mill throughput had increased to 135,800 tpd by 2000 from 38,250 tpd in 1975. Ball charge volume in the SAG mills has been increased to 12% v/v. Valley and Lornex ores can be processed separately, but are normally blended in coarse ore stockpiles. Whereas the autogenous mills operate best on a predominance of Valley ore (porphyritic monzonites and granodiorites) which constitutes about 80% of plant feed, higher proportions of the less competent Lornex ore (quartz diorite, more altered, and with quartz porphyry and aplite dykes) can cause circulating loads through the pebble crushers to reach 600tph from the more normal 200 to 300 tph (Lornex Staff 1972, Wright Engineers 1972, McManus 1979, Meekel 2001). Other operations that followed a similar approach to testwork scope are Copperton, Collahuasi, El Teniente, Mt. Keith, and Mt. Isa Copper Plant. BatuHijau Restrictions on sampling for comminution testwork influenced the methodology for sizing the SAG mills, ball mills, and pebble crushers for a dual line, 120,000 tpd nameplate plant for this low grade copper-gold operation on the island of Sumbawa, Indonesia. Sampling criteria was dictated by minimal environmental disturbance and the project schedule prior to completion and approval of the feasibility study. This meant that only drill core could be utilized with insufficient quantities being available to permit pilot plant testwork for grinding. Underground development and bulk sampling, with storage on surface that would have required extensive disturbance of rainforest and many months’ duration, was precluded. Consequently, 5.11 km of core was obtained in stages with PQ, supplemented by HQ core at depth, and a total of 105 intervals were selected according to lithology and alteration types. Each interval was sampled for Bond’s low energy impact crushing, rod milling, and ball milling work indices, and the abrasion index. These indices were composited according to mining depth and bench level for each mine year up to year 10 for mill sizing using a power-based model. Since the relationship Wic < WiRM > WiBMwas consistent, pebble crushing was designed into the circuits (Barratt 1996). The predominant ore types are coarse-grained diorite and tonalite, coarse-grained volcanics, and fine-grained volcanics, and each ore type is predicted to become harder over time. Accordingly, a future expansion was factored into the plant design. Following commissioning and ramp-up, design throughput was attained during the second year of operation. Optimization of SAG mill shell linedlifter and grate design, SAG mill grinding ball size, and pebble crushing circuit design contributed to this achievement and continues to offer upside potential. Grinding is in sea water (MacLaren 2001). Other operations that followed a similar approach to testwork scope include Alumbrera, Ernest Henry, and Cadia Hill. Freeport One of the unique applications of semi-autogenous grinding is at Freeport in Indonesia. Mined ore is fed to the SAG mills after passing through 1,000 m of vertical ore passes through which primary crushed ore degrades to 100 mm top size. The variable speed SAG mills, one 10.4 m (34 ft dia.) and the other 11.6 m (38 ft dia.), together process 168,000 tpd and have been operating with high ball charges, 17% v/v and 19% v/v respectively, to accommodate the low ore charge volume of finely sized ore. Needless to say, one of the prime reasons for selecting SAG/Ball mill circuits for the C3 and C4 expansions was to construct as much grinding capacity as possible in a limited

553

0

available space, with the largest proven unit sizes of milling equipment available at the time (Staples 2001, Coleman 2001). Henderson Single-stage SAG mills in closed-circuit with cyclones have been processing a granitidgranodiorite hosted molybdenum ore for over twenty-five years. The circuit was designed following extensive pilot plant testwork. Five grinding lines have been operating with progressive optimization of ball charge volume and feed rate according to prevailing market conditions for molybdenum. Rather than shut a grinding line down, it is more profitable to reduce feed rate and realize higher recoveries of molybdenite due to the finer circuit product sizings in flotation feed. Continuing improvement in operating efficiency was realized by testing the effects of changes in ball charge volume, ball size, mill speed, and shell linedlifter design (Hinken 1982, Wood 1996). Inco Clarabelle As part of the Mills Rationalization Program (1988 - 1991), production in Sudbury, Ontario from the Frood Stobie mine was transferred to the Clarabelle Mill and the FroodStobie Mill was shut down, The additional throughput was achieved by the addition of a SAG mill, screen classification, and pebble crushing circuits to the existing rod mill - ball mill circuits. Built-in flexibility of the design allowed the sharing of existing ball mills between the rod mills and SAG mill for a total milling capacity of 40,000 tons per 24 hours. Pilot plant test results for SAG milling and pebble crushing were supplemented by projections from analysis of the existing grinding operations to generate design criteria for sizing the SAG mill. Subsequent operation has shown that production objectives can be met and that rationalization of pebble handling and crushing is continuing (Wright Engineers Limited Reports 1991).

Lead and Zinc Modern practice for lead-zinc and copper-lead-zinc ores can be considered as conventional with variations on a theme. TeckCominco’s Red Dog operation in Alaska utilizes SAGBall milling and is the world’s largest producer of zinc concentrates (Lee 2001). The most recent operation at Antamina in Peru has a capacity of 80,000 tpd and has one 11.6 m (38 ft dia.) SAG mill and three 7.3 m (24 ft dia.) ball mills, all with variable speed gearless drives designed to cope with variations in ore competence. Brunswick Mining and Smelting has recently converted to AG/SAG milling as a cost saving method (Larsen 2001). Mineralogy and grain size play a large part in the development of flowsheets, especially with the more complex ores. Since flotation is the most common method of concentration, most operations strive to float the lead from as coarse a feed as possible to avoid sliming, and rely on regrinding to upgrade concentrates for either lead, zinc or both. Some Canadian operations find it necessary to conduct primary copper or lead flotation and regrind the tailing prior to scavenger flotation followed by zinc flotation. Gold, Silver, Mercury, and Platinum Group Semi-autogenous grinding of precious metals has been practised in the USA since 1975 and extended into Canada in 1984. Installations at McDermitt (mercury), Nevada, DeLamar, Idaho, and in the Carlin Trend were designed to process ores of high clay content and thereby avoid the problems associated with fine crushing of clay ores. Dry grinding in a dual compartment rotator prepares feed for roasting refractory gold ores in the Carlin Trend at Newmont’s Gold Quarry operation and at Barrick Goldstrike (Thomas 2001). The Escalante Silver Mine in S.W. Utah was designed in 1980 to use semi-autogenous grinding on the basis of realizing capital and operating cost savings for a 500 tpd plant. The ore was competent and very siliceous. Other operations that have since taken advantage of SAGBall milling, with or without pebble crushing, are TeckCominco’s Hemlo and Williams (with Barrick), North American Palladium, Barrick’s Holt McDermitt in Ontario, Inmet’s Troilus in Quebec (Sylvestre 2001), Stillwater

554

Platinum in Montana, Kennecott’s Ridgeway in South Carolina, and Fairbanks Gold Mining Company at Fort Knox (Magnusson 2001), Alaska, and in Australia, Newcrest’s Cadia Hill (Hart 2001) and KCGM’s Fimiston (Karageorgos 2001). In contrast, ROM autogenous grinding and pebble milling has been practised in South Africa on conglomerate gold ores from underground and finer feed top size for many decades, with more recent conversion to SAG milling, using low aspect (mill diameter: length ratio) mills. Platinum mines in the Transvaal employ autogenous or SAG milling in closed-circuit with fine screens and stage grinding with ball mills, alternating with stages of flotation, as a means for optimizing liberation and minimizing overgrinding.

Tungsten, Tin, Chromium, Tantalum, and Niobium The minerals wolframite, scheelite, cassiterite, chromite, tantalite, and pyrochlore are not only friable compared to quartz they are also denser. Specific gravities range from a maximum of 7.9 for wolframite down to 4.2 for pyrochlore. Two properties, friability and gravity govern the design of grinding and classification circuits for the preparation of feed to subsequent concentration processes. The principal objective in processing ores of these minerals is to eliminate slimes generation other than that by natural degradation. Removal of the liberated mineral as soon as it is generated is of prime importance. However, due to the effect of high specific gravity on hydraulic classifiers, which would result in the recycling of liberated minerals back to the grinding circuit, screens are almost universally used to close the grinding circuits. The Canada Tungsten operation in the Northwest Territories of Canada treats ore that contains scheelite in iron and copper sulphides, limestone, and pyroxene. This plant has recently (February 2002) been re-commissioned after fourteen years shutdown and is now owned by North American Tungsten Corporation Ltd. The crushing plant is a traditional three-stage with jaw crusher, standard cone crusher and short head cone crusher with primary and secondary screens. The grinding circuit consists of twostage rod milling with cyclones and screens closing the second stage. The primary mill is opencircuited and discharges into the cyclone feed pump box together with the secondary mill discharge. Cyclone underflow is sized on a trommel screen fitted with a 50 mesh cloth. Screen oversize reports as feed to the secondary mill. Most of the power used in this circuit is applied to the primary mill (450 hp) with 250 hp being applied to the secondary mill. This is the reverse of what might be expected, but it ensures that the power is applied in the first stage to achieve particle liberation of scheelite at about 50% minus 65 mesh before sliming becomes a problem. The choice of a smaller rod mill for the second stage ensures that a short residence time in the mill mitigates against slimes generation. An interesting feature of the calculation of primary milling power is that, with a reduction ratio of 40:l and the expected low power efficiency, the power requirement and product size distribution are exactly as would have been predicted from standard calculations, provided the high ratio of reduction factor for rod milling, EF6, is ignored. The tailing from gravity concentration is reground in a ball mill prior to recovery of copper sulphides and remaining scheelite by flotation (Bolu 1984). At Tantalum Mining Corporation of Canada Limited, Bernic Lake, Manitoba, the original grinding circuit consisted of a single-stage rod mill in closed-circuit with DSM screens at the head of an all-gravity concentrator. Another smaller rod mill was available as a regrind mill. Over the years, the ore grade decreased and the tantalum minerals became more finely disseminated. The primary rod mill was converted to a grate discharge ball mill and a middling regrind ball mill was installed to take up to 25% of the feed as sand table middlings. It is important in such circuits to remove tramp metal and abraded iron prior to concentration in order to maintain concentrate grades (CIM 1978). The Niobec Inc. concentrator at St. HonorC, Quebec, Canada, employs a traditional rod millball mill circuit using DSM screens instead of cyclones to close the ball mill circuit. Rod mill feed

555

is prepared by an open-circuit crusher with no surge capacity between the two. Concentration by flotation follows desliming in cyclones (CIM 1978).

Diamonds HPGR is used extensively as a re-crusher following pre-concentration from kimberlites. In three applications, Argyle, Western Australia, Ekati, N.T., Canada, and Premier, South Africa, HPGR is used for both primary and secondary crushing in the extraction of diamonds (Klymowksy 2002). In one operation located in Siberia, a 9.75 m (32 ft) dia. rubber-lined SAG mill is processing a diamondiferous kimberlite, often in frozen ore conditions (Dubiansky 2001). Laterites Laterites vary widely from one location to another but share many common characteristics that are of interest to the comminution engineer. Laterites are weathered ores that contain a high percentage of ultra fines, usually with high amounts of moisture (both surface and combined) and with various quantities of solid rock inclusions. Such rock is either rejected or processed depending on its mineral content. Even if boulders are disposed of, they still have to be crushed, scrubbed or dried to facilitate handling and to maximize mineral recovery. The sticky nature of the ore is the main consideration when designing crushing and screening facilities. One plant uses wobbler feeders and a rubbler, which is an autogenous mill fitted with peripheral grate discharge and low lifters. This design avoids breaking the boulders and cleans their surfaces to release higher grade fines for processing. Residual boulders are discarded to waste through the open-ended trunnion (Ferguson 1979). In some lower grade cases, boulders are crushed to release higher grade fines and so increase feed grade at a reduced content of MgO. Thixotropic properties of some laterites have dictated operation at a specified elevated temperature; e.g., Murrin Murrin in Western Australia, ahead of hydrometallurgical treatment. In the alumina industry, bauxite is first crushed then wet ground in ball mills using the return caustic liquor from the evaporators as the liquid medium. More recent operations utilize SAG mills; e.g., Alcoa at Wagerup in Western Australia. Limonitic or goethitic iron ores are processed into direct shipping lump ores or sinter feed fines. Proper selection of screens and crushers and careful design of the material handling components (bins, transfer points) is of utmost importance. The ores of Venezuela’s Ferro-Minera Orinoco are screened and crushed in cone crushers. In the wet season screening and handling can be a problem; e.g., to screen at 9 mm, the feed is first dried to approximately 5% moisture. CONCLUSIONS The factors which influence the selection of comminution circuits are varied and are dependent upon the nature of the project, whether it is a greenfields plant or an expansion, as well as on a through understanding of ore characteristics and scoping of testwork at each stage of study. The comminution engineer has many test procedures to draw upon, some of them relatively new, such as QEM*SEM or QemSCAN to define mineralogy and liberation characteristics of an ore, and the Minovex SPI to delineate variability within an orebody. Consequently, much useful information for defining a grinding circuit can be made available economically and on a timely, cost-effective basis.

556

Crushing and grinding circuit options that have formed the background of this paper are as follows:

Table 1 Crushing and grinding circuit options Traditional Circuits Crusher, Rod Mill, Ball Mill Crusher, Rod Mill, Pebble Mill Crusher, Single-Stage Ball Mill Crusher, Multi-Stage Ball Mill

Conventional Circuits Autogenous Semi-Autogenous Single-Stage Autogenous Single-Stage Semi-Autogenous Autogenous, Ball Mill Semi-Autogenous, Ball Mill Autogenous, Ball Mill, Semi-Autogenous, Ball Mill, Crusher Crusher Autogenous, Pebble Mill Pre-Crushing, SemiAutogenous, Pebble Mill, Autogenous, Ball Mill, Crusher Crusher

Since the predecessor paper on this subject in 1982, 200 grinding operations in non-ferrous and ferrous mines have been designed on the basis of some form of semi-autogenous grinding circuit (Jones 2001), with special situations involving autogenous grinding, pebble milling, and high pressure grinding rolls. In a few instances, traditional circuits have been employed because of more favourable technical and economic reasons. The factors which influence the selection of comminution circuits are summarized in Table 2. Brief descriptions of selected grinding plants and the factors which influenced circuit selection are provided to illustrate operating practice for uranium, iron, copper, nickel, molybdenum, lead, zinc, gold, silver, mercury, platinum group, tungsten, tin, chromium, tantalum, and niobium ores, diamond and laterite processing.

ACKNOWLEDGEMENT The authors acknowledge the contributions of the many companies and individuals who are referenced in the paper, and who have collectively advanced the art of comminution.

557

Table 2 Summary of factors of interest, information gained, and effect on circuit selection Factors of Interest Geological Interpretation of Drill Core and Bulk Samples

Information Gained Identification and Relative Mineral Constituents Degree of Dissemination Types of Lithology Types of Alteration Degree of Oxidation Geotechnical Competence Hardness.

Effect on Circuit Selection Abundance

of a

a

Mineralogical Analysis

Identification of Ore and Gangue Minerals, and Middling Associations Liberation and Modal Analyses QEM*SEM/QemSCAN Analyses.

Chemical Analysis

Identification of Metallics, Non-Metallics, and Acid-Generating Constituents.

Physical Properties

Hardness, Blockiness, Friability, Quantification of Primary Fines and Clay Content Specific Gravity of Mineral Constituents.

Circuit Feed Parameters

ROM Top Size Parameters Primary Crusher Discharge Size Analysis Throughput Requirement and Schedule Mining Plans, Schedules, Methods,

Determines Ratios of Reduction, Feed, and Product Size Analyses in Primary, Secondary, and Regrind Circuits.

a

Determines Requirement for Pre-Washing the Ore.

Provides a Guide to Potential Problems in Crushing, Screening, and Grinding the Ore with respect to Selection of Traditional, AG, or SAG Milling, Over-grinding and Avoidance of Slime Generation with respect to the Softer Minerals of Sn, W, Ta, Nb, Pb. Cr. a

and

Provides a Guide to the Type(s) of Circuit(s) Required, and the Types of Samples Required based on Precedent Determines the Necessity of Separate Plants to Process Sulphide and Oxide Ores Provides a Guide to the Selection of Autogenous Grinding.

Determines Selection of Primary Crushers, and Necessity for Pre-Crushing can Influence this Selection by Determination of the Product Size at the Required Throughput Rate.

Factors of Interest Sampling Requirements

Effect on Circuit Selection

Information Gained Preliminary Drill Core for Resource Definition and Split for Bond Work Indices, WiRM and WiBM,and Twin Samples for SPI Whole Core for Autogenous Media Competency Index, Impact Crushing Work Indices, Wi,-, and WiRM,WiBM,Ai, PLI, UCS, RQD, and Fracture Frequency Bulk Samples, Large Diameter Drill Core (6-in. to 8-in.), Open Pit or Underground, for Pilot Plant Testing.

*

Preliminary Assessment of Grinding Requirements and Ore Variability Power - based Methods for Mill Sizing, using Results from Bond Impact and Grinding Work Indices and Interpretation of Autogenous Media Competency Index, SPI, Bond Abrasion Index, and Geotechnical Parameters, can Distinguish AG from SAG Milling, and Assist in Definition of Pilot Plant Test Program and Ore Variability Characteristics.

Contiguous Properties

Definition of Equipment Characteristics.

Feed and Product Specifications

Definition of Requirements Comminution Stage.

Each

Influence of “Mine-to-Mill” and Choke Feeding the Primary Crusher on AG/SAG Mill/Pebble Crusher, Secondary Grinding Circuit Performance (Power Efficiency, Throughput, Development of Critical Size, Circulating Loads, etc.) Maximum Feed Top Size in Relation to High Aspect and Low Aspect Primary Mills UseofHPGR.

Bond Work Indices, Abrasion Index, and Specific Power Consumptions

Calculation of Specific Power Consumption at each Comminution Stage for Different Ore Types and Composites, either from Bond Work Indices, or Pilot Plant Testwork, or Both Assessment of Ore Variability (along with SPI) Checks on Pilot Plant Test Data Assessment of Risk or Contingency using SPI and Bond Work Indices results based on Samples selected according to the Mine Plan.

Confirmation of Specific Power Consumptions and Contingencies for Process Design Criteria Calculation of Estimates for Media and Liner Wear and Potential Investigation of AG vs SAG Milling in Pilot Plant Estimation of Mill Power Requirements and the Distribution of Power between Crushing and Grinding Stages to Account for Variation of Operating Conditions, Transfer Sizes, and Production Targets.

at

Determines the Utility of Equipment with respect to its Inherent Operating Behaviour; e.g., Autogenous Mills grinding to a Natural Grain Size, SAG Mills Breaking Across Grain Boundaries, and Rod Mills Minimizing the Creation of Fines.

Factors of Interest

Effect on Circuit Selection

Information Gained

Circuit Selection

0

Metallurgical Efficiency

0

0 0

Assessment of Overall Power Requirement and Power Efficiency for Different Circuit Options Assessment of Overall Operating Availability for Different Circuit Options Determination of Unit Power Cost (and its Predicted Escalation) and Demand for Different Circuit Options.

Determination of the Most Economic Option on the Basis of NPV of Capital and Operating Cost (not Necessarily on the Basis of the Most Power-efficient Circuit), and Circuit Availability for a Fixed Revenue Rate. Power Efficiency should be Minimized in Design for Each Circuit Option Considered; e.g., whether a Pebble Crushing Circuit is Included or Not.

Definition of Optimum Comminution Configuration Definition of Feed Rate Variation Selection of Grinding Media.

Determination of Necessity for Stage Grinding and Stage Concentration to Optimize Mineral Liberation and Recovery, use of Autogenous Grinding to Liberate Discrete Grains Quantify the Effect of Variation in Feed Rates with AGISAG Milling on the Metallurgical Efficiency of Downstream Processes Realization of Potential Benefits of Autogenous Grinding vs Steel Media on Concentrate Grades and Recovery from Complex Sulphide Ores Use of Rod Mills to Minimize Generation of Fines.

0

0

Cost Considerations

0

Definition of Largest Practical Equipment Size and Design Differences between Comminution Options. 0

0

Effect on the Efficiency of Crushing and Grinding Equipment; e.g., Choke Feeding Crushers and Feeding Arrangements, Separation of Screening Plant from Crushing Plant Avoidance of Short-circuiting in Large dia. Ball Mills and SAG Mills Advantages of Large Fabricated Open-ended Discharge Mills on Grinding Efficiency in Certain Situations compared to Possible Limitations of Trunnion-Mounted Mills Lower Operating Cost of Pebble Mills vs Ball Mills must be balanced against Larger Pebble Mill Sizes for Equivalent Power vs Ball Mills Advantages of Variable Speed Mills to Accommodate Variations in Feed Rate and Ore Hardness, and their Effects on Sizing Equipment for Downstream Processes.

Factors of Interest

Information Gained

Effect on Circuit Selection

Water Supply

Definition of Process Alternatives.

Determination of Plant Location vis A vis Mine Location. Applicability of Dry Grinding, Pre-concentration, and use of Sea Water

Fine Grinding

Definition of Test Requirements, Batch- and/or Pilot-Scale Tests.

Determination of Optimum Location(s) of Fine Grinding Application within Circuit, and Definition of the Type(s) of Machine used

Plant Layout

Definition of Geographic Location, Climatic Conditions, Accessibility Definition of Relative Location of the Mine vs Plant Definition of Operating Schedules and Manpower Requirements Definition of Expansion Potential.

Determination of Wet and Dry Processes Determination of the Physical Sizes of Equipment and Footprint of the Plant, the Number of Building (Unit, if uncovered) Components and their Locations Determination of Built-in Contingencies that allow for Future Expansion in the Case of Small Plants, and Site Layout Considerations for the Addition of Equipment Lines in the Case of Larger Plants; e.g., Coarse Ore Stockpile Layout.

CJl

P

REFERENCES Barratt, D.J., and Sochocky, M.A. 1982. Factors that Influence the Selection of Comminution Circuits. In Design and Installation of Comminution Circuits, eds. A.L. Mular and G.V. Jergensen 11, Chapter I. New York: SME-AIME. Ashley, K.J. 2002. Sampling a Mineral Deposit for Feasibility Studies. Proceedings 2002 Mineral Processing Plant Design, Operating Practice, and Control Conference, eds. A.L. Mular, D.N. Halbe, and D.J. Barratt, Paper A-1. Barratt, D.J., and Allan, M.J. 1986. Testing for Autogenous and Semi-Autogenous Grinding: A Designer’s Point of View. Miner. Metallur. Process. 3 : 65. Barratt, D.J. 1992. Design of Pebble Crushing Circuits. SME Annual Meeting. Amelunxen, P., Bennett, C., Garretson, P., and Mertig, H. 2001. Use of Geostatistics to Generate an Orebody Hardness Dataset and to Quantify the Relationship between Sample Spacing and the Precision of the Throughput Predictions. Proceedings International Autogenous and SemiAutogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, IV : 207. McDermott, W.F., Lipovetz, G.J., and Peterson, H.R. 1972. The Dollars and Sense of Autogenous Grinding. SME-AIME Annual Meeting. Dance, A. 2001. The Importance of Primary Crushing in Mill Feed Size Optimization. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 189. Sherman, M. 2001. Optimization of the Alumbrera SAG Mills. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 59. Parker, B., Rowe, P., Lane, G., and Morrell, S. 2001. The Decision to Opt for High Pressure Grinding Rolls for the Boddington Expansion, Proceedings International Autogenous and SemiAutogenous Grinding Technology, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I11 : 93. Sochocky, M.A., and Mok, J.K. 1972. Coarse Grinding in a Ball Mill. SME-AZME Annual Meeting. Rowland Jr., C.A. 1970. Applying Large Grinding Mills. Proceedings Pacijic Southwest Minerals Conference. Flavel, M.D., and Rowland Jr., C.A. 1981. Selecting Circuits to Prepare Beneficiation Circuit Feed from Primary Crusher Product. SME-AIME Fall Meeting. Rowland Jr., C.A., and Kjos, D.M. 1980. Rod and Ball Mills. In Mineral Processing Plant Design, eds. A.L. Mular and R.B. Bhappu, 2”ded., Chapter 12. New York: SME-AIME. Barratt, D.J., and Brodie, M.N. 2001. The “Tent” Diagram, What it Means. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, IV : 368.

562

Barratt, D.J. 1989. An Update on Testing, Scale-up and Sizing Equipment for Autogenous and Semi-Autogenous Grinding Circuits. Proceedings Advances in Autogenous and Semi-Autogenous Grinding Technology, eds. A.L. Mular and G.E. Agar, 1 : 25. Hanks, J.T., and Barratt, D.J. 2002. Sampling a Mineral Deposit for Comminution and Metallurgical Testing. Proceedings 2002 Mineral Processing Plant Design, Operating Practice, and Control Conference, eds. A.L. Mular, D.N. Halbe, and D.J. Barratt, Paper A-2. Powell, M. 2002. South African Progress on Closing the Design Gap between High - and Low Aspect SAG Mills. Proceedings 341h Annual Meeting of Canadian Mineral Processors. 12 : 189. CIM. JKMRC Commercial Division. 2001. JK SimMet Steady State Mineral Processing Simulator Version 5.1. MacPherson, A.R. 1977. A Simple Method to Predict the Autogenous Grinding Mill Requirements for Processing Ore from a New Deposit. Transactions, Vol. 262. SME-AIME. Barratt, D.J. 1979. Semi-Autogenous Grinding: A Comparison with the Conventional Route. CIA4 Bulletin. 72 : 74. Bassarear, J.H. 1981. Autogenous and Semi-Autogenous Grinding Practices. SME-AIME Fall Meeting.

Kjos, D.M. 1979. Grinding Circuits Current Status and Projected Future Developments, an Allis Chalmers publication. Proceedings 52"' Annual Meeting of the Minnesota Section. AIME. Flavel, M.D., and Rimmer, H.W. 1981. Particle Breakage Studies in an Impact Crushing Environment. SME-AIME Annual Meeting. Rey, M., and Formanek, V. 1960. Some Factors affecting Selectivity in the Differential Flotation of Lead-Zinc Ores, particularly in the presence of Oxidized Lead Minerals. Proceedings 5" International Mineral Processing Congress. IMM. Thornton, E. 1973. The Effect of Grinding Media on Flotation Selectivity. Proceedings 51hAnnual Meeting of Canadian Mineral Processors. CIM. Iwasaki, I., Reid, K.J., Lex, H.A., and Smith, K.A. 1981. The Effect of Autogenous and Ball Mill Grinding on Sulphide Flotation. SME-AIME Fall Meeting. Morrell, S. 2001. Large Diameter SAG Mills Need Large Diameter Ball Mills - What are the Issues? Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I11 : 179. MacPherson, A.R., and Turner, R.R. 1980. Autogenous Grinding from Test Work to Purchase of a Commercial Unit. In Mineral Processing Plant Design, eds. A.L. Mular and R.B. Bhappu, 2"d ed., Chapter 13. New York: SME-AIME. Edwards, C.R. 2002. The Cigar Lake Project - Mining, Ore Handling, and Milling. Proceedings 341hAnnual Meeting of Canadian Mineral Processors. 40 : 675. CIM.

563

Patzelt, N., Klymowsky, R.I.B., Burchardt, E., and Knecht, D. 2001. High Pressure Grinding Rolls in AG/SAG Mill Circuits - The Next Step in the Evolution of Grinding Plants for The New Millenium. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I11 : 107. Klymowsky, R.I.B. 2002. E-mail communication. Jones Jr., S.M. 2001. Appendix of AG-SAG Installations World Wide. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 373. van Heerden, J.J. 1996. Development of Autogenous Milling at Palabora. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, I : 123. Brown, C.M. 1974. Island Copper Mine, Milling for Copper and Molybdenum, Western Miner. Brown, C.M., and Pipke, M.A. 1985. The Island Copper Grinding Circuit - A Progress Review. SME-AIME Fall Meeting. Staff, Lornex Mining Corporation Ltd. 1972. Western Miner. Wright Engineers Limited Staff. 1972. Minutes of Fall Meeting. Canadian Mineral Processors (B.C. Section). CIM. McManus, J. 1979. Grinding Copper Ores in British Columbia, Proceedings Autogenous Grinding Seminar, Volume 2. Trondheim, Norway. Meekel, W., Adams, A., and Hanna, K. 2001. Mill Liner Development at Highland Valley Copper. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I11 : 224. Barratt, D.J., Matthews, B.D., and DeMull, T. 1996. Projection of SAG/AG Mill Sizes, Mill Speeds, Ball Charges, and Throughput Variation from Bond Work Indices. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, I1 : 541. MacLaren, D., Mitchell, J., Seidel, J., and Lansdown, G. 2001. The Design, Start-up and Operation of the Batu Hijau Concentrator. Proceedings International Autogenous and SemiAutogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, IV : 316. Staples, P., Siewert, H., Stuffco, T., and Mular, M. 2001. SAG Concentrator Improvements at PT Freeport Indonesia. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 91. Coleman, R., Nugroho, S., and Neale, A. 2001. Design and Start-up of the PT Freeport Indonesia No. 4 Concentrator. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 76. Hinken, W.R. 1982. Single-Stage Semi-Autogenous Grinding at the Henderson Mine. Proceedings XIV International Mineral Processing Congress. I : 1.1.

564

Wood, C. 1996. Single-Stage SAG Grinding Experience at Henderson. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, I11 : 200. Wright Engineers Limited. 1991. Reports to INCO. Lee, K.A., and Kojovic, T. 2001. Comparison of Rubber and Steel Liners at the Red Dog P b E n Concentrator. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I11 : 288. Larsen, C., Cooper, M., and Trusiak, A. 2001. Design and Operation of Brunswick’s AG/SAG Circuit. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, IV : 350. Thomas, K.G., Buckingham, L., and Patzelt N. 2001. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 174. Sylvestre, Y.,Abols, J., and Barratt, D.J. 2001. The Benefits of Pre-Crushing at the Inmet Troilus Mine. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I11 : 43. Magnusson, R., Hollow, J., Mosher, J., and Major, K. 2001. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 159. Hart, S., Valery, W., Clements, B., Reed, M., Song, M., and Dunne, R. 2001. Optimization of the Cadia Hill SAG Mill Circuit. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 11. Karageorgos, J., Skrypniuk, J., Valery, W., Ovens, G. 2001. SAG Milling at the Fimiston Plant (KCGM). Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I : 109. Bolu, M., Bouma, P., and Paterson, J. 1984. Rod Mill Grinding for Improved Gravity Recovery at the Canada Tungsten Scheelite Concentrator. SME-AIME Annual Meeting. Milling Practice in Canada. 1978. In CIM Special Volume, ed. D.E. Pickett, Chapter 9. Montreal. CIM. Dubiansky, G., and Ulrich, P. 2001. Logistics of Designing, Transporting and Installing a Large Autogenous Mill in a Siberian Diamond Mine. Proceedings International Autogenous and SemiAutogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I1 : 217. Ferguson, B .A., Camposano, G., and Aponte, J. 1979. Falconbridge Dominicana Ore Handling and Preparation. International Luterite Symposium, eds. D.J.I. Evans, R.S. Shoemaker, and H. Veltman, Chapter 8. New Orleans. SME-AIME.

565

Types and Characteristics of Crushing Equipment and Circuit Flowsheets Ken Major, HATCH Associates Ltd., Vancouver,BC, Canada

ABSTRACT Crushing flowsheets and equipment are selected to prepare ore for downstream processes. The crushing equipment standard to the minerals industries has been jaw crushers, gyratory crushers and cone crushers. Recent trends in processing have allowed mine operators to target lower grade ores and refractory ores. Low metal prices have required operators to search for lower cost processing opportunities through higher equipment efficiencies and new equipment developments. The realm of crushers in the minerals industry has expanded to include waterflush cone crushers, vertical and horizontal impactors and high pressure grinding rolls. This paper will look at application of the various types of crushers and how they are incorporated into recent crushing circuits. INTRODUCTION Crushing is an integral portion of the comminution flowsheet for mineral processing operations and is critical for the preparation of ore for downstream processing. The selection of the right crushing equipment for a specific application is influenced by many factors some of which are upstream of the crushing plant (blasting pattern and mining method) and others which are downstream of the crushing plant (heap leach or mill, grinding circuit selection). For most applications there is a flowsheet previously designed that will match the requirements. The cost to recover metal from ores continues to increase and at the same time fewer high grade deposits are being discovered. Designers and manufacturers, plant operations and maintenance personnel and plant designers continue to look for opportunities to enhance the plant operation with better equipment designs and/or unique applications that improve the economics of the operation. With the improvement in materials of construction larger crushers with more horsepower, higher speed and higher throughput have been designed. The equipment designers have attempted to make the new designs compatible with the smaller, older models to make it possible to upgrade existing facilities and minimize capital cost. With the growth in SAG mill grinding circuits the cone,crusher was eliminated from most comminution flowsheets. This trend has been reversed with the addition of a cone crusher in a SAG circuit (SABC) to crush the recirculating pebbles. These pebbles tend to be more resistant to impact breakage in the SAG mill and the crusher creates ore surfaces more conducive to breaking and grinding in the SAG mill. When compared to feed for a single stage ball mill circuit the cone crusher product would be conducive to ball mill grinding. In most cases the crusher has been retrofitted to the circuit and neither the means to feed the material to the ball mill nor sufficient horsepower exists. A crusher has a more efficient transfer of applied power to the breakage of rock than a grinding mill. This has been seen with the development and acceptance of the SABC circuit and also led to Nordberg’s (Metso Minerals) development of the “Waterflush” cone crusher. The application of Waterflush technology has resulted in the production of a finer product than normally achievable with a shorthead cone crusher.

566

The high pressure grinding rolls (HPGR) have been included in most diamond processing flowsheets as they have been effective in crushing the Kimberlite ore to liberate the diamonds while minimizing diamond breakage. The HPGR has had recent successes in the iron ore industry as a substitute for the AG or SAG mill in preparing ball mill feed. The equipment suppliers are working very hard to adopt this technology for base metal and precious metals operations. The primary application proposed for the HPGR will be to provide one more stage of crushing prior to the rod mill or single stage ball mill reducing the unit energy consumption.

FACTORS AFFECTING CRUSHER SELECTION The factors that affect the selection of size and type of crusher for a specific application include:

0

0 0 0 0

Plant throughput, ore delivery schedule Sizeof feed Desired product size for downstream processing Ore characteristics; hard rock, clay, gravel, variability, etc. Climatic conditions Downstream processes

Plant Throughput and Ore Delivery Schedule

Plant throughput and ore delivery schedule will form the baseline for flowsheet design and equipment selection. From this information the size, type, number of stages and number of crushing units per stage required for an application can be identified. For example, a primary jaw crusher will be better suited for a conventional underground mining operation because: 0 0 0

Tonnages are typically lower Feed material size is smaller, more fines are generated in blasting and mucking Less headroom and a smaller excavation is required

Higher throughput, open pit operations will typically use one or more gyratory crushers for the primary crusher application and product will be delivered to the mill feed stockpile for delivery to a SAG mill grinding circuit. Throughput limits for the design of secondary and tertiary crushers have a significant impact on the capital costs and operating costs for higher tonnage operations because of the number of crusher units per stage required to address capacity issues. As a result as plant throughputs increased the fewer the 3-stage crushing plants that were designed. Process operations are normally designed for nominal throughputs. Grinding circuits generally achieve availabilities of 94% (SAG) to 98% (Rod Mill / Ball Mill). Crushing circuit availability will typically be 75% to 85% depending on the complexity of the circuit and application of surge piles and/or bins. Availability is a measure of time that the crushing circuit is in operation. The two components that impact the availability are scheduled downtime and unscheduled downtime. Scheduled downtime is typically attributed to routine plant maintenance including preventative maintenance, crusher rebuilds and clean-up and maintenance on ancillary equipment such as conveyors, chutes and screens. Unscheduled downtime is an interruption in the continuous operation that might result from equipment failures or plugged chutes. The installation of surge capacity between unit operations permits the continuous independent operation of the equipment upstream or downstream of the surge pile or bin. The time period available for continuous operation will be dependent on the capacity of the surge. Without the surge capability unscheduled downtime on one stage of the crushing circuit will impact the other stages. For a three stage crusher circuit with no surge capacity between stages the overall plant availability will be lower than what is capable from any one unit operation. If one assumes that each unit operation has an 80% design availability based on 10% each for scheduled and unscheduled maintenance and assuming scheduled maintenance is conducted on all units at the same time, the overall plant availability could be 65%. To achieve the design throughput in these conditions larger crushing and ancillary equipment will need to be selected. This will result in

567

higher equipment and installation costs. The inclusion of surge capacity in the circuit design will also impact the size of the crews required to maintain the equipment. If all of the equipment has to be maintained at the same time larger crews will be required. Whether or not to include surge piles or bins in the circuit design will be a capital and operating cost trade-off of increased costs for equipment versus cost of installing surge capacity for each application. There are also occasions where the scheduled delivery of the ore will have a significant impact on crusher sizing and selection. Examples will be mines that are developed on the basis of delivering ore 2 shifts per day or 5 days per week. With the expectation of maintaining continuous feed to a downstream processing plant there will be a need to install larger crushers with surge capacity.

Feed Size For different mining methods and different ores, feed sizes to the crusher can vary significantly. The crusher selected must be sized for throughput and also the largest piece of rock that will be expected from the mine. The smaller the crusher the smaller the dimensions of the feed ore that can enter the crushing chamber. A balance is often required to ensure that the plant capacity and size of crusher are matched. The installation of a 48” jaw crusher underground for an operation that the throughput only requires a 30” crusher because the largest expected piece of ore generated from blasting requires further evaluation to determine if the blasting or drilling concepts should be changed or a method of scalping oversize for secondary breakage should be installed. In multi-stage crushing circuits the product of the preceding stage will be a determining factor in the selection of the type of crusher and the configuration of the crusher liners. Product Size The target product size required from the crushing circuit will determine the number of crushing stages and types of crushers to be used for a specific application. For a SAG mill grinding circuit application requiring a coarse, 150 mm feed a single stage primary crusher can be used. For a low tonnage rod mill / ball mill or single stage ball mill application requiring a 15 mm feed size a two stage crushing circuit using a primary jaw crusher and a secondary cone crusher in closed circuit with a vibrating screen may be appropriate. A similar, higher tonnage operation will require a three stage crushing circuit with the third stage using shorthead cone crushers to produce the 15 mm product. The ability to crush finer has been required for specific applications. The “WaterFlush” cone crusher concept was developed by Nordberg (Metso Minerals) using shorthead cone crusher designs to produce finer feed for grinding circuits allowing increased throughput at lower capital costs than costs typical of increasing throughput by adding grinding capacity. Water is added with the feed to the cone crusher. The water acts as a medium to flush fine particles from the crushing chamber and prevent packing enabling the crusher to be set with a tighter closed side setting. For fine product sizes in dry process applications flowsheets have incorporated vertical impact crushers operated in closed circuit with vibrating screens. At Newmont’s Muruntau project in Uzbekistan vertical impactors were used as a quaternary stage of crushing to produce a 6 mesh product to optimize recovery from the heap leach process. Ore Characteristics There are many ore characteristics that must be considered when sizing and selecting the equipment for inclusion in the crushing flowsheet including hardness, toughness, abrasiveness, moisture content, mineralization. The work done by the geologist in defining the ore body and the types of mineralization will provide information that can help define the approach to circuit design. The geologist should be able to provide some indication as to the different rock types and relative amounts of the different rock types that will be processed during the life of the mine. Together with the mining engineer a mine plan should be developed to indicate the short and long term schedule for delivering the ore. It is not unusual to find a situation that the best crushing circuit design for the early years of mine operation will not be the best circuit for later years requiring significant changes in how the circuit is operated and often requiring additional equipment.

568

In evaluating the different ore characteristics an ore can be hard but can be relatively easy to crush because the rock will break along mineral boundaries. Similarly, a soft ore can be tough to break because there are no defined mineral boundaries. The minerals in the ore will provide an indication as to how abrasive the ore will be, the higher the silica content the higher the wear rate on crusher liners, transfer chutes, etc. High moisture content in the ore may lead to higher equipment downtime because of equipment plugging from fine material. Similar situations will also result if the ore contains significant quantities of clay or clay type minerals. For hard ores with minimal fines conventional types of crushers that rely on gravity to draw the ore through the crusher, such as jaw, gyratory and cone crushers, can be used successfully. For softer ores with higher fines or moisture contents the use of crushers that propel the ore through the crushing chamber, such as rotary breakers (MMD), rolls crushers, horizontal impactors, may be a better selection. Rotary breakers have been used very successfully for fine soft ores with high clay contents. The application of traditional crushing equipment with high clay ofes typically requires the use of water to help move the fines through the crusher or the installation of a wet screening plant before the crusher circuit to remove the clay. The rapid development and acceptance of the SAG mill was in a large part a result of soft, clayey ore conditions which were not amenable to multiple stage crushing because of screen, crusher and chute blockage.

Climatic Conditions Climatic conditions have a different type of impact on selection of a crushing plant flowsheet. If the plant is located in a dry warm climate a crushing plant can be designed for an outside installation. In a wet climate location the crushing plant will need to be enclosed for operator accessibility and equipment protection. Depending on the ore type the wet conditions can also result materials handling and crushing problems. In very cold climate conditions the plant will need to be enclosed for operator accessibility. There are significant capital and operating cost implications when the crushing circuit has to be enclosed. A large building enclosure will be required to be heated and ventilated. This in turn will impact the design of the dust collection system. Poor climatic conditions will favor the installation of a SAG mill grinding circuit requiring only a single stage of primary crushing to prepare the ore feed size. Operating costs for a SAG grinding circuit may be similar to a three stage crushing, ball mill grinding circuit but the capital costs will reduce significantly with the elimination of the crusher building and ancillary equipment favoring the selection of the SAG mill. Downstream Processes The downstream recovery processes will also influence the selection of crushing equipment. Whether there is a heap leach or milling circuit. If there is a heap leach, the crusher product size will be specified for optimum recovery. The type of grinding circuit will influence the number of stages of crushing. APPLICATIONS The crusher application is generally defined by the position in the crushing circuit flowsheet, the feed size and the product size. Typically a crushing flowsheet for a mineral processing plant will have from one to three stages of crushing. There are some cases where the process requires a fine dry product and a quaternary stage of crushing will also be included. Primary Crushing Run-of-mine ore is delivered to the primary crusher. The purpose of the primary crusher is to reduce the ore to a size amenable for feeding to the secondary crusher, the SAG mill grinding circuit or as feed to a heap leach stockpile. The primary crusher is operated in open circuit. Typical crushers used for primary crusher applications are gyratory, jaw, horizontal impactors and rotary breakers. Primary crushers are open circuit applications. Primary crushers come in arrange of sizes that relate to the largest feed rock size and to the throughput. The ratio of reduction through a primary crusher will be about 8 to 1. For jaw crusher and impact crusher installations some form of scalping screen may be installed in front of the

569

crusher to increase the overall capacity of the circuit by removing product size material. This concept was also included for the design of early gyratory applications. With the size and capacities of the gyratory crushers it was found that the grizzlies that needed to be installed to create a similar impact were very large and expensive to install and that multiple reclaim systems were required to withdraw ore from both the grizzly and the crusher increasing both capital and operating costs. More recent gyratory plant designs have not included prescreening or scalping. Secondary Crushing. A secondary crusher is included in the flowsheet in order to produce an intermediate or final product. Feed to the secondary crusher will typically be between 75 mm and 200 mm dependent on the feed rate to, and opening of, the primary crusher. For example, a low tonnage underground operation, lo00 tpd, using a primary jaw crusher with a setting 75 mm will feed a secondary standard cone crusher to produce a 16 mm feed for a rod mill / ball mill or single stage ball mill circuit. Similarly, a loo00 tpd concentrator using a gyratory crusher with a crusher setting of 125 mm will feed a secondary standard cone crusher to produce 37 mm feed for a tertiary crushing circuit. In order to optimize the operation of the secondary crusher a vibrating screen is often installed ahead of the crusher to scalp off the product sized ore prior to the crusher. In some applications this could allow for a smaller crusher to be installed and in other applications the crusher could be closed up for a tighter setting to produce a finer product size distribution. The standard cone crusher has been the crusher traditionally selected for the secondary crusher application in mineral processing crushing plant flowsheets. Alternative applications have included horizontal impact crushers and more recently the high pressure grinding rolls (HPGR). HPGR are the crusher of choice for diamond applications. The pressure on the rolls can be set up to crush the diamond host minerals but to minimize the crushing and resulting financial losses that result from diamond breakage. A significant amount of successful work has also been completed in the iron ore industry where roll crushers have been installed to augment or replace existing autogenous grinding circuits.

Tertiary Crushing The tertiary crushing stage in a mineral processing plant typically produces the final product required for downstream processing. Feed to the tertiary crusher will be about 37 mm and the product about 12 mm. For mineral processing projects the traditional crusher selection has been a shorthead cone crusher. The shorthead cone crusher has a longer crushing chamber than the standard cone crusher and is generally suited to providing a more even product size distribution. To maximize the performance of the tertiary cone crushers the crushers are operated in closed circuit with vibrating screens. Product from the secondary and tertiary crushers is collected on a conveyor and fed to a screen. Screen undersize, product, is transported to fine ore storage while coarse oversize is fed to the tertiary crushers. The tertiary crushing application is another application where the HPGR crushers have been tested. Recent installations have also looked at the HPGR to produce a smaller feed size from the tertiary crusher product in order to enhance the production through the grinding circuit. Although the HPGR crushers have been installed as a quaternary application they generally are not expected to produce the fine product size typically expected from a quaternary crusher. In this application they enhance the operation of the tertiary crushers. The application of a crusher in the recycle load of a SAG mill is another example of a tertiary crusher application. The top feed size can be coarser than that expected from a standard cone crusher however the feed is clean of fines and the crusher can be set up to produce a 12 mm product. The successful application of a cone crusher in the SAG circuit will be mostly dependent on removal of tramp metal from the recycle conveyors using magnets. Nordberg’s “WaterHush” cone crushers have been applied to tertiary crusher locations because of fine or sticky ore conditions that would create problems with packing in conventional cone crusher applications. They have also been used to augment existing 3-stage crushing circuits to provide a finer feed to the grinding mills. Typically, the “WaterFlush” crushers will accept a nominal 100 mm top size and crush the material to provide a nominal top size of 10 mm with a P ~ofo 4 111111.

570

Quaternary Crushing The quaternary stage of crushing is generally included in flowsheets where a fine dry product is required for downstream processing. Vertical impact crushers have been installed for this application at Newmont’s heap leaching operation in Uzbekistan. The vertical impactor is a high speed crusher that uses the high speed impact to effect particle size reduction. The Nordberg (Metso) Gyradisc crusher uses a combination of impact and attrition to effect particle size reduction. The Gyradisc crusher has been applied to the industrial mineral and sand industry producing finished product to 800 microns. CRUSHER DESCRIPTIONS Jaw Crusher The jaw crusher shown in Figure 1 has a stationary jaw plate and a moveable jaw plate. The opening at the top of the jaws will be the limiting factor with respect to the maximum size of the rock that can be delivered to the crusher and on the capacity of the crusher. Jaw crushers can be found in a large range of sizes as the smaller models are frequently found in laboratories. For industrial applications the crusher sizes range from 450 mm to 1600 mm with capacities from 50 tph to lo00 tph. The jaw plates and liners for the crushers provide a progressive crushing cavity, Figure 2, with the rock dropping through the crusher each time the moveable jaw swings open. Jaw crushers have been the primary crushing equipment of choice offering simplicity of operation and maintenance and low head clearance to minimize the underground excavation requirements.

Figure 1: Jaw Crusher

571

Figure 2: Jaw Crusher Gyratory Crushers The gyratory crusher, Figure 3, provides a crushing chamber similar to the jaw crusher providing a progressive crushing cavity. The crushing action is provided through an eccentric that swings the bottom of the crusher mantle with respect to the bowl and concaves. Gyratory crushers have the largest unrestricted opening when compared to other crushers. Standard crusher feed sizes range from 1067 mm (42”) to 1829 mm (72”). The 1067 mm gyratory crusher has essentially the same capacity as the 1600 mm (1600 x 2000) jaw crusher. Gyratory crushers tend to offer more flexibility than most other crushers with respect to moderating feed rates. Gyratory’s are generally fed end dump from haul trucks into the crushing chamber.

Figure 3: Gyratory Crusher

572

Horizontal Impact Crushers Impact crushers, Figure 4, accomplish material breaking and size reduction through impact of the material with fixed or free swinging hammers that revolve about a central rotor. The product gradation will be a function of rotor speed and ore friability. In the impact crusher particles of an ore with a common impact value will break according to initial mass, larger pieces will be subjected to more severe impact and will break more readily than finer particles. Impact crushers when compared to conventional compression crushers; jaw, gyratory, cone, will have a lower installed capital cost per ton of capacity. Because of the rotating speed of the crusher the operating and maintenance costs and equipment downtime will be higher for the impact crusher. Two mining applications that used the horizontal impact crusher were the Atlas Gold Bar operation near Eureka, Nevada and Barrick’s Eskay Creek Mine in northern British Columbia. In both cases the impact crusher was selected because of the fine, sticky nature of the ore and concerns with backing in conventional crusher applications. At Atlas Gold Bar the impact crusher was used on SAG mill feed. The ore had a very high clay content and water was added to the crusher. At Eskay Creek primary crushing was done with a jaw crusher and secondary crushing to -2” was done in an open circuit impact crusher.

Figure 4: Horizontal Shaft Impact Crusher

Rotary Breakers The rotary breaker style of crusher has been successfully applied in clay or soft rock conditions. The rotation of the feeder arms promotes the movement of ore through the crushing chamber. Figure 5 provides a cross section of an MMD breaker. The breakers can be supplied in a large range of sizes. With this particular machine the highest wear will be on the teeth. The teeth are provided with replaceable wear caps. Examples of large installations in North America can be found at Suncor’s Canadian Oil Sands operations where a single stage rotary breaker produces a product for pumping to the process plant and at the US Borax plant at Boron California where a two stage circuit is installed to crush borate.

573

Figure 5: Rotary Breaker

Roll Crushers The high pressure grinding rolls (HPGR) have been used very successfully in diamond processing because of the ability to crush the host rock while minimizing diamond breakage. Crushing in the high pressure grinding rolls takes place between two counter-rotating rolls, Figure 6, one of which is fixed and the other moveable. The ore is choke fed into the crushing chamber between the rolls which promotes crushing in the bed of particles, Figure 7. Compared to conventional roll crushers where crushing occurs between and in contact with the two rolls, contact between the rock and the surface of the roll is limited reducing abrasion of the rolls. HPGR technology has also been used successfully in the iron ore industry and is undergoing further development for application in the gold and copper industry. Preliminary development work for the gold industry has indicated that the micro-fracturing that occurs in the HPGR may promote heap leaching kinetics andlor recoveries.

574

Figure 6: Roll Crusher

t Figure 7:

575

Cone Crushers Cone crushers are typically installed for secondary and tertiary crushing applications. The operation of a cone crusher is similar to that of a gyratory crusher in that the cone or mantle travels eccentrically with respect to the bowl. The significant differences are that the cone operates with a higher speed and that the cone travels through a much larger distance. The configuration of the cone and bowl provide a much flatter crushing angle than the gyratory, Figure 8. The two main crusher configurations for cone crushers provide for standard and short-head configurations, Figure 9. For each of these there are different liner configurations; coarse, medium and fine, for both the cone and the bowl. Initial configurations will be determined based on plant throughput and top size of feed. Operations and maintenance personnel then have the ability to assess equipment wear, maintenance and operating characteristics to determine whether better performances can be achieved by changing the shape of the liners. The “Water Flush” crusher technology previously discussed is a cone crusher application with modifications to the crusher required to allow the addition of water with the ore feed.

Figure 8: Cone Crusher Cavity

Figure 9: Cone Crusher

576

Gyradisc Crusher Gyradisc crushers (Nordberg) are specialized reduction machines that are designed to produce large quantities of cubical product from stone, gravel, ores, and nonmetallic minerals. Gyradisc crushers, differ from conventional type cone crushers because comminution utilizes a combination of impact and attrition. These crushers operate best with a choke level in the cavity and by controlling the feed rate to the crusher to maintain a constant power draw. Vertical Impact Crushers Vertical impact crushers are similar to the horizontal impact crushers in that they are high speed units that rely on lunetic energy to develop the crushing force. Ore is fed through the top center of the crusher into the rotor. The high speed rotor throws the ore to the outside of the crusher. Crushing results from interparticle collisions and from ore striking the hammer or material bed that forms at the periphery of the crushing chamber. These crushers can be supplied with mechanical wear components and liners or for autogenous operations as shown in Figure 10 and Figure 11.

Figure 10: Vertical Shaft Impactor

577

Figure 11: Vertical Shaft Impactor APPLICATIONS With most of the high tonnage mining operations the design of the crushing circuit is straightforward, pick the gyratory crusher or crushers required to match the throughput of the operation and produce SAG mill feed. Of more importance is the location of the crusher, whether the crusher is mobile or semimobile in case operating conditions warrant relocation of the crushers, the system of feeding the crushers and the system to reclaim the crushed product. To produce a fine crushed product, -19 mm, for rod mill / ball mill and single-stage ball mill grinding circuits feed, multiple crushing steps will be required. Similar product size material has also been required to optimize recovery in gold heap leach operations. Sizing and selection of the equipment for the crushing circuit flowsheet needs to be based on optimizing the power draw, the crusher component wear rate and crusher equipment availability. All crushing equipment suppliers have tables and curves that relate to the general performance expected from a specific piece of equipment. This information is available to the end user and can be interpreted to provide a preliminary crushing flowsheet design. The actual performance of the crushers will be dependent on the many variables discussed previously. When confirming the flowsheet the equipment vendor will be able to provide best design for maximum ore information. Because of the size and accessibility to the deposit and the representativeness of the samples designs will be subject to variations due to specific ore conditions.

578

There will be applications where multiple flowsheet alternatives are available. It may possible to select a single-stage crusher that can deliver the final product using a high reduction ratio. However, better selection may be two-stages of crushing to more efficiently use the installed power. The selection of the right crusher for a specific application is important but if the ancillary equipment selection does not match the requirements the circuit will not work properly. The main process components that need to be considered include: conveyors feeders (belt, vibrating) bins vibrating screens control package Non-process ancillary equipment will include dust collection and heating and ventilation requirements. The design of the conveyors and transfer chutes must account for surge loading and the crusher circulating loads (where applicable). The use of variable speed belt or vibrating feeders to regulate the feed to the crushers will ensure the crusher can be operated with a “choke” feed chute feeding optimizes power utilization and minimizes component wear. The use of intermediate storage bins will be important to maximizing equipment availability and for maintaining choke feed conditions to the crusher for optimum power utilization. Vibrating screen selection will be evaluated to maximize screen efficiency minimizing the amount of product size that will be recycled back to the crusher feed. The control package will provide the medium to measure what will be happening at each step of the crushing flowsheet and to effect changes to setpoints to optimize plant throughput.

Flowsheet Examples (Multi-stagecrusher applications) A number of flowsheet examples showing typical applications for the various crushers follows.

BIBLIOGRAPHY 1. Edward H. Wipf, SAG 1996. Volume 3. To SAG or Not to SAG. 2. Kurt O’Bryan, SAG 1996. Volume 3. Dealing with Critical Size Material: Application of Conical Crushers. 3. Jerome C. Moty, Mineral Processing Plant Design. AME, 1978. Crushing. 4. Wolfgang Baun et al, Comminution Practices. SME, 1997. Metallurgical Benefits of High Pressure Roll Grinding for Gold and Copper Recovery. 5 . R.E. McIvor, Comminution Practices. SME, 1997. High Pressure Grinding Rolls - A Review. 6. E.S. Burkhardt, Design and Installation of Crushing Circuit. SME, 1982. Primary Crushers: Factors that affect Capacity. 7. M.D. Flavel, Design and Installation of Crushing Circuit. SME, 1982. Selection and Sizing of Crushers. 8. J.C. Motz, Design and Installation of Crushing Circuit. SME, 1982. Types and Characteristics of Common Crushing Flowsheets. 9. Metso Minerals. Various Marketing Information 10. Krupp Polysius. Various Marketing Information 11. MMD Sizers. Various Marketing Information

579

PORGERA BELL CREEK -LOWSHEFI # I .

TWO STAGE CRUSHING (RNE PRODUCT)

FLOWSHEET #2.

TWO STAGE CRUSHING (COARSE PRODUCT)

FLOWSHEET (PS.

THREE STAGE CRUSHING

580

ESKAY CREEK ATLAS GOCD BAR

OR FOX MINE

FLOWSHEET

RUUAN GIBWUTER BRENDA

THREE STAGE CRUSHING

FLOWSHEET

THREE STAGE CRUSHING

PINTO VMLM UN MPNUEL SIERlTh

FLOWSHEET

TM)

SlAGE CRUSHING WATERFLUSH.

AND

581

THREE STAGE CRUSHlffi (GOD HEAP LEACH)

H A Y " HILL SLEEPER

WATERFLUSH CRUSHING

582

KIDSTON

1 HUCKLEBERRY

LONE TREE

i ESCONMQA

FLOWSHEEI 11.

THREE STAGE CRUSHING WlTH MRTlcAl SHAFT IMPACTORS

583

Selection and Sizing of Primary Crushers Ronald W. Utley, General Manager, Crushing Division, FFE Minerals, Bethlehem, PA, USA

ABSTRACT The selection of the primary crusher is the key to the success of any operation that involves size reduction. Each of the various types of primary crushers is discussed. The material hardness, impact strength and abrasive index define which primary crushers are acceptable. Required capacity, feed and product sizes are considered to narrow the selection and define the physical size of the crushers. Once the acceptable primary crushers have been identified, the selection is further defined by above ground or underground location of the plant if the crusher is to be stationary or with some degree of mobility. DEFINITION The primary crusher selection is the key to the success of the mining, quarrying or industrial minerals operation that involves the reduction in size of rock, ore or minerals. The crushing plant can be provided with almost any type of primary rock crusher. The rocklore determines the type of crusher The plant capacity determines the size of the crusher As the term “primary” implies these crushers are used in the first stage on any size reduction cycle. These crushers take blasted, run-of-mine, or run-of-quarry feed up to 1500 mm (60 inches) and produces a product ranging in size from -300 mm (12”) for conveyor transport, or -300 mm (8”) for SAG mill feed; to -38 mm (1 %’) when crushing soft, nonabrasive materials such as limestone. The primary can produce these sues at a rate of 150 to 12,000 MTPH per hour depending on the feed characteristics, crusher setting and crusher size. The family of primary crushers include the:

-

Gyratory Double toggle jaw Single toggle jaw High speed roll crusher Low speed sizer Impactor Hammermill Feeder breaker

HISTORY Crushing and brealung of rock, ore and minerals is one of the oldest industries undertaken by man. The earliest crushing known was by hand on “native” ores (gold, silver, and copper) or on ores containing lapis lazuli, garnet, diamond, jade, etc. Glass making may have required some brealung of agglomerates. If hand crushing was impractical a heavy rock or weight was raised by men or animals with a rope and allowed to drop onto the rock to be crushed. Irrigation led to invention of

584

waterpower to raise the weight. The invention of the cam and cogwheel allowed the operation to be continuous. Crushing is the essential function in the treatment of all rocks and minerals, whatever is their end use. The primary crushing of rocks does not appear compatible with high tech engineering of the 21” Century. However, we have come a long way fiom the beginning of the 1800’s when crushing was carried out by hundreds of men and women equipped with sledgehammers. The earliest US.patent for a crushmg machine was issued in 1830. It covered a device, which, in a crude way, incorporated the drop hammer principle later used in the famous stamp mill, whose history is so intimately llnked with that of the golden age of American mining. Eli Whitney Blake invented the first successful mechanical rock breaker - the Blake jaw crusher patented in 1858. Blake adopted a mechanical principle familiar to all students of mechanics, the powerful toggle linkage. That lus idea was good is attested to by the fact that the Blake type jaw crusher is today the standard by which all jaw crushers are judged. In 1881, Philters W. Gates was granted a patent on a machine that included in its design all of the essential features of the modem gyratory crusher. An interesting sidelight of these early days occurred in 1883 at Meriden, Connecticut, where a contest was staged between a Blake jaw crusher and a Gates gyratory crusher. Each machine was required to crush nine cubic yards of stone, the feed-size and discharge settings being similar. The Gates crusher finished its quota in 20 % minutes, the Blake crusher in 64 !4 minutes. For some years after these pioneer machines were developed, requirements remained very simple. All mining and quarrying, whether underground or open-pit, was done by hand; tonnage’s generally were small, and product specifications simple and liberal. Even the largest commercial crushed stone-plants were small, consisting of one crusher, either jaw or gyratory, one elevator and one screen. When demand grew beyond the capabilities of one crusher, it was generally a simple matter to add a second machine. When the business outgrew the capacity of this sort of plant, it was not unusual to double up, either in the same building, or by erecting an entirely separate plant adjacent to the original one. The steam shovel began to change the entire picture of open-pit mining. With the steam shovel came the really “huge” gyratory crusher, with its 18 inch receiving opening. This tum toward really large primary crushers started just a few years before the turn of the century, and in 1910 crushers with 48 inch receiving openings were being built. About h s time the jaw crusher suddenly came back to life and stepped out in front with a great contribution to the line of mammoth sized primary crushers: the 84 inch x 60 inch machme built by the Power and Mining Machinery Company for a trap rock quarry in eastern Pennsylvania. During the same years the industry was concerned with development of larger and still larger primary crushers, another member of the family was born: the single, sledging roll crusher. The machine quickly achieved a high degree of popularity, and although its field of application was relatively limited, quite a number of these machines were installed for primary crushing service. The line was expanded to also include the 84 inch x 60 inch machine. The double roll crushers had limited popularity in the mining industry. In 1919, Traylor Engineering manufactured the largest gyratory crusher yet built. T h s was a 60 inch gyratory sold to Michigan Limestone and Chemical Company in Rogers City, Michigan. This size stood until 1969 when Traylor Engineering built the first 72 inch gyratory crusher. This crusher was sold to Johns Manville of Canada. It was in service as the world’s only operating 72 inch machine until 2001 when the mine went underground and the gyratory was replaced with low speed sizers. By 1920, the hammermill had been developed to produce a fine product in a single pass machine. These machines employ the impact principle of breaking stone. The hammermill is a simple mechanism. The machine comprises a box-like frame, a centrally located, horizontal shaft

585

rotating element on which hammers are mounted., and a set of circumferentially arranged grates in the lower part of the housing. The shaft rotates at high velocity, whch break the stone by impact. From the 1920’s through the 1950’s the hammermills were modified to include impactors that broke the rock with fixed breakmg bars and the elimination of the grates on certain machines. The double toggle jaw crusher was supplemented by the single toggle jaw crusher where the pitmadtoggle arrangement was eliminated and the motion was derived from an eccentric drive shaft. In 1960, the feeder-breaker was developed for underground coal mining to follow the continuous miners. These are low-headroom machines that consist of a flight bar feeder and a rotating breaking drum with teeth or picks. The material is transferred to the tailend of the feeder, is dragged toward the headend and crushed in-line as the material passes under the drum. These machines have now been modified and upgraded to handle limestone and other non-abrasive materials. Some manufactures identify the rock machines as impact horizontal roll crushers. In the early 1980’s, low speed sizers were introduced. This represented one of the only fundamental developments in primary crushers in three-quarters of a century. Conventional roll crushers are suitable for crushing soft to medium strength materials and can provide accurate reduction of >10:1. Roll crushers, both smooth and toothed, are generally designed with high peripheral speeds, and uneven wear can be a major problem. The main feature of the low speed sizer, which can broadly be considered a variety of toothed roll, is that it exploits the fact that the ratio of compressive strength to tensile and shear strength in the majority or rocks is approximately 1 O : l . The low speed sizers breaks the rock in tension or in shear by its “snapping” and chopping action rather than in compression as conventional crushers do. Additionally, the positioning of the teeth on the rolls allows undersize to fall directly through the machine resulting in high throughputs at very low rotational speeds that leads to greatly reduced wear, energy savings, better control of discharge size in three dimensions, and greatly reduced fines.

1850

1900

1950

2000

Gyratory DT Jaw ST Jaw Double Roll Low Speed Sizer

Impactor Hammermill Feeder Breaker

MECHANICAL REDUCTION METHODS There are four basic ways to reduce a material - by impact, attrition, shear or compression - and most crushers employ a combination of all these crushing methods.

586

Compression. As the name implies, crushing by compression is done between two surfaces. Gyratory and double toggle jaw crushers using t h s method of compression are suitable for reducing extremely hard and abrasive rock. However, single toggle jaw crushers employ attrition as well as compression and are less suitable for abrasive rock since the rubbing action accentuates the wear on crushing surfaces.

As a mechanical reduction method, compression should be used as follow: - when the material is hard and tough - when the material is abrasive - when the material is not sticky - when a uniform product with a minimum of fines is desired - when the finished product is to be relatively coarse, i.e. -38 mm (1 size - when the material will break cubically

%,’)

or larger top

Impact. Refers to the sharp, instantaneous impingement of one moving object against another. Both objects may be moving, such as a baseball bat connecting with a fast ball, or one object may be motionless, such as a rock being struck by hammer blows. There are two variations of impact: gravity impact and dynamic impact. Material dropped onto a hard surface such as a steel plate is an example of gravity impact. Material dropping in front of a moving hammer (both objects in motion) illustrates dynamic impact. When rock is crushed by dynamic impact, the material is unsupported and the force of impact accelerates movement of the reduced particles toward breaker blocks and/or other hammers. Dynamic impact has definite advantages for the reduction of many materials and it is specified under the following conditions: when a cubical particle is needed. when the finished product must be well graded and must meet intermediate sizing specifications as well as top and bottom specifications. when ores must be broken along natural cleavage lines in order to free and separate the mineral from waste. when materials are too hard and abrasive for hammermills, but where jaw crushers and gyratory crushers cannot be used because of particle shape requirements, high moisture content or capacity. Attrition. A term applied to the reduction of materials by scrubbing it between two hard surfaces. Hammermills operate with close clearances between the hammers and the screen bars and they reduce by attrition combined with shear and impact reduction. Though attrition crushing consumes more power and exacts heavier wear on hammers and screen bars, it is practical for crushing the less abrasive materials such as low silica limestone or coal. Attrition crushing is most useful in the following circumstances:

-

when material is friable and/or non-abrasive when a closed-circuit system is not desirable to control top size. when a maximum of fines is required.

Shear. This consists of a trimming or cleaving action rather than the rubbing action associated with attrition. Shear exploits the fact that the ratio of compression strength too tensile and shear strength in the majority of rocks is approximately 1O:l. Low speed sizers break the rocks in tension and shear by its chopping action. Shear crushing is normally called for under these conditions:

587

-

when material is somewhat friable and has relatively low silica content. when the material is soft to medium hardness. for primary crushing with a reduction ratio of 6 to1 . when a minimum of fines is desired. when a relatively coarse product is desired (usually no finer than 38 mm (1 %”) top size).

PRIMARY GYRATORY CRUSHER Gvratorv Crushers. A conical shaped element is supported in a flared shell or frame creating a chamber wide at the top and narrow at the bottom. The center element is caused to gyrate about its fulcrum point causing it to advance and retreat with relation to the shell. Rock introduced at the top is broken as it passes through the crusher chamber.

Typical gyratory crusher capacities are 350 to 10,000 MTPH. The primary gyratory crusher is known for its high capacity and low maintenance. The main capacity advantage offered by a primary gyratory crusher is centered around Archimedies principal of ‘TI‘’. a) The round crushing chamber cylinder provides more effective volume than a rectangular volume. b) The shaft grating speed adds a third dimension to crushing as opposed to twodimensional crushng. On this basis, capacities up to 10,000 MTPH can be achieved with this design with 60” x 110” machines.

Advantages: 1) 2) 3) 4) 5) 6)

Designed for direct dump from trucks up to 300 tons. High capacity ratings. Lowest maintenance per ton processed of any design crusher. Highest availability of any crusher design. Can handle crushing ore hardness up to 600 Mpa (90,000 PSI) compressive strength. Easy handling of tramp material with hydraulic relief system

588

Disadvantages: 1) Highest installed capital cost of any crusher design. However, since the ROI for the crushing and conveying system is tied to capacity and the ability to handle hard and abrasive ores, the gyratory crusher is usually the machine of choice. The machines are typically furnished with feed size openings of 1.07 m (42”), 1.37 m (54’7, and 1.52 m (60”). The largest feed opening of any unit operating in the world is a 1.83 m (72”) Traylor crusher. The capacity of even the smallest standard unit, the 1.07 m (42”) gyratory, approaches a capacity range of 2500 MTPH; the 1.52 m (60”) unit will crush more than 10,000 MTPH, depending on crusher design, characteristics of the material to be crushed, and desired product. DOUBLE TOGGLE DESIGN

DT Jaw Crusher. Consists of a stationary and a moving jaw so positioned as to provide a wide opening at the top and narrow at the bottom. The moveable jaw is fixed at the top and the motion is generated through the toggle arrangement. The maximum motion is at the bottom of the jaw. The rock is squeezed with each motion of the moving jaw, is broken and drops by gravity to a lower position where the process is repeated until all will pass the narrow opening at the bottom.

The swing jay of the standard DT (Blake) crusher pivots from an overhead shaft. A Pitman hung from an eccentric shaft transmits motion through a pair of toggles at the bottom of the swing jaw. In this type crusher, swing jaw movement is greatest at the discharge opening. The hinge pin is located behind the centerline of the crushing zone and it causes the swing jaw to move in a perpendicular to the fixed jaw. This arrangement provides twice the force in crushing. Typical duty for a double toggle jaw crusher is 350 Mpa (90,000 PSI). Advantages: 1) Lower installed cost than a gyratory crusher. 2) Can handle high abrasion with low maintenance. 3) Can handle tough crushing applications up to 600 Mpa (90,000 PSI), nickel ores, iron ores, etc. Disadvantages: 1) Same capacity limitations as the single toggle jaw crusher. 2) Substantially higher installed cost than a single toggle jaw crusher. 3) Same crushing size limitation as single toggle jaw crushers.

589

SINGLE TOGGLE DESIGN

ST Jaw Crushers. Consist of a stationary and a moving jaw so positioned as to provide a wide opening at he top and narrow at the bottom. The motion is created by the eccentric shaft at the top of the jaw. The maximum motion is at the top of the jaw. The rock is squeezed with each motion of the moving jaw, is broken and drops by gravity to a lower position where the process is repeated until all will pass the narrow opening at the bottom. Q

The “single toggle” is a light to medium duty crusher, capable of crushing ores up to 200 Mpa (27,500 PSI). The rotation of the eccentric shaft causes the swing jaw assembly (attached to the rotating eccentric shaft), to move in an elliptical path. Maximum movement of the swing jaw assembly occurs at the top of the crushing chamber, with minimum movement at the discharge opening. At all points in the crushing chamber the crushing action has both vertical and horizontal components. The larger motion at the receiving opening accentuates the shock loads on the bearings of the eccentric shaft. Such loads create high bearing pressures, which have destructive potential. Due to the rubbing action of this type of jaw, jaw plate wear is accelerated, and power efficiency is lowered because the swing jaw assembly is lifted on every stroke. shovel teeth, etc. The same product size, capacity limitation, and safety features as stated for the double toggle jaw crushers apply to the single toggle jaw crushers. Advantages:

1) Lower installed cost than a double toggle. 2) Lower power usage than a double toggle. 3) Can handle sticky, muddy ore easier than a double toggle jaw or gyratory crusher. Disadvantages: 1) Maximum capacity is 1000 MTPH, although normal economical, maximum capacity is 750 MTPH. 2) Duty of crushing is for light or medium hard materials, up to 27,500 PSI compressive strength. 3) Does not handle h g h abrasion material as well as a double toggle jaw. 4) Primary crushing only. Requires additional crushing equipment unless downstream process is a SAG mill circuit. 5) Requires feeder.

590

JAW CRUSHERS Single Toggle versus Double Toggle Single and double toggle jaw crushers on first appearance seem to overlap their application as a primary crusher; however, they both have their specific advantages. The single toggle generally has a larger angle of nip than a double toggle unit. The larger the angle of nip, the harder to grip the material. Typical Ore

Verv Hard

Single

Double

Double

30"

25"

20"

1/3rdof the way down the jaws 26"

25"

20"

At top ofjaws

With the single toggle jaw crushers, the greatest movement is at the top of the jaw. With double toggle jaw crushers, the greatest movement is at the bottom. The movements of the jaws of single toggle jaw crushers are in a downward rolling direction. This action gives the crusher a force feed action that is of particular benefit when handling sticky materials and does much to compensate for the greater angle of nip at the top of the jaws. Due to the crushing action of the single toggle, the life of the jaws is less than that of the double toggle. The double toggle has a direct squeeze type action at 90" to the crushing faces reducing scuffing. Double Toggle

Single Toggle

Due to the smaller angle of nip of the double toggle unit, it can grip harder material than the single toggle. Production drops off markedly due to slippage of the jaw when very high compressive strength material is being crushed.

59 1

DOUBLE ROLLS

Double Roll Crusher. Heavy cylinders or rolls are mounted parallel in a frame with a controlled gap between the rolls. Rolls are individually powered or coupled by gears. Crushing is primarily by compression as rock is passed through the gap between the rolls. An element of impact is imported in the material due to high speed of the rolls.

LOW SPEED SIZER

Low Speed Sizer. Two toothed rolls revolve in a chamber at low speeds and high torque. The teeth are so arranged that the crushing is accomplished by shear.

The low speed sizing principle is the employment or a commnanon or mgn torque 1 IOW roll speeds and tooth profiles to arrive at specified end product sizes with a minimum of fines production. In essence the interaction of tooth, spacer and roll set up a “sized Void” which in turn “Sizes” oversize material when fed through the machine rolls. Undersize material basically “free falls” through the unit with little or no further reduction. Low speed sizers are used for hard non-abrasive, sticky types of materials up to 200 Mpa (27,500 PSI) and up to 400 Mpa (60,000 PSI) when a mixture of these materials is expected. Examples: medium hard limestone, bauxite, kimberlite, gypsum, clay, shale, schist and gold ore. In the primary role, the low speed sizer is not particularly sensitive to abrasion if the reduction ratio is low because the wear parts can be replaced or refurbished in-situ economically. Low speed sizers are used on many different materials and when the material can vary from soft to hard and any combination of the two extremes i.e. overburden and when high outputs are required in the 3,000 to 10,000 MTPH. Advantages: 1) The low speed sizer can handle very high tonnages. Peak loads of 12,000 MTPH have been recorded. 2 ) The compact sizer reduces the installation costs and allows sizers to be considered in locations where no alternative can be found. Lowest head room position. 3) The capital cost of mobile units using a low speed sizer is considerably lower than any other type of crusher. 4) Fines production is very low. 5 ) Power consumption is very low for a given output. 6) Oversize feed can be rejected without loss of production by employing side discharge gates.

592

Disadvantages: 1) Low reduction ratio. 2) A good electrical power supply is required (peak power loading can be up to eight times the installed power). 3) Not economical for low tonnages unless the material is very difficult to handle.

IMPACT CRUSHER

Impact Crusher. One or two heavy rotors carrying fastened projections, revolve inside a casing. The projections break the rock by primary impact and propels the rock against the case where there it is broken by secondary impact.

Impactors are utilized in soft, non-abrasive applications or when crushing availability and maintenance can be economically offset against capital cost.

Advantages: 1) 2) 3) 4)

An impact crusher can handle a large size reduction - one meter to 75 mm. High reduction ratio for amount of investment. Impactor provides a high degree of fines. Can handle up to 2500 MTPH.

Disadvantages: 1) Higher silica ore, plus 8% cause increased wear. 2) Power consumption is hgher as more fines are produced. 3) Cannot handle tramp metal. 4) Requires feeder.

593

HAMMERMILL Hammennills are similar to impact breakers except hammers are pivoted on the rotor and usually only one rotor is used. They may have a grate at the discharge, the spacing of which determines the product size.

FEEDER BREAKERS Feeder breakers are utilized in soft applications to medium hard applications when the requirement is to coarsely break material for belt conveying. Frequently used for overburden and underground duty.

Advantages: Avoids costly site preparation and civil work The feeder breaker can transfer and crush material in a single machine. Reduces relocation time, expenses and lost production during moves. Handles wet materials with ease. Very low headroom. Reduces dump height and the need for ramps and large hoppers. Often self-propelled on crawlers or wheels. Can handle up to 2000 MTPH. Disadvantages: 1) Very low reduction ratio. 2) Crushing takes place on breaker bars and chains, whch cause wear. PRIMARY CRUSHER SELECTION CRITERIA When selecting the type of crusher, the following criteria must be considered:

594

0

0 0 0 0 0 0

0 0 0 0

0 0 0

0

Will it produce desired product size at the required capacity? Will it accept the largest feed size expected? What is its capacity to handle peak loads? Will it choke or plug? Is the crusher suited to the type of crushing plant design? Is the crusher suited for underground or in-pit duty? Can it pass uncrushable debris without damage to the crusher? How much supervision of the unit is necessary? What is the crusher’s HP demand per ton per hour of finished product? How does the crusher resist abrasive wear? Does the crusher operate economically with minimum maintenance? Does the crusher offer dependable and prolonged service life? Does the crusher have acceptable parts replacement costs? Does the crusher have easy access to internal parts? How does the initial cost of the machine compare with its long term operating costs?

TYPE OF MATERIAL TO BE CRUSHED Defines the type of crusher to be considered. The family of primary crushers can be subdivided into three classes by the material they are best suited to handle. a) Gyratory and double toggle jaw - tough, abrasive (high silicate), non-sticky, types of material having compressive strengths up to 600 Mpa (90,000 PSI). Examples of these materials include Taconite, trap rock, granite, hard limestone, porphyry copper, high silica gold ores. b) Single toggle jaws, and low speed sizers - medium hard, non-abrasive, and sticky materials having compressive strength up to 200 Mpa (27,500 PSI). Examples include medium hard limestone, bauxite, kimberlite, low silica gold ores. c) Impactor, roll crusher, feeder breakers and hammermills - soft friable, non-abrasive (low silicate) materials having compressive strengths below 115 Mpa (16,500 PSI). Examples include limestone, phosphate rock, gypsum, trona. RUN OF MINE FEED (R.O.M.) The maximum feed in nominal terms. The 100% passing gradation is used to determine feed opening of the crusher. PRODUCT SIZE REQUIRED The product size required in nominal terms. The 80% passing gradation is used to determine the size of the crusher and calculate the horsepower. ABRASIVE INDEX Pennsylvania Crusher Abrasive Index The abrasive index of the ore to be crushed can be used as a guide in selecting the crusher.

595

Material

Typical Abrasive Index

Aluminum Oxide Sandstone Quartzite Gold Ore Granite High Silica Limestone Tungsten Ore Hematite Limestone Clay

14,000 13,000 11,000 8,000 7,000 5,000 3,000 600 500 25

CRUSHER SELECTION BASED (BURBANK) ABRASIVE INDEX Abrasive Index

Crusher

0-35,000

Gyratory Double Toggle

0-6000

Gyratory Double Toggle Single Toggle Low Speed Sizer

0-2000

Gyratory Double Toggle Single Toggle Low Speed Sizer High Speed Sizer

0- 1000

Gyratory Double Toggle Single Toggle Low Speed Sizer Impactor Hammermill Feeder Breaker High Speed Double Roll

IMPACT VALUE The impact value is a measure of the hardness of the material to be crusher. The work index can be calculated from the impact value, which in turn is used to calculate the approximate power required to crush the ore.

596

Rocks can be divided into different classifications on the basis of its resistance to crushing as follows:

VI V IV I11 I1

I

Impact Strength

Compressive Equivalent

Designation

Jfoot-pounds Der inch)

(PSI)

Extremely hard Very hard Hard Medium Soft Very soft

above 24 20-24 16-20 12-16 8-12 below 8

above 33,000 psi 33,000 psi 27,500 psi 22,000 psi 16,500 psi 10,000 psi

CRUSHER SELECTION BASED ON IMPACT STRENGTH Impact Ranpe

Crusher

Above 24

Gyratory Double Toggle Jaw

20-24

Gyratory Double Toggle Jaw

16-20

Gyratory Double Toggle Jaw Single Toggle Jaw Low Speed Sizer

12-16

Gyratory Double Toggle Jaw Single Toggle Jaw Low Speed Sizer High Speed Roll Feeder Breaker

8-12

Gyratory Double Toggle Jaw Single Toggle Jaw Low Speed Sizer High Speed Roll Feeder Breaker Impactor Hammermill

0-8

Gyratory Double Toggle Jaw Single Toggle Jaw Low Speed Sizer High Speed Roll

597

Compressive Equivalent A = +250 230 190 150 115 70

Feeder Breaker Impactor Hammermill

PRIMARY CRUSHER SELECTION BASED ON CLAY CONTENT Clay or sticky clay like materials represents a difficulty for almost every type of primary crusher. The crushing action of the gyratory, jaw and double roll machines is by compression. When the feed to these machines contains a significant amount of clay, the material will pack. The material flow through the gyratory and jaw crushers is by gravity. When the clay packs in the chamber of the crusher the capacity is reduced and in some cases completely stopped. Clay will pack between the teeth of the primary double roll crushers. In severe cases the rolls will become significantly packed to prevent any crushing to take place. The downward action of the single toggle jaw crusher provides some degree of improvement over the double toggle design. Impactors and hammermills cannot be used for clay bearing materials as the chamber will buildup with they clay and prevent impact crushing from taking place. The crushmg chamber will quickly pack. Feeder Breakers represent a fair to good alternative. The flight bars on the feeder will move the material along the pan toward the breaker drum. The rotating drum breaks the material as if is forced along by the flight bars. There is no chamber in which the clay will pack. Clay, however, can cause maintenance problems with the flight bars and there is a possibility of clay packing between the picks of the rotating drum. The low-speed sizer is the only primary crusher that can handle the clay effortlessly. The two shafts of the low-speed sizer rotate inward at low speeds. There is no impact that will cause the material to pack in the chamber. The low speed sizer can be provided with scraper bars that are located between the rows of sizing teeth to keep the toothed shafts clean at all times. PRIMARY CRUSHER SELECTION BASED ON UNDERGROUND SERVICE In addition to the typical selection criteria, the underground primary crusher must be designed to handle blocky, hard and wet material, steel in the form of drill steel and roof bolts and typically be able to operate in an area of limited head room. The impact machines - impactors and hammer mills can not be used underground if any steel is present. The steel will not go through the crusher but will become lodged in the chamber causing the crusher to stop. The double roll crusher is not designed for blocky hard material. The DT and ST jaw crushers can provide very large feed openings for crushmg block cave materials. Wet materials do not adversely effect the crushers as long as the material is not sticky. The jaw crushers can handle the steel quit well. Capacities are limited to about 1000 MTPH and a 1000 MTPH machme is quite large. These machines are easily sectionalized for underground installation. The gyratory crusher is equally good for the service. The gyratory crusher will handle much larger capacities than the jaw, will handle steel but the bottom of the crusher incorporates arms that can become a resting-place for steel. The gyratory crusher can be direct feed. The jaw crushers require a controlled feed. The feeder breaker is designed to operate underground with wet and steel laden material. Most of the feeder breakers are operating on coal or low abrasive materials. The feeder breaker is a low reduction ratio machine and will produce a course product if crushing blocky material. The low-speed sizer provides distinct advantages for underground applications. The machine is the lowest headroom design available. The machine is not adversely affected by water and can handle clay materials if present. The shafts roll at slow speeds inward and in most cases

598

will pass the steel through the center of the m a c h e . If the steel does stop the sizer, the unit can be reversed and the steel expelled automatically through side discharge gates. The machine will handle large blocky ore. The low-speed sizer is limited to materials of about 200 Mpa (27,500). PRIMARY CRUSHER SELECTION FOR MOBILE CRUSHING PLANTS All of the family of primary crushers can be used on mobile plants. Impactors and hammermills are compact and generate a high reduction ratio. The h g h speed of the machines requires special attention to dynamic forces. The jaw crusher are very good for small tonnage plants and the ST jaw has the advantage of being lighter in weight than the DT jaw. Large capacity jaw crushers result in large crushing plants. Double roll crushers are very large machines. The two rolls rotate inward so the out-ofbalance forces are minimized. These machines are limited to relatively soft and non-abrasive materials. The gyratory crusher is the work hours of the semi-mobile crushng plants. The selection of the gyratory crusher for this application is based on capacity. For most of the in-pit, mobile, semi-mobile and movable crushing applications in hard rock mining the capacities exceed 3500 MTPH. This criterion limits the choice of crushmg machine to gyratory and low-speed sizer. The gyratory crusher has been the crusher of choice for large capacity applications since the early 1900’s. The low-speed sizer has only been available since 1980. The gyratory crusher can crush material to 600 Mpa (90,000 PSI). The low-speed sizer is limited to 200 Mpa (27,500 PSI). The gyratory can be direct dump. The low-speed sizer must be control fed. The gyratory crusher by its very nature generates a significant out-of-balance force. The sizer shafts rotate toward each other and there is very limited out-of-balance generated. The gyratory crusher is the tallest of all primary crushers. The sizer is the shortest of all primary crushers. The feeder breaker is designed as a mobile crushing plant. However the applications of this machine are limited. QUICK SELECTION CHARTS FOR PRIMARY CRUSHER APPLICATIONS

MTPH

1500

3000

6000

Gyratory DT Jaw ST Jaw

Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker Hammermill 2500 MTPH with grate, 3000 MTPH without grate.

599

9000

12000

MM

500

1000

1500

2000

Gyratory DT Jaw

ST Jaw Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker

MM

100

200

300

Gyratory DT Jaw

ST Jaw Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker

600

400

2500

MPA

0

100

300

200

400

500

600

tiyratory DT Jaw ST Jaw Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker

BURBANK

0

800

16000

Gyratory DT Jaw

ST Jaw Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker

601

24000

32000

MTPH

1500

3000

6000

9000

12000

Gyratory DT Jaw ST Jaw Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker

QUICK SELECTION CHARTS FOR PRIMARY CRUSHER APPLICATIONS

Poor

Fair

Good

Very Good

Excellent

Gyratory DT Jaw ST Jaw Double Roll Low Speed Sizer Impactor

N/A

Hammermill

N/A

Feeder Breaker

* Impactors and

Hammermills cannot be used to crush clay, as the clay will plug the

crusher.

602

Poor

Fair

Good

Very Good

Excellent

Gyratory DT Jaw ST Jaw Double Roll Low Speed Size Impactor

NIA

Hammermill

NIA

Feeder Breake

* Impactors and Hammermills are unacceptable for use underground due to the inability to handle drill steel, roof bolts, etc.

Poor

Fair

Gyratory DT Jaw ST Jaw Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker

603

Good

Very Good

Excellent

GLOSSARY OF TERMS MOH Scale Relative hardness of material compared to: 1-Talc, 2-Gypsum, 3-Calcite, 4-Fluorite, SApatite, 6-Feldspar, 7-Quartz, %Topaz, 9-Corundum, and 10-Diamond.

Angle of Nip The angle formed between the moving surface of a crusher roll or jaw plate and the stationary plate surface, at which point the material will be pinched. Angle varies with machine size and material lump size.

Nominal Describes product size, usually denoting that at least 80% of product is smaller than size indicated.

Breaker Block (Breaker Plate) The steel anvil surface of a crusher against which material is crushed by impact or pressure.

Oversize Material too large to pass through a specific size of screen or grizzly opening.

Bridging Blocking of crusher opening by large pieces of material.

Packing A compacted or compressed condition of the material in the crusher, characterized by a complete or nearly complete absence of voids.

Burbank Abrasion Test A standard method of comparing the relative abrasiveness of rocks, minerals and ore.

Plugging Restriction of material flow through a crusher.

Choke Feed Operating the crusher with a completely filled crushing chamber.

Primary Crusher The first crusher in a rushing system to which material is fed.

Choke Point Place in the crushing chamber having the minimum cross section. All compression type crushers have choke points but this does not necessarily mean that choking is likely to occur.

Reduction Ratio The ratio of the top size of feed material to the top size of crusher discharge.

Feeder A device that regulates and distributed material into the crusher.

Reversible Crushers Hammermills and impactors with rotors that can be run both clockwise and counterclockwise.

Fines Material with particle size smaller than a specified opening.

ROM Run of Mine - material from a mine that has not been crushed or screened.

Finished Product The resulting material after it has been processed.

ROQ Run of Quarry - material from a quarry that has not been crushed or screened.

Hammers Free-swinging or fixed metal impact surfaces attached to the rotor assembly of an impactor or hammermill crusher.

Top Size The largest allowable particle size in a product. Tramp Iron A bolt, shovel teeth, picks and other uncrushable metal that is often present in crusher feed.

604

REFERENCES Handbook of Crushing. 1995. Pennsylvania Crusher Corporation, Broomall Pennsylvania. Brownell McGrew. 1953. Crushing Practice, Allis-Chalmers, Appleton, Wisconsin. Ib Finn Petersen. Crushing, F.L.Smidth & Company. N S Copenhagen, Denmark

605

In-Pit Crushing Design and Layout Considerations Ken Boyd, Manager, Material Handling, AMEC Mining & Metals, Vancouver, BC, Canada Ronald W Utley, General Manager Crushing Division, FFE Minerals, Bethlehem PA, USA

ABSTRACT The selection of the crusher and the layout of the in-pit crushing plant requires a great deal of design based on the mine plan and operations input. The evolution of in-pit crushing and the various plant layouts will be discussed along with the sizing and selection of the crusher. The mine production plan, safety and environmental considerations, geographical location, climatic conditions, the expected life of the operation, types of ore or rock, minimization of capital costs, future expansion possibilities, operational and maintenance design requirements are all major design considerations in sizing and laying out an in-pit crushing system.

INTRODUCTION To establish and maintain competitiveness in international markets for minerals and stone products, it is indispensable to adopt the latest proven technology and economical systems in open cast mining. In today’s markets, overburden of increasing thickness has to be stripped and dumped, distances to the stockpile are getting longer and longer, depths of mines and quarries are increasing, ore grade is decreasing, and costs for energy and labor are continuously escalating. Trucking of waste and ore from quarries and pits is a very flexible material handling transportation system. Mine planners, especially at the start of a Greenfield project, find that trucking is the easiest transportation system to design and plan for. As the pit or quarry becomes deeper or further away from the delivery point then mine planners and designers have to review the transportation system. This will insure that their operation will continue to have the best and most economical material handling system for their operation, flexibility is no longer the prime driver. As the pits get deeper, In-Pit Crushing and Conveying, has become the transportation of choice of most mine planners. IPCC has been with us for some time now and many mine planners would have been selecting t h s design option over trucking. However, just as the IPCC option may become economically more viable then the truck manufacturers come up with larger trucks. When long term planning is possible IPCC has been the method of choice for the material handling transportation system. There are three main steps in designing a good crushing plant: process design, equipment selection and layout. The first two are dictated simply by production requirements and other design parameters, but the layout can reflect the inputs, preferences and experience of a large number of parties. These can include the owner’s engineering staff; operations and maintenance personnel; equipment manufacturers; and especially the mine planners. The types of in-pit crushers that are being reviewed by mine planners are: fixed plants, semimobile gyratory, jaw, impact and roll sizer crushers. The two types of conveyor systems being selected for these systems are the conventional conveyor and high angle conveyor systems. The high angle conveyor system has yet to put in operation in high tonnage mining operation.

IPCC DEVELOPMENT In 1954, the fKst “mobile” or self-propelled crusher was installed in a limestone quarry in Hover, Germany. Use of these earliest mobile crushers solved the major problem of wet, soft ground conditions that did not permit the use of haulage trucks due to the high cost of building and maintaining haulage roads. Quarry operators also wanted to take advantage of continuous belt conveyor haulage systems and the resulting cost savings.

606

European mines, particularly in the coalfields, are frequently characterized by soft materials, allowing the use of bucket wheel excavators followed by a continuous haulage system of belt conveyors. In the late 1950's, contemporary quarry operators with this background knowledge of belt conveyors in mining easily accepted the conveyor concept once in-pit mobile crushing solved the problem of run of quarry size reduction required for conveyor application. True in-pit crushing started as early as 1969-1970 in Alcoa's bauxite mines in Western Australia. In the late 70's the first fully mobile gyratory crushers (54-74) were installed in South African open cast mimes (Palabora, Grootegluk). All above crushing plants were fully mobile crushers (walking feet) however, one was used with dump trucks and stayed in position for several years at a time the others were working at the face in conjunction with shovels and moved several times a shift. Three events took place at the dawn of the 1980's that laid the foundation for worldwide acceptance of IPCC and changed primary rock crushing forever. 1) The oil embargo of 1979 put mine operators on notice that they should not tie the future to conventional truck haulage systems. Approximately 50% of truck operating costs are related to fuel, lubrication and tire consumption. These products are either directly or indirectly dependent upon the cost and availability of petroleum products. In the late 1970's and early 1980's, petroleum costs (and thus the cost of their derivative products) as well as availability were unpredictable because of cartel pricing policies and an extremely volatile world market. Severely increasing truck haulage costs, coupled with potential fuel shortages, forced greater emphasis on the development of haulage systems that either eliminated the need for trucks or at least greatly reduced truck haulage distances.

2) Duval Corporation installed in the company's Sierrita copper-molybdenum open-pit mine located near Sahuarita, Arizona, a movable, indirect feed crushing and conveying system that incorporated a 60" x 89" primary gyratory crusher with a rated capacity of 4000 DSTPH. 3) The United States Bureau of Mines presented their highly published treatise entitled MOVABLE CRUSHER SYSTEMS - CONCEPTS AND APPLICATION at the American Institute of Mining Engineers In-Pit Crushing and Conveying Symposium in Salt Lake City, Utah. This combination of need, visibility and credibility built on twenty-five years of related experience was the springboard for change. In the last twenty years, the mining industry has made great strides in improving efficiency. In-pit primary crushing and conveying systems have significantly contributed to increased profitability through reduced production costs. Installation of in-pit crushing and conveying systems has achieved marked reduction in overall mining costs by limiting truck haulage to short distance material transport between the face and the crusher and, in some cases, by completely replacing trucks with front end loaders.

ADVANTAGES OF IN-PIT CRUSHING AND CONVEYING Advantages of belt conveyor haulage as compared to truck haulage include: 1) Stationing of the crusher in the pit reduces cost by shortening the haulage distance between the loaders and the crushing plant. 2) Reduction of operating costs associated with fuel, tires, and lubricants. These products tend to increase in price at a rate that exceeds the rate of monetary inflation.

3) Reduction of manpower costs. Although most in-pit systems, either operating or planned, use truck haulage, the haulage distance is shorter and the number of trucks required is reduced by as much as 75%. This reduction produces a corresponding reduction in operators and maintenance personnel.

607

Mining ventures are long-term, and in-pit crushers and conveyors offer greater predictability for future costs. Dependence on the availability of fuel is greatly reduced. Conveyors can traverse grades of up to 30 degrees versus approximately 10 to 12 degrees for trucks. This ability facilitates shorter haulage distances and reduced haulage road construction. Conveyors can easily cross roads, railways, waterways, and other obstructions. With the reduction of haulage costs, lower-grade ore bodies can be mined economically. This is particularly important since many of the established ore bodies are decreasing in grade with depth. Downhill conveyors can produce regenerative electrical power instead of dissipating heat as in trucks. 10) Conveyors are more energy efficient than trucks. 11) Conveyors require less skilled labor for maintenance than trucks. 12) Maximum operational availability of the equipment resulting from greater independence from weather conditions such as fog, rain, snow and frost. 13) The cost of haulage road maintenance is significantly reduced by using conveyors. 14) Continuous flow of material can be maintained by using conveyors. 15) With the availability of technologies such as finite element analysis and computer simulation, in-pit crushing stations have been refined to a point where their performance and integrity is equal to that of traditional crusher stations. This refinement allows for far greater long term mine planning flexibility by allowing for future relocation as the mine expands.

DISADVANTAGES OF IN-PIT CRUSHING AND CONVEYING The two principle drawbacks of belt conveyor haulage as compared to truck haulage are:

1.

Reduced Short Term Flexibility The great mobility of trucks allows mine managers extreme flexibility in the mine plan. Once an overland conveyor is installed, it is prohibitively expensive to move it as part of a mine plan change. In mines where ore blending is important, truck flexibility provides an added advantage.

2.

Lump Size Limitations Once blasted, ore and waste in hard rock mines can be loaded directly into a truck and hauled out of the pit. Generally, it is necessary to crush the blasted ore or waste for conveyors. In the case of ore, mines must not only look at an overland conveyor for haulage, but they must also evaluate crushing of the ore, which generally requires moving the primary crushing stage into the open-pit mine. In the case of waste, the advantage of conveyor haulage must take into account crushing costs that would not otherwise be incurred if truck haulage were to be used.

608

DESIGN PARAMETERS The principal design parameters that drive IPCC selection and configuration include: production requirements truck sizes capital cost ore characteristics safety and environment project location (climate, geography, terrain) life of mine I expansion plans operational considerations maintenance requirements. Each of these is addressed in the sections that follow.

PRODUCTION REQUIREMENTS The process design criteria define the project’s production requirements, and typically include: Process Description:

0 0

General Ore Characteristics: 0

0 0

0 0 0

Operating Schedule:

0 0

0 0

0

General Primary crushing Maximum rock size Ore specific gravity Ore bulk density Ore moisture, wet season Ore moisture, dry season Angle of repose Angle of withdrawal Days per year Hours per day Nominal annual throughput Mining shifts per day Crushing plant shifts per day

The flowsheet specifies the nominal production flowstream requirements and the equipment suing to handle those flows. Crusher station receiving hoppers have to be designed to handle as quickly as possible at least one truck capacity minimum and in some cases up to three trucks capacity. The initial semi-mobile gyratory crushers were fed with an inclined apron feeder which allowed the overall height of a station to be contained within two bench heights and would allow for instantaneous dumping of material into the hopper. Some of the semi-mobile crushing stations now being designed are direct dump stations with high speed receiving conveyors taking the material direct to the out of pit transport conveyor or into an in-pit surge bin.

CAPITAL COST Large semi-mobile primary crushers cost plants can be very costly especially if they include the inclined apron feeder. It is dangerous to estimate crusher installation costs based simply on equipment price plus a contingency allowance for other costs. The following direct costs, including installation manhours, must all be taken into account: earthworks and civil engineering In-pit construction planning to not interrupt ongoing mining operations concrete

609

structural steel architectural mechanical electrical and instrumentation. Indirect costs can be at least half as much again as direct costs, and include: engineering, procurement and construction management (EPCM) start-up and commissioning construction equipment construction indirects spare parts freight taxes escalation owner’s costs (relocation, hiring and training personnel, permits, licensing fees, etc).

ORE CHARACTERISTICS Ore characteristics are a critical element in both crusher and conveyor selection. Dry ores require greater provisions for dust suppression and collection. Wet, sticky ores can plug chutes, plug crushers, reduce surge capacity, and misalign belts. For mines at which ore characteristics change over time, it can be costly to initially design a plant with the necessary flexibility. Some owners stipulate that initial capital investment be kept to a minimum, with design modifications paid for out of the operating budget. This is not always easy to achieve. SAFETY AND ENVIRONMENT Safety is a must. The modem plant includes safety rules, which must be followed and kept up to-date. Ongoing safety training of plant personnel is imperative. Crusher and conveyor operators must be part of the mining departments operating team.

PROJECT LOCATION A project’s geographical location, topography, remoteness and climate will all have an affect on crusher plant design. Construction costs are generally much greater at high altitudes, in cold climates and at remote sites. Modular construction and subsequent transportation to site can improve the economics of a project. Geography dictates what material can be best used economically in a particular region. A flat quarry operation lends itself to having the conveyor installed in one position for very long periods of time. A deep copper pit will sometime require that the crushing station and receiving conveyor may need to be moved from time to time. Naturally it would be best to find a wall that is requiring no more set backs and the conveyor could be installed either up this face with a high angle conveyor or a slot designed to install a conventional conveyor. Another alternative would be like the Island Copper Installation on Vancouver Island where the mine installed a conveyor in a tunnel up at 15 degrees out of the pit. Most in-pit crusher stations have not been totally enclosed structures with only the operator’s cab, electrical and lubrication rooms being enclosed.

61 0

LIFE OF MINE / EXPANSION PLANS The life of the mine is a key element in the design of any crushing plant. The selection of a fixed crusher versus a semi-mobile plant is a very important design consideration in the overall life of a mine. Moving a crushing plant and adding feed conveyors to feed to the takeaway conveyor can be very expensive. The costs per move can be up to $2,000,000. Any expansion plans for most IPCC systems should be built into the crusher and conveyor systems at the start of a project. Conveyor system tonnage’s can be increased in the future, simply by speeding up the conveyor, and if required, adding additional drives.

OPERATIONAL CONSIDERATIONS It’s important to provide a comfortable, well-ventilated workspace with drinking water and restroom toilet facilities nearby. The operator should also be able to see all the main parts of the crushing facility under his control, either through a good window or by means of TV cameras and monitors. Vibration and noise at any crusher station must be kept to a minimum. The conveyor should have vehicle access along at least one side.

MAINTENANCE REQUIREMENTS Keeping maintenance requirements to a minimum helps achieve higher overall operating availability. Scheduled preventive maintenance at the crusher station and conveyor involves a number of elements, including: crusher wear parts feeder wear parts conveyor skirting and adjustment oil and lubrication conveyor belt repair electrical and instrumentationadjustments visual inspections. Provisions must be made for either jib or mobile cranes to remove and replace crusher wear parts, concaves and main shaft. Trolleys, jib cranes and pull points should be designed to facilitate equipment maintenance. Oil and lubrication systems should be centralized and designed for easy automatic changes, with provisions for well-ventilated centralized lubrication rooms where possible.

PROCESS DESIGN CRITERIA Design Criteria Information Typically, the information required to develop an IPCC crusher system design criteria includes: geographic data climatic data process design data (process description, ore characteristics) civil design criteria structural design criteria mechanical design criteria electricalhstrumentation design criteria.

61 1

PLANT LAYOUT AND DESIGN A carefully designed layout can save big dollars, since it is structures and infrastructure, rather than major equipment items that represent the major cost element of the crushing plant. The mine planner and plant designer must prepare a layout that meets the needs of the design criteria, flowsheet and select the equipment in the most economical possible configuration. It's important to keep structural costs down, to design for ease of maintenance and operation, and to combine best practices with advances in fabrication and erection. Most in-pit crushing plants are designed by crusher manufactures, so it is imperative that the designer works closely with the selected equipment supplier. Remember to impress on the manufacturer that the production, process, economical and operational design needs come first. There are many new advances in in-pit crusher plant design, fixed plant, semi-fixed, semi-mobile feeder fed and semi-mobile dump fed. Out off pit conveyor systems or overland conveyors are very forgiving in design with the newer soft start systems and the selection of better designed idlers to assist in minimizing the power requirements of some systems. 3 D cad systems tied up with the mine planning 3D modeling assists greatly in being able to visualize the finished and phased effects of any IPCC installed material handling transportation system.

Tyoical IPCC Layout

PRIMARY CRUSHER SELECTION Primary Crusher Selection The crusher is the key to success with in-pit crushing and conveying systems. The in-pit crushing plant can be provided with almost any type of primary rock crusher. Selection of the primary rock crusher is based on three fundamental considerations. The type and characteristics of the ore determine the type of crusher required. Plant capacity determines the size of the crusher. Plant layout and design

History of Primary Crushers As the term "primary" implies, primary crushers are used in the first stage of any size reduction cycle. The gyratory crusher is the workhorse of the hard rock crushing industry. Primary gyratory crushers are capable of taking blasted run-of-mine and run-of-quarry feed in size up to 1500 mm (60 inches) and producing products ranging in size from 2 to 12 plus inches. This type of crusher can sustain production at rates between 350 to 10,000 tons per hour depending upon the feed characteristics, crusher setting and crusher size of the application.

612

The primary gyratory crusher is only one of a family of primary crushers which include the: Gyratory Crusher Double Toggle Jaw Crusher Single Toggle Jaw Crusher High Speed Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker All of the family of primary crushers can be used on mobile plants. Impactors and hammermills are compact and generate a high reduction ratio. The high speed of the machines requires special attention to dynamic forces. The jaw crushers are very good for small tonnage plants and the ST jaw has the advantage of being lighter in weight than the DT jaw. Large capacity jaw crushers result in large crushing plants. Double roll crushers are very large machines. The two rolls rotate inward so the out-of-balance forces are minimized. These machines are limited to relatively soft and non-abrasive materials. However, since the return on investment for crushing and conveying systems in the mining industry is heavily dependent upon both high capacity and the ability to handle hard and abrasive ores, the gyratory crusher has been the machine of choice throughout the evolution of IPCC.

MTPH

1500

3000

6000

9000

12000

Gyratory

DT Jaw ST Jaw Double Roll Low Speed Sizer Impactor Hammermill Feeder Breaker Primary gyratory crushers are typically furnished with radial feed openings of 1.07 m (42”), 1.37 m (54”),and 1.52 m (60”). The largest radial feed opening of any primary gyratory crusher operating in the world is a 1.83 m (72”) Traylor@crusher. The capacity of even the smallest standard unit, the 1.07 m (42”) gyratory, can be sustain in the range of 2500 DMTPH. The 1.52 m (60”) gyratory can crush more than 10,000 DMTPH, depending on the crusher design, ore characteristics, and desired product size. The first 60“ gyratory crusher was manufactured by Traylor Engineering & Manufacturing Company in 1919. At the time this machine was manufactured, the largest haulage trucks available had a 35 short ton payload and shovels were manufactured to match. In 2001 the operator has at his disposal 90 cubic yard shovels and haulage trucks with a 360 short ton payload. Haulage trucks with a payload of 400 short tons are on the drawing board. Truck manufacturers have advised that 500 ton trucks are a possibility. Tires are the only limitation. The result of larger haulage trucks is an obvious mismatch between the top size of ore fed to the crusher and the largest radial feed opening available. The consequence of this mismatch is bridging of two or more large lumps fed to the crusher at the same time.

613

Bridging has been partially compensated for by the use of hydraulic rock breakers installed on pedestal mounted booms. In some installations, the hydraulic rock breaker is employed up to one-fifth of the total time the crusher is in operation with 20% spent breaking oversize rock and 80% spent breaking bridges. LOW SPEED SIZERS In the early 198O’s, low speed sizers were introduced. This represented one of the only fundamental developments in primary crushers in three-quarters of a century. The main technical feature of the low speed sizer, which can broadly be considered a variety of toothed roll, is that it exploits the fact that the ratio of compressive strength to tensile and shear strength in the majority or rocks is approximately 1O:l. The low speed sizers breaks the rock in tension or in shear by its “snapping” and chopping action rather than in compression as conventional crushers do. Additionally, the positioning of the teeth on the rolls allows undersize to fall directly through the machine resulting in high throughputs at very low rotational speeds that leads to greatly reduced wear, energy savings, better control of discharge size in three dimensions, and greatly reduced fines. Low speed sizers are used for soft to hard non-abrasive, sticky types of materials up to 200 Mpa (29,000 psi). Examples: medium hard limestone, kimberlite, gypsum, clay, shale, schist and gold ore. The sizers are also used to crush bauxite and overburden where the host rock is relatively soft and the inclusions range up to 400 Mpa (60,000 psi). The low speed sizer is not particularly sensitive to abrasion if the reduction ratio is low. Low speed sizers are fabricated and the frame can be designed to accommodate material larger than 1500 mm (60”) if the trucks continue to grow and ROM feed continues to increase.

Types of In-Pit Crushing Systems The in-pit crushing systems developed and operated to date have varying degrees of mobility ranging from fully mobile units to permanently fixed plants which resemble traditional in-ground crushing plants. The following tenns are presented to help distinguish the range of mobility within the generic term of “in-pit crushing systems”.

Fully Mobile Crusher The fully mobile crusher is mounted on a steel platform and is self-propelled. The platform houses all auxiliary equipment and sub-systems to operate the crusher. The platform is self-supported and rests on the mine floor. The crusher is located at the working face to minimize truck or front-end loader haulage. Wheels, crawlers or pneumatic pads are integrated into the platform and drive power to move the equipment is included on board. The planned frequency of moves for a fully mobile crusher is between one day and one week.

Semi-Mobile Crusher The semi-mobile crusher is mounted on a steel platform. The platform houses all of auxiliary equipment and sub-systems to operate the crusher. The platform is self-supported and rests on the mine floor. The crusher is located near the working face to minimize truck or front-end loader haulage. Wheels or skids are inhgrated into the platform; however, drive power to move the equipment is supplied externally. The planned frequency of moves for a semi-mobile crusher is between five days and three weeks.

Movable Crusher The movable crusher is mounted on a steel structure. The structure houses all of the auxiliary equipment and sub-systems to operate the crusher. The structure is self-supporting and rests on the mine floor either with or without footers. The crusher is located near the centroid of the working portion of the mine to minimize truck haul distance. Bulkheads are built into the structure to allow for movement of the structure by commercially available transport equipment. The planned frequency of moves for a movable crusher is between one and ten years.

614

Semi-Fixed Crusher The semi-fixed crusher is mounted on a steel structure. The structure houses some or all of the auxiliary equipment and sub-systems to operate the crusher. The structure rests on a concrete foundation. The crusher is located at or near the edge of the pit. Some degree of disassembly is required to move the structure. The planned frequency of moves for a semi-fixed crusher is no less than ten years.

Fixed Crusher: Stationary Rim Mounted Crusher The stationary rim mounted crusher is installed in a concrete structure which is part of or attached to the bench wing wall. A portion of the structure may be fabricated steel and could be moved. The stationary rim mounted crusher would be installed for fifteen or more years.

Fixed Crusher: Stationary In-Ground Crusher The stationary in-ground crusher is installed in a concrete structure below grade. The crusher is usually located external to the pit. The stationary in-ground crusher is never moved.

Operation: Mobile and Movable Crushers The majority of fully mobile, semi-mobile and movable in-pit crushing stations utilize an apron feeder to lift ore into the feed opening of the primary crusher. The use of an apron feeder allows for the crusher station to either operate at grade or to utilize a single low bench. Additionally, the use of a truck dump hopper at the apron feeder creates a surge pocket between the mine and the crusher, making the flow of ore through the crusher more uniform and continuous. Advantages of indirect feed using an apron feeder include: 0 0 0

0

Low bench height for dumping ore Reduced truck cue time due to the surge pocket Improved control of oversize material fed to the crusher Reduced crusher downtime due to bridging of large lumps

615

Disadvantages of indirect feed using an apron feeder include: 0 0

0

Increased total capital cost Increased maintenance costs associated with adding an apron feeder Increased maintenance costs associated with the crusher resulting from using an apron feeder. Due to the nature of the feeder, ore tends to impinge upon small areas within the crushing chamber, causing premature localized wear of the concaves and mantles. The use of alloy steels has mitigated the problem, however, the cost of alloy steel components remains higher than manganese steel and availability is limited.

Operation: Semi-Fixed Crushers Semi-fixed crushers are the mining industry's attempt to incorporate the advantages of limited mobility while eliminating the need for an extremely expensive inclined apron feeder. Semi-fixed plants have incorporated both indirect feed using a horizontal apron feeder as well as various forms of direct dump arrangements. Due to the high capita1 and maintenance costs of apron feeders as well as the availability of high payload haulage trucks capable of sustaining high crushing capacities, the majority of semi-fixed crushing plants installed in the last five years have incorporated direct dump arrangements. In a semi-fixed crushing plant, a portion of the station is fabricated from steel. The direct dump feed hopper, crusher support structure and control rooms are almost always fabricated. Differences in design are related to the degree where the lower portion of the plant is concrete or steel. Advantages of semi-fmed crushers with direct feed arrangements include: 0

0 0 0

0 0

0

Traditional plants with simple configurations are easily adapted for in-pit crushers Reduced maintenance cost due to deletion of the apron feeder High crushing chamber throughput Reduced capital cost due to limited degree of mobility Increased long term flexibility due to limited mobility Reduced maintenance cost due to greater amount of crushing in the upper portion of the chamber and decreased localized abrasive wear Greater capacity and finer product size due to the weight of the ore column

Disadvantages of semi-fixed crushers with either indirect or direct feed arrangements include: 0 0

Difficult to move Greater overall height due to the higher the dump point bench level

Operation: Semi-Fixed and Fixed Rim Mounted Crushers Semi-fixed and fixed rim mounted crushers are increasingly incorporating traditional direct dump arrangements. With these designs, the hopper above the gyratory crusher is designed to hold 1.5 to 2 times the capacity of the largest truck which will dump into the crusher during operation. Discharge surge bins have traditionally been sized slightly larger than the feed hopper to accommodate any unusual fines condition. In order to reduce overall height, and thus capital costs, discharge apron feeders have been replaced by impact resistant discharge belt conveyors. The trend away from discharge apron feeders to discharge belt conveyors has allowed for wider belts with greater capacity. In conjunction with high capacity discharge belt conveyors, the typical capacity of the surge bin below the crusher has decreased dramatically. Even with removing the discharge apron feeder and reducing surge bin capacity, direct dump arrangements result in tall structures. With rim mounted in-pit crushers, this tall overall height requires wing walls to support and reinforce the structure. Traditional fixed crushers are installed below grade and fed at grade. Recent in-pit installations such as at Kennecott Copper in Bingham Canyon, Utah and El Abra in Chile are partially below grade and partially above grade to accommodate a single bench height. The latest generation of dump pockets are arranged for either two or three dump positions. With a two position dump pocket design, the two dump points are set 90 degrees apart from each other. The spider

61 6

orientation is either in-line with the centerline between the two dump positions or at 90 degrees to the centerline between the two dump positions. Either position is mechanically satisfactory to the gyratory crusher. With a three position dump pocket design, it is universally accepted that the spider is orientated in-line with the centerline of the central dump position. Advantages of fixed rim mounted crushers with direct feed arrangements include: 0 0 0

0 0

Traditional plants with simple configurations are easily adapted for in-pit crushers Reduced maintenance cost due to deletion of the apron feeder High crushing chamber throughput Reduced capital cost due to limited degree of mobility Increased long term flexibility due to limited mobility Reduced maintenance cost due to greater amount of crushing in the upper portion of the chamber and decreased localized abrasive wear Greater capacity and finer product size due to the weight of the ore column

Disadvantages of fixed rim mounted crushers with either indirect or direct feed arrangements include: Difficult to move Greater overall height due to the higher the dump point bench level

Conclusion In-pit crushing and conveying systems with primary gyratory crushers have been the answer to escalating energy and labor costs from 1980 through 2000 and have the potential to continue to substantially increase the profitability of open pit mining. The type and configuration for virtually any site-specific application have been designed and built by the world's leading manufacturers and have operated for sufficiently to prove their viability. As evidenced by recent installations such as El Abra and Radomiro Tomic in Chile which operate above 8500 DMTPH, there is a definite trend toward increased capacity for large primary gyratory crusher installations from 1982 to 2001. Similarly, there has been a steady increase in the size of widely available haulage trucks during the same period from 165 tons to 360 tons. In contrast, there has not been an increase in the radial feed opening of the primary gyratory crusher since the first Traylor@60" crusher was built in 1919 to the present. In-pit crushing stations have decreased in mobility from the filly mobile and movable designs dating from 1982 to 1985 to the semi-fixed and fixed rim mounted stationary designs of 1998 to 2001. For example, the 54" movable indirect feed crusher at Minera Escondida in Chile was built in 1990. When the company installed a 60" crusher in 1995, a semi-mobile direct dump design was chosen. Similarly, the 60" movable indirect crushers at Codelco Chuquicamata in Chile dating from 1984 have been rebuilt, relocated and reinstalled in new semi-fixed direct dump installations. The 90 cubic yard shovel and the 360 ton haulage trucks (and soon to be available 400 ton haulage trucks) have created a gross mismatch between feed size and feed opening with the primary gyratory crusher. Consider that a single bucket load weighs more than the total volume of ore which can be held within the largest primary gyratory crusher currently in operation. Further, a single truckload weighs three times the total weight of the very common 42" primary gyratory crusher. It is physically possible for a dipper to pick up a single rock 2.5 m (100") in diameter and deposit it into a crusher with a 42" radial feed opening. Change is required in the comminution flow from fragmentation in the pit to the size reduction in the IPCC. The top size of ROM feed to the IPCC must be controlled or the industry needs to consider alternative crushing technologies to compensate for lack of top size control and still be i n 9 position to take advantage of the economies of scale with shovels trucks and IPCC.

617

The Future: Beyond 2000 IPCC PROCESS CONTROL There exists growing recognition in the industry of the affect mining practices have on the efficiency of mineral processing operations. Among numerous variables, size distribution of ore is a variable widely accepted to have a significant effect on the throughput and recovery achieved in mineral processing. Using .available digital imaging technology, improved process monitoring and modeling provides the opportunity not only to identify these variables but also to monitor and control these variables throughout the entire mining process in real time. The majority of mine designs employ crushers to reduce size of the material originating from the mine. The primary crusher is the link between the chemical comminution (blasting) and the beginning of the mechanical comminution circuit (crushing and milling). As such, an in-pit crusher is not only a key point in the process to apply a measurement monitor, but also is a key resource to be optimized. Feed to the primary crusher from the mine can be measured and monitored to establish blasting performance. Size information associated with each haul truck can be traced back to location in the mine plan and use to help design future blasting practices. The product of the crusher is usually the beginning of the mineral processing circuit that involves more energy consumption to further reduce fragment size. In the short-term, how the crusher performs is the responsibility of the crusher operator who, through the use of digital imaging, now has a record of the quality of the crusher product size and can make adjustments to the crusher to keep the product “in spec” as required by the design of the remainder of the comminution circuit. In the long term, archives of size trends of crusher feed and product as related to other key performance indicators such as blast fragmentation, crusher/mill throughput, crusher reduction ratios, Bond Work Index, energy consumption and efficiency expose management level optimization decisions. Opportunities for optimization include: how to tailor blasting to feed the stationary mechanical comminution circuit; how to load the crusher; how to establish better pro-active maintenance of the crusher; keeping tighter specs on the feed and products of the various stages of comminution. By applying digital imaging technology, a technology widely used in numerous other manufacturing industries, new innovative solutions are available to the mining industry. Applying imaging technology can generate volumes of size information not previously possible for use in long-term studies, as well as in short-term process control as well as create long-term operating savings as compared to manual sizing methods or worse yet no measurement at all! An engineer with over thirty years of mining experience was quote to say the following about image analysis, “At our mine we utilize digital image analysis systems to provide quality, quantitative fragmentation information on our blasting and integrate the fragmentation information into our operations database as a quality control mechanism within our ongoing continuous improvement program. An important step to controlling costs is controlling your process. Basically, any company that has a product that requires control of particle size and is concerned with profitability needs this valuable information.” ALTERNATIVE IPCC COMMINUTION EQUIPMENT The standard 360 ton truck of the year 2000’s will feed semi fixed and fEed crushing plants utilizing 72“ primary gyratory crushers or some other type of primary crusher capable of size reduction of ROM material larger than 1500 mm (60”). Efficiency will increase as ore grade decreases. Capacities will increase as a result of the larger crushers while, at the same time, operators will take advantage of the larger crusher to obtain the economy of scales. The type of in-pit crushing plant has evolved from fully mobile to semi-mobile to moveable to semi fixed and in many cases to rim mounted stationary structures in the period 1980 to 2001. This progression from fully mobile in the 1970’s and early 1980’s to less mobility throughout the balance of 1980’s and into the 2lStCentury would seem like a step backwards, considering that the original premise of in-pit crushing was to reduce dependence on truck haulage. The evolution if the in-pit crushing plant can not be considered by itself but must be considered in relation to other changes in the size reduction of ROM ore in the pit and delivering to the secondary stage of comminuation. The platform for the development of IPCC was the “typical” multi-bench mine. The premise was that the design of the plant should be one bench high, i.e. 45 feet for slope stability, economy of dump pocket design and safety.

618

Inserting an apron hopper between the dump pocket and the primary crusher could attain the dump height. In 1980 the thought was that the crushing plant would be moved every time that the pit became deeper by one or two benches. The “typical” mine would be in a position to move the crushing plant in less than 5 years. The plant move could be accomplished in as few as 3 days. The truck size in 1980 was 165 tones. The development of the 240-ton, the 300-ton, the 350-ton truck and maybe the 400-ton haul truck has had a profound effect on the in-pit crushing plant designer and operation of in-pit crushing systems. With the larger trucks, the dump pocket was no longer a simple structure. The economies of scale and efficiencies of the new larger haul trucks negated the original parameters for moving the plant for every one or two benches of depth, or every few years of operation. The operational advantages for the use of apron feeders to elevate the ROM ore was largely gone due to the cost of the dump pocket and the limited number of move required in the life of the mine. Maintenance cost that showed that the operating costs of the inclined apron feeder was equal to or more than the maintenance cost of the gyrator crusher. The capacity of the gyratory crusher was actually less than when direct dump the crusher. The produce from the gyratory crusher is finer when direct dumped than when apron feeder fed. Concave and mantle costs are higher when the crusher is fed with the apron feeder. Average capacities increased from 3000 to 9000 MTPH. The net effect of these changes was the progression that evolved into rim mounted stationary plants and to the semi-fixed designs. Both of these designs are based on: Plant designs that make use of the slope topography of the pit which reduces structure cost and maintains some measure of low bench height. Plant relocation of 10 years of more Capacities to 9000 MTPH. Direct dump formats Gyratory crushers

This mismatch in shovel size, truck size and crusher size will likely manifest itself in fundamental changes to the primary crusher andor improved and innovative crushing plant designs rather than the installation of the larger and heavier gyratory crushers. Three of the eight types of primary crushers have achieved operating capacities of 10,000 MTPH. Gyratory crushers are operating at or near these capacities on copper oxide ores in Chile. A 10% increase in capacity will be achieved with the next size of primary, assuring continuous operation at the 10,000 MTPH level. Low speed sizers have attained capacities of 14,000 MTPH operating in oilsands projects in Canada. Double roll crushers have also attained capacities of 14,000 MTPH operating in oilsands projects in Canada. Gyratory crushers can crush material with compressive strengths as hard a 600 Mpa (90,000 psi). Low speed sizers are crushing material with compressive strengths of 200 Mpa (27,500 psi) on a continuous basis. The giant double roll crushers are currently limited to 100 Mpa (14,500 psi).

619

The designers of the IPCC crushing plants are constantly working toward shorter, more cost effective facilities. The low speed sizer with its inherent low profile design and proven high capacity in crushing materials with characteristics similar to copper oxide, appear to provide the most realistic prospect of transfer of technology for future mining IPCC facilities. Comparison of the “next larger size” gyratory with a low speed sizer. ROM - 10,000 MTPH Top Size - 1500 mm (60 inch) Product - 80% - 9”

Gyratory Crusher 72 - 124

Low Speed Sizer 161350

Weight: 635 tons

Weight: 155 tons

ACKNOWLEDGEMENTS We are grateful to*TomBob0 of Split Engineering for his input in Digital Imaging. In addition, we thank Man Takraf for the conceptual drawings used in this paper.

620

Selection and Sizing of Secondary and Tertiary Cone Crushers Gary Beerkircher', Kurt O'Bryan', and King Lim3

ABSTRACT The selection of crushers for moderate to hard materials is the primary focus of this review. Factors affecting machine selection will be presented. Modern crushers have increased in size and performance. Requirements for today's crushers have evolved to a greater focus on the desired quality of the crushed product. Aggregate production today must meet more stringent shape and gradation requirements. In crush to leach applications, proper size reduction results in better recoveries. In mill feed preparation, the generation of fines and total top size reduction are key to maximum mill productivity. Proper understanding of a crusher's capabilities will minimize both installation and operating costs. CONE CRUSHERS Cone crushers today have increased performance capabilities as compared to the first cone crushers developed in the mid-1920's by Edgar B. Symons. Not only do the cone crushers today have more power capability; they are larger in size with higher capacities, offer better product shape, and a higher percentage of fmal product yield. In recent years, the mechanically loaded spring type cone crushers have been replaced by safer more reliable hydraulic clamping and clearing systems to protect the cone crushers from uncrushables and overload conditions. The adaptation of hydraulic setting adjustment system in the cone crusher's design has also improved the overall efficiency of the crushing operation. The new generation cone crushers provide ease of operation, simple maintenance, uniform production throughout the liner life, and high availability. Development in cone crusher technology has evolved to include computer controls to maximize and optimize the cone crusher performance based on application requirements. Modem solid state devices provide real time feedback from the cone crusher such as power draw, cavity level, crushing force, lubricating oil flows, temperatures, pressures, and filter conditions. The information prxjvides inputs into computer controls which in turn can vary the feed rates and crusher setting to maximize the cone crusher performances accordingly. Application tools are available to assist in the flow sheet development and equipment selection. However having an understanding of the crusher selection criteria is the key to a successful installation of secondary and tertiary crushing equipment.

' Metso Minerals Industries, Inc., Milwaukee, Wisconsin.

2

3

Metso Minerals Industries, Inc., Milwaukee, Wisconsin. Metso Minerals Industries, Inc., Milwaukee, Wisconsin.

621

CONE CRUSHER SELECTION CRITERIA Gathering the correct application information is the key to the proper selection of cone crushers for a specific application. The information needed includes the capacity requirements with consideration for expected availability of the overall crushing circuit. Information on expected feed gradation and product size is required along with characteristics of the material being crushed. The material characteristics should include the specific gravity or bulk density of the material, impact work index, moisture content, and abrasion index. Information on how the material breaks is helpful as well. Actual data from an existing operation if available, is a valuable aid in the selection of the secondary and tertiary crushmg equipment. For "Greenfield" projects, information on the material characteristics should be obtained from either laboratory tests on small size samples or a full scale pilot crushing test can be performed to obtain the required material characteristics. CONE CRUSHER DESIGN LIMITS Before one can properly select a cone crusher for a given application, three design limits of a cone crusher must be understood. The design limits are as follows:

0

VolumeLimit PowerLimit ForceLimit

Volume Limit The volume limit of a cone crusher is the maximum rate of feed to the cone crusher without overfilling the cone crusher feed hopper. The volume limit is a function of the cone crusher speed, closed side setting (CSS), head angle, and material density. Generally, the capacity tables in most manufacturers' catalogs will reflect the cone crusher% volumetric capacity for a specific closed side setting with a standard bulk density of 1.6 t/m3. Most manufacturers indicate a capacity range at a given closed side setting in their catalogs. Seldom do the manufacturers indicate the precise conditions at which a more specific capacity can be achieved. Feed gradation, crushing chamber configuration, transportation of material through the crushmg cavity, and fragmentation characteristics of the material are among the many variables that define the volumetric limit of a cone crusher for a specific application.

Power Limit The cone crusher power limit is reached when the average power draw (kW) of the cone crusher exceeds the installed motor power on the cone crusher. Ore of high impact work index or strong resistance to fragmentation tend to reach or exceed the power limit more easily. High ratio of reduction (within the available crushing force) will also tend to reach or. exceed the power limit. Without sufficient crushing force or volumetric capacity the cone crusher will not have the ability to reach its theoretical installed power limit. A common misconception in cone crusher application is that all cone crushers are capable of producing capacities as stated on manufacturers' brochures at the installed power. Plant designers should receive an analysis from manufacturers as to expected power consumption at specific production rates and reduction ratios. If possible, a pilot crushing test can provide the power consumption (kWh/t) information, then the power consumption information can be used directly to determine the cone crusher capacity at a specific closed side setting.

Force Limit The force limit of a cone crusher is reached when the combined forces exerted during crushing exceed the force available on the machine to hold the desired closed side setting. For the older

622

cone crushers this would be the spring force holding the adjustment ring to the mainframe. Newer cone crushers utilize hydraulic cylinders to hold the adjustment ring to the mainframe. In the case of spider type cone crushers with the mainshaft supported by a hydraulic piston, it is the force maintaining the mainshaft position. Force limits may be exceeded due to un-crushable material (tramp steel, rubber, wood) entering the crusher cavity, operation at too small a closed side setting, packing of wet and sticky material, high power draws, or incorrect crushing cavity configuration. Exceeding the crusher force limit for any extended period of operation will ultimately lead to cone crusher damage and component failure. The crushing force limit for a specific type of cone crusher can be correlated readily to the overall mass of the cone crusher. The higher the crushing force limit, the greater the applied loads on the structural components of the cone crusher. Therefore, a high crushing force cone crusher must be heavier than a cone crusher of similar design with a lower crushing force. A cone crusher with a low crushing force limit may be unable to sustain continuous high power consumption. A low crushing force limit cone crusher design may not have the flexibility in achieving high reduction ratios or high productivity. Vibration switches can be installed on the cone crusher adjustment ring to detect when the crushing force is exceeded (adjustment ring movement). A signal is sent to the crusher operator when the force limit is exceeded. The crusher operator should then correct the root cause. Often the problem is too high a reduction ratio. On cone crushers that have an automation package as part of the controls, the vibration switch signal can be used to automatically open the closed side setting or, if the vibration levels are severe, interrupt the feed to the cone crusher.

CONE CRUSHER SIZES AND CAPACITY RANGES For a preliminary estimate on the size of a cone crusher, manufacturers have published capacity data for their cone crushers. The latest trend in data published is a capacity range (Table 1 shows Nordberg cone crusher sizes and capacity ranges at two different closed side settings). There is no standard applied for stating performance data amongst the manufacturers. A detailed review of the application should be conducted by the equipment supplier and plant designers. Capacity adjustment should be considered for high reduction ratio applications or materials with a very high impact work index. Table 1 Nordberg cone crusher sizes and capacity ranges Power Max. Feed Capacity @12 mrn HeadDia. Model Rating (kW) Size (mm) Closed Side Setting (mm) HP 100 HP200 HP300 HP400 HP500 HP800 MP800 MP 1000

700 900 1100 1300 1500 1800 2100 2400

75 150 225 300 375 600 600 750

141 183 210 301 350 353 378 378

120 - 150 115 - 140 140 - 175 175 - 220 260 - 335 495 - 585 615 - 730

Capacity @ 32 mm Closed Side Setting

190 - 235 250 - 320 325 - 430 405 - 535 545 - 800 932 - 1145 1160 - 1500

SECONDARY CONE CRUSHER SELECTION With an understanding of the cone crusher design limitations and in conjunction with the information presented in Table 1, one can start the cone crusher selection process for a specific application. Typically feed to the secondary (also commonly referred as Standard) cone crusher is scalped if the percentage of the desired product exceeds 10 to 15% of the new feed. The secondary

623

cone crusher generally operates in an open circuit arrangement with the product going to a tertiary screening process prior to tertiary crushing. Several considerations are taken into account for selecting the proper secondary cone crusher. First, ensure that the feed material to be crushed does not exceed the acceptable maximum feed size for a specific size cone crusher. Next is to determine the proper size cone crusher that will meet capacity requirements at a given closed side setting based on a 4 to 6: 1 reduction ratio. For example, what sue cone crusher is needed as a secondary cone crusher to meet the following requirements: maximum feed material is 200 mm and minimum capacity of 500 tons per hour? Referring to Table 1, a HP300 is suitable to accept a maximum feed size material of 200 mm. Once we have the minimum cone crusher size determined, we can proceed to see if the capacity requirements can be achieved using a HP300 cone crusher. An appropriate closed side setting has to be selected prior to meeting the minimum capacity requirement. Using a 6: 1 reduction ratio, the proper cone crusher closed side setting (in a secondary crushing application) with a 200 mm feed size is approximately 32 mm. Now, referring to Table 1 again, we see that the HP300 at 32 mm closed side setting is unable to achieve the required minimum capacity of 500 tons per hour. Hence, the proper size cone crusher according to Table 1 is a HP500 to meet the outlined requirements of accepting maximum 200 mrn feed material and achieving minimum capacity of 500 tons per hour. For simplicity the example described did not take into consideration the cone crusher design limits described earlier in this review. Often overlooked in the selection and sizing process of either a secondary or tertiary cone crusher is the correct cavity configuration. The cavity configuration has to suit the feed gradation so that maximum crushing performance and liner utilization is achieved. Several cavity configurations are available for secondary cone crushers to maximize the performance of the cone crusher. In addition, an improper liner configuration applied can create high crushing forces leading to adjustment ring movement, exceeding crusher force limit.

Examples of Secondary Cone Crushing Applications When a new flow sheet for a “Greenfield ” project is developed, both the secondary and tertiary crushing equipment can be selected to maximize the overall crushing operation performance. Opportunities exist in current operations to increase the performance of the overall crushing operation by upgrading the secondary crushing section of the plant. The following examples of existing crushing plant upgrades demonstrate the results (high reduction ratio and higher final product yield) possible in a secondary crushing application. Western USA Copper Mine. Three MP800 cone crushers were installed to replace three SYMONS 7 ft cone crushers. This is a two-stage crushing plant producing rod mill feed with the scalping screen undersize and secondary cone crusher product. Refer to Figure 1 for performance data, ore characteristics, and gradation chart on this application. Copper Concentrator in Mexico. A large copper operation chose to replace existing SYMONS EHD 7 ft cone crushers with MPlOOO cone crushers to increase the overall plant capacity and to produce a higher percentage of the ball mill feed (-13 mm product) in the secondary crushing application. Refer to Figure 2 for performance data, ore characteristics, and gradation chart on this application. Granite Quarry Operation in North America A granite quarry operation in North America has installed a MP800 cone crusher as a replacement for a SYMONS 7 ft cone crusher to increase capacity and improve final product yield in the secondary crushing application. Refer to Figure 3 for performance data, ore characteristics, and gradation chart on this application.

624

~

PERFORMANCEDATA Production 635tph css 19mm 140mm F80 P80 18mm RR80 7.80 % Passing CSS 83.0% Energy Consumption (kWh/t) 0.94 Operational Availability 90 - 95%

100

90 80 70

.-p

60

50 n

s

40

30 20

ORE CHARACTERISTICS Ore Type Copper Ore Avg. Work Index 16.5kWh/t Max Work Index 19.1kWh/t Specific Gravity 2.50 Bulk Density 0.378gms Abrasion Index 14gms/t Wear Metal Rate

10

0

1

100

10

IOOC

Size (mm) -.o- Feed Gradation +Product

Gradatton

PERFORMANCE DATA 770 870tph

-

P80 RR80 'Yo Passing CSS

24mm 6.46 76.5%

f

50

P

$ 40

30 20 10 0

Copper Ore

Ore Type I

Max. Work Index Specific Gravity

14kWh/t 2.60

I

10

1

100 Size (mm)

+Feed

Gradation +Product

Gradation

1

Figure 2 Copper concentrator in Mexico, secondary crushing application

PERFORMANCE DATA

P80 RR80 o/o Passing CSS

1.23in 7.32 86.8% 20 10 0 0.01

ORE CHARACTERISTICS Avg. Work Index Max. Work Index Specific Gravity Bulk Density

13.7 25.1 N/A 761b/ft3

0.1

1

1oc

10

Size (inches) [ L A s s u r n e d Feed Gradation +Product

Gradation

Figure 3 Granite quarry operation in North America, secondary crushing application

625

1

TERTIARY CONE CRUSHER SELECTION Tertiary crushing (third stage of crushing) in most cases is the final crushing stage. Feed sizes to a tertiary (also commonly referred to as Short Head) cone crusher are typically between 150 mm and 25 mm. Again it is important to have the correct cavity configuration to suit the feed gradation so that maximum crushing performance and liner utilization is achieved. With an understanding of the cone crusher design limits, one can then proceed to select a flow sheet and equipment for a specific application keeping the overall tertiary crushing stage reduction ratio in the range of 4 to 6:l. The feed to a tertiary cone crusher shall be pre-screened to remove the finished product sizes and to provide void space for the crushed particles produced in the cavity. The tertiary cone crusher can operate in an open circuit environment with the cone crusher product being combined with the screen undersize to become the final product. In many circuits, the tertiary cone crusher is operated in a closed circuit environment with the pre-screen prior to the tertiary cone crusher. In a closed circuit environment, the screen undersize becomes the final product. Maximum production will be obtained when the cone crusher operates at or near full horsepower load continuously. Good feed control via a bin and feeder ahead of the tertiary cone crusher to allow for choke feeding will maximize the feed rate, power draw and desired product production. For full utilization and sound mechanical operation of the cone crusher a closed circuit installation is preferable to an open circuit installation. Examples of Tertiary Cone Crushing Applications Just as with secondary crushing, tertiary crushing performance can be maximized with good flow sheet design on "Greenfield" projects. Nevertheless, many opportunities exist in current operations to upgrade the tertiary crushing portion of a plant. The following examples of existing crushing plant upgrades demonstrate the results (maximum fines generation) possible in a tertiary crushing application. Copper Heap Leaching Operation in Chile. A new "Greenfield" operation in Chile utilized nine MPlOOO cone crushers (three secondary cone crushers & six tertiary cone crushers) operate in a large heap leach copper mine to produce 130,000 tons per day of a nominal 19 mm heap leach feed material. Refer to Flow sheet 1 for secondary and tertiary crushing arrangement. QTY. 3

MPlOOO st s.!mommn

QTY. 6

MPlOOO sh S.ni"l11 mn

Flow sheet 1 Copper heap leaching operation in Chile

626

Copper Mine in Poland. An example of a tertiary cone crusher upgrade is at a large copper mine in Poland, where a HP700 short head cone crusher was installed to replace a hammennill that was used to prepare rod mill feed. The HP700 cone crusher allowed the rod mill circuit to achieve a 20% gain in energy efficiency by reducing the rod mill feed from 80% passing 30 mm to 80% passing 14 mm. Refer to Figure 4 for performance data, ore characteristics, and gradation chart on this application. Production

css 'F80 P80

PERFORMANCE DATA 360mtph 1 Ilrnrn 51mm 13mm

RRBO . .. .- -

% Passing CSS Energy Consumption (kWh/t) Operational Availability

100 90

80 70

4 25

.-p

60

76.3% NIA >95%

s

40

3 n

50 30

20 10 0

10

1

100

Size (rnm) 'Feed Gradation

Product Gradation

Figure 4 Copper mine in Poland, tertiary crushing application Other operations have used new high performance cone crushers to upgrade existing operations in both the secondary and tertiary crushing, including:

0

Large copper mine in Peru where MPlOOO & MP800 cone crushers have been installed in secondary positions and HP700 & HP800 cone crushers in tertiary positions. Large SW USA copper operation where five MP800 cone crushers were installed in both the secondary and tertiary crushing positions to replace existing SYMONS 7 ft cone crushers.

CONCLUSIONS Cone crushers have evolved to new levels of performance. The physical size, power, and crushing force capabilities have increased. Automation or computer controls offer the ability to maintain maximum performance levels of the cone crusher with feed rate or cavity level control, remote adjustment capability, liner wear detection, and monitoring of the crusher's vital signs. Remote monitoring of cone crusher operation, coupled with historical trending provides the user with valuable operating data. Application tools via computer simulation programs assist plant designers with equipment selection and the ability to perform many "what if'' scenarios. The ongoing emphasis on high tonnage heap leach projects will take advantage of the latest technology on crushing to minimize both operating and initial installation costs. Larger, higher capacity cone crushers result in fewer units being required. Upgrades of existing crushing facilities will continue to present opportunities for the installation of new crushing technology.

627

Selection, Sizing and Special Considerations for Pebble Crushers Kurt O'Bryan' and King Lim'

ABSTRACT Pebble crushers serve a unique role in the comminution flow sheet. The application of crushers, relative to the crushing of critical size material, has broadened to crushing critical size material both before and after the primary grinding mill. Selection and sizing of pebble crushers is impacted by special considerations associated with linkage of a crusher to a grinding mill. These special considerations include: variable feed rates, variable feed size distribution, wet and sticky feed, tramp contaminants, and the production benefit of maximum fines generation in a single pass. Along with the selection and sizing of pebble crushers, a thorough discussion of the special considerations common in pebble crushing circuits is provided. THE ROLE OF CRUSHERS Crushers were immediately affected by the development of autogenous or semi-autogenous grinding mill circuits. Initial flow sheets made no consideration for crushers. However, in a short period of time, the need for supplemental crushmg, particularly with hard competent ores became evident. Operators quickly discovered the production benefits of extracting from the grinding mill what has been technically described as the "critical size" material for open circuit crushing and subsequent return to the primary mill for grinding. Relative to ore characteristics, productivity gains as high as 50% could be achieved with simple open circuit crushing of the critical size material. Critical size material is commonly referred to as pebbles, and crushers selected for crushing pebbles are commonly referred to as pebble crushers. Pebbles are typically rounded and without sharp edges. The rounded characteristic of mill discharge pebbles is a result of attrition grinding in the primary mill. A pebble crusher fractures the material at a more efficient and rapid rate due to compressive forces applied to the pebbles by the crushing surfaces or due to impact breakage as generated by the inertia of an impact crusher. Most common and the focus of this paper will be the role of compression crushers which is the commonly accepted practice for pebble crushing. The role of compression crushers has continued to evolve to significantly higher levels than simple open circuit crushing in a mechanically loaded spring cone crusher first developed in the mid-1920's by Edgar B. Symons. Today, we have new experiences with two-stage crushing and pre-crushing. Also today, in most cases, the spring type crusher typical of the SYMONS cone design (Figure 1 shows a SYMONS cone crusher) has been replaced by a crusher with more suitable features of hydraulic clamping, clearing, and setting adjustment (Figure 2 shows a HPSeries cone crusher equipped with hydraulic clamping, clearing, and setting adjustment). Coupled

' Metso Minerals Industries, Inc., Milwaukee, Wisconsin. Metso Minerals Industries, Inc., Milwaukee, Wisconsin.

628

with advances in automation and greater knowledge of the overall comminution process, the major role cone crushers serve in crushing pebbles is widely accepted and applied.

Figure 1 Symons cone crusher

Figure 2 HP series cone crusher THE PRE-CRUSHING OPTION The most recent evolution for pebble crushing finds a basis in the presumption that the most appropriate primary mill feed contains a minimal amount of critical size material. The initial feed to the primary mill should dominantly consist of coarse and fine material. The coarse material has sufficient mass to serve a role as impact media and the f i e material would transport through the mill to downstream processes. In pre-crushing, typically the target is to convert the middling size fractions to fines. Such is the case at a Canadian installation. At Troilus Mine, (Sylvestre, Abols, and Barratt. 2001) the 150 mm by 50 mm size fraction is pre-crushed with a HP700 cone crusher. Productivity increased and operating costs decreased. An earlier pre-crushing success in Australia (Needham and Folland. 1994) achieved excellent results with pre-crushing. At Kidston Mine, the primary crushed ore is pre-screened simply to remove fines. All +50 mm oversize is crushed at

629

maximum reduction ratio to deliver maximum fines. Both methods used by Troilus and Kldston have proven effective in boosting milling productivity and lowering operating costs. Installing a Pre-Crusher Installation of a pre-crusher is typically after the primary crusher and prior to the mill (Flow sheet 1 shows a typical pre-crusher installation). Primary crushers typically operate intermittently due to direct feed by truck dump, therefore, to dampen these mine induced surges, a feed bin or surge pile should be installed prior to the pre-screen. The screen oversize or the middling size is conveyed to the pre-crusher for crushing. Ideally, the pre-crusher should also have a bin and feeder. A bidfeeder arrangement is even more critical if there is a requirement to operate more than one precrusher. With the bin and feeder arrangement, a controlled feed to the crusher can result in a full cavity. A full cavity ensures both maximum mechanical and process efficiency (maximum capacity and size reduction) of the cone crusher.

Gyratory Crusher

Primary Surge Pile Screen

Pre-Crush Cone Crusher SAG Mill

Flow sheet 1 Typical pre-crusher installation

Sizing a Pre-Crusher The selection of the size of the pre-crusher should be based on the production needs of the grinding circuit. A typical utilization factor with today's modem crushing plants is 75 to 80%. The utilization factor should be combined with the maximum expected primary mill feed rate to determine the appropriate size of the pre-crusher. For example, if the maximum mill feed rate is 900 metric tons per hour, the average capacity of the pre-crushing system should not be less than 1200 metric ton per hour. The actual size requirement for the crusher would be defined by the quantity and size distribution of the material to be crushed at this 1200 metric ton per hour rate.

630

For a preliminary estimate, manufacturers publish capacity data for their crushers. The latest trend in data published in brochures is a capacity range (Table 1 shows Nordberg cone crusher sizes and capacity ranges) and there is not a consistent standard applied for stating performance data amongst the manufacturers. A detailed review of the application should be conducted by the equipment suppliers and plant designers. Capacity adjustments should be considered for high reduction ratio applications or materials with a very high impact work index.

'able 1 Nordberg cone crusher sizes and capacity ranges Head Dia. Power Model

HP 100 HP200 HP300 HP400 HP500 HP800 MP800 MP 1000

900 1100 1300 1500 1800 2100 2400

150 225 300 375 600 600 750

120 - 150 115 - 140 140 - 175 175 - 220 260 - 335 495 - 585 615 - 730

183 210 301 350 353 378 378

190 - 235 250 - 320 325 - 430 405 - 535 545 - 800 932 - 1145 1160 - 1500

CRUSHING AFTER THE MILL Crushing after the SAGIAG mill is common milling practice employed throughout the industry. Due to the nature of the milling environment, the crusher faces a greater challenge than in many other applications. For example, greater exposure to tramp material is a certainty. To address the tramp material issue, common solutions are trash screens, steel removal by magnets, metal detectors, and bypass conveyors. Cone crushers are also adversely affected by excessive moisture. Excessive moisture reduces both the throughput and size reduction capability of the crusher. Excessive moisture is a result of inefficient size classification following the primary mill. In designing a pebble crushing circuit, expect the quantity and size of the pebbles to vary widely. The variation will result from the ore body characteristic and the worn condition of the primary mill discharge grates. The pebble crusher must be able to deliver continuous and unintempted performance under extreme (tramp material in feed and excessive moisture) and unbalanced conditions (varying quantity and size of the pebbles).

Extracting the Pebbles Primary mills have two primary alternatives for size classification, trommel screens attached to the mill trunion and/or vibrating screens located directly at the mill discharge. Of the two options, a vibrating screen is clearly preferable due to higher capacity and efficiency. Double deck vibrating screens mounted in low profile or horizontal configuration are preferred. Screening surfaces are generally slotted polyurethane or rubber. In large mills, the plant designer may make provisions for a second standby screen. The screens would be supported by rails so the standby screen can be quickly moved to the operating position, and the formerly operational screen would be prepared by maintenance for the next operating cycle. Tramp Removal Once the size classification process is complete, the extracted pebbles are conveyed typically to the crusher or to a storage bin. During the conveyance phase, the pebbles will pass under magnets to remove steel balls. A final inspection with a reliable metal detector is critical. Large metal tramp, not removed by the magnets, should be bypassed back to the mill or even rejected from the circuit.

631

Feeding the Crushers A simple arrangement for a single pebble crusher is provided (Flow sheet 2 shows a simple pebble crusher arrangement). This simple arrangement is a low cost method of installing a pebble crusher but incorporates the most critical components for safe and reliable crushing. These components include: 1) Metal removal by magnets. 2) Metal detection for rejects. 3) Diversion chute. 4) Variable speed feed conveyor to pebble crusher. 5) Bypass conveyor for surplus material, rejects and crusher downtime.

Metal Detector

Metal Reject5

Pebble Cone Crusher

New Feed

Magnet

5AG Mill To Downstrear Processes Metal Rejects

Flow sheet 2 Simple pebble crusher arrangement In this simple arrangement, the crusher should be of sufficient capacity to limit the quantity of excess material onto the bypass conveyor. In order to achieve the target of a fully choke fed crusher, crushing all pebbles at the highest possible reduction ratio is preferable to continuous bypass. A single large crusher is preferable over several smaller crushers due to higher availability and lower maintenance costs. When two or more crushers are required, then a bin and feeders should be added to facilitate proper feeding of multiple crushers. Special care must be taken in designing the bin and feeder layout to prevent size segregation within the bin. Size segregation compromises the performance of cone crushers and can reduce both the mechanical and process efficiency of the crushers.

632

Stability of continuous pebble crushing operation is required to prevent surging of the downstream processes. The crusher supplier prefers the crusher to be fully choke fed at all times, however, variable production of pebbles coupled with the critical need for downstream process stability typically will not allow a crusher to be consistently choke fed in most cases. One solution to both process stability and crusher optimization is a stockpile for the extracted pebbles (Flow sheet 3 shows an ideal pebble crushing arrangement). A large stockpile will allow one or more crushers to operate continuously at full capacity. While a stockpile is more costly, the investment cost in a stockpile can be recovered quickly by the improved performance of the overall comminution circuit.

New Feed

SAG Mill

Stockpile

To Downstream Processes

Pebble Cone Crushers

Flow sheet 3 Ideal pebble crushing arrangement Two-Stage Crushing Various forms of two-stage crushing have been applied in pebble crushing, specifically: 1) Cone crusher followed by a vertical impact crusher. 2) Cone crusher followed by a WaterFlush cone crusher. 3) Cone crusher followed by a high pressure roll.

Common objectives are either to produce a greater amount of material which can be fed directly to the ball mill versus return to the primary grinding mill; or to increase the overall fines generation in the pebble crushing process prior to returning the crushed material to the primary mill.

633

Two-stage crushing typically includes additional screening and material handling. Also, because of more unit processes, the full operational availability can not match the more conventional one-stage process. The success of new high performance cone crushers has made two-stage crushing a less competitive option versus one-stage crushing. Vertical impact crushers are generally of the autogenous impact design (Figure 3 shows a vertical impact crusher). Vertical impact crushers rely on both acceleration and mass to generate force for material fragmentation. As a second stage of crushing, the impact crusher's performance is impaired by the fineness of the incoming feed. An impact crusher requires frequent maintenance. Tramp material can result in severe damage to the internals of an impact crusher.

Figure 3 Vertical impact crusher WaterFlush (Figure 4 shows a WaterFlush cone crusher) cone crushers have the advantage of the proven durability of a cone crusher in the pebble crushing environment. The addition of water permits wet screening after the crusher and the WaterFlush crushers coupled with wet screens in a closed circuit are able to generate a screen undersize product that can be fed directly to the ball mills. Full success in WaterFlush crushing requires control of both the water addition and the crusher feed. Pumping of the crusher discharge should be avoided. Extensive pumping of the screen undersize should also be avoided.

Figure 4 WaterFlush cone crusher

634

The high pressure roll (Figure 5 shows a high pressure roll) has been applied successfully as a second stage of crushing following a 7 ft SYMONS cone crusher (Dowling, Korpi, McIvor and Rose. 2001). The high pressure roll process requires a continuous full feed to achieve desired performance. At Empire Mine, the SYMONS cone crusher delivers sufficient top size reduction to provide an acceptable feed gradation for the high pressure roll. Coarse feed will accelerate the wear rates of the high pressure roll surfaces, therefore adequate feed preparation is critical. The high pressure roll surfaces are composed of wear materials which are not forgiving of excessive tramp material. Extra care must be taken to ensure tramp material larger than the operational gap of the rolls is rejected.

Figure 5 High pressure roll Two-stage crushing has yet to achieve popular acceptance. Cone crushers, as a single unit operation, remain the preferred option for most pebble crushing applications. Reliability, simplicity, and tolerance to variability favor cone crushers, but alternative technologies focused on maximizing fines generation have proven successful in maximizing milling efficiency.

CONCLUSIONS The methods for dealing with pebbles continue to expand with both new processes and flow sheets. Cone crushers continue to be applied with indisputable acceptance. The ability to fragment rapidly and efficiently hard competent materials under adverse conditions is key to success in pebble crushing applications. After the primary mill, pebbles are to be freed of moisture and tramp materials. Vibrating screens, belt magnets, bypass systems, and feed control are recommended to maximize efficiency and mechanical reliability. Pre-crushing has proven effective in addressing the effect of hard competent ores on primary mill performance. Two-stage crushing has yet to achieve acceptance, but some success has been achieved with new technologies coupled with cone crushers.

REFERENCES Sylvestre, Y., Abols, A., and Barratt, D. 2001. The benefits of pre-crushing at the Inmet Troilus Mine. International Autogenous and Semi-Autogenous Grinding Technology. Volume 111: 43-62. Dowling, E.C., Korpi, P.A., McIvor, R.E., and Rose, D.J. 2001. Application of high pressure rolls in an autogenous-pebble milling circuit. International Autogenous and Semi-Autogenous Grinding Technology. Volume 111: 194-201. Needham, T.M. and Folland, G.V. 1994. Comminution Circuit Expansion at Kidston Gold Mine. Proceedings of SME Annual Meeting.

635

SELECTION AND SIZING OF HIGH PRESSURE GRINDING ROLLS Rene Klymowsky, Norbert Patzelt, Johann Knecht and Egbert Burchardt

'

ABSTRACT High Pressure Grinding Rolls are well established in the cement industry for the grinding of clinker, limestone, slag and other relatively non-abrasive materials. Minerals are however some 20-100 times as abrasive as cement raw materials, thus acceptance by the minerals industry has required the development of special wear protection surfaces and rapid change-out procedures for the rolls. The range of application of HPGRs extends from the grinding of coarse ores < 75 mm, to the grinding of fine concentrates < 100 pm with moisture contents up to 12%. Machines are available with capacities up to 3000 t/h, and with installed power up to 6000 kW. This paper gives a description of the new technology, design aspects of the machine, key parameters of the process, test procedures, scale-up and sizing considerations, together with examples from existing and potential future applications. 1. HISTORY The HPGR is an outcome of fundamental studies of fracture physics conducted on single particles and multiple layers of particles in a packed-bed by Prof. K. Schonert (1979-80). These showed that significantly lower specific energies would be required for comminution if the material were first treated in a HPGR, and opened the possibility for plant capacity expansion without the necessity of making major alterations. The first HPGRs were installed in the cement industry for the -grinding of clinker and raw materials (1985-86). This was followed by the installation of the first HPGRs-in the diamond industry (1987) - rolls 2.8 m in diameter x 0.5 m wide. Sales proceeded rapidly from that point 35 amounting to a total of over 500 units world-wide to date. The appeal to the cement industry was the reduction in costs achieved through energy savings, increases in throughput, low space requirements, and long service life of the roll units. Improved liberation and product recoveries was the appeal in the diamond mineral industry. Some early 0 attempts were made to apply HPGRs to 1985 1990 1985 m 2010 other minerals, but these failed largely YewdhsQlaticn because of excessive wear of the roll surfaces. The next development was the Figure 1 Growth in diamonds and iron ore introduction of studs into the rolls. The first sale of studded rolls was to the iron ore industry (1 994) for the grinding of pellet feed. This was soon followed by further sales into the iron ore industry for coarse ore crushing (1997). Growth of HPGRs in the diamond and iron ore industries is shown above. 1 Polysius AG, Beckum, Germany

636

The first application of a HPGR in hard rock copper (UCS > 300 Mpa) was at Cyprus Sienita (1995) - the world’s largest unit, 2.4 m in diameter x 1.4 m wide. This unit processed over 7 million tonnes of ore before it was de-commissioned in 1996. The unit was designed to test different wear protection surfaces - from NiHard liners to various grades of tungsten carbide studs. Lessons learnt from this operation have allowed considerable improvements to be made to the design of the machine, roll surface wear protection, as well as in plant layout to increase the availability of HPGRs to levels exceeding 92% in high wear situations. BRIEF DESCRIPTION OF THE UNIT 2. The HPGR consists of two counter-rotating rolls mounted in heavy-duty frictionless bearings, enclosed in a strong frame. Pressure is applied to one of the rolls by means of a hydro-pneumatic spring system, while the other roll is held in a fixed position in the frame. The ‘‘free” or “floating” roll is allowed to slide (or float) on frictionless pads, reacting to the forces acting on the roll by the material and the spring system. Feed to the rolls is provided by means of a hopper mounted above the rolls equipped with level control to ensure that the rolls are continuously choke-fed. Normally freeflow from the hopper is sufficient to exert a separating force on the rolls. Special attention to the design of this hopper is required in applications where the feed is fine and moist. The rolls are driven by separate motors connected to the roll shafts through gear reducers. The rolls can be operated at fixed speed or at variable speed depending on the Figure 2 A cutaway view of a high pressure rolls demands of the process. A torque reaction system is included to prevent the gearboxes from turning and to divert any differential forces away from the frame. The rolls can be solid, or can be made up of special wear resistant segments or tyres mounted on the roll shafts. The surfaces can also be protected by layers of hard metal welded onto the surfaces. In most mineral applications, the surfaces are protected by implanting tungsten carbide studs. These help to form an autogenous wear layer on the rolls, and improve the drawing of the material into the rolls.

Roll diameters of industrial and semi-industrial units vary from 0.8 - 2.8 m. The forces applied range from 2,000 - 20,000 kN. Pressures in the gap between the rolls range from 80 - 300 Mpa. Most ores and minerals have compressive strengths between 50 and 280 Mpa. Capacities range from 50 to up to 3000 t/h. Energy consumption is between 1 and 3 kWh/t. The machines are compact, noise and vibration free, and thus require smaller foundations than conventional crushing and grinding equipment. Any dust created in the process can reazly be controlled.

637

3. MECHANICAL ASPECTS In the 20 years of history of HPGRs, the design of the unit has been steadily optimised, and the mechanical durability of the major components such as bearings, gear reducers and the drive train has advanced to a such degree that they offer the highest availability. Main Components of HPGRs 0

0

0 0

0

0

2 Roll units each consisting of o 2 bearing blocks and bearings o 1 shaft o 1 wear protection Hydraulic system o hydraulic aggregate o hydraulic rams o nitrogen accumulators Machine frame 2 Drive trains each consisting of o 1 gear box o 1 cardan shaft or v-belt pulley o 1 safty coupling o 1 main drive motor Feed arrangement consisting of o external feed chute I hopper o internal material guide plates Lucrication system@)

Figure 3.1 High-pressure grinding roll with main coml onents The chief factor which determines the operating costs and above all the availability of a HPGR is the wear protection. Several types of wear protection have been developed to meet the various requirements of cement and mineral applications. A breakthrough in wear protection technology for mineral applications was the adoption and development of studded liners that can offer wear lives of >5000 hours even for very hard and abrasive ores.

L/D ratio For many years there has been an on-going debate as to whether it is more advantageous to design rolls with smaller diameters and larger widths (high L/D-ratio), or to design larger and narrow rolls (low L/D-ratio). The decision as to which approach to adopt is capital. It has an impact not only on the performance of the HPGR but also a major impact on the design of individual components, and on the general layout of the unit. The minimum roll diameter is prescribed by the outside diameter of the bearings and the thickness of the bearing block. The bearings themselves are sized according to the installed grinding force. In order to design HPGRs with high L/D-ratios, it is necessary to select bearings with the smallest outside diameter, i.e. cylindrical roller bearings.

Gear box to st!aft interface

Bearing

Bearing block ~

Figure 3.2 Roll unit with gear box

638

The size of the bearings also determines the shaft diameter and pre-determines the manner in which the gear box and shaft are to be connected. If the power to be transmitted is large, the gear boxes have to be located on opposite sides, or they will touch if they were located on one side. Larger rolls with low LID-ratios offer greater freedom in selecting the most appropriate bearings. The larger roll diameters make the connection between the shaft and gear box simpler to execute, and allow large gear boxes to be located on one side to save space and facilitate maintenance. Wear Protection Systems A wear protection system is characterised by the roll design and the type of wear protection surface applied. A wide range of wear protection systems has been developed to match the abrasiveness of various feed materials, lifetime expectations and local conditions. Roll Design Three different roll designs have been successfully applied: solid rolls, rolls with tyres and rolls with segmented liners. Criteria for selecting the optimum roll design are: 0 0

0

the balance between operating and investment costs the acceptable lifetime and frequency of replacement the tolerable downtime for liner replacement

Solid rolls are made as compound castings or forgings. Forgings require an additional surface protection, e.g., hard facing. The solid roll design has been applied mainly in the cement industry for the grinding of hot clinker. The wear rates in these applications are up to 100 times lower than in mineral applications. The long wear life, low investment cost and possibilities of refurbishment make this choice attractive for this industry. However, solid rolls have not found their place in the Mining Industry as the service life would be too short for most of the ores treated. Tyres have been successfilly installed in all industries. Compound, bainite and Ni-hard castings as well as forgings are used as tyres. The largest diameter tyres in service are 2.8 m, and the largest widths are 1.8 m (but not on the same machine). Tyres for rolls of 2.8 m diameter and 1.6 m length are considered to be mechanically feasible. Segments have been installed in nearly all industries but with mixed success. Practice has shown that they are only applicable in low pressure applications. They failed, for example, in the Cement Industry where high grinding pressures are required. Nevertheless, segmented roll liners performed quite well in several other applications such as grinding of chromite and iron ore concentrates, as well as in the grinding of diamond ores.

Figure 3.3 Clamped segments

Materials of construction. Steel and Ni-hard castings are used for segmented roll liners. The largest segments built to-date have been 2.25 m in diameter by 1.4 m wide. Changeout times. In general, both tyres and segmented roll liners could be applied in the Mining Industry. The major arguments in favour of tyres are lower maintenance and longer life-times. Segmented liners offer shorter downtimes for liner replacement, 1 to 2 days, and easier handling of the segments.

639

The downtime for tyre change out depends on several factors, such as, machine design, machine layout and plant infrastructure. It is possible to shorten the time for change-out of the roll units to less than 2 days, if complete spare roll units are available. After installation of roll units with new tyres, the worn out tyres can be replaced at a workshop on-site or elsewhere. The various advantages of the tyre design are listed on the right.

Wear Protection of roll surfaces The different roll design options can be combined with various types of wear surface protection. Forgings must be protected with either a hard facing, hard metal tiles, or studs. Hard or compound castings do not need any further surface protection. The surface itself, may be smooth, profiled (Figure 3.4), grooved or studded (Figure 3.5). Thus a comprehensive matrix of different wear protection options is available for various applications.

Figure 3.4 Welded on profile

1

sono of sasments *higherinvesbnentast joints betweensegnents require mxe maintenawe due t o m out

Wear Protection Surfaces Base material Surface material

Surface type

forging

0

0

hard facing

0

hard casting (Bainite, Nihard IV)

smooth welded-on profiles

0

hard metal studs

autogenous

0

hard metal tiles

grooved

not required

smooth welded-on profiles 0 grooved

0

0

Figure 3.5 Studded (autogenous) roll surface

Profiles are applied in order to improve the nip-in characteristics and to increase the throughput rates of the rolls. Concurrently, the wear rate is reduced because of lower slip and extrusion on the roll surface compared to smooth rolls. These profiles are applied mainly for clinker and limestone grinding but are also used for diamond ores. They need to be renewed frequently for abrasive ores.

Lifetimes of hard faced rolls and hard castings. Typical lifetimes of hard faced rolls in cement applications are more than one year, but only 6 to 12 weeks in operations grinding kimberlite. Segmented Ni-hard liners, used for the grinding of lamproite, have lifetimes of 8 to 16 weeks depending on the thickness of the segments. Compound cast tyres for the grinding of limestone have lasted longer than 40,000 hours.

640

Lifetimes of studded rolls. The best choice for abrasive ores is a studded (autogenous) roll surface. The studs are made of tungsten carbide and are extremely wear resistant. The softer base material is protected by the autogenous material layer built up between the studs, so that both wear at about the same rate. This form of surface wear protection provides the highest throughput rates and the longest lifetimes. The properties of studs differ depending on chemical composition, grain size, hardness, toughness, etc. Stud qualities need to be selected to confer sufficient wear resistance with a minimum of stud breakage. Generally, the wear resistance is a function of the hardness of the studs. However, increasing hardness may lead to unacceptable levels of stud breakage. The lifetimes that can be achieved with studs varies with the material treated. On diamond ores, lifetimes of more than 6,000 hours has been obtained on studded rolls, with expectations of achieving more than 10,000 hours. Lifetimes of more than 10,000 hours have been achieved on bulk iron ores, and more than 20,000 hours on iron ore concentrates. Operating results on a very hard and abrasive copper ore indicated lifetimes of 4,000 to 5,000 hours with the selection of the appropriate studs. Bearings The roll shafts may be mounted in either self-aligning bearings or cylindrical roller bearings. Selfaligning roller bearings are grease lubricated. If required, cooling water is applied. Cylindrical roller bearings need oil lubrication, which is also used for cooling. Furthermore, they need grease lubrication for the bearing seals. The quantities of grease used are similar in both cases. Cylindrical roller bearings have a smaller outside diameter and allow the installation of smaller diameter rolls for a given grinding force. In principle, both type of bearings are fit for purpose. Nevertheless, the smaller bore diameter of the cylindrical roller bearings limits the self aligning cylindrical thickness of the shaft, resulting in higher Figure 3.6 Bearing design of HPGRs deflection and lower safety factors on the shaft. Figure 3.7 shows the calculated L10 vs the actual lifetime achieved with selfaligning roller bearings for different sizes of HPGR. The actual lifetime achieved with these bearings is in most cases significantly higher than the calculated lifetime. Normally the lifetime is based on the installed grinding force, which may not always be applied in actual operation. Practically all bearings achieved more than 50,000 operating hours.

100.000

.-,80.000

c,

.-cE g

60.000 40.000

J

20.000 D

0

1

2

3

4 5 Machine type

6

7

a

g

Figure 3.7 Lifetimes of self-aligning roller bearings

64 1

Hydraulic System The grinding force is introduced into the material bed via the floating roll. It is generated by a hydropneumatic spring system, consisting of one or two nitrogen accumulators, and one or two hydraulic cylinders with piston rods on each side of the roll. Nitrogen accumulators may be piston or bladder type accumulators. The hydraulic aggregate consists of a tank, pump, filters, temperature elements, valves and switches in a single unit controlled from a separate, or from a plant control system. The hydraulic pressure is adjustable during the running of the machine.

Hydraulic cylinder

accumulator Nitrogen

I

Hydraulic oil

Initial nitrogen and hydraulic pressures are pre-set before starting the HPGR. As Opt1 Opt2 Opt3 Opt4 Diameter of hydraulicpiston [dm] 4 4 4 4 material is fed into the unit, the floating Volume of operating accumulator (total) [ I] 25 25 50 25 roll is pushed back opening the gap. Volume of s a w accumulator (total) 5o initial gap mm 1 I0 I0 I0 I0 When the hydraulic system is closed, the Initial pressure in operaangaccumulator [bar ] 30 55 55 35 Initial pressure in safety accumulator [bar 1 0 0 0 120 nitrogen is compressed and the pressure Initial hydraulicpressure [bar] 70 70 70 70 increases according to the change in the volume of the nitrogen accumulator. IStiffness of hydro-pneumaticspring system1 180 The stiffness of the hydro-pneumatic 170 spring is adjustable, and a function of the 160 volume of the nitrogen accumulator, the 150 .E 140 pressure pre-settings as well as the 5 130 number and diameter of the cylinders. D 120 ep 110 The second nitrogen accumulator is -.2 100 mainly used as a safety accumulator, and E 9 0 is usually larger than the first, to limit the x80 70 pressure increase after the pre-set 60 pressure in the second accumulator has 50 10 20 30 40 50 6o 0 been exceeded. Figure 3.8 shows a range W,,kinn in spring stiffhesses used in various operations with one and two nitrogen Figure 3.8 Spring characteristic of hydraulic system accumulators. Roll diameter Roll width

Data of

1400 mm 800 mm

a.m.

system

...m

Machine frame

The hydraulic cylinders as well as the roll units are held in position in the machine frame. The frame design can affect the change-out time of the roll units and crane capacity required. This is especially important in cases where wear protection tyres are installed and need to be replaced frequently. Different designs are available:

e a

The “completely closed frame” requires the complete removal of the upper frame in order to lift out the roll units. The “one end open frame” needs only one end of the upper frame to be lifted at the nonhydraulic side to provide sufficient clearance to enable the roll units to be pull out. The “hinged frame” avoids a bolted connection between the frame and the floating roll unit. The rolls may be pulled out either at one end, or at both ends depending on the actual design. Elimination of the bolted connection reduces downtime for roll unit change out. However, the savings in downtime have to be justified by a higher investment cost.

642

Figure 3.9 One end open frame

Figure 3.10 Hinged frame design

Drive train Differences in design details of the drive train have an impact on mechanical reliability and maintenance of HPGRs. ~~

~~~~

~~

~

Gear boxes and shafts Drive trains with Planetary type reducers are used for A) Shaft mounted gear box HPGRs. These can be floor-mounted gear box is bolted to or shrink fitted onto shaft or attached directly to the roll shafts. gear box and motor are connected with a v-belt pulley or a low torque cardan shaft Each roll has a separate drive. Only an overload coupling is mounted few smaller size HPGRs have been between gear box and cardan shaft on the high speed side equipped with a single drive B) Floor mounted gear box Floor-mounted reducers require Shaft and gear box are connected high torque couplings or cardan shafts with a high torque shaft (tooth gear coupling) as a connection between the gear gear box and motor are connected reducer and the roll shaft. This is with a flexible coupling working as a safty coupling usually done when the gear reducers are too large for mounting directly on the roll shafts. Failures of high torque couplings have resulted on floating rolls as a result of the additional stress imposed by the angle of displacement. Shaft mounted reducers may be s h n k fitted or bolted onto the shaft. Shrink fitting involves the lowest cost and allows the attachment of fairly large gear boxes to small diameter shafts. Shafts with small diameters may not have sufficient space to accommodate the bolts required to transmit the torque : . ;s safely. Then shrink fitting must be employed. The shrink fit may run into severe problems if during operation the reducer seizes on the shaft. Then the shaft may have to be cut to disconnect the gear reducers from the roll unit. Bolting is the safest and fastest method of attaching gear reducers to the shafts. This type of attachment may ------ I be done directly to the shaft, or through an intermediate flange, and eliminates Figure 3.1 1 Bolted gear box shaft connection any risk of damage to the components.

1

rn-11-A

643

A:--

4. PROCESS DESCRIPTION Comminution takes place in high-pressure grinding rolls either exclusive as interparticle comminution in a material bed or as interparticle comminution superimposed by single particle comminution (Figure 4.1). Interparticle comminution occurs when the maximum particle size in the feed is smaller than the working gap between the rolls. Single-particle comminution occurs when the maximum feed size is larger than the working gap. In this case, larger particles are nipped directly by the rolls (nip-angle asp)and are pre-broken before entering the compression zone. The compression zone is defined by the nip-angle qP.Frequently the product is the form of a compacted cake.

Bulk inaterial

Compression zone

a,, : nip angle (interparticle)

qp: nip angle (single particle)

Figure 4.1 Compression zones in operating gap In the compression zone the feed material is compacted from a bulk density (yf) to a cake density (SJ. The cake density is typically in the range from 70% (for fine materials with high moisture) to 85 % of the real density (for coarse materials). The nip-angles for interparticle and single particle comminution are calculated according to the following equations: 1.1 1.2

(xip

asp

1 1

asp

[ [

s 6, x,,

[mm] [Urn’] [mml

aip

: : : : :

= =

arccos (1 - ( 6, / yf-l)* s / (1000 * D)) arccos (1 - (xmx/s- 1)* s / (1000 * D)) nip-angle for interparticle comminution nip-angle for single particle comminution cake thickness D density of compacted cake Yf maximum particle size

[m] : [tw :

roll diameter feed bulk density

Average s/D ratios for studded rolls are between 2-3% of the roll diameter. (Moisture and very narrow sized feed can reduce these ratios.) From the above, the largest particle nipped between the rolls in the compression zone would be about 1.5 times the gap. Larger particles would cause the rolls to separate and the compression zone to collapse, reducing grinding efficiency.

644

Compaction of the feed material in the compression zone results in a pressure increase towards the narrowest gap between the rolls (Figure 4.2). At that point, the maximum pressure is reached. After passing the gap, the pressure in the material bed is released and the cake expands slightly.

1 Compression angle

Figure 4.2 Pressure profile in compression zone

(( IP

JAngleof relaxation

A pressure profile also exists along the width of the rolls (Figure 4.3). The roll width may be divided into a centre and edge zone. The material in the roll centre is exposed to the highest pressure and is fully pressed (centre material). The pressure profile fades out towards the roll edges. The material in this zone is either crushed (real edge material) or passes through the gap at a higher speed than the roll speed without any significant comminution (internal bypass). An additional external bypass may occur if the cheek plates are worn or rockboxes are used in the place of cheek plates. The shape of the pressure profile depends very much on the roll width and the tightness of fit of the cheek plates. Rolls with very narrow widths exhibit very peaked profiles.

Center zone Roll width

Figure 4.3 Comminution zones along the roll width

5. KEY PARAMETERS The objectives in sizing HPGRs are to meet the throughput requirements and to achieve the desired product fineness. The throughput of a HPGR is mainly a function of the roll dimensions, type of roll surface used, and the feed material properties. For a given material and roll dimensions, the throughput is controlled by the roll speed. The product fineness is controlled by the grinding force applied to the material bed between the rolls. The grinding force creates the pressure in the material bed which causes micro-cracks and breakage of the particles. The correlation between particle breakage and the grinding force required needs to be determined for each material. The key parameters are therefore the specific throughput rate and the specific press force to be applied to obtain the desired comminution result. Definition of these terms will be given later in the section. The main drive motors are then sized to provide the torque required to turn the rolls. The drive power is a function of the grinding force applied.

645

Throughput Calculation The continuity equation forms the basis for the throughput calculation of a HPGR:

M

(1) M s 6

L*s*urn*6*3.6

=

[tph] : [mm] : [t/m3] :

throughput working gap material density in the gap

L u,

[m] : [ds] :

roll width

material velocity in the gap

The throughput of a HPGR is given by the volume flow through the operating gap between the rolls (L * s * urn)times the average density of the discharge material. The discharge consists of pressed, broken and bypassed material. Therefore the average density of the material passing the gap is the weighted average density of the various constituents. The material speed is normally set equal to the roll speed in the throughput calculations. However, exceptions may occur. In some cases the feed material may go through the gap at a speed higher than the roll speed - it may be squeezed through the gap (extrusion), fall through the gap and be accelerated by the rolls (internal bypass), or fall outside of the rolls (external bypass) without being nipped by the rolls. In other cases the material speed may be lower than the roll speed if the material slips on the rolls (slippage). These cases can be identified by testing. For scale-up it is assumed, that for a given material and operating conditions, the working gap (s) of HPGRs scales linearly with the roll diameter (D). This assumption has been proven in practice for rolls of different size.

A specific throughput rate, mc , can be defined which is proportional to the gap-to-diameter ratio, s/D:

.

mc

(2) mc 6c

D

[t*s/(m3*h)] : [t/m3] :

[ml

rmml

S

(s/D)

=

*

*

6,

3.6

specific throughput (calculated from the cake) density of the pressed material roll diameter working gaD

The continuity equation can then be re-written as follows: (3) M D u

M [tphl [m]

[ds]

=

m* D throughput roll diameter roll speed

: : :

*

L

*

u m

[t*s/m3*h)] [mml

L

:

specific throughput roll width

The specific throughput rate, m , depends on the particular feed material and operating conditions. It can be determined by test work from operating results: (3.1)

mf

=

M / ( D * L * u )

The specific throughput rate calculated from the cake (2) and from the feed rate (3.1) may / m;’ provides information on the material behaviour in the differ in many cases. The ratio of ‘‘m working gap. A “m, I m;’ - ratio of lower than 1 indicates extrusion in the compression zone or internal and/or external bypass. In both cases, the material goes through the gap at a speed higher than the roll speed. A “m, / m;’ - ratio higher than 1 indicates that the whole width of the rolls may not be utilised. The material flow at the roll edges may be restricted. Also, a ‘‘m / m;’ - ratio higher than 1 may indicate slippage.

646

The expected working gap can be calculated for a given roll diameter from the specific throughput as follows: =

(2.1) s

. .

(h * D) / (6, * 3.6) * (m ,/ m f) f

(A * D) / (6, * 3.6) * c

=

The factor c is calculated from the ratio of the specific throughput rates coarse ore applications, the factor c is between 0.85 and 1.

‘‘m/ mi’.

In most

Calculation of Grinding Pressures The grinding pressure acting on the material bed controls the product fineness. However, it cannot be measured directly. Various parameters are available for quantifying the grinding pressure in the material bed. One such parameter is the specific grinding force cp. The specific grinding force is the grinding force divided by the projected area of the rolls:

(4) cp

D

cp

=

[N/mm2] [ml

F/(1000*L*D) :

[kN] [m]

F L

specific grinding force roll diameter

: :

grinding force roll width

The specific grinding force is particularly suitable for establishing correlations between the grinding pressure in the material bed and the achievable product fineness, and for comparing grinding forces between HPGRs of different size. The average pressure paveacting on a material bed is defined as the grinding force divided by the area of compression on the rolls: (5) Pave

pave

=

F/(1000

[MPaI

1

clip

[

F D

[kNI [ml

:

*

L

*

D/2

*

a,,)

=

2

*

cp /

(xi,

average pressure in material bed nip-angle for interparticle comminution (refer to formula 2.1) grinding force L [m] : roll width roll diameter

The maximum grinding pressure pmaxin the material bed between the rolls was defined by Prof. Schonert as follows:

(6) Pmax

k

pmx

=

[MPaI : [ I

F/(1000

* k * D * L * ai,)

=

q/(k*aip)

5

*

9/qp

maximum pressure in material bed material constant (0.18 - 0.23)

In practice the material constant k is quite difficult to determine. Studies done on smooth rolls have shown that the peak pressure is some 40-60 times the specific press force, at 2 - 6 N/mmz. Studded rolls have larger nip angles, therefore the maximum pressures are lower at same specific grinding forces than for smooth rolls. The pmx - formula is mainly used in the cement industry. In the minerals industry, the application of the specific grinding force has proven to be sufficient for quantifying grinding pressures.

647

Power The motor power (P) required to drive the rolls is proportional to the applied grinding force (F). The action of the grinding force (F) on one of the rolls is shown in figure 5.4.The point where the force is acting on the rolls is determined by the force acting angle fl. The grinding force may be resolved into a radial component and a tangential component (Ft). The tangential component gives rise to the torque which has to be provided by the main drive motors to turn the rolls. The motor power required is calculated for a given roll speed according to equation (7):

(7)

-

PR

PR T F

P

[kW] : [m] : [kN] : ["I

o*T

=

I I

Figure 5.1 2 *z

Force acting angle p

Action of grinding force on roll

* 11/60 * Dl2 * sinp * F

motor power per roll rolltorque grinding force force action angle

[lk] : [rpm] : [m] :

w

n D

angular roll speed roll speed roll diameter

The total motor power P is then:

(8)

-

p

P D F

[kW] : [m] : [kN] :

2*PR

=

z

* n / 30 * D * sin p * F PR

Totalmotorpower roll diameter grinding force

[kW] : [rpml [I :

n

P

motor power per roll roll speed force acting angle

The specific energy wSpabsorbed by the feed material can be calculated according to the following equation which is derived from equations 3,4and 8: (9)

wSp

=

P/M

=

2000* sinp /

r;l *

cp

Equation (9) indicates a linear correlation between specific energy and specific grinding force. often vary with the However, the force action angle p and the specific throughput rate specific grinding force. Thus one can obtain non-linear relationships. From equation (9), high specific throughput rates lead to less energy consumption at a given grinding force.

648

6. TESTING The main objectives of material testing are to determine: 0 the general suitability of the ore to high pressure grinding 0 the key parameters required for sizing (specific throughput, specific grinding force and resulting energy input) 0 the achievable product size distribution, and 0 the abrasiveness of the ore I Test Data Recording and Analysis The test programme should include 1) Ore Characteristics 2) Machine Parameters investigation of the effect of: specific gravity of the ore speed of the rolls 0 grinding pressure total press force 0 bulkdensity (at least 3 pressures) feed size distribution 0 netpower 0 moisture 0 oil and nitrogen pressures 0 moisture (if applicable) cakedensity zero gap and operating gap 0 feed size distribution 3) Performance Requirements for data recording as well as 0 specific energy consumption performance and material analysis are 0 throughput product size distributions (edge, center, total discharge) summarised in the table on the right. 0 proportion of center to edge product obtained A mineralogical analysis of the ore is usually mostbeneficial; and should be included. Grinding tests on the feed and HPGR products to determine the overall energy reduction are also recommended. For coarse products, these tests are best camed out in a mill large enough to handle the products without any further size reduction.

Laboratory High Pressure Grinding Rolls These rolls may be used for preliminary scouting tests, when limited samples are available, or for tests on fine materials such as iron ore concentrates. Process data obtained from these tests allow preliminary sizing of an industrial unit. Top feed size : < I 2 mm : 0.25 or 0.30 m Roll diameter 0.10 or 0.07 m Roll width Roll speed 0.20 to 0.90 m l s -Wearprotection : tyres Wearsurface : smooth, profiles, studs Sample requirements 30 kg per batch test run

-

LABWAL

Semi-industrial (Pilot scale) High Pressure Grinding Rolls A number of mobile semi industrial (pilot scale) units are available from different suppliers for testing in the laboratory or field. Most of them are equipped with studs. Specifications for these units are given below: Top feed size Batch testing : <45 mm Continuous < 30 to 35 mm Roll diameter : Roll width Roll speed Throughput : Sample requirements

0.7 to 0.9 m 0.2 to 0.3 m 0.3 to 1.2 m l s 30 to 90 tph 70-1 50 kg (per batch test)

649

Wear Testing The abrasiveness of ores varies widely with the physical properties of the material and operating conditions. Several abrasion tests are available in the minerals industry. However none are based on high pressure as the principle for comminution, and therefore cannot be used for reliably predicting wear rates in a HPGR. Inclusion of this principle in the test procedure is a precondition for determining the abrasiveness in a HPGR. The picture at the right shows an ATWAL abrasion testing high pressure grinding unit. The rolls are 100 mm in diameter by 30 mm wide. The unit can be equipped with tyres made of different wear materials. The tyres are weighed before and after a test run. The wear rate in glt is calculated from the weight loss divided by the amount of material treated. Typically 100 kg is sufficient for a test run. 7.

.

PERFORMANCE

ATWAL

Factors influencing Throughput Machine related

7.1 Factors Influencing Throughput The throughput depends on several factors: the ore, the roll surface and operating conditions, as well as to the feed conditions. A summary of the factors influencing throughput is given in the table on the right. The impact of the individual factors varies with the type of feed material (bulk ores, scrubbed kimberlites, fine and moist concentrates). Some materials are quite sensitive to even small variations in these parameters.

m roll surface

grinding force roll speed m L/D ratio Ore related

oredensity feed moisture feed size distribution feed size m compressive strength

Feeding conditions

m filling level in feed hopper

flowability in feed hopper segregation of feed amount of bypass material

Roll surface The roll surface has of major impact on the throughput (Figure 7.1). Studded rolls yield higher specific throughput rates - 50 to 100 % higher - than either grooved or smooth rolls. Furthermore studded rolls are less sensitive to higher moisture, grinding force and roll speed. 350

350

300

300

250

250

200

200

150

150

100 Gold ore

0

2

4

6

8

10

50

12

0

Specific grinding force [Nlmml]

Figure 7.1 Spec.throughput vs roll surface

1

2

3

4

5

6

7

Specific grinding farce [N/mm7

Figure 7.2 Spec.throughput vs grinding force

Specific grinding force has only a limited effect on throughput for most types of material except for moist, fine feed, such as iron ore concentrates (Figure 7.2).

650

Next Page

The roll speed has a moderate influence on the specific throughput as shown in Figure 7.3. The absolute throughput can be seen to increase linearly with increasing roll speed over a wide range of relative roll speeds, even though the specific throughput decreases at the same time. However the rate of increase is less than proportional to the increase in the roll speed. The L/D ratio can affect the specific throughput by 510%. Larger diameter rolls with low to moderate L/D ratios provide more favourable draw-in conditions for coarse feed.

Figure 7.3 Throughput vs relative roll speed

Ore characteristics Ore density has a significant effect on specific throughput because the HPGR behaves mainly as a volumetric device. Higher material density results in higher specific throughput rates. The effect of the feed moisture varies considerably with the type of material and roll surface (Figure 7.4). In fine material, moisture affects the friction between particles to a greater extent than on coarse particles resulting in a lower separation force and in low specific throughput rates. Nevertheless, fine concentrates can be treated with high levels of moisture (8 to 12 %). However, there is critical level of moisture for each type of fine concentrate which should not be exceeded.

1

0

2

4

6

8

! 1

0

1

2

Moisture content [Oh H,O]

Figure 7.4 Spec.throughput vs feed moisture The feed size distribution has a major effect on the specific throughput rate of bulk ores. The more narrow the feed size distribution, the lower the specific throughput rate (Figure 7.5). Material with fewer fines to fill the voids between particles can be compacted to a greater degree, thus reducing the gap and throughput. 250

1W M

10

0

1

0

0.010

50

0,100

1.000

10,000

100.000

Particle size [mm]

~ _ ~ ~ _ _ _ _ F - T m m

0

r0-40 r--

mm

+6-40 mm

+12-40

+18-40

+0-18

mm

mm

rnm

+4-18 i

I R ~

t l 8 . 4 0 m m

-6-40mm -c-.O-l8mm

-. -12-40mm -.--. 4 - 1 8 -

j

,

.-

Figure 7.5 Specific throughput rates for different feed size distributions Hard ores with h g h compressive strengths have narrower feed size distributions due to the deficiency of fines and yield lower specific throughput rates. Also ores which have been prescreened to remove the fines (truncated feed) yield lower specific throughput rates. There may be benefits in removing the fines from greater compaction and more efficient energy utilisation, but these are often out-weighed by the resulting higher wear.

651

Crushing Plant Design and Layout Considerations Ken Boyd, Manager, Material Handling, AMEC Mining & Metals, Vancouver, BC

ABSTRACT In mining operations, the layout of crushing plants and ancillary equipment and structures is a crucial factor in meeting production requirements while keeping capital and operational costs to a minimum. This paper addresses the critical design parameters as well as the consideration of ore characteristics, geographical location, climatic conditions, expected operational life, expansion potential, safety, environment, and operability and maintainability. INTRODUCTION The fundamental goal for the design of a crushing plant is an installation that meets the required production requirements, operates at competitive cost, complies with today’s tough environmental regulations, and can be built at a reasonable price despite the rising costs of equipment, energy and construction labor. The following industry trends must be taken into account: Equipment suppliers are offering ever-larger primary crushers, with 1,800 mm (72 in) gyratories expected soon, as well as secondary and tertiary machines of up to 3,000 mm (120 in). Rising energy costs are causing owners to increase the integration of mine and mill design, so that they can identify ways of reducing overall electrical power consumption. Electronic control of crusher discharge opening and feed rate. With adjustment of a crusher’s discharge opening, as the production continues through an on-line coarse size analysis of the crushed product (digital image analyses). Dance, A. 2001) More attention is being paid to the impact on crushing circuit design caused by variations in ore characteristics, size distribution, moisture content, ore grade and climatic conditions. Operators have always dreamed of reducing the need for crushing equipment; when SAG mills were first introduced, it was hoped that they would eliminate secondary and tertiary circuits. As it turned out, designers are now adding secondary or pebble crushers to SAG circuits, on both greenfield and retrofit projects, to increase feed rate to the SAG mill. In other words, crushing plants, from primary to quaternary circuits, are here to stay. There are three main steps in designing a good crushing plant: process design, equipment selection, and layout. The first two are dictated by production requirements and design parameters, but the layout can reflect the input, preferences and operational experience of a number of parties. These can include the owner’s engineering staff, safety personnel, operations and maintenance personnel, equipment manufacturers, and the engineering consultant. Ideally, the consultant combines his knowledge and experience with an understanding of all parties’ needs, to provide a balanced, workable, safe and economic plant design.

669

DESIGN PARAMETERS The principal design parameters that drive crushing plant selection and configuration include: Production requirements Ore characteristics Project location Operational considerations Climatic conditions

Capital cost Safety and environment Life of mine/expansion plans Maintenance requirements

Each of these is addressed in the sections that follow.

Production Requirements The process design criteria define the project’s production requirements, and typically include those shown in Table 1. Table 1 Production requirements Process Description General Ore Characteristics General Maximum rock size in the feed Primary crushing Ore types, compressive strengths Fines crushing and abrasion indices Ore specific gravity Storage & reclaim Ore bulk density Ore moisture, wet season Ore moisture, dry season Angle of repose Angle of withdrawal Angle of surcharge

Operating Schedule Days per year Hours per day Nominal annual throughput Mining shifts per day Crushing plant shifts per day System availability and utilization

The flowsheet specifies the nominal design, peak production flow rate, and equipment sizing to handle those capacities. Manufacturers provide ratings for their equipment, preferably based on testwork and/or experience, so a project flowsheet specifies tonnage requirements and the equipment is selected to meet or exceed the capacities. Design criteria can be calculated from a simple spreadsheet as shown in Table 2. Mine haul-truck capacity is an important factor at primary crusher installations, because it is cost-effective to integrate truck cycle time at the crusher station with mine/shovel operations. If a primary crusher dump pocket is undersized and unable to handle the mine’s trucks, then operators must slowly meter the ore into the receiving hopper.

670

-

Table 2 Production Requirements Typical Example: 60” x 89” primary crusher & mill feed conveyor system Operating schedule calculation - for 3 x 8 hours per shift Days per year 365 Tonnes per year 32,850,000 Metric tonnes per day 90,000 TOTAL TIME AVAILABLE 8,760 Hours per year UNPLANNED DOWNTIME (Subract planned or known downtimes) 0 Industrial Electrical - grid 24 Weather 48 Holidays 0 Major scheduled maintenance 416 lx 8 hr maintenance shift /wk Crusher maintenance 78 1 concave change box 1 , 2 months changes 24 x 2 Minor scheduled maintenance 0 Shift changes 183 10 minuteslshift Total lost time 749 PRODUCTION TIME 8,011 Hours per year (Time system is available) System availability percentage 91 Production time/total time UNPLANNED DOWNTIME (Subtract unplanned downtimes) No ore 40 1 5% of production time (8 hrs/wk no trucks delivering ore or other reasons) Crusher plug 160 2% of production time Chute plug 200 2.5% of production time Stockpile full 80 1% of production time Safety switch 200 2.5% of production time Metal on belt 52 Approx. 1 hrlwk Belt repair 240 3% of production time Electrical 200 2.5% of production time Mechanical Repair 200 2.5% of production time Others 0 Subtotal Unplanned Downtime Hours 1,734 RUN TIME (Operating Time) 6,277 Prod. time minus unplanned downtime Total yearly downtime 2,483 Planned and unplanned hours System utilization % 78 Run time/prod. time Average hours per shift 6.24 Hours per day13 3 shifts, hours 18.72 Utilization % x 24 hrs System availability % 72 Runtime hours/total time available NOMINAL OPERATING RATE 4,808 90,00O/hours in 3 shifts (Average tonnes per hour) Conveyor design rate - tph 5,769 (1.20 * operating rate) 20% factor added for conv. selection

671

Capital Cost Direct Costs. The largest primary gyratory crushers cost US $2 million or more, while overall crushing plant costs can be as high as $18 million. It’s necessary therefore to estimate crusher installation costs based on equipment costs plus the following direct costs, including construction contractor indirects: Earthworks Concrete Structural steel Architectural

Mechanical Electrical Instrumentation.

Indirect Costs. Indirect costs can fall within a range of 40 to 60% of the direct costs, and include: Construction indirects Construction equipment Spare parts/first fill Engineering, procurement and construction management (EPCM)

Startup and commissioning Freight Taxedduties Owner’s costs (relocation, hiring and training personnel, permits, licensing fees, etc).

In addition to the above, a contingency to cover unforeseen costs will be in the range of 10 to 20% of the sum of the direct and indirect costs. The designer must be aware of the project-specific costs of all such elements, so that he can monitor costs and promote methods of reducing total installation costs. In some locations, for example, labor and material costs could make a gabion wall more expensive than a poured concrete wall, which has minimal structural backfill. Ore Characteristics Ore characteristics are a critical element in both crusher selection and plant design. Dry ores require greater provisions for dust suppression and collection, including dust enclosures around screens, sealing on conveyor skirts, and vacuum and wash-down systems. Wet, sticky ores can plug chutes, reduce surge capacity, and decrease the live storage capacity of bins and silos. To address this problem, chutes must be easily accessible for clean-up, and large feeder openings must be provided for bins, silos and tunnels. If it is practical to obtain representative ore samples, it is prudent to have testwork conducted to establish ore flow properties, which will influence design parameters. At virtually all mines, ore characteristics change over time, and it can be costly to “design in” the optimal flexibility required to handle such changes. Some owners stipulate that initial capital investment be kept to a minimum, with design modifications paid for out of the operating budget. This is not always easy to achieve. Safety and Environment Safety must be designed into all mining facilities. North American mines must comply with local and national regulations such as OSHA, MSHA, the Mines Act and the WCB. The modern plant includes safety guards around all moving equipment, and emergency pull-cords on both sides of any conveyors with personnel access. The maintenance department and safety officer must keep these safeguards in working order. Ongoing safety training of plant personnel is imperative, and is considered to be one of the most vital and monitored feature of most mining operations.

672

Dust emissions must comply with the latest regulations for the jurisdiction. Designers must make provisions for the installation of dust abatement, suppression or collection equipment. Spillages from feeders, chutes and conveyors must be minimized. Spill collection can be “designed in” on feeder installations; chute designs can minimize spillage at receiving and discharge points; and conveyor belts can be widened to be more forgiving (e.g., skirting internal back-to-back width can be reduced to allow the belt more side travel.) Skirting should be extended a minimum of three belt widths past the load point. Rules for conveyor and load point design should be used for guidance only, with transfers custom-designed to suit a particular project. It goes without saying that clean plants have lower operating costs. Crushers, screens and dust-collection fans all contribute to high noise levels. Air-cooled lubrication systems are not only noisy, but often leak oil. Well-balanced, choke-fed crushers, dustenclosed screens and dust collector fans with silencers can keep noise levels under control. Recirculating water can be used to cool crusher lubrication systems. Project Location A project’s geographical location, topography, geotechnical conditions, remoteness and climate can all affect crusher plant design. Construction costs are generally much greater at high altitudes, in cold climates and at remote sites. To improve the economics of such locations, modular and pre-assembled structures and plant facilities are used prior to transportation to site. Local labor costs often dictate what material can be best used economically in a particular region; for example, cement structures are much cheaper to erect in Mexico than in Alaska. Remote projects can suffer from difficulties in obtaining spare parts on short notice. Crushing plant design should accordingly provide for laydown and workspace for onsite equipment refurbishment and repair. Where possible, equipment manufacturers should be encouraged to stock and provide spare parts close to the mining operation. Good geotechnical information is essential to crushing plant siting and design. Installing a primary crushing plant on solid rock reduces the cost of concrete and structural steel.

Life of Mine/Expansion Plans The life of the mine is a key element in the design of any crushing plant. Short-term mine lives (three to eight years) require a very careful approach to design, layout and construction. Since the crushing plant’s structure and enclosure can represent the largest single cost element in a primary crushing plant, it is imperative to optimize these structural and construction costs to suit the life of the operation. Perhaps a steel-supported, modular design will be best for short-term operations, since the equipment can be relocated and re-used; while for long-life mines, large concrete structures with fully insulated enclosures might be more economical. In conducting trade-off studies, short-term operations should aim for lower capital cost, while long-life installations should be designed to minimize operating costs and emphasize maintainability. “Operating availability is a function of the design of the processing lines and the ease and type of their maintenance” (Shoemaker and Gould, Modern Mill Design, 1980). Again to quote Shoemaker and Gould, “Increased production of the final product is often more easily and economically attained through expansion than by increased recovery.” Planning for expansion is therefore a consideration in all but the shortest-lived operations. Even at mines with expected lives of only five or six years, it may be necessary to select equipment that can handle anticipated throughput increases. Expansion plans for most crushing plants can be incorporated in the early planning stages at much lower cost than waiting until the mine is up and running before deciding to expand. More and more, operators want to increase primary crusher throughput, especially when they incorporate larger trucks into their mine planning or operations. One manufacturer has modified its 1,067 mm (42 in) and 1,371 mm (54 in) primary crushers to allow for larger rocks and increased

673

tonnage to pass through (larger openings at the top of each crusher), with minimal changes to the receiving hopper structure.

Operational Considerations Designers of new plants must be aware of ways of making a plant simple and economical to run; many plant modifications and additions can be justified by reductions in operating costs. Operation rooms should provide a comfortable, well-ventilated workspace with potable water and toilet facilities nearby. The operator should also be able to see all the main parts of the crushing facility under his control, through good direct visibility and by means of TV cameras and monitors. Although spills cannot be avoided, plant layout must facilitate quick and easy cleanup. Provisions should be made for suitable plant cleaning equipment. Wash-down hoses should be located within easy reach throughout the plant. Water pressure should be sufficient to wash down hard-to-access areas. Some operators regularly wash their crushing plants from top to bottom to eliminate dust build-up on the structural steel and equipment. Build-up on structure steel members tends to filter down throughout the plant during operation. Conveyors should have adequate clearance above the floor to permit access to spillage by shovels or plows. Crushers, chutes and belts are all subject to extensive wear, and wear parts and plates can be heavy. The designer should keep the weight of replacement parts, which must be manhandled to within 27 kg (60 Ib) for ease of installation. Monorails and hoists should be provided for ease of maintenance. Maintenance Requirements Plants must be designed for ease of access and maintainability if they are to meet their production goals. Keeping maintenance requirements to a minimum helps achieve higher overall operating availability. Scheduled preventive maintenance at crushing plants involves a number of elements, including:

0

0

Crusher wear parts Feeder wear parts Oil and lubrication Visual inspections

0

0

Screen decks Conveyor skirting and adjustment Conveyor belt repair Electrical and instrumentation adjustments.

Provisions must be made for overhead cranes to remove and replace crusher wear parts. Supports must be provided for gyratory and conveyor main shafts and laydown space for the cone crusher bowls is essential. Some operators carry a complete spare screen and change out for major screen maintenance. Trolleys, jib cranes and pull points should be designed to facilitate equipment maintenance. Oil and lubrication systems should be centralized and designed for easy automatic changes, with provisions for well-ventilated centralized lubrication rooms where possible. (e.g., a line of fine cone crushers should have a central oil receiving area, with piping to and from each crusher lube package for quick and easy oil changes.) Conveyor head chutes should be designed for easy access (not just through an inspection door, but through a man door in the chute). Conveyor belt change areas should be provided. Maintenance personnel should have easy visual and rapid access to screen decks for panel replacement. Designers should work with the screen manufacturers to ensure that covers provide good access for working on screens. Screening facilities must meet rigid dust emission requirements, but many off-the-shelf screen dust covers have not kept pace with these requirements. It may be necessary to custom-design covers that minimize emissions and provide easy access to the screen.

674

Climatic Conditions Building for cold-weather operations is very challenging, as is designing a plant in a desert environment. This is particularly true when year-round operation is required. Seasonal variations can change ore moisture content, so the crushing plant must be adaptable to changes in the material flow characteristics. Higher moisture requires greater angles of withdrawal, and stoneboxes must be designed to avoid plugging. The crushing plant equipment itself must be adjustable to climatic changes; for example, screen decks must be designed to maintain production, possibly by using wire mesh during the wet season and plastic during the dry. (Vary screen deck types dependent on seasons and material characteristics to achieve maximum passing through deck openings. Climate also dictates the type of plant enclosures required as shown in Figures 1 and 2. Many crushers in milder weather climates or desert areas are installed with an open face and have no enclosures at all.

Figure 1 Teck Cominco, Red Dog Operations, Alaska

-

Figure 2 Teck Cominco, Red Dog Operations, Alaska 42” x 65” gyratory

675

PROCESS DESIGN CRITERIA Design Criteria Information Typically, the information required to develop crusher process design criteria includes:

0

Geographic data Civil design criteria Structural design criteria Mechanical design criteria

Climatic data Process design data (process description, ore characteristics) Electrical/instrumentation design criteria.

Flowsheet Some sample flowsheets are provided in Figures 3, 4, and 5 showing crusher circuits. Figure 6 shows a typical three stage closed crushing circuit with its ancillary equipment. I’RIMARY CRUSHER

COARSE ORE ,STIK‘KPILE

CLIIED CIRCUIT

DOUBLE DECK SCREEN

SECONDARY CONE CRUSHER

Ol’lIN CIRCUIT PRODUCI

Figure 3 Two stage opedclosed circuit

PRIMARY CRUSHER

COARSEORB STOCKPILL

UOUHLE DECK SCREEN

!

SFCONDARY CbNE CRUSHER

3 STAGll CLOSED CIRCUll

DOUBLE DECK

DOUBLB DECK

SCREEN

SCREEN

,TERTIARY CONE CRUSHER

OPEN ,.m,TIm

TERTIARY )CONE

’CRUSHER

OPEN CIRCUIT I’RIIDUC‘I

Figure 4 Three stage opedclosed circuit

676

PRIMARY CRUSHER

COARSE OHE L

ST0CKPII.E

IIOUHLE DECK SCKEEN

nouH1.E DECK SCREENS

E

I1IIUHI.E DECK . SCREEN

SECONDARY

CONE ! CWUSHER

PRIMARY

SECONDARY

TERTIARY

QUATERNARY

OPEN

CLOSED

Figure 5 Four stage crushing circuit EQUIPMENT SELECTION Crusher Types The choice of crusher depends on the type and amount of material to be crushed. Gyratory and jaw crushers represent the bulk of primary crushers used at mining operations today, although some operations use roll impact crushers, low-speed roll sizers and feeder breakers. Cone crushers remain the most popular for fine crushing applications, although some mines use vertical impact crushers for tertiary and quaternary crushing. Major Equipment The major equipment in a primary crushing circuit usually includes only a crusher, feeder and conveyor. Secondary and tertiary crushing circuits have the same basic equipment items, along with screens and surge storage bins. Additional and Optional Equipment Other equipment items in crushing circuits can include: Rock breaker Overhead crane Freight elevator Service air compressor Sump pumps Air vacuum clean up systems Rock grapple Conveyor belt magnets Conveyor belt metal detectors Belt monitoring systems Belt feeders Screw feeders Bin ventilators

Apron feeder to the primary crusher Dust collection/suppression system Eccentric trolley removal cart Man-lift elevator Air cannons Water booster pumps Service trolleys Conveyor gravity take-up service winch Conveyor belt rip detector Conveyor belt weigh scales Vibratory feeders Limekement silos Sampling stations.

677

TF'J&P

IRON

MGNB lOCK 1REAKER VIBRATING GRIZZLY

ATMOSPHERE

PRIMARY

DUST COLLECTOR

JAW CRUSHER

__

..

SUMP PIMP

A P W

CRUSHER MAlNTENANCE

COMPRESSOR

SECONDARY CRUSHER CRANE

REWM CONMYOR

CRUSHER CRUSHER

TERTWRY CRUSHER CRANE

RECWM

RECLAlM

SUMP

ADS COMPRESSOR

SECONDARY' SCREEN

ADS COMPRESSOR

ADS

COMPRESSOR

ADS SYSTEM

$k

RECEIMR

AD.S. SYSTEM FOR PRlMARl

WMP POCKET

BIN

SECONDARL CONE CRUSHER,

~ T E M %PRmoR

RECENER

VlBRATNG FEDER

SCREEN

ADS. SYSIEM FOR REClAlM FEEDERS

T E M W E CRUSHDl

SECONDPRY DISCHARGE

comm

Figure 6 Three stage crushing closed circuit

' W HOUSE 'COUECrn SURGE

SECONDARY CRUSHER SUMP

RERlRN CONMYOR

CRUSHER SUMP

PLANT LAYOUT AND DESIGN A well-designed plant layout balances the capital versus operating cost over mine life. Buildings, infrastructure, and major equipment items, represent the major cost elements of a crushing plant. The designer must prepare a layout that suits the design criteria, flowsheet and selected equipment in the most economical possible configuration. It’s important to keep structural costs down, to design for ease of maintenance and operation, and to combine best practices with advances in fabrication and erection. Input from an experienced mining plant structural engineer can be very helpful. Crushing circuits and ancillaries have not changed a great deal over the years, so “Keep It Simple” is still the best way to design a plant. Owners may wonder why the design of head chutes hasn’t changed in decades, but the explanation is simple: it’s because the old, well-proven approaches still work best. On the other hand, it’s dangerous to assume that a layout that works well at one mine will work just as well, or at all, at another. Provisions must be made for the replacement of wear parts (e.g., install man-doors on head chutes with flood lighting inside the chute.) Faster part replacement means less downtime. Layout tools can include cut-and-paste arrangements, 2D arrangements fitted onto site topography, or 3D CAD to superimpose the design on the selected site. The choice of tool depends on whether the work is being done at the prefeasibility, feasibility or detailed engineering level, as well as on the accuracy required of any associated cost estimate. The best designs are developed using basic approaches and tools: site visits, discussions with mine personnel, sketches, and cutand-paste layouts. This writer believes that only after the initial concepts have been developed and optimized does 3D CAD have a role to play. Different industries have different approaches to crushing plant design. The standard approach in the oil sands industry is to use Microstation 3D CAD from the start; in some cases, the finalization of a system design (hopper, feeder, sizer crusher, and takeaway conveyor) has taken as much as two years, because of the uniqueness of the application. A similar design in the hard-rock mining industry takes from four to six months. THE PRIMARY CRUSHER Primary crushers, no matter what type, must all meet the design parameters described earlier. Design details that are fundamental to the layout of gyratory crusher plants are listed in the sections that follow. Some of these details are applicable to other types of crushers as well. A typical in-ground gyratory crusher layout is shown in Figures 7 and 8. Figure 9 breaks this plant down into major areas that are identified as “project specific” or “necessary”.

679

Figure 7 60” x 89” gyratory installation, Freeport, Indonesia

Figure 8 60” x 89” gyratory installation, Freeport, Indonesia

680

PROJECT SPECIFIC

NECESSARY

NECESSARY

PROJECT SPECIFIC

PROJECT SPECIFIC

NECESSARY

Figure 9 Gyratory broken into necessary process portions and project specific portions

68 1

A typical jaw crusher plant is shown in Figure 10 and Figure 11 shows a typical underground jaw crusher layout. A typical low speed roll sizer plant is shown in Figure 12.

Figure 10 Typical jaw crusher plant enclosed

Figure 11 Typical underground jaw crusher installation

Figure 12 Typical low speed roll sizer installation

683

Upper Superstructure

It is always a challenge to size the crane. Should it be used to install the crusher, or to just service the components of the crusher? The main service hook doesn’t need to travel any further than the top of the crusher; beyond this point, slings can be added to lift anything at lower levels. Crane main hook speed should be slow, for inserting the main shaft. Always prepare a hook coverage plan to check all areas of crane service. Crane maintenance access should include stairs and a platform to service an overhead bridge crane. Choose the type of crane, overhead bridge, jib, gantry or mobile crane required to meet project requirements. A well-ventilated crusher control room is required. There must be a washroom, with or without potable water. The electrical room can be located at the upper or lower level; keep high-voltage runs short. The location of the dust collector should consider operating lengths of ductwork.. Two main shaft supports are required, for the shaft in use and a spare. Storage must be provided for both. The spare shaft may be stored near the crusher or in the truck or maintenance shop. Provision may be required for a furnace and zincing, although most of today’s crushers use epoxy instead. Provide a maintenance laydown area. Locate the plant air compressor in a room away from dusty areas. Receiving Hopper Area

Determine whether a grizzly is required in the receiving hopper area. This is very expensive in gyratory installations, but is frequently used in jaw crusher installations. A splitter may be required in the receiving hopper area, to reduce impact on the spider, particularly in the expectation of large run-of-mine material. Some crusher manufacturers request that protection be provided from direct rock impacts on the spider. The design of a splitter remains a very controversial design subject and has to be reviewed for each project, remembering that any splitter installation can be very expensive. Investigate the design of the receiving hopper relative to where the material impacts as it leaves the truck. Well-hatch covers should be fixed-hinge on one side so they won’t accidentally drop to a lower level. To minimize dust emission, a vertical dump hopper spray system is best, with up to ten sprays per header. This provides greater distance for dust to pass through. A receiving hopper dust-hood plenum is required for plants using dust collection. Care should be exercised in determining whether receiver hopper liners are required, and if so, how many and what type. These items are costly. Often it is preferable to install only the steel inserts in the concrete for attaching liners. When the hopper deadbeds are formed, liners need to be applied only on exposed areas. Design a simple, easily removed, drop-in circular steel seal for the crusher and dump hopper. A rubber lining will prevent or minimize water leakage. Installing a receiving hopper access door at the crusher seal level allows for quick access when concaves are being changed.

684

The ability to easily dig out the upper crusher pocket is critical. A long length of stud link chain can be placed on the pocket floor with one end exposed for lifting out and breaking up the stone boxed material in the pocket. Rock breaker location is critical to provide reach in all areas of the hopper, and also for concave removal. Ensure that the breaker impact head can be parked in the vertical position out of the way of truck dumping. Determine whether the rock breaker should be remotely and/or locally controlled. The control room operator should be able to see down into the dump box, preferably through a window installed to the floor. He should be able to see approaching trucks. The operator should have access to washroom facilities. Truck bollards can be of concrete, old tires or tree trunks. The bollards should be located so they will always deflect the truck body away from any structural columns. Crusher Floor Level 0 0

0

A walkway platform should be installed around the crusher for easy spider removal. A built-in circular monorail under the concrete floor level at the top of the crusher will provide support for a trolley support air hammer, which can be used to remove or install the spider nuts Choose countershaft, in-line drive versus V-belt drive and clutch. The spider lube system should normally be located at the crusher tloor level. The air-seal compressor can be located on the crusher floor. The balance cylinder should be located close to the crusher, with maximum pipe bend radii for ease of quick response. Provisions must be made for oil relief collection.

Surge Bin Area 0

0

0

0 0

0

0

0

0 0

0

The surge bin access door should be split horizontally on one side, to allow easy manaccess to one side without opening the whole door. For the crusher discharge opening, tapered concrete with a welded AR plate is preferred. Air cannon discharge points must be pre-designed. Locate the air receivers and valves out side the surge bin. Level detectors should be installed at high and low points in the surge pocket. For the eccentric trolley, a hung design is best, complete with a man-access platform at the top of trolley to permit servicing the crusher hydraulic cylinder and eccentric. Surge pocket withdrawal opening liners should be made in a minimum number of large pieces for ease of removal. Drop-in design is best, without bolts. Liners can be lifted with a sling from the service crane hook through the crusher . The surge pocket opening slot should be made as long as practical to maximize live capacity . At a minimum, one truckload’s worth of live capacity should be provided, but a capacity of 1.5 or 2 truckloads is better. Some plants are designed with no surge bin under the crusher, with a wide, high-speed take-away conveyor to take the surge (flush rate) to a nearby stockpile or to a external surge bin. Install floodlights in the surge pocket to facilitate inspection and maintenance. Provide two pick-up openings through the crusher floor level for dust collection at the back of the crusher. Provide low-level protection for the surge feeder, possibly gamma detectors. This maintains a bed depth of material to protect the feeder from material falling directly on it.

685

Crusher Lubrication 0 0

0

0 0

The crusher lubrication package must have maintenance access at all points. Crusher oil piping must consider protection and ease of maintenance. Lube piping routing is critical; correct slopes must be maintained as per the manufacturer’s recommendations. Pickling of the lube lines can be a problem in some jurisdictions; stainless steel may be the best solution. Automatic drain and re-supply piping must be provided. The crusher lube package must be in a well-ventilated area, and in a closed room if possible. Choose an automatic or manual system for crusher oil changes. Choose water- or air-cooling for the crusher lube system.

Feeder Area from Surge Bin 0

0 0

0 0

0

Choosing a suitable feeder to draw material from the surge bin is always interesting. Some common types are apron feeders, hydrastrokes, belt feeders, magnetic or mechanical vibratory feeders. Dust collection facilities should be installed at the feeder discharge point. Dribble chutes should pass spillage onto the receiving belt, with the feeder discharge being designed to minimize the distance the rock will fall. This loading point should also ensure that under all material handling conditions, material will not stone box or build backup into the feeder. The design of feeder skirting should always provide relief in the direction of flow. Design a clean-up chute, which will allow spilled material from the floors above to pass down onto the take-away belt. To assist in maintenance of the feeder, provide for equipment pull points in the structures near the feeder.

Maintenance Items Gyratory crusher pre-fitted concave liner platforms can provide rapid concave replacement if pre hung on installation platforms. Chipping off existing concaves and letting them fall to the surge bin below facilitates the installation of new concaves. Remove old concaves from the surge bin at the same time as new concaves are being installed. Provide air hammer support (spider bolt removal) from a circular monorail. Pull points should be located in a manner that provides maximum assistance for equipment maintenance. Provide a hoist and trolley for lifting the crusher and feeder motors. A man-elevator is always useful in a gyratory crusher plant, but is often eliminated to reduce capital costs. Provisions should be made for installing one at a later date. Service air and water stations should be located throughout the plant. Include hoses and nozzles at predetermined washdown stations. Take-Away Conveyor 0

The take-away conveyor should have easy access under it for cleanup. Try to support the conveyor stringers from the floor above. There should be easy man-access into the feeder discharge chute.

686

0

0

0 0

If possible, allow for access by a “Bobcat” for cleanup on both sides of the conveyor and at the lower floor level. Provide dust collection measures at the conveyor (hood, plenums or sprays). Try to have the floor slope down and self-drain to an outside sump, to eliminate sump pumps within the plant structure. Keep good clearance under the conveyor tail pulley (a minimum of 400 mm). Provide walkway access to service the conveyor skirting.

Electrical 0

The electrical MCC/transformer electrical rooms can be located at the top or bottom of a crusher plant. Electrical cable tray routing and orientation should be checked by the mechanical process engineer. Vertical trays should be used to eliminate collection of spilled material in the trays. If the substation/electrical room is at the surface level, ensure there is no possibility of damage from impact from haul trucks.

Structural Considerations Provide easy stair access to each level. Provide access to both sides of the take-away conveyor. Braces and structures must be located away from equipment service and maintenance areas. Primary crushing plant enclosure costs can represent up to two-thirds the capital cost of a crusher station. It is therefore very important to select the most economical structure for the support of the crusher and ancillary equipment. There are many approaches: total concrete structures, round concrete structures, a mixture of concrete and steel, and reinforced earth structures with steel levels. The designer should spend considerable effort on selecting a structure that best suits the design parameters outlined above.

Dust CollectiodSuppression

0 0

0

There are many choices for dust collection/suppression systems, including bag filters, scrubbers, cartridge collectors, surfactants, water sprays and sonic fog. Whether one system or a combination is selected, care must be taken to provide service and maintenance access. Control, lube, compressor and electrical rooms should all be well-ventilated. Surge bins must include collection hoods. The take-away conveyor should have provisions for dust collection/suppression. Determine whether dust-collector air must be preheated in cold climates. A collector fan silencer should be considered, as fan noise can be excessive in closed areas. If an aircooling system is selected for crusher-oil, it will require venting for hot-air evacuation.

Crusher Installations A summary of previous gyratory crusher installations is shown in Table 3.

687

Table 3 Primary crusher plant installations Mine

%

Bethlehem Copper Endako Granisle B.C.Moly Marcopper

Highland Valley, BC Endako,BC Topley.BC Alice Arm.BC Phillipines

Bell C o p p e r Afton E q u i t y Silver Gibraltar

B a b i n e Lake.BC Kamloops,BC Houston.BC McLeese Lake,BC Lynn L a k e , M B Aznalco1lar.Spai

Ruttan Aznalcollar

--

n

GibraltarlPit

" E

Location

Similkameen

z No.2

Ranger Uranium Cadomin Quarry Highmont Similkameen No. I Palabora Mining Twin KennecottlRay Anaconda/Twin Butes Bouganville

z Inspiration

Cons. C l i m a x Moly

McLeese Lake.BC Princeton.BC

Normal Capacity

Service Crane stons

42 x 5 4 AC

1

1962

6"

700

40/5

51

600

3280

42 x 54 AC 42 x 54 AC 42 x 5 4 AC 42 x 54 AClKobi 42 x 54 AC 42 x 5 4 AC 42 x 5 4 AC

1 1 1 1

1964 1966 1967 1969

7 7" 5" 7

1500 1000 750

50/5 40/5 40/5 3515

74 52 56 74

1400 1150 1300 2500

1811

I I I I

1972 1977 1980 1972

6.5"

8" nla

1500 2000 nla

70 70 70

2300 1680 1600

2561 2120 2054

7 I*

3000

45/11 35/5 35/5 75/10

82

2600

3260

54 x 7 4 AC 5 4 x 74 AClKobi 5 4 x 7 4 AC

I I

1972 1977

6" 8"

1800 1760

50/10

II-Jun

81 81

2100 1950

2388 3270

I

1979

7"

3000

pit c r a n e

63

1662

1500

I500

Approx. Approx. Crusher Concrete Flr sq. It.@ cu. y d s

nla nla nla

1

1979

9 'I

2880

mobile

79

I200

nla

I

1981

6.5"

1400

60110

86

2100

nla

Hint on. AB

54 x 7 4 A C

I

1982

6"

2000

44/5

u/g

nla

nla

Highland Valley,BC Princeton.BC

5 4 x 7 4 AC

1

1990

8"

3000

75/10

81

2000

1500

5 4 x 7 4 AC

1

1972

8.5"

I500

82 m o b i l e

59

I200

nla

Phalaborwa. South Africa Ray,AZ Sahuarita,AZ

54 x 7 4 AC

2

1966

8"

1265 e a .

75/15

75

6000

8566

54x74AC 54 x 74 AC

I

1966 1968

8"

8.5"

I200 2250 ea.

60110

2

35/mobile

61 nla

2500 nla

nla nla

Papua.New Guinea 1nspiration.AZ

54 x 74 AC

2

1972

6"

1750ea.

60l10

102

5300

9500

54 x 74 AC

1

1972

7"

1500

75 m o b i l e

76

1100

1550

Henderson,CO

54 x 74 Traylor 5 4 x 7 4 AC

1

1976

8"/9"

3000

75/15

Ill

6100

9475

1

1977

8"

1265

80/20

101

circular

7680

5 4 x 7 4 AC

I

nla

6"

2000

75/15

139

5800

n/a

60 x 89 AC 60 x 89 AC 60 x 89 AC

I I I

1969 1976 1980

7" 7" nla

3500 3200 nla

100/10 90119 88/11

87 98 100

4200 1800 2200

4300 nla nla

60 x 89 Traylor 60 x 89 AC 60 x 109 AC 6 0 x 89 AC g r i z z l y feed 6 0 x 89 Traylor 60 x 89 AC 60 x 89 AC 60 x 89 AC (grizzly feed) 7 2 x 93 Traylor

I

1982

nla

4500

100130

96

2400

4238

2

I

1962l72 1955 1971

7" 8.5" 5.5"

75/10 130 75/20

97 1665 122

10000 circular 4200

nla 11000 1280

1

1971

8 '*

4000 e a 3000 5000 combined 3500

80l20

86

2100

4874

1 2

I

1972 1976 1976

10" 6" 8"

4000 2500ea. 4000 combined

75 70120 80/20

85 89 125

2800 4400 5500

6255 8000 8400

I

1972

7.5"

100120

106

2100

nla

-

c! Cities Service

Miami.AZ

Lornex Duval S i e r r i t a S o u t h e r n Peru

Logan Lake.BC Tucson.AZ Cuajone,Peru

Cdn. J o h n s -

Asbestos.QE

W

(I) (2)

Product Size

54 x 7 4 AC

=

$ Manville

stph

Approx. Overall Height It.

5 4 x 74 AC

Year Installed

5 4 x 7 4 AC

Jabiru.Australia

Palabora Phalaborwa, Mining South Africa KennecottlBonn Salt Lake eville City,UT -Brenda Mines Peachland,BC Cerro Verde Arequipa.Peru Rio Tinto Rio Tinto.Spain Minera C y p r u s Mining T h o m p s o n Corp. Creek.lD Iron O r e C o . Lab-City,NF R e s e r v e M i n i n g Babbitt.MI AnacondalButte Butte,MT

-

C r u s h e r Mfg Qty

1

W r i g h t E n g i n e e r s L t d . I n t e r n a l s t u d y of P r i m a r y C r u s h e r D e s i g n , 1982 M c Q u i s t o n F.W. & S h o e m a k e r R . S . " P r i m a r y C r u s h e r D e s i g n " " A m e r i c a n I n s t i t u t e o f M i n i n g , M e t a l l u r g i c a l and P e t r o l e u m E n g i n e e r s " J a n . 1978

688

Underground Crushers Gyratory, jaw and roll sizers have all been installed underground to act as primary crushers prior to the transportation of the ore to the surface. There are a few gyratory crusher installations in hard rock mines (the latest, 1998, being Phelps Dodge Henderson mine 1,371 mm x 1,880 mm [54 ft x 74 in] in Colorado) but the greatest percentage of crushers working underground are jaw crushers. Until a few years ago, double toggle crushers were the underground crushing choice, however, these are now being replaced with more efficient and less expensive single toggle crushers. A typical underground jaw crusher installation is shown in Figure 11.

FINE CRUSHING (Secondary, Tertiary and Quaternary) Fine crushing circuits can be more challenging to design than primary crusher installations. There are more equipment options, and each has different installation and maintenance requirements. The process flowsheet dictates the expected performance of the items in the flow stream. The designer must configure the equipment and structures into a balanced, economical plant design. Screens, feeders, stockpiles, bins, conveyors and crushers must all be interfaced with the most economical supporting structures and buildings. An open-circuit crusher is easier to design and lay out than a closed-circuit design, since it has fewer equipment items and structures. Figures 13, 14, and 15 show some typical secondary crusher open circuit layouts. Figure 16 shows some typical open circuit secondary and tertiary crushers. Provisions should be made for possible future conversion of an open-circuit plant to a closed-circuit version. Cone crushers remain the choice for most secondary and tertiary operations, with some gyradisc and vertical-impact crushers also utilized on certain ore types. Water-flush cone crushers have been introduced in secondary and tertiary installation, which requires careful design of the water systems to and from the crusher. Most open-circuit secondary and tertiary crushers include scalping screens to remove fine material prior to the secondary crushers. Closed-circuit crushers use tertiary screens to control the final product size. (See Figure 17 and 18 for typical closed circuit secondary and tertiary crushers.) In some larger installations, the secondary and tertiary crushers are located in one plant area and the screens in another. (See Figure 18f.) Crushers and screens in these plants have common bins feeding to the multiple crushers or screens. Most plants now have the screen feeding the fine crusher, providing for easier access to service the screens. As with primary crushers, fine crushers must meet the design parameters listed above. Design details that are helpful to fine crusher plant layout area listed in the sections that follow.

689

Figure 13 Open circuit (1) secondary crusher

Figure 14 Open circuit (2) secondary crushers

LEGEND BIN SCREEN CRUSHER CONVEYOR

1?a

1sd

15b

15c

15e

151

Figure 15 Typical open circuit secondary crusher options

690

16e

Figure 16 Typical open circuit secondary and tertiary crushers

LEGEND BIN SCREEN CRUSHER CONVEYOR

I7b

17a

I7c

Figure 17 Typical closed circuit secondary crushing

691

18b

18c

18d

Figure 18 Typical closed circuit secondary and tertiary crushers Coarse Ore Stockpile

0

The primary crushing circuit is normally separated from the secondary crushing circuit by provision of a coarse ore stockpile, which retains the primary crushed ore. For mines with short lives and small tonnage rates, operators may eliminate the coarse ore stockpile to reduce costs. Stockpile live capacity is a source of controversy. Coarse ore stockpiles were originally designed for three days’ capacity (a long weekend), but this is impractical for some of today’s mines with high daily throughputs. Now, it is possible to size a stockpile by simulation, using pile specific criteria. There are many types of coarse ore stockpiles (e.g., conical, elongated, radial, covered or open, heated) with just as many types of withdrawal arrangements and feeders, (See chapter on bins, stockpiles, and feeders.) The stockpile provides improved overall system utilization by de-linking the primary and subsequent crushing operations. Coarse ore stockpiles also provide the capability to continue operation of the secondary crushing facilities should the primary crusher become inoperative, by bulldozing the stockpile into the reclaim openings.

Feed to Screens or Direct to Secondary Crusher

0

The conveyor or feeder transporting ore directly to a secondary crusher should be retractable (or tilt at the head end) to permit crane access to the crusher. (See Figure 17c.) The conveyor feed chute to a secondary screen should have sufficient height to allow the material to feed to the total width of the screen as rapidly as possible. This is especially critical when feeding from a conveyor to the new wider banana screens. The installation of a bin feeding to a single secondary crusher or screen can allow for the future installation of an additional secondary crusher and screen. (See Figures 14, 15c, 15d and 15e.)

692

Screens 0

0

0

Screens are being manufactured wider so the feed must have more height to allow the material to spread out across the width of the screen. The chute feeding the screen should have easy man-access for replacement of liners and easy removal of material build-up. Most screen installations should be totally enclosed. Screen manufacturers have yet to develop a cover that allows for ease of maintenance and access. Adequate platforms should be provided for access around the entire screen, to facilitate rapid inspection and changing of screen decks. The discharge chute from the screen should seal against the screen, and be designed as one piece for ease of removal. When the chute is removed, there should be sufficient opening in the floor to lift out the crusher bowl and head. The discharge chute should have a full man-door. Permanent floodlighting should be installed inside the chute for ease of inspection.

Crushers

0

Some cone crushers require servicing and removal of components from the bottom. The design and layout should allow for such service requirements. There have been many changes recently in the manufacture of cone crushers. A careful assessment of each suppliers’ requirements for lubrication, water and air services to the crusher is mandatory. Automatic bowl adjustment is now common on most cone crushers and is definitely going to be used for on-stream electronic control adjustment to maximize input and product size control.

Tertiary Feed Bins Most manufacturers ask for controlled ore feed rate to their tertiary crushers. The secondary crushed material (and in the case of a closed circuit, the recirculated material) is usually stored in a bin, and fed to the tertiary screen or directly to the crusher with a variable speed feeder. (See Figure 19.) When the feed comes from a bin via a feeder directly to a crusher, the feeder design should be retractable so that the tertiary crusher head and bowl may be removed by a service crane. Feed to the tertiary feed bins is dependent on the number of crushers being fed. A single point discharge will be adequate with three crushers, although feed distribution is not good. Five crushers can be fed with a two point feed discharge system, with the main feed conveyor feeding directly to the bin or using a flop gate to a fixed horizontal conveyor to feed to the other discharge point (or a tripper conveyor can be utilized) See Figures 18d, 18e and 18f. Feeders from a Tertiary Bin 0

A belt feeder will maximize live volume in the bin Vibratory feeder/magnetic/mechanical is less expensive but provides less live storage i n the bin.

693

Lubrication 0

0

Some lubrication systems are mounted on skids which can be as large as large as 4 x 2.4 m ( 1 2 x 8 ft), therefore sufficient access space must be provided. Other fine crusher lubrication considerations are the same as for primary crushers.

Maintenance Items

Pull points should be located in a manner that provide maximum assistance for equipment maintenance. If an overhead crane is provided, it should be able to service all main equipment items. For crushers without building enclosures, a modified gantry crane or mobile crane can be used. A man-elevator is always useful in large crushing and screening plants, and is a necessity in very large capacity plants. Service air and water stations should be located throughout the plant. Include the hoses and nozzles at predetermined wash-down stations. Adequate water pressure should be provided for wash-down and cleaning of all areas of the plant. Allowance of adequate working width on each side of the lower-floor conveyors, and sufficient clearance under conveyors for easy cleaning Man-doors should be provided on all chutes for maintenance and replacement of wear parts. Inspection doors should not be used to provide access for wear plate replacement and unplugging of chutes. Clean-up chutes should be provided at various levels, to enable spilled material to be passed to a receiving conveyor at the lower levels. Conveyors Conveyor design considerations for fine crushing circuits are the same as for primary crushers. Electrical 0 0

0

Locate the electrical/MCC room centrally to minimize long cable runs. Electrical cable tray routing and orientation is critical and should be reviewed by mechanical process personnel. Vertical trays are preferred to horizontal (vertical trays allow no collection of spilled material in tray) Allow for plenty of lighting in all areas of a plant, including inside chutes and bins.

694

695

Q

a V

I

k QJ

t

. YI

Y

a

C

a

Structural Considerations

0

Provide easy stair access to each level. Provide access to both sides of the take-away conveyor. Braces and structures must be kept away from equipment service and maintenance areas. As with primary crushers, structural costs for fine crushing circuits are very high. It is therefore very important to select the most economical structure for the support of the crusher and ancillary equipment. The designer should spend most of his effort on selecting a structure that best suits the design parameters. The selection of the types of flooring in a crushing plant is always controversial. Some selections are grating, checker plate and concrete. Checker plate flooring allows for easier clean-up and does not allow for smaller rocks or spillage to any floors below. (The most suitable flooring for conveyor galleries in northern climates is wood.)

Dust collectiodSuppression

0 0

0

0

There are many choices for dust collectiodsuppression systems, including bag filters, scrubbers, cartridge collectors, surfactants, water sprays and sonic fog. Whether one system or a combination is selected, care must be taken to provide service access. Control, lube, compressor and electrical rooms should all be well-ventilated. Bin aiddust evacuators are required. In cold climates, determine whether dust-collector air must be preheated. A collector fan silencer should be considered, as fan noise can be deleterious in closed areas. If an air system is selected for crusher-oil cooling, it will require venting and hot-air evacuation.

CONCLUSION Primary crushing will see the introduction of bigger, 1,800 mm (72 in) gyratory crushers, and lowspeed, high-tonnage roll sizers will become more generally accepted. These larger crushers are required to handle the higher throughputs, and 400- to 500-tonne capacity trucks, expected on some future projects. Fine crushing has already seen the introduction of 10 ft cone crushers. Much wider banana screens, with sizes up to 4 x 8 m, are also being introduced. Vertical impact crushers are still finding their place for certain crushing applications. Safety and the working environment remain the two areas of plant design, which require more attention. Noise abatement and the reduction of dust emissions remain the goals of most operators and crushing plant designers. As designers become better trained and familiar with 3D software, then 3D plant design may become the method of choice to optimize the most economical plant designs. The biggest change, which is now being introduced into the design of crushing plants is the design of the mining process as one complete system, from mine quarry and pit to the final product. Digital image analysis is now allowing electronic monitoring of the size distribution of the material being handled at any point in the product stream. This monitoring and analysis of the size distribution from the pit face to the mill or heap, now allows for the adjustment of the crusher discharge openings as production continues. We can now plan for a more uniform product, increased production, less wear, and longer mine life.

696

ACKNOWLEDGEMENTS The author would like to thank Bill Dowall for reviewing the paper, Andrew Solkin and Shirley Ernewin for editorial assistance, and David Ng for his assistance in preparing all the CAD figures. REFERENCES McQuiston, F.W. & Shoemaker, R.S. 1978. “Primary Crusher Design” Book for “American Institute of Mining, Metallurgical and Petroleum Engineers”. Boyd, K.L. & Dowall, W. Wright Engineers Ltd. 1982. Internal study of primary crusher design Could, D.G. & Shoemaker, R.S. 1980. “Mineral Process Plant Design”, Chapter 40, Society of Mining Engineers, “American Institute of Mining, Metallurgical and Petroleum Engineers”. Dance, A. 2001. “The Importance of Primary Crushing in Mill Feed Size Optimisation”, International Autogenous and Semiautogenous Grinding Technology 2001, pp 1-201.

697

Types and Characteristics of Grinding Equipment and Circuit Flowsheets M. Ian Callow' and Anthony G.Moon2

ABSTRACT This paper describes grinding equipment and grinding circuits typically used in the ferrous and non-ferrous mineral industry. The technology and equipment is specialized because of the need to grind mostly siliceous, highly abrasive ores. , INTRODUCTION This paper mainly describes North and South American and Australian practice using predominantly two-stage SAG and ball mill grinding circuits. The SAG mills are typically large diameter and short length compared to the squarer South African lump mills. The technology is fundamentally similar with minor variations to suit local conditions. Most technology and equipment used in other mineral industries such as cement and limestone is generally not suitable for more abrasive ores. There has been however some adaptation of cement industry equipment, notably vertical grinding mills and high-pressure grinding rolls, that has resulted in power savings in some applications with harder rock. Ball mills (and a few related pebble mills) remain, after one hundred years, the central component and workhorse of most grinding circuits. Circuit differences are mainly in the method of preparing the feed to the ball mill. Wet grinding is almost universally practiced with the notable exception of dry-grinding, air-swept, double-compartmentball mills (double-rotators, also adapted from the cement industry) that grind refractory gold ores prior to roasting. OVERVIEW OF BALL MILL FEED PREPARATION SYSTEMS Ball mills are used as: Preparation of run-of-mine (ROM) feed to the ball mills is conducted in stages as follows: 1. Crushing the ROM feed by primary crusher to a top size of about 300 mm at crusher

settings of I50 to 200 mm to permit transport by conveyor. 2. Further size reduction of the primary crusher product by either: 0 Two or more stages of crushing by cone crushers to a ball mill feed size of 10 to 15 mm. 0 Two or more stages of crushing plus rod milling to a ball mill feed size of 10 mesh. 0 Semi-autogenous grinding (SAG) to a ball mill feed size of 80% passing 1 to 4 mm. There are other less common combinations of size reduction equipment using mineral sizers, and high-pressure grinding rolls. Over the last 30 years, as concentrator capacities have increased, SAG mills have become the preferred method of preparing ball mill feed. This is mainly because the capacity of secondary and tertiary crushers has not kept pace with the increase in plant capacity. Currently the largest cone I

Bechtel Corporation, Santiago, Chile.

' Rio Tinto Technical Services, Salt Lake City, Utah, USA. 698

crusher commonly in service is the Nordberg MPlOOO driven by a 1,000 hp motor. A 50,000 tpd secondary and tertiary crushing plant is therefore complex with many crushing lines plus associated screens, conveyors, bins and support equipment. This leaves the SAG mill as currently the only practical system to prepare ball mill feed at medium and high tonnage rates. SAG mills (or autogenous mills) have been installed in virtually all the mineral processing grinding circuits built in the last 20 years. TYPES AND MAIN CHARACTERISTICS OF GRINDING EQUIPMENT Ball Mills Selected recent large ball mill installations are listed in Table 1. Following the poor performance of 1 8 4 diameter ball mills installed at Bougainville some years ago, there was much speculation that the limit of ball mill size had been reached. It was subsequently proven that operating conditions were the cause of the observed reduced grinding eficiency and it was not the effect of ball mill size. Similar ball mills at Pinto Valley operated with success. Since that time ball mill sizes have continued to increase and there is currently no evidence to suggest that efficiency drops as diameter increases. Some words of caution are being raised on the possibility that very high slurry velocity through large mills may be a factor to consider in the fhture if, as expected, ball mills sizes continue to increase. (') Table 1 lists the largest ball mills as three-25 ft dia x 40 ft long mills, driven by 18,000 hp gearless motors, currently being installed in the Escondida Phase IV expansion. TWO-26ft dia x 38 ft long mills, driven by 20,770 hp gearless motors, are being considered for the Collahuasi expansion. Table 1 Selected large diameter ball mills PROPERTY No. Dia.ft ShellL. ft HP(Ea) HP/ft 196 6 18 28 Kennecott 1-3 5,500 191 34 6,500 2 18 Ray

CS% 73

Drive

L/D 1.56 1.89

Robinson Freeport 118 Exp. Ernest Henry La Candelaria I1 Century Zinc Kennecott 4 Exp. Alumbrera Escondida 3 Exp

2 2 2 2 2 2 4 2

20 20 20 20 20 20 20 20

31 30.5 28.75 30.5 32.5 30 30.5 33.5

8,750 8,500 7,500 7,500 9,000 7,500 8,000 9,000

282 279 261 246 277 250 262 269

75

78

1.55 1.53 1.44 1.53 1.63 1.50 1.53 1.68

Los Pelambres

4

21

33

9,500

288

75

1.57

Batu Hijau Collahuasi Cadia

4 2 2

22 22 22

33.5 36.5 36.5

9,727 12,869 12,064

290 353 33 1

75 75 74.5

1.52 1.66 1.66

Freeport 190 Exp. 4 3 Antamina Escondida 3.5 Exp 1

24 24 24

30.5 36 34.5

14,000 15,000 14,000

459 417 406

78 72/90 76

1.27 1.50 1.44

Escondida4Exp

25

40.5

18,000

444 74 to 90

3

699

73 76 74

Gearless

1.62

Ball Mill Characteristics Feed and Product Sizes. Typical feed and product sizes are listed in Table 2. Feed size in primary grinding ball mills is sensitive to ore hardness. Typically softer ores can be fed at coaxser size and visa versa. If the feed size is too large, particularly with very hard ores, the ball mill will discharge an unacceptable amount of coarse tramp oversize or “scats.”

Duty As primary mills grinding crushing plant product As secondary mills grinding SAG mill product As regrind mills regrinding concentrate

Feed Size Minus 10 to 15 mm

Product Size 80% - 100 to 200 microns

80% - 1 to 4 mm

80% - 100 to 200 microns

80% - 100 to 150 microns

80% - 20 to 40 microns

Ball Load. Ball load varies with the size of the mill and the bearing support system. Traditionally in smaller ball mills, say up to 16.5 ft diameter, the ball load has been in the range of about 40% of total mill volume. However, as mill sizes have increased, the size of the bearing has also increased to support the additional load. This means that, in trunnion-mounted mills, the trunnion diameter has also increased relative to the overall mill diameter resulting in a lower percentage of the mill volume being available for grinding steel. Mills above 20 ft diameter typically have ball loads in the range of 30 to 35%. Some operators insert filler rings in the trunnion discharge to increase the fill volume and increase power draw. Shell mounted ball mills, as installed at Collahuasi, do not have this load restriction as the diameter of the discharge opening is independent of the bearing requirements. Ball Size. Typical ball sizes are listed in Table 3.

Duty As primary mills - grinding crushing plant product As secondary mills - grinding SAG mill product As regrind mills - grinding concentrates

Ball Diameter u p to .loo mm Up to 80 mm 25 to 40 mm

Critical Speed. The critical speeds listed in Table 1 show a range of 72 to 78%. The 72% is for Antimina is not typical, being the lowest speed of a variable speed gearless drive to grind unusually soft ore. A more normal conservative selection is 74%. Some mills run at 78%. In the case of Freeport No. 4 expansion the higher critical speed of 78% allowed slightly shorter mills to be selected to fit into the available space. Experience with a variable speed 18 ft diameter ball mill at Bougainville in the mid-1980s showed that speeds above 74% critical were not advantageous and grinding performance deteriorated at speeds above 80%. (’) Drive Systems. This subject is addressed in more detail in other papers at this conference but a brief summary of the current status is included here for completeness. In the past the extra cost of variable speed drives on ball mills has not been justified. An early such drive was installed on an 18 ft ball mill at Bougainville on copper ore to determine the optimal speed for later additional ball mills and the recommendation at that time was that there was no justification for a critical speed higher than 74%. The first major deliberate selection of variable-speed ball mill drives is on the three 24 ft dia x 36 ft long 15,000 hp ball mills at Antamina as illustrated in Figure 1. The required 15,000 hp is within the capabilities of a less expensive gear and twin pinion drive but pilot test work showed that the ore had such a wide range of hardness that the inherent variable-speed feature of a gearless

700

Figure 1 Antimina 24 ft dia x 36 ft 11 mW ball mills drive would be an advantage. The mills are currently in early operation and results are awaited with interest. This was not the reason for selecting gearless drives for the three 25 ft dia x 40 ft 18,000 hp ball mills currently being installed in the Escondida Phase IV expansion. A less expensive gear and twin pinion drive at 9,000 hp on each pinion approaches the upper limit for this type of drive but is still feasible. It was believed that a gearless drive requires less maintenance than a gear and pinion drive. Length to Diameter Ratio. Length-to-diameter ratios listed in Table 1 range from 1.25 to 1.65. Previous practice has generally been to select a higher ratio for a finer product. For example, South African tube mills grinding gold ores to a finer product size than most flotation feeds typically have a ratio of around 2.0. The Freeport No. 4 ball mills are of particular interest because the 24 ft dia x 30 ft long mills have a low L:D ratio of 1.25. They were selected to fit the space available but there is some evidence, based on the operating work index (Wio), of a slight increase in grinding efficiency compared to the ball mills in their earlier installations that have a higher ratio. (3) Liner Systems. Ball mill liners have not received the same attention in recent years as SAG mill liners but there is some indication that t h s is about to change. Plans to study the ball motion in the large ball mills being designed for the Collahuasi expansion are planned using computer simulation. The most popular liner remains the cast white iron double wave liner. A variation of this liner, colloquially known as the “hump and bump” is a variation on this theme with one wave larger than the other. The Noranda wave design comes in a distant third in popularity. Rubber liners have not found much acceptance in larger ball mills and in the few where they remain installed, t h ~ sis often a function of the plant design not allowing for the use of liner handlers!

70 1

SAG Mills Selected recent SAG mill installations are listed in Table 4.

Table 4 Selected SAG mills PROPERTY No. Robinson 1 Collahuasi 2 Huckleberry 1

Dia. ft 32 32 32

Lengthft 14.75 15 14.5

EGLft 13 13.25 13

HP(Ea) 10,000 10,724 10,000

Drive 2~5,000VS 2~5,400VS 2~5,000VS

Ray Freeport 118 Exp. Ernest Henry

2 1 3 1 1 1 1

34 34 34 34 34 34 34

17 15.3 17 19 19 17 17

15.25 13.5 15 17 17 15.25 15.5

12,000 12,000 12,000 13,800 14,000 14,200 14,800

2~6,000VS 2~6,000VS 2~6,000VS 2~6,900VS 2~7,000VS Gearless 2~7,400VS

Fimiston Exp. La Candelaria I1 Century Zinc Kennecott 4 Exp. Alumbrera Escondida Exp. Batu Hijau St.Ives Los Pelambres

1 1 1 1 2 1 2 1 2

36 36 36 36 36 36 36 36 36

16 16.7 17 18.7 19 19 19 19.5 19

14 15 16 17 17 17.25 17 17.75 17

16,000 16,000 16,000 16,000 18,000 18,000 18,000 18,000 17,000

2~8,000VS Gear1ess Gearless Gearless Gearless 2~9,000FS Gearless Gearless Gearless

Escondida 4 Freeport 190 Exp. Olympic Dam Antamina

1 1 1

1

38 38 38 38

22.5 20 25 21

20 18 23.5 19

26,000 26,000 24,120 27,000

Gear1ess Gearless Gearless Gearless

Cadia

1

40

22

20

26.810

Gearless

Kemess Ft.Knox Kennecott 1-3 Olympic Dam

SAG mill size and power continue to increase. The largest SAG mill remains the 40 ft dia. x 20

A (EGL) long 20 mW mill at Cadia as illustrated in Figure 2. There are three 38 ft diameter mills in operation and a fourth under construction. A larger 40 ft x 22 ft (EGL) mill driven by a 21 mW gearless drive is proposed for an upcoming expansion at Collahuasi. Vendors have already completed preliminary designs for 42 and 44 ft diameter SAG mills driven by up to 30 mW motors. Again some words of caution are being raised on the possibility that the very high slurry velocity through the large mills with softer material may be a factor to consider in the future particularly in the design of the grates and pulp discharge system. Freeport No. 4 increased grate aperture size from 38 mm to 50 mm and modified the pulp discharger design to meet the nominal design capacity of 95,000 tpd through a single 38 ft diameter mill. (3) The Escondida Phase IV SAG mill is designed to handle 1 10,000 tpd of new feed through a single 38 ft diameter mill.

702

Figure 2 Cadia 40 ft dia. x 22 ft 20 mW SAG mill Feed and Product Sizes. The top size of the feed to most SAG mills is controlled by the setting of the primary crusher whose main function is to reduce run-of-mine ore to a s u e that can be readily conveyed, stored and then reclaimed as feed to the SAG mill. Primary crusher closed side settings are in the range of 150 to 200 mm resulting in a top size of up to 300 to 350 mm. In some cases, for example at Copperton, Escondida and Freeport, the run-of-mine ore can be very fine and

703

the top size is considerably less. Alumbrera, as part of an optimization program to handle harder ores, targeted a smaller feed size with an 80% passing size of down to 80 to 100 mm. (4) In the last few years there has been increasing interest in determining optimum feed size for operating plant under the “mine-to-mill” concept. The basic idea is that it is cheaper to blast than to grind and savings have been demonstrated at a number of operations. The emphasis points to the production of fines during blasting and not just to the reduction of top size material. In some cases there has been a deliberate attempt to further reduce the size of the feed to the SAG mill by additional pre-crushing. This concept has been successful applied at Troilus and Asarco Ray where, in both cases, the run-of-mine ore became significantly harder. (5) (6) The result was a serious drop in capacity as the circuits became SAG mill limited. In both cases, pre-crushing increased capacity to the original design rating. However pre-crushing has limited application and, for example, an ori inal grass roots installation at Fimiston, based on the results of simulations, was not successful. Imaging systems have become available in recent years and they have proven to be a very useful tool to determine not only the size distribution of the primary mill feed but also the size distribution of blasted bank material and primary crusher feed as it is dumped. The product size from autogenous and semi-autogenous grinding can vary from 80% passing 150 microns in single-stage Illy-autogenous operations on hard ores, to 80% passing 4,000 microns with semi-autogenous grinding on “soft” ores. Ball Load. The ball load in primary mills varies between zero in autogenous mills to up to 20% in semi-autogenous mills. A typical ball charge ranges from 8 to 12%. Higher percentage ball loads are normally associated with a more unusual SAG mill operation in which the feed does not provide sufficient large rock to build and maintain an adequate grinding charge in the mill. An example is Freeport No. 4 in which the SAG operation has been described as similar to a semi-ball mill because of the 20% ball load. It is important to evaluate the expected ball load in the SAG mill in the early stages of design so that a predicted ball load can be specified for the structural design of the mills. Most specifications call for a 15% design ball load but Freeport, for example, recognized their unusual application and designed for a 21% ball load. Ball Size. Top ball size has traditionally been 4 to 5 inch diameter. Larger balls, up to 6 inch, have become available in the last few years and have been beneficially used in some applications, for example at Alumbrera, on harder ore that requires greater impact breakage. (4) Critical Speed. Virtually all autogenous and semi-autogenous mills have variable-speed drives. A typical mill design speed range is 60 to 80% of critical. Mills normally operate in the range of 74 to 80% critical and may vary outside this range as ore types change and lifters become worn.A key feature of speed control is to minimize damage to the liners that results from overthrowing the charge. Drive Systems, This subject is addressed in more detail in another paper at this conference but a brief summary of the current status is included here for completeness. As stated, virtually all autogenous and semi-autogenous mills have variable-speed drives. There have been a few fixed speed drives that have either been justified by a very homogenous unchanging feed or by the belief that cost saving will accrue. The latter justification has almost universally been proven incorrect because of higher liner wear due to breakage and lack of operating flexibility. A previous fixed speed drive at Escondida Phase 3.5 was subsequently converted to variable speed. Liner Systems. The understanding of the effect of liner and lifter design in SAG mill operation has received a major boost in recent years with the development of the Millsoft software. This system has been responsible for a major shift in thinking that has resulted in operating improvements at some properties. Previous traditional designs were based almost exclusively on the empirical assumption that the number of rows of lifters was twice the diameter of the mill in feet. Also the lifters were either “hi-lo” or “hi-hi’’with the choice being somewhat nonscientific, oriented around the perceived problem of packing between the liners.

8,

704

The traditional designs have been replaced typically with fewer rows of lifters that have a shallower face angle. The choice of number of rows of lifters when relining existing mills is limited because the mill shells were drilled for traditional lining systems. Vendors advise that new enquiries now seek drill patterns that will allow maximum flexibility in liner and lifter design.

Rod Mills Rod mills are restricted in size and capacity by a limit on the length of grinding rod (about 20 ft) that can be easily handled and that do not bend or break in the mill. No operator regrets the demise of the rod mill with its high maintenance and potential for nightmare rod tangles. There remains a small nitch use for rod mills in smaller capacity circuits in which a close size distribution is required with minimal production of fines. Due to their very limited application rod mills are not considered further in this paper. Double-rotators Downstream processing can dictate that dry grinding is employed to prepare the feed. Gold roasting applications for refractory ores requires a dry ground product and two of the most recent installations have used cement industry experience with double rotator mills. These mills incorporate the advantages of both an air swept mill and a diaphragm mill. The new feed enters the drying chamber with the necessary high volume of hot air, and the resulting dried feed passes through a grate diaphragm into the frrst or coarse grinding chamber. The coarse chamber discharges peripherally through the mill shell and the ground material is transported mechanically into the classification system. The large portion of the classifier oversize is fed into the second fine grinding compartment whose discharge combines with the coarse chamber discharge. The balance of the classifier oversize joins the new feed to the mill where it improves the efficiency of grinding in the coarse grinding chamber. Dry grinding circuits are characterized by relatively complex classification and transport systems when compared to slurry systems. Vertical Mills Vertical mills use a double helix screw to stir the grinding media, which can be steel balls, ceramic media or pebbles. Feed enters the top of the mill and an external recycle pump provides an upward rise velocity in the grinding chamber. Grinding is by attrition and abrasion. Less energy is required when compared to conventional tumbling mills. The primary application for vertical mills has been for regrinding base metal concentrates and for production of slaked lime. However vertical mills have been successfully applied to secondary or tertiary grinding applications ahead of flotation and a notable installation is four 1,250 hp units as a third stage of grinding at Chino. (9) Their use is restricted to relatively small tonnage operations since the maximum size currently designed is 1,500 hp. The mills are described in more detail in another paper in this symposium. High Pressure Grinding Rolls A previously undocumented comment by Rob Morrison of the Julius Kruttshnitt Mineral Research Centre is very pertinent - “What this industry needs is a good 5 or 10 mW (secondary and tertiary) crusher.” The comment was prompted by a discussion about the lower efficiency of a SAG/ball mill circuit compared to equivalent size reduction by crushers and single-stage ball mills. A step towards the “big” crusher is the high-pressure grinding roll whose adaptation from the cement industry to the hard rock industry has been the subject of much interest. Early problems of rapid shell wear have been the subject of considerable experimentation with some success. The potential energy saving was recently recognized by Boddington in an expansion study. (lo) They predicted a saving of about 5 kWh/ton over the equivalent SABC circuit by replacing lowcapacity tertiary crushers with high-capacity, high-pressure grinding rolls followed by single stage ball mills. Both cone crushers and lugh-pressure rolls are covered in more detail in other papers at this symposium and are not discussed further in this paper.

705

CHARACTERISTICS AND APPLICABILTY OF DIFFERENT CIRCUITS Secondary and Tertiary Crushing plus Single-stage Ball Mill This circuit (and the following rod millhall mill circuit) is the original pre-1975 conventional circuits that were at one time almost universally in use throughout the industry. Variations include replacing ball mills with pebble mills or tube mills (particularly in the South African gold industry), and deleting tertiary crushing (typically with softer ores or low tonnage operations). The three-stage crushing and single-stage ball circuit remains one of the most energy efficient circuits compared to later autogenous and semi-autogenous circuits. However, as previously discussed, crusher size did not keep pace with industry requirements leading to the demise of this type of circuit. PRODUCT

PRIMARY CRUSHER FEED

SECOND R Y

CRUSHPNG

TERT 1 A@ CRUSH I NG I N CLO D CIRCULT

Figure 3 Single-stage ball mill Secondary and Tertiary Crushing plus Rod Mill and One or Two Stages of Ball Mills PRDDUCT

7

Figure 4 Rod mill ball mill Single-stage Autogenous Mills Early installations in the iron ore industry have operated most successfully for many years. Similar early installations in the copper industry were not so successful, typically grinding too fine at low tonnage rates. An exception is Palabora that has also operated successfully for many years with the later addition of a pebble crushing circuit. (") Some circuits were successfully modified by the addition of ball mills and by adding balls to the primary mills thereby converting from autogenous to semi-autogenous operation.

706

Single-stage SAG Mills There are a few single-stage SAG circuit installations operating successfully, notably at Henderson that has operated for 20 years. (”) This type of circuit was also well suited to uranium sandstone deposits in which uranium coatings were released for subsequent leaching. Autogenous plus Ball Mill A notable autogenous plus ball mill installation in the copper industry is at Cyprus Bagdad. However, with secondary and tertiary pebble crushing included in the circuit, a significant proportion of the size reduction is conducted outside the autogenous mill. The result is a very low operating cost due to power efficiency (that supports the earlier comment about the advantages and power efficiency of crushing) (I3) and zero grinding media costs. Outokumpu have developed a totally autogenous (“Outogenius”) circuit that extracts pebbles from the primary circuit for use in the second-stage pebble mill. (I4) SAG Mill plus Ball Mills This circuit (and as supplemented with pebble crushing illustrated below) has become, the workhorse of the industry in the last 20 years. The circuit is the main subject of this paper and others at the conference. Selected equipment examples are listed in Tables 1 and 4. A more complete recent listing, plus a summary of the historical development, is available in a paper by Svedala in the proceedings of the International Autogenous and Semiautogenous Grinding Technology 2001 symposium. (I5)

BALL MILL

SAG M I L L

Figure 5 SAG Mill plus ball mill SAG Mill plus Ball Mills plus Pebble Crushing MAGNETIC SEPARATOR

PEBBLE CRUSHE PRODUCT

SAG MILL OR AG MILL

BALL MILL

Figure 6 SAG mill ball mill with pebble crushing

707

Double Rotator Dry Grinding This circuit has been adopted at two gold roasting operations in Nevada, one at Newmont and the more recently at Barrick Goldstrike. (I6) The circuit was adapted fkom practice in the cement industry and combines drylng with two stages of grinding. The handling and classification circuit for the dry ground material is relatively complex consisting of airslides, bucket elevators, dynamic and static classifiers, and product recovery baghouses associated with each classifier.

PRODUCT PRIMARY CRUSHER FEED HOT GAS

DOUBLE ROTATOR GRINDING M I L L

Figure 7 Double rotator dry ball mill

REFERENCES (1) S. Morell. Large Diameter SAG Mills Need Large Diameter Ball Mills - What are the Issues? International Autogenous and Semi-autogenous Grinding Technology 2001. (2) P. Plavina and G. Clark. The Selection, Commissioning and Evaluation of a Large Diameter, Variable Speed Ball Mill at Bougainville Copper Limited. 13” CMMI Congress. (3) Rick Coleman, Setia Nugroho, Andrew Neale. Design and Start-up of the PT Freeport Indonesia No. 4 Concentrator. International Autogenous and Semi-autogenous Grinding Technology 200 1. (4) Mark Sherman. Optimisation of the Alumbrera SAG Mills. International Autogenous and Semi-autogenous Grinding Technology 2001. (5) Yvon Sylvestre, Jennifer Abols, Derek Barratt. The Benefits of Pre-Crushing at the Inmet Troilus Mine. International Autogenous and Semi-autogenous Grinding Technology 200 1. (6) Steve McGhee, John Mosher, Mark Richardson, Dean David, Rob Momson. SAG Feed Precrushing at Asarco’s Ray Concentrator: Development, Implementation and Evaluation. International Autogenous and Semi-autogenous Grinding Technology 200 1. (7) M. Nelson, W. Valery Jr., S. Morell. Performance Characteristics and Optimisation of the Fimiston (KCGM) SAG Mill Circuit. International Autogenous and Semi-autogenous Grinding Technology 1996. (8) Michael Dennis, Dr. Raj Rajmani. Evolution of the Perfect Mill Simulator. International Autogenous and Semi-autogenous Grinding Technology 2001. (9) J. L. Vanderbeek. Grinding Circuit Evolution at Chino Mines Company. International Autogenous and Semi-autogenous Grinding Technology 1996.

708

(10) Brendan Parker, Peter Rowe, Greg Lane, Steve Morell. The Decision to Opt for High Pressure Grinding Rolls for the Boddington Expansion. International Autogenous And Semiautogenous Grinding Technology 2001. (1 1) Johan J. van Heerden. Development of Autogenous Milling at Palabora. International Autogenous and Semi-autogenousGrinding Technology 1996. (1 2) Carl D. Wood Single Stage SAG Grinding Experience at Henderson. International Autogenous and Semi-autogenous Grinding Technology 1996. (13) Bruce K.Glements, Tom Bender, Bruce Apland. Grinding Circuit Evolution at Bagdad Copper Corporation. International Autogenous and Semi-autogenous Grinding Technology 1996. (14) Seppo Rantanen, Martti Lahtinen, and Warren Schumacher. Operation of “Outogenius” Type Grinding at Forrestania Nickel Mines. International Autogenous and Semi-autogenous Grinding Technology 1996. (1 5) Stuart M. Jones. Autogenous and Semiautogenous Mills 2000 Update. International Autogenous and Semi-autogenous Grinding Technology 200 1. (16) K. G. Thomas, L. Buckingham, N. Patzelt. Dry grinding at Barrick Goldstrike’s roaster facility. International Autogenous and Semi-autogenous Grinding Technology 200 1.

709

Selection of Rod Mills, Ball Mills and Regrind Mills Chester A. Rowland, Jr.

I.

INTRODUCTION

Comminution is the breaking of materials from a large size to a smaller size. In minerals processing it is generally a feed preparation stage for concentration or agglomeration processing stages. In minerals processing, the fine product production of comminution requires a large capital investment and usually is the processing stage with the maximum usage of energy and wear resistant materials. Grinding is most frequently done in tumbling drums utilizing loose grinding media, lifted by the rotation of the drum,to break the ores with combinations of impact, attrition and abrasion to produce the specified product. Grinding media can be the ore itself (autogenous grinding primary and secondary), natural or manufactured non-metallic media (pebble milling) or manufactured metallic media (steel rods, steel or iron balls). This chapter describes grinding mill design and the methods to determine the type and size of the following tumbling grinding mills. 1. Overflow Rod mills as shown in Figure 1 2. Peripheral Discharge Rod mills as shown in Figures 2 and 3 3. Compartment mills 0 Rod / Ball mill as shown in Figure 6 a Ball / Ball mill as shown in Figure 6a 4. Pebble mill as shown in Figure 6b 5. Overflow Ball mill as shown in Figure 8 6. Diaphragm (Gate Discharge) Ball mill as shown in Figure 9

11.

GENERAL MILL DESIGN

1. Liners

The interior surfaces of grinding mills that are exposed to grinding media and/or the material being ground are protected from wear and corrosion with liners made of rubber, metallic, or a combination of rubber and metallic, or non-metallic wear resistant materials. 2. Drives Economics and available energy sources at the time of plant design and mill purchase determine the type of mill drive to be used. The simplest drive used is the low speed synchronous motor with speeds in the range of 150 to 250 RPM connected to the mill pinionshaft by either an air clutch or a flexible. Grinding mills essentially draw constant power, thus are well suited for synchronous motors with power factor correction capabilities as drive motors. Approximately 120 to 130% of running torque is required to cascade the charge in these mills. The pull-in torque is about 130 to 140% with the pullout torque, to keep the motor in-step (in-phase) generally in excess of 150%. 710

When mills are started across-the-line the starting and pull-in torques result in the inrush currents exceeding 600% that can result in possibly high voltage drops. To deliver 130% starting torque to the mill the motor design must take into account the maximum anticipated voltage drop. Motor torque decreases as the decimal fraction of the voltage available squared. E.g., a motor rated 160% starting torque with a 10% system voltage drop will deliver 160% x (100 - 10)2/100= 129.6% torque to its output shaft. When it is not possible or practical to start a fully loaded synchronous motor across the line it is possible to utilize the motor’s pullout torque to start the mill. By using a clutch, between the motor and the pinion shaft on the mill, the motor is brought up to synchronous speed before the clutch is energized. If the motor has an adequate amount (175% or greater) of pullout torque the pullout torque starts the mill, without a major disruption of the electrical system. Since the energy release at initial cascade of the mill charge is an inverse function of acceleration time, a minimum acceleration time of 6 to 10 seconds or more is recommended to prevent damage to the mill or the mill foundation. Using a speed reducer between the motor and pinion shaft permits using motors having speeds in the range of 600 to 1000 FWM. In this speed range, if a power factor correction is not required, induction motors can be used; squirrel cage when there is no restriction on inrush current; slip ring when a slow start and low inrush current is required. Clutches can also be used to eliminate the starting problems with squirrel cage motors. In some areas of the world induction motors and starters are less expensive than synchronous motors at a sacrifice of motor efficiency and power factor correction. Dual drives, that is two motors and pinions driving one gear mounted on the mill, can be required for ball and autogenous mills drawing more than 3500 to 4000 horsepower (2600 to 3000 kilowatts). Single pinion drives with ratings of 6000 horsepower is a practical limit for single motor and pinion drives. Low frequency, low speed synchronous motors with the rotor mounted on the mill shell or an extension of the mill trunnion are available for autogenous and ball mills requiring 7500 horsepower and larger motors. These are known as wrap around drives with the mill shell becoming the motor shaft. This design eliminates the gear and pinion mill drive. The critical speed, of tumbling mills, is the speed at which the centrifugal force is sufficiently large to cause the small particles, next to the shell liners, to adhere to the shell liners for the complete revolution of the mill. The percent of critical speed is one of the major factors in determining the power that a grinding mill draws. Critical speed is determined from the following: (1) Nc = 42.305 /Do-’ Where: D = mill diameter inside liners specified in meters Nc = critical speed in RPM When D is specified in feet: Nc = 76.63 Peripheral speed, which is the distance that the surface of the mill shell travels in one minute, is a factor which effects liner wear and to an extent media wear, and has to be considered in mill design. It can be determined by the following either as meters per minute or as feet per minute:

71 1

Np=nDN Where: Np = Peripheral speed D = Diameter inside liners N = Mill speed in RPM To relate critical speed and peripheral speed as mill diameters increase, the average recommended speed as percent of critical speed are shown in Table 1. These are guidelines for initial plant design. Actual speeds can differ from these to suit specific ore and economic conditions that apply to a specific ore. Table 1 verage YOof Critical Speed

Mill Diameter Inside Liners Meters 0.9 1- 1.83 1.83-2.74 2.74-3.66 3.66-4.57 Greater than 4.57 111.

% of Critical Speed

12-15

Rod Mills 76-73 73-70 70-67 67-64

Ball Mills 80-78 78-75 76-74 less than 75 less than 75

75-72 72-70 70-68 68-64

78-75 75-72 73-70 70-68

ROD MILLS

When grinding to a coarse product size in the range of 80% passing 2.0 mm to 80% passing 0.5 mm (sometimes finer) rod mills are used. The feed size can be as coarse as 80% passing 20 mm and as fine as 80% passing 4 mm. A rod mill is a tumbling mill in which rods are grinding media, see Figures 1 , 2 and 3. Rod mills are usually used in wet grinding applications. For finer rod mill products such as ball mill feed, wet overflow (Figure I ) rod mills are used. For the coarser products, where a minimum amount of extreme fines is desired, center peripheral discharge rod mills (Figure 3) are used. Dry grinding in rod mills is generally not recommended. Dry materials flow poorly, which causes swelling of the rod charge leading to rod breakage and tangling. Dry grinding rod mills are used for special applications such as grinding coke breeze in iron ore sintering plants and grinding cement clinker (an energy saver but high capital cost). Dry grinding rod mills are usually designed for end peripheral discharge (Figure 2) but can be center peripheral discharge (Figure 3). Except in cases such as cement clinker, dry rod mills are inefficient in the use of power and have mechanical problems particularly rod tangling. To prevent most conditions leading to rod charge tangling, the recommended relationship of rod length to mill diameter inside liners is 1.4 to 1.6. When this ratio becomes less than 1.25 the risk of tangling increases. When planning to use rod mills the availability of quality rods has to be determined. Table 2 gives rod length to mill diameter ratios for the larger diameter rod mills. The practical limit on the length of good quality rods is 6.8 meters (20 feet). This is the length that rods will stay straight in the mill and will break into small pieces that will discharge from the mill when worn. This length is a function of rod quality and production limits established by the rod manufacturers. The mill length inside end liners measured along the surface of the shell liners should be 0.10 to 0.15 meters (4” top 6”) longer than the rods, so that the rods will fit in the length of the grinding chamber without tipping or laying across the charge. A slope as steep as possible, perpendicular

712

FIGURE 1: Overflow Rod Mill

FIGURE 2: End Peripheral Discharge Rod Mill

713

FIGURE 3: Center Peripheral Discharge Rod Mill to the shell should be used on rod mill head (end) liners to prevent the ends from protruding from the charge and being broken under by impact from other rods. The rod specifications given in Table 3 can be considered as a minimum acceptable rod quality specification. This rod specification reduces breakage, and allows rods to wear to a smaller size, both of which can reduce rod operating cost. Better quality rods are available and are recommended when using 100 mm (4”) diameter rods and/or in larger diameter rod mills. The feed ends of rods wear into a long tapered “spear-shaped“ profile, while the discharge ends wear into more of a conical shape. Approximately the middle two thirds of the rod length eventually wears into an elliptical shaped section. Small pieces of broken rods can accumulate in the mill before being discharged. The tapered wear and accumulation of broken rods reduces the bulk density of the mill charge, and thus mill power. The rod charge bulk density given in Table 4 can be used to determine the power draw of rod mill with a worn-in charge draw. Bulk density is variable, subject to care given a rod charge “culling” out broken and thin rods. Experience indicates that mill diameter also has an effect on bulk density of the worn-in charge. The larger the diameter of the rod mill the less practical “culling” of the charge becomes, thus more broken and worn rods in the charge reducing the bulk density of the charge. Rod mills normally carry a rod charge from 35 to 40% of mill volume. They can carry up to a 45% charge. The limits on charge level are: keeping the feed end trunnion open so feed will go into the mill and keeping the rod charge low enough so rods will not work into the discharge end trunnion opening, where they can tip and cause rod tangling.

714

Table 2 Rod Mill Diameter Rod Length Rod Length Mill Diameter Inside Liners L = 1.25 D L = 1.40 D

-

I

4.27 4.42 4.57 4.72 4.88

14.5 15.0 15.5 16.0

5.71 5.90 6.20

18.8 20.0

'

5.75 5.98 6.19 6.40 6.61 6.83

19.6 20.3 21.0 21.7 22.4

Table 3 Minimum Rod Specifications Grinding mill rods should be hard enough to remain straight throughout their entire life, yet they cannot be so brittle as to break up at coarse sizes. When rods are too soft, they are subject to bending in the mill. Bending causes premature breakage and tanglement of rods. Tangled rods make mill cleaning difficult and hazardous, and cause costly downtime. Rods made with the following chemical analysis is recommended: CHEMICAL ANALYSIS Carbon 0.85 to 1.03% Manganese 0.60 to 0.90% Silicon 0.15 to 0.30% Sulphur 0.05% max. Phosphorous 0.04% max. PHYSICAL REQUIREMENTS Rods should also have the following physical requirements: Rods are to be special commercial straightened 0 Rods are to be hot sawed to length when steel mill facilities Permit. If hot sawing is not possible, use an abrasive cutting Wheel or machine cut both ends to proper length. 0 All grinding mill rods should be 152 mm (6 inches) shorter in length than the working length of the rod mill.

Spout feeders normally feed rod mills as shown in Figure 4. To get the proper flow of feed into the mill a minimum head of 1.5 meters ( 5 feet) is required above the mill centerline to the bottom of the feed hopper to which the feeder is attached.

715

Table 4 Bulk Density Worn-In Rod Charges KG per Cubic Meter

Pounds per Cubic Foot

New Rods in New Charge

6247

390

Worn-In Charges Mill Diameter Meters 0.91-1.83 3-6 1.83-2.74 6-9 2.74-3.66 9 - 12 3.66-4.57 12 - 15

5847 5766 5606 5446

365 360 350 340

FIGURE 4: Spout Feeder

716

Heavy-duty single wave shell liners cast of either alloy steel (manganese steel is not recommended) or wear resistant alloyed cast iron are most frequently used in rod mills. The number of lifters to the circle is usually equal to approximately 6.6 x D in meters. For D in feet, the number of lifters is 2 x D. These liners have 65 mm (2.5”) to 90 mm (3.5”) high waves. The thickness of new liners under the waves is from 65 to 75 mm (2.5 to 3 inches). Rubber backing can be used between the liners and shell to protect the shell and heads from washing and corrosion. However, with rubber backing, care must be taken with the liner bolt, sealer and nut assembly to assure the liners will stay tight and not move in the shell. This creates leaky liner bolts and causes the boltholes in the shell to wear into an elongated shape. There are modifications such as the two-piece liner and lifter design that can be used instead of the single wave liner. Rubber shell liners have been successfully applied in the smaller diameter rod mills running at slow speeds. When using rubber liners care must be given to using good quality rods and culling broken and thin rods from the charge. Rubber liners can help reduce the level of the sounds emanating from a rod mill. End liners are generally a thick, smooth liner cast of alloy steel. Impacting from the rod charge, which has a lateral movement in the mill, can break wear resistant cast iron end liners. Rubber end liners can be used in rod mills. Rubber end liners can be damaged from the sharp ends of worn rods. Except when using rubber end liners there should be a rubber backing between the end liners and the heads (ends) of rod mills. End liners used in rod mills should be smooth with no waves or lifters as waves and lifters on end liners can disrupt the rod action and cause rod tangling. The general design trend is to have a minimum slope on end liners. Some mills less than 3.35 meters in diameter have vertical end liners to help keep the rods straight in the mill. In larger mills the slope is from 3 degrees to 7 degrees to keep the weight of the liners to a weight that can be supported in the mill. One mill manufacturer has used end liners with a 20-degree slope. Overflow rod mills can be equipped with trommels to remove broken pieces of rods and tramp oversize from the rod mill discharge. The discharge end of an overflow rod mill can be enclosed in a housing that will help contain the noise and splash coming from the mill. A large door should be provided at the end of the housing that can be opened for entry into the mill and for charging new rods into the mill. Sufficient clear space at the discharge end of the mill should be provided for charging rods to the mill to replace worn out rods. See Figure 5. The following equation, published by Fred Bond, is used to determine the power that a rod mill should draw. - 5.4 Vp) C, KW (Rod Mill) = 1.752D0,34(6.3

(3)

Where KW D VP

cs

= = = =

Kilowatts per metric tonne of rods (1 000 kg = 2 104 pounds) Mill diameter inside liners in meters Fraction in percent of mill occupied by rod charge Fraction in percent of mill critical speed

In terms of mill diameter in feet and rod charge in short tons (2000 pounds) the equation becomes: KW (Rod Mill) = 1.7

(6.3 - 5.4 V,) C,

(3a)

Table 5 lists many of the common size rod mills giving speed, loading and power data. The power is in horsepower at the mill pinion shaft. For different lengths of the same diameter, rod mill

717

power varies directly as length of the rods. For difference between new and worn liners, increase power draw by 6%, and for a worn-in rod charge, adjust for changes in the bulk density per Table 4.

FIGURE 5: Rod Mill/Ball Mill Installation; shows space for rod charging

FIGURE 6: Rod/Ball (RODPEB) Compartmented Mill

718

FIGURE 6a: Ball (COMPEB) Compartmented Mill

FIGURE 6b: Pebble Mill with Stone Lining

When wet rod milling a non-abrasive material to prepare feed for a wet grinding open circuit ball mill, and when the rod mill and ball mill have the same diameter, it is feasible to make the two mills into one two-compartment mill. The rod compartment of a two-compartment rod-ball mill, see Figure 6, is the same as an overflow rod mill. Such mills have been used to wet grind cement raw material and to grind bauxite in a caustic solution

719

Table 5 Rod Mill Power at Mill Pinionshaft (Horsepower) Rod Length

Mil 1 Diameter

(L 1

Length

M

I Ft I

0.91 3.0 1.22 1.22 4.0 1.83 1.52 5.0 2.44 1.83 6.0 3.05 2.13 7.0 3.35 2.44 8.0 3.66 12.59) 8.513.661 2.74 9.0 3.661 2.89 9.5 3.96 3.05 10.0 4.27 3.20 10.5 4.57 3.35 11.0 4.88 3.51 11.5 4.88 3.66 12.0 4.88 3.81 12.5 5.49

I I

4.42 14.5 6.10 4.57 15.0 6.10

I

I

I

I I

6 8 10 11 12 12 12 13 14 15 16 16 16 18

20 20

3.20 3.51 3.51 3.51 3.81 4.11 4.42

IL/O I Mill SDeed

I

Bulk Oensi ty

I

I

Rod Charge Weight

Inside

Rod Metric Tc nes Short Tc Charge - % Volumetric oading % Volumetric 45 35 40 RPM I%CSIFPM kg/+ Ilb/ftJ 35 I 40 I

I

10 5 1 62 21 0 69 9 428 I11:511:53 I19:4 I69:3 1457 11.5 1.44 18.7 69.0 470 11.5 1.38 17.9 67.5 470 12.5 1.41 17.4 67.6 483 13.5 1.44 16.8 67.0 493 14.5 1.47 16.2 66.4 501

I

I

I

5766 5766 5766 5766 5606 5606 5606

365 365 365 365 360 360 360

Mill Power I Dia (0)

I

1

20 0 22.8 29:O 33.2 33.0 37.7 36.0 41.1 42.7 48.8 51.5 59.0 61.4 70.1 72.5 82.8 79.7 90.7 82.7 99.8 104 119 120 137 130 148

1.27 2.9 8.89 16.8 25.6 37.4 42.5 45.5 54.9 63.8 78.9 93.5 103 112 134 154 166 190 204

1.1 2.48 7.62 14.4 22.0 32.0 36.4 39.7 47.1 56.8 67.7 79.9 87.8 91.1 115. 132 143 162

1.25 1.4 3'2 2.84 8.76 9.8 18.5 16.5 25.1 36.6 46.8 41.6 45.3 50.1 53.8 60.5 65.0 77.3 91.3 100 110 131 151 163 183 209 186

;g:!

:::

1' 1

I

I

5;l 23

l:25 6

8 0.76 2.5 6411.371 26 1.07 43.55

114 181 2751 318

122 128 1.68 194 204 1.98 2951 3iolz.zg1 341 359 2.44 369) 38812.551 416 3441 446 470 2.70

1385 1580 1715 11853

I I

I

5.5 6.5 7.5 8.0 8.35 8.85

1486 1562 3.92 12.85 1695 1783 4.07 13.35 1840 1935 4.22 13.85 19881209114.37114.35

The calculations relative to the rod mill compartment are the same as for a separate rod mill. With two compartment rod-ball mills it is necessary to charge replacement rods through the feed end of the rod compartment. This makes it more difficult to charge replacement rods to the rod compartment than to a separate rod mill. The time intervals between shut downs to add new rod is greater than for separate rod mills. Some additional power must be added to the rod compartment to compensate for the loss in power as the rod charge wears. Each manufacturer of rod mills has their own method for determining the power that rod mills draw. They all come close to the same calculated power draw as given in Table 5 . IV.

BALL MILLS

Ball mills are tumbling grinding mills in which metallic balls, conical shapes or cylindrical shapes are used as the grinding media. Most frequently the grinding media are balls made of cast steel, forged steel or cast iron.

CLASSiFlER W A F T -

PROOW1 DIICHARGE PORT

URlVtN $>

'VARICBLC SPCLO DQIVE

.

_ _ I .

FEED SWVT RZKLEB

summ ROLLER

HOUSING

WRTED AIR RIW.

-

-

LOAGING ROO (ONE OF THREE,

-

GLS PLLYLl WtARINC RING

-

nm oh$

IRLCLLILED f40M K1111 PREHEATfR OR COOLCR)

4ilTATllrS G914UIN; YABlF-

HVORAULIC LOAD'NG Crl IhDfR lOYE OF W R E E

HCGH SPCiD 5 H N T SPEEO

REGLQR

-

FIGURE 7: Roller Mill

Ball mills are most frequently used to grind products finer than 80% passing 0.5 millimeters. When dry grinding non-abrasive materials to these finer sizes ring roller mills such as shown in Figure 7 are used. Stirred energy mills have recently been used for wet grinding of minus %-inch 721

materials. High-pressure crushing roll have been developed for dry and wet coarse grinding to prepare ball mill feed. Figure 8 shows a cross section of an overflow ball mill and Figure 9 shows the cross section of a diaphragm (grate) discharge ball mill. Ball mills can be used for either wet or dry grinding. Normally either following mineral processing units or the ore itself will determine whether wet or dry grinding should be used.

FIGURE 8: Overflow Ball Mill

FIGURE 9: Diaphragm (Grate) Discharge Ball Mill, Dry Grinding Type

722

The feed to dry grinding ball mills must be dry containing less than 1% moisture by weight. There is a loss in efficiency when the feed contains sufficient moisture by weight. There is a loss in efficiency when the feed contains sufficient moisture to slow down the flow rate or cause coating of the grinding media and/or mills liners. Drying of the mill feed can be accomplished in one of the following ways: A. B.

C.

D.

Separate dryer Drying in the mill by drawing hot gases through the mill with partial or total air sweeping Combined dryer and ball mill with the dryer being the first compartment and the ball mill the second compartment. The ball mill compartment is air swept because gases from the dryer are pulled through it. Drying in the air separator where the feed goes to the air separator. Hot gases are drawn through the air separator drying the new feed. Drying also occurs in the bucket elevator used to convey the mill discharge, which is hot, and the new feed to the air separator.

Except for fully air swept mills, dry grinding ball mills are supplied with low-level discharge diaphragms, see Figure 10. With air swept and partial air swept mills, the air volume and velocity will be that required to carry the coarsest particle, desired in the ball mill product, from the mill. With closed circuit grinding the largest particles to be swept from the mill will be larger than the desired product size. Diaphragm discharge mills require a sufficient air draw, through the mill, to keep the mill under negative pressure to prevent dust from leaking from the mill around the feeder and discharge housing. Dry grinding ball mills use spout feeders, see Figure 4, with air seals. A rule-of-thumb for determining the air required for dust control is: Closed Circuit - 5.5 cubic meters per hour per horsepower of mill power. Open Circuit - 5.0 cubic meters per hour per horsepower of mill power. Being free of the limits imposed on rod mills by the rods, ball mills have more variations in length to diameter ratios, ranging from slightly less than 1:1 to some greater than 2: 1. There are no fixed rules on the proper L/D ratios to use as the ratio to use varies with the circuit used, ore type, size of the feed, the ratio of reduction and the specified fineness of the desired product. Table 6 gives some guidelines on UD ratios used in the application of ball mills.

Table 6 Ball Mill L/D Ratio - Application General Guidelines Type o f Grinding

Feed 80% Passing S i z e

1

Top

Ball Size

L/D R a t i o

Mi 11 imeter

Micrometers Wet

5.000 t o 10,000

Wet

900 t o 4.000

Wet o r Dry

Fine Feeds

- Regrind

Wet o r Dry

Fine Feeds

- Open C i r c u i t

Dry

5,000 t o 10,000

0r y

900 t o 4,000

60 t o 90

2.5 t o 3.5

1:l t o 1.25:l

40 t o 50

1.8 t o 2.0

1.25:l t o 1.75:l

20 t o 30

20 t o 50

314 t o 1-114 ,314 t o 2

1.5:l t o 2.5:l 2.O:l t o 3.0:l

60 t o 90

2.5 t o 3.5

1.3:l t o 2:l

40 t o 50

1.8 t o 2.0

1.5:l to 2 : l

Depending upon the size of the largest balls in the ball charge, adverse ball segregation, that is large balls accumulating at the discharge end of the mill and small balls at the feed end, can occur as the mill L/D ratio becomes larger. This begins to occur as the size of the largest balls, in the ball charge, become larger than 65 mm (2.5”).

723

G R A T E SECTION (OR WEARING PLATE)

CONE R I N G LINER

\

DISCHARGE

w1 NG

(LIFTERS:

\

F ‘ L L E R RING

DISCHARGE HEAD

FIGURE 10: Discharge Diaphragm Grinding balls can be made of forged or cast steel or cast iron. The quality depends upon the source of supply. While not always true, frequently the better quality balls are forged steel balls. The bulk density of steel balls varies between 280 and 300 pounds per cubic foot with cast iron balls having a bulk density in the area of 170 pounds per cubic foot. Generally balls are spherical but they can be in various cylindrical, conical or irregular shapes often simulating worn balls. Balls vary considerably in hardness with soft balls having a Brine11 hardness in the range of 350 to 450, while hard balls have a hardness in excess of 700. A rule-of-thumb subject to argument is: “the harder the ball the better its life, provided it is not too brittle and breaks or becomes too highly polished and too smooth to nip the material being ground”. Local economics and the specific need for grinding are the deciding factors in selecting the size and type of balls used. The balls giving the lowest operating cost and best grinding performance are generally selected. This need not be the ones giving the lowest wear rate, but can be a compromise between the two extremes in wear rates. Balls should be solid with a reasonable uniform hardness through the cross section of the ball. They should wear in a relatively uniform pattern. An indicator of good ball wear is when the worn ball discharging from the mill are approximately 16 mm (5/8”) or smaller in size and have the shape of a polygon with from 8 to 12 surfaces, which can be slightly concaved. Some broken balls can be found in the worn balls discharged from a ball mill. Broken balls can be found in the

724

shape of circular discs, half rounds and crescents. Broken balls with holes in them, indicates poor quality balls with sand inclusions, and/or blowholes and/or hollow centers. For determining the power that ball mills draw forged steel balls have a bulk weight of 4,806 kilograms per cubic meter, (300 pounds per cubic foot), cast steel balls have a bulk weight of 4,646 kilograms per cubic meter (290 pounds per cubic foot) and cast iron balls have a bulk weight of 4,6 15 kilograms per cubic meter (260 pounds per cubic foot). Ball mills normally carry a ball charge occupying from 40 to 45% of the mill volume, but can carry up to a 50% or slightly higher charge. Figure 11 shows the relationship of mill power and volumetric loading. For plant capacity and design purposes, ball mills are frequently selected based upon carrying a 40% ball charge with the mills and drives designed for the power necessary to carry higher charges if required.

20

25

30

35

40

45

50

55

60

% OF MILL VOLUME OCCUPIED BY BALL OR PEBBLE CHARGE

FIGURE 11: Grinding Mill Power vs Loading Dry grinding ball mills normally carry from 35 to 40% ball charge with air swept mills carrying lower charges nearer to 30% of the mill volume. This is to provide mill volume needed for air sweeping and dust control. Plant design, capital costs and operating costs influence the selection of feeders for ball mills. Spout feeders, as shown in Figure 4, are the most common style of feeders to feed open and closed circuit, wet and dry grinding ball mills. When wet grinding ball mills, with ball charges occupying 45 to 50% of the mill volume, are operated in closed circuit using rake or screw classifiers, double scoop feeders, as shown in Figure 12, are used to feed the mill. With open circuit mills where there is insufficient headroom for spout feeders, drum feeders, as shown in Figure 13, are used. Spout feeders used with wet grinding mills allow for the use of classifiers where the classifier underflow flows by gravity into a hopper to which the mill spout feeder is attached. The classifier must be installed above the mill at an elevation so the coarse material discharge from the classifier has sufficient head to flow, by gravity, into the mill spout feeder. This requires high pumping heads to pump the mill discharge to the classifier. Figure 14 shows an installation of closed circuited ball mills with a double scoop feeder where the cyclone classifiers are installed at above the horizontal centerline of the mills. Double scoop feeders consume from 25 to 30 kilowatts of power (30 to 40 horsepower). This arrangement

725

reduces the pumping head and pumping power, which must be balanced against the high maintenance and capital costs for the double scoop feeders.

FIGURE 12: Single Scoop Feeder with Ball Charging Drum (Double Scoop is available)

FIGURE 13: Drum Feeder

726

FIGURE 14: Single Stage Ball Mill Installation Make-up grinding balls are fed to the mill, as required, with the material being ground, through the mill feeder. Single compartment mills are not stopped for charging balls. Balls are fed to ball mills with spout feeders through the feeder. Balls should not be fed into a scoop feeder housing because of jamming of balls between the scoop feeder housing and the scoop feeder which can cause major damage to the feeder and housing. Scoop feeders usually have a pipe for feeding balls directly into the feed trunnion of the mill or they are supplied with a ball-charging drum for feeding balls into the discharge from the scoop feeder. There are many different designs and styles of ball mill liners. Operating costs and grinding mill performance determine the design and material to use. The initial set of liners is rarely the final design and material used. Based upon individual experience, mill superintendents develop preferences for liner designs. The following is a guide for selecting the initial set of liners. A. For 65 mm (2.5”) diameter and smaller top size balls and cast steel shell liners use double wave liners with the number of lifters to the circle 13.1D, when the mill diameter is specified in meters and 13.1D/3.3, when the diameter is specified in feet. The wave height from 40 to 65 mm (1-5/8 to 2-1/2”), above the liner thickness, which is from 40 to 50 mm (1-5/8 to 2”).

Rubber liners of the integral molded design, made to the same design as double wave cast metal liners are well suited to use in ball mills with small balls. They can reduce operating costs and the noise levels in ore processing plants. Rubber liners should not be used in dry grinding ball mills where the heat from grinding exceeds the temperature limits for using rubber. Single wave liners can also be used in ball mills. When single wave liners of either the integral wave design or the replaceable lifter bar design, made from either metal or rubber, are used, the number of lifters should be 6.6D+2 when the diameter is specified in meters or 2D+2 if the diameter is specified in feet. When the largest ball size is less than 65 mm (2.5”) and the mill speed is less than 72% of critical speed, liners cast of alloyed wear resistant cast iron can be used.

727

Next Page

The use of double wave liners, particularly when using 50 mm (2”), diameter or larger balls, may show a loss of about 5% on the mill power until the waves wear in and the balls can nest between the lifters. At times, ball mill shell steel, particularly cast steel double wave liners wear with circumferential grooves. When this occurs in indicates slipping of the charge and warns of accelerated wear. A reduction in the mill power draw when the liners are near the end of their life indicates slipping of the ball charge. B. Single wave liners are recommended for all mills using larger size balls 65 mm (2.5”) and larger. The number of waves or lifters to the circle is 6.6D+2 when the diameter is specified in meters and 2D+2 when the diameter is specified in feet. The liners are from 50 to 65 mm (2 to 2.5”) thick with the waves or lifter from 65 to 70 mm (2.5 to 3”) above the liners. There could be up to a 10% loss in power draw when using rubber liners. Rubber liners are not recommended when the largest balls are larger than 75 mm (3”). Because of the impacting from the large balls, single wave liners for ball mills are usually made of alloyed steel or special wear and impact resistant cast irons. Because of the difficulty of balancing growth and wear with work hardening, manganese steel is not recommended for grinding mill liners. There are cases where double wave liners are used in ball mills with large balls as replacements for single wave liners. This requires a study of wear patterns, mill power draw, mill capacity and operating costs.

C. Classifying liners have been used in ball mills to put the larger balls at the feed end and the smaller balls at the discharge end of the ball mill. Spiral shell liners, with an advanced spiral, such as shown in Figure 9, have been used to obtain the proper segregation of the ball in ball mills. The square mill liner also known as a classifying liner gives a square configuration to the inside of a ball mill. There are a series of off setting circumferential rows, which retard the movement of the ball and allow a better material flow and filling of the ball charge. There have been successes and failures with this type of ball classifying lining. It does reduce mill volume, which causes a reduction in the power drawn by the mill. In some cases a reduction in power consumed per ton of ore ground occurred, improving grinding efficiency. In other cases there has been a corresponding reduction in capacity with the reduction in power drawn by the mill.

D. End liners for ball mills conform to the slope of the mill head and can be made of rubber, alloyed cast steel or wear resistant cast iron. To prevent racing and excessive wear, end liners, for ball mills, are furnished with integral radial ribs or with replacement lifters or both. E. When a grate discharger is used, see Figure 10, the grates and wear plates normally are perpendicular to the mill axis while the discharge pans conform to the slope of the mill head. The grates and wear plates are normally made of alloyed wear resistant cast steel or rubber. They are ribbed to prevent racing and excessive wear. The discharge arms and pans are generally made of wear resistant cast iron, rubber or wear resistant fabricated steel. Slot plugging can be a problem in grate discharge mills. Whether the grates are made of metal or rubber the slots should have ample relief tapered toward the discharge side. The angle of the relief normally is from 7 to 10 degrees, 3.5 to 5 degrees on each side of the slot. Metal grates often have a small lead-in pocket or recess that can fill in with peened metal rather than have the slot be partially shut with peened metal. With the proper combination of metal internals and rubber surfaces, rubber grates have the flexibility that tends to make them self-cleaning and yet strong enough not to fail due to

728

Selection and Sizing of Autogenous and Semi-Autogenous Mills Derek Barran’ and Mark Sherman’

ABSTRACT The technology for sizing autogenous and semi-autogenous grinding mills has evolved over the last twenty years from the use of basic equations and the determination of specific power requirements to an understanding of the effects of feed preparation and size analyses, dynamics in the mill charge, the flow of pulp through the mill, the use of simulation programs, and grinding efficiency studies. This paper reviews the traditional methods of selecting and sizing autogenous and semiautogenous mills and comments on some of the recent developments that can in€luence the determination of the applied power, such as mill shell liner design and feed preparation.

INTRODUCTION Since publication of the precedent chapters on aubgenous and semi-autogenous grinding @or and Bassarear 1982, Turner 1982), some 200 primary mills (mostly semi-aubgenous mills) larger than 8.535 m (22 ft) diameter, and including the world’s largest 12.192 m (40 ft) diameter mill, have been installed world-wide (Jones 2001). Until 1996, the largest mill was 10.973 m (36 ft) in diameter with a 13425 kW (18,000 hp) gearless drive. The largest gearless drive to date is 19395 kW (26,000 hp) on the 40 ft dia mill at Cadia Hill. The first gearless drives of this type were odered by Codelco - Chuquicamata in 1986 for two 32 ft dia mills, 8205 kW (1 1,000 hp each), followed by Codelco - El Teniente with an 11190 kW (15,000 hp) drive on a 36 ft dia. mill ordered in 1987. This group of mills are grinding ores of copper, gold, nickel, lead, zinc, silver, platinum, palladium, molybdenum, iron,diamonds, aluminum, and limestone. This expanded use of autogenous (AG) and semi-autogenous (SAG) technology begs the question as to whether it is “selected” in comparison to “old technology” (multi-stage crushing, rod mills/ball mills or single-stageball milling). There have been recent (2000-2001) examples of new projects or plant expansions for which AG/SAG grinding has not been selected. Instead, the multi-stage crushing and single-stage ball milling option has been favoured. While these projects are at the basic engineering stage, the largest sizes of crushers, screens and ball mills and, in one case, high pressure grinding rolls, HPGR (Parker et alia 2001) are being utilized. Whereas current perceived “limitati~ns’~ with respect to the manufhctwe of SAG mill designs and sizes larger than 40 ft dia and ball mills larger than 24 ft dia or 26 ft dia have in some cases governed this rejection, the capital cost and operating cost of the multi-stage crushing, screening and ball milling operation also enters into the , circuits can be cost competitive, and can be especially so equation. In certain c i r c ~ c e sthese if the AG/SAG option is not very powerefficient, even after all the necessary OptimiZation steps have been taken. In one case (as yet unpublished), the capital cost of an additional primary crusher, required to ensure the specified feed size distribution (as determined by pilot plant testwork) to a pre-crushing and screening circuit ahead of SAG milling, was sufficient for that SAG milling expansion to be rejected in favour of the largest “old technology” machines.

’ DJB Consultants,hc., North Vancouver,B.C.,Canada

* Sherman and Associates, Pty Ltd., Canberra ACT,Australia

755

The various components of AGISAG technology are now better understood and risk assessment, with respect to its use, can be minimized with confidence provided certain basic principles are adhered to, namely: 0

Definition of the applied net power which will cover the range of operating conditions that will be encountered during the time required to recover the capital investment of the project at an acceptable rate of return on that investment (payback period), and to continue to generate an acceptable rate of return on mill throughput for the foreseeable life-of-mine, even if a planned plant expansion is necessary. Success will depend on the degree to which the economic cycle, or more probably, the gradual decreasing trend in real commodity prices coupled with escalating operating costs affects ongoing confirmation of the mineable resource and ore grades. However, from an engineering aspect, the predicted range of specific power consumption ( k W t ) and power sharing between primary and secondary grinding constitutes the design basis for the project and both depend entirely on a thorough understanding and interpretationof: Project geology, geological model, different lithologies, mineralizations, and alteration types, mineable ore zones - Geotechnical parameters: point load index (PLI), uncontined compressive strength (UCS), rock quality determination @QD), and fracture frequency (FF) throughout the mineable resource - Mine blasting practice - Comminution testwork on predominant ore types in the mine production plan, whether by bench-scale tests, pilot plant tests, or a combination of both, coupled with an assessment of the variability of ore grindability - Assessment of autogenous grinding vs semi-autogenous grinding Methodologies for scaling up testwork results. Definition of the primary mill size and aspect ratio which will ensure that the required throughput range can be processed and the required power CM be drawn ( k W t x t/h) over a defined range of ore hardnesses, mill speeds, and other operating conditions: Primarycrushersetting - The necessity for pre-crushing part or all of the primary mill feed - The necessity for pebble crushing Ball charge volume and top ball size - Total mill charge volume - Assessment of voids between ore/balls and pulp - Pulpdensity - Discharge classificationtype(s) - Mill charge size distribution - Circulating load variation - Shell liner and lifter geometry and design - Grate open area requirement Pulp discharger design. Application of known engineering and manufacturing procedures and practices with respect to ensuring the structural integrity and operability of both the mill and motor at the most economical cost without compromising standards, quality assurance, quality control and inspection, and engineering co-ordination between the mill supplier, motor supplier, and engineer-of-record with third party review.

-

-

-

-

0

This paper deals particularly with a summary of the types of testwork required for assessing AGISAG grinding and ore variability, methodologies for scaling up testwork results, definition of the applied net power, and criteria required for definition of the primary mill size and aspect ratio.

756

In assessing the applied net power, the question of a contingency often arises and, while its impact is not as noticeable for smaller AG/SAG mills, the principles govemhg its application are the same and have to be quantified for the larger mills, namely: 0

0

0

0

0

Whether an AG mill is to be converted to a SAG mill Definition of the maximum opemting ball charge volume in combination with the total mill charge volume to give the maximum operating mill charge density Definition of the maximum operating mill speed for power-efficient grinding of harder ores Definition of the optimum shell linermer design for pomr-efficient grinding of harder ores using modem design and simulation concepts Definition of the required grate open area and pulp discharger capacity and their design to efficiently discharge the anticipated maximum flow of pulp and pebbles. These parameters, particularly grate open area, can determine the optimum mill diameter.

Therefore, the applied contingency is made up of a number of intrinsic components and it does not necessarily have a predetermined value. Whereas a lot of dry autogenous grinding mills have been converted to wet semi-autogenous grinding, this paper concentrates on wet grinding. It comments briefly on the sizing of secondary mills, whether they are ball mills or pebble mills.

TESTWORK AND SCALE-UP Testwork methodology, for both bench-scale and pilot plant, is described in the relevant chapters elsewhere in this book. Requirements for the determinationof the applied net power to the primary AG or SAG mills for the purposes of sizing these mills are dictated by the results from such testwork and the quality of thattestwork. This paper discusses these requirements in the context of the basic engine&g phase and issue of technical specitications preparatory to equipment pmhase and detailed engineering. These requirements can be subdivided into: 0 0

Those benefiting from a pilot plant program Those for which a pilot plant program was not possible or practical.

The basic parameters which are used to calculate the applied power to the primary mills are: 0

0

0

The size analysis of the mill feed, whether by primary crushing alone or primary crushing supplemented by some form of pre-crushing The specific power consumption, k W t , or range thereof for different ore composites, lithologies, mineraliz.ations, alteration types, mineable ore zones, or operating conditions: ball charge volume/total mill charge volume combinations, top ball size, mill speeds, pebble crushing, and pulp densities The size analysis of the transfer product from primary grinding as new feed to secondary grinding.

In some methodologies without pilot plant testwork it is quite common to calculate the total specific power consumption for primary and secondary grinding first, then select a transfer product size (80% passing TW, microns) and lastly, calculate the split in specific power consumption between primary and secondary grinding. At least one of these methodologies can be used to incorporate the results of pilot plant testwork on primary grinding and predict the specific power consumption for secondary grinding. Contingencies can be applied either to primary grinding, secondary grinding, or both, for instance, a contingency could be applied to primary grinding k W t to account for production of a fmer transfer size than sampled in a pilot plant, as well as a contingency applied to secondary grinding k W t to account for a coarser feed (transfer

757

size). This approach is a common one for porphyry copper ores in which there is usually a contrast between softer (weathered or secondary mineralized zones) and harder (silicified or primary minzones) ore types, as well as differential grinding phenomena in ball milling (e.g., accumulation of harder porphyritic feldspar grains and variable circulating loads), as distinct fiom normal variations in new feed rate. Pilot Plant Testwork Basis A pilot plant test program will have been run either in an accredited facility or in a custom-built plant that is located in a remote exploration camp. A composite bulk sample, or a number of samples, will have been selected for testwork. For the principal bulk sample, tests will have been mu to stable operating conditions, recorded, and sampled for selected circuit types on the recommendationsof the supervising engineer to the project development team.These basic circuit types, from which to choose, are: Single-stage or two-stage circuit to finalproduct sizing Autogenous, with and without pebble crushing Semi-autogenous,with and without pebble crushing F’re-crushing of AG or SAG test feed, this is particularly important for application to harder ores in low aspect primary mills with limitations of feed sizing, or in high aspect mills if perceived limitations on mill size and c o d o n have been reached, e.g., 40 ft dia. currently 100% recycle of crushed pebbles to the primary mill feed, or whole/parhl advancement to secondary grinding, use of high pressure grinding rob (HPGR) in this context secondary grhdiq as pebble milling, preferably in closed4cuit with a cyclone Secondary grinding as ball milling, although this is optional in a pilot plant test program if it is decided to size the ball mills by power-based methods using Bond work indices. If crushed pebbles are advanced to secondary grinding, then it should be piloted. A pilotscale ball mill should be closed with a cyclone to establish circulating loads, size analyses of streams, pulp densities for optimum operation, and power efficiency. Following tests with autogenous grinding, which may or may not be applicable depending upon ore competency, semi-autogenous grinding tests will have been run to establish the followingparameters for primary milling: Optimum ball charge volume (expressed as the volumetric percentage of the total mill volume % v/v) and top ball size, most pilot mills are limited to a 12%v/v ball charge Optimum sizing for new feed, most pilot mills are l i i t e d to a top size of 200-

(8-inche~) optimum classitication method, whether by a coarse screen (9 mm to 13 mm), medium screen (5 mm to 9 mm), fine screen (400 mimns to 5 mm), or by a combination of screen and cyclone Optimum pulp density Optimummillspeed Stable opedon by feeding weighed aliquots of defined size W o n s , e.g., +125 mm, -125 mm +50 mm, and -50 mm; this is to ensure blending, avoidauce of segregation, and regularity in feeding the appropriate proportions of came, medium, and fine sizes, and controlling feed rate to a pre-set load cell value for stable operation, which is based on the ball charge volume, top ball size, and the optimum total volumetric fUkg of the pilot mill Optimum sizing of classified product, screen undersize or cyclone overflow Optimum specific power consumption, k W t .

758

The graph in Figure 1 is typical of the relationship between specific power consumption and primary circuit product size for a hard gold ore in diorite (Barratt et alia 1999). It covers a range fi-om: Autogenous with pebble crushing (ABC),Test Nos. 1A and 1B (slightly coarser feed) Semi-autogenous@ 8% v/v ball charge, Test 6 Semi-autogenous @ 8% v/v ball charge with pebble crushing (SABC),Test Nos. 2A and 2B (coarser feed) Semi-autogenous@ 12% v/v ball charge with pebble crushing (SABC),Test Nos. 4A (i) and 4A (ii) Semi-autogenous@ 8% v/v ball charge with pebble crushing and with partial recycle of cyclone underflow, Test 8A (20% recycle) and Test 8B (40% recycle).

0

0 0

0

0

8

F'igure 1 Primary pilot mill: specified power consumption vs primary circuit transfer product sizing (Barratt et alia 1999) The benefits of pebble crushing were considerable in the case of autogenous grinding (not shown at 3 2 . 6 k W t to Tm= 114microns), and semi-autogenous grinding, (18.7kWb/t), to 13.8/14.4 kWh/t for ABC and 11.4111.6 k W t for SABC. Partial recycle of cyclone underflow produced a finer product at a higher specific power consumption, particularly as the percent recycle increased above 20%. In the interest of Ininimisr.;.~ the specific power consumption and project risk, the SABC option was selected with a coarser product as feed to secondary grinding, Test 2A, making an adjustment for a coarser feed, and adding a contingency of 5% to give 12.13 k W t and Tm= 1,573 microns. Such contingency could be higher, depending upon the quality of the testwork (stability, feed control, interruptions, power measurement, etc.). In this case, a ball charge volume of 12% v/v and a total mill charge volume of 26% v/v were selected for design criteria at W/Wand an Operating speed of 76% C.S. 72%

759

Pilot-scale testwork for secondary grinding was not conducted for this project. Instead, the reduced recovery method (or “phantom cyclone”) was used to establish the specific power consumption for secondary ball milling (Barratt 1989). This method uses the reduced recovery (efficiency factors) for cyclone operation and establishes a size analysis of feed for secondary grinding after segregation of finished product (e.g., 80% passing size as flotation feed) h m a primary mill test product. The rationale for this approach is recognition of the bi-modal size distribution of a primary mill circuit product, particularly in the case of SAG milling, which departs h m the more natural straight-line slope that prevails in ball milling circuit feed and products. The specific power consumption ( k W t of original new feed to primary grinding) is calculated using a power-based method, in this case “GRINDPOWER” (developed by Flwr Daniel Wright Ltd.), Bond work indices for rod milling and ball milling, and the calspecific power consumption for an imaginary single-stage ball milling circuit (SSBM) for reference (Barratt 1989) which includes an applied contingency. The results h m these calculations are shown graphically in Figure 2 for the pilot test products in Figure 1. Design criteria for secondary ball milling was selected for the coarsest circuit product as feed at Tso = 1,573 microns, 12.14 kWh/t to 80% passing 75 microns, including an applied contingency of 10% in the SSBM calculation.

12.00

7m

I

t

Figure 2 Secondary milk calculated specific power consumptionvs new feed sizing (from pilot mill primary circuit product Barratt et alia 1999)

For any project, it is important to recognize the dif€erences between industrial-scaleand pilotscale mill circuit operation on the size distribution of the primary circuit product. The transfer size (Tso), top size, and the percentage of final product in this stream are all af€ected by the screen aperture and conditions in the primary mill, more particularly in SAG milling with variations in

760

ball charge volume, total mill charge volume, top ball size, and mill speed. The limitations that are prevalent in pilot milling are usually logistic: maximum ball charge volume, pebble port size, discharge screen aperture vs successhl pumping of screen undersize (which is the enough for transfer to a secondary grinding circuit), top size in the mill feed, and fixed mill speed. All of these limitations can be removed in industrial-scale operation and usually are (e.g., 15% v/v balls vs 12% v/v; 19 mm trommel screen aperture vs 9 or 12 mm pilot screen aperture). Generally, coarser apertures for grates and screens will naturally produce a coarser product, whereas ball charge size and composition in the larger mill can sometimes compensate for that.The length: diameter ratio in the larger mill is usually higher than in the pilot mill and, coupled with the vastly higher population of balls, can contribute to a finer product AU of these issues create a balance so that, invariably, the transfer size is not too different from that produced in the pilot mill, especially if the pilot screen aperture (e.g., 12 mm) is not constrained by the logististics of pumping the undersize to a secondary grinding circuit. Another factor which could influence scale-up from pilot plant test results is the comparison of impact forces and residence time in a pilot mill with those in a full-scale production mill; i.e., "Can the specific power consumption which is derived from pilot plant testwork be relied upon for scale-up in the same manner as that calculated for secondary ball milling using Bond work indices and Rowland's efficiency factors (Rowland 1982)T' The answer, Surprisingly, is yes in spite of the fact that the forces involved a~ proportional to mill diameter and the residence time is proportional to the differences in D L ratio and inversely proportional to the ratio of the square root of the diameters. The production mill will generally see a coarser top size in the feed and have a shorter residence time than the pilot mill, so the comparison has to be based on the rate of breakage vs applied energy in different size classes. This, of course, is the principle that forms the basis of the JK Tech comminution simulations (JK SimMet). Scale-up from pebble milling tests follows the same principles as those for primary milling; i.e., accepting the measured specific power consumption to the r e q W product size distribution and adding a designated contingency. Again, this contingency is dependent upon the stability of the tests and their duration, as well as the stability of pebble consumption rates to satisfy design criteria.

BenchSeale Testwork Basis Prediction of the specific power consumption for primary and secondary g c i d q from benchscale testwork is, in many respects, still an art and is dependent upon the methodologies that are used. In the context of selecting and sizing semi-autogenous mills for basic engineering and equipment purchase, power-based models are commonly used in situations for which pilot plant testwork is not possible or practical. In the absence of pilot plant testwork, it is the authors' opinion that power-based models should only be used to predict specific power consumption for autogenous grinding if they can be supported by a database of criteria which is specific to rock strength and hown commercial autogenous grinding operations. Such a database would include results from the autogenous media competency test (AMCT), which is available from Amdel (Siddall 1996), analysis of rock strength data (unconfined compressive strength, either measured or scaled up h m point load indices, coupled with hcture frequency), breakage rates by size class fiom JK Tech drop weight tests and simulations, and Bond low energy impact work indices by size class as part of the AMCT. The database would make a distinction between operations with lower mill speed, high aspect mills which favour impact and those with higher mill speed, low aspect mills which grind more by abrasion (Powell 2002). For semi-autogenous grinding, the Bond work indices for low energy impact crushhg (Wk), rod milling (Wiw), and ball milling (WiBM)can be used in power-based models to predict the specific power consumption for primary grinding without direct reference to other specific operations. Testwork results would be obtained in the normal manner as input to the model in use. Rock strength data and JK Tech drop weight test data would be obtained specifically for the sample of interest

761

Confidence in specific power consumptions which are predicted by power-based models depends upon their precision which, in turn,depends upon the basis of the underlying equations and the database that supports them. Their precision carries more importance in respect of the capital cost of large-scale projects and the size of the applied contingency than it does for smaller projects. Power-based models are offered by Orway Mineral ~ n s u l ~ t Minproc, s, and Fluor (“GRINDPOWER”). “GRINDPOWER” is considered by many to be the most versatile in view of its mill sizing capabiities for both autogenous and semi-autogenous grinding, pebble milling, and secondary ball milling, the ability to input pilot plant results, and to expeditiouslytest sensitivities of mill operating parametem for both primary and secondary grinding, changes in ore type, maximum power capability, and the power split between primary and secondary grinding (Matthews and Barratt 1991). Its “CHARGEPOWER” subprogram can produce a series of spread sheets,each of which details net power draw across a range of ball charge volumes and total mill charge volumes for a paaicular mill speed, and which in aggregate covers a range of mill speeds. This capability becomes useful when specific power consumptions for diffkrent ore types (lithologies, minedkations, alteration types, and mineable ore zones) are converted to applied net power draws for a particular mill size (see “tent” diagrams in Figure Nos. 7 to 11).

20.0 18.0 16.0

14.0 n

.g 12.0

i Crusher Rod Mill Ball Mill

i,

v

10.0

X

a,

2 -

8.0

6.0 4.0

2.0 0.0 1

2

3

4

5

6

7-10

Year

Figure 3 Relationship between Bond work indices for design of the Batu Hijau Project (Barratt et alia 1996)

762

The effects of ore variability (grindability) on the specific power consumptions for primary and secondary grinding can also be assessed using power-based models. "GRINDPOWER" was used to design the 120,000 tpd plant for Batu Hijau (Barratt et alia 1996, MacLaren et alia 2001) on the basis of testing 105 composites for the three Bond work indices (Wii, Wim, and WiBM) with a contingency for secondary ball milling to accouIlt for the effects of pebble crushing. These Bond work indices, composited by mine production year fiom weighted bench composites, are illustrated in Figure 3 in which the relationship between the higher values of Wim vis B vis Wi, and WiBM is fairly consistent. Pebble crushing was included in grinding circuit design as a direct result of Wi,< Wim > WiBM.Subsequent testwork on a new set of infill composites has shown a correlation of 0.985 between new feed rate t/h and Wim at constant power draw in the primary and secondary mills with a variable grind to flotation. Such a good correlation is specific to Batu Hijau and the opportunity to test its efficacy on other projects has not yet arisen. JK SimMet simulations were run on a few macro lithologicdalteration composites to support the circuit design. Pilot plant testing for AG/SAG grinding on this project was not practical due to environmental and logistical reasons. The Minnovex SPI test (SAG power index) is now widely used to test for ore variability in SAG circuits and to predict mill throughput based on a simple test procedure (Kosick and Bennett 1999). The SPI represents the specific power consumption ( k W t ) required to produce an 80% passing 10 mesh (1,680 microns) product. Values of SPI can be obtained from drill core and fitted into the geological model. Mill throughput can be predicted from mill power draws, either for the circuit as a whole, or if the SAG mill is limiting, or if the ball mill is liiting. The program has been extended to include economic criteria, CEET (Kosick et alia 2001). The SPI test and its associated software is a powerful tool that, like JK SimMet simulations, can complement pilot plant testwork and the more accurate power-based models for mill sizing. The net result from a testwork program as input to design criteria and mill sizing will include a range of applied net power draws based on: Design values for specific power consumption for primary grinding and operating criteria in terms of ball charge volume, top ball size, total mill charge volume, mill speed, ore specific gravity, mill charge density, pulp density, feed sizing, and product sizing Design values of specific power consumption for secondary grinding, ball charge volume, mill speed, pulp density, and product sizing, and (for pebble millii) total mill charge volume, mill charge density, pulp density, and pebble consumption Criteria, similar to that preceding, which represents a practical range for variation in ore grindability for primary milling and, separately, secondary grinding, both to indicate maximum and minimumvalues A decision to use fixed speed or variable speed mills.

PRIMARY MILL SIZING AND ASPECT RATIO Given that the testwork program and process simulation exercises have defined the necessity for a single-stage or, more commonly, a two-stage grinding circuit and ranges for the applied net power draw in the primary mill, it is necessary for the grindins circuit designer to be in possession of a method for relating mill size to mill power draw for the range of operating conditions which are detailed in the mill purchase specification. Traditionally, mill suppliers have provided data for net mill power draw in tabular and graphical form for a mill size, which is: Either specified by the engineer-of-record to a specified range of operating conditions, mill speeds and power draws that are scaled up k m testwork Or is selected by the prospective supplier to satisfy the same range of specified operating conditions, mill speeds,and power draws.

763

More often than not, the predicted power draws from d i f f m t mill suppliers do not agree with those of the engineer-of-record for a specified mill size and, alternatively, the mill size selected by a supplier does not agree with that calculated by an engineer-of-record for agreed power draws. These disagreements have arisen due to differences in approach that the various vendors and design parties have adopted when surveying various mills and interpreting their power draw, resulting in differences in calculating e x p t e d power draws. If one of the vendors has information on a similar ore type in their database, then this information is very beneficial to the design team. However the design team could also be inheriting the vendor’s design prejudices as well, so care needs to be taken when assessing vendor supplied power draw information The answer usually lies in a commercial compromise with respect to both mill size and motor power, usually in comparison with an installed list, which may result in a limitation on the operability of the mill. Such a limitation could take the form, with a gearless drive for insta~ce,of: Selection of a mill with a diameter which is inadequate with respect to drawing power but, more importantly, also with respect to passing pulp or pebbles; i.e., having mflicient grate open area 0 Selection of a mill with other than the optimal aspect ratio; i.e., the effective grinding length is either too long or too short to be compatible with the selected range of operating conditions. This length obviously affects the calculation of the effective mill volume, but not nearly so much as does the mill diameter 0 Selection of a motor with less than the required torque output over the desired mill speed range; i.e., selection of a motor with other than the optimal base, or rated design speed, will leave the motor with insufficienttorque output at mill speeds that are higher than the rateddesign speed, even if the rated design power of the motor is adequate. This aspect of milYmotor selection is critical for grinding harder, more competent ores at the desired feed rates. These points will be discussed and illustrated in more detail later in the paper. 0

Power Equations The question of which comes first in the sequence of mill design, applied power or mill size, can best be approached by referring to some basic (but similar) power equations. Traditionally, net mill power is expressed as

kW

=

Where: N

=

z

=

Where:

................................................................................ I11 2n.N.z millspeed,rpm torque developed by the mill charge as its centre of gravity changes position b m rest W.E.Sin9 3,403.5

w

=

mill charge weight, kg

E

=

9

=

radius to the mill charge centre of gravity, m angleof E tothevertical

:.

kW

=

2 n . N .W .E . Sin0 3,403.5

764

.............................................................

121

But this only expresses power as a function of the mill charge weight in a mill of a given size. It does not, as Morrell points out (Morrell 1992, 1996), take into account the contributions to power draw made by: The imprecise position of the centre of gravity of the mill charge; i.e., it is never known, physically Attrition Abrasion Friction and rotation within the mill charge Heat and sound emanating h m the millcharge and within the mill itself Wind loss Bearing fiction Size analysis of the millcharge (and new feed) Shape of the mill charge.

In Morrell’s methodology, and as output from JK SimMet simulations: kW(gr0ss)

=

kW (no-load)

+ K .D2’ .Le . p, .a. 6

.........................

- Effective mill diameter (inside shell liners), m - Effective grinding length expressed as the length of an

Where: D Le

r31

equivalent mill cylinderto include the volume of the conical ends, m density ofthe total mill charge, t/m3

Non-linearfunctions of total mill filling and mill speed, respectively constant selected h m a d2ltabase to reflect losses associated with movement of the mill charge 2.474 . . .Nc) 0.861

...........................................

[41

fiaction of critical speed So far, from these equations, mill power is derived from a mill size and its contents. Such mill power is very dependent upon the value of “K”, which is selected h m JK Tech’s database of operating mills. It is conceivable that the mill size could be predicted with confidence h m a calculated power value (based on grinding testwork) if a reliable value of “K” was available to the user. Mill sizing Mill suppliers use similar but proprietary equations that estimate net power draw, or in reverse estimate mill size. Much depends upon the exponent which is applied to the “Diameter, D and also which “Diameter“ is used: diameter at the mill shell plate or, more usually, an effective mill diameter which takes into 8ccounf the thickness of the shell liner plates plus the thickness of the shell lifters averaged (or “smoothed”) over the mill circumference, and for which the exponent 2.5 is developed h m applied mathematics and is unalterable in modem technology. Also important in this context are the equations that are used to calculate the mill charge density and whether changes in mill speed are considered to be directly proportional and independent of total mill charge volume. For dry grinding, the following formula has been proposed (Tuner 1982): kW (Net)

=

2.208 . p . D 2.5. L .Nc

765

.....................................................

151

Note that Turner specified dry mills to run at 84.5% C.S., and also that NCis an independent factor. The value 2.208 looks like a power number. Note that the combinationof K, a , and 6 in equation 3 looks like a power number. Another approach and, in the authors’ opinion, a more accurate and although an empirical one is based on Loveday’s definition of power number (Loveday 1978). Power number is derived f+om the following equation:

where

and:

=

pN.p,

P, D L

=

mill charge density, tlm3

=

PN

=

effective mill diameter (inside shell hers), m effective grinding length (EGL) expressed as the length along the mill cylinder at the mill shell in between the feed end liner plate and the inboard side of the discharge grate plate, m Power number calculated k m mill power measurements and which takes into account mill speed and all aspects of the charge motion within the mill cylinder and both conical ends: impact breakage, attrition,abrasion, friction and rotation within the mill charge, losses due to heat and sound emauating from within the mill, wind loss, shape of the mill charge, the imprecise position of the centre of gravity of the mill charge, size analysis of the millcharge, and no-load power.

=

.@ 5 .

................................................................

kW (Net)

% vlv balls

P,

% VIV total

L

x p (balls) +

% VIV ore

% vlv total % voids

100

Where:

x p(o=)

I

x

............................

% v/v

=

volumetric percentage of the total mill volume

P

=

specificgravity

[61

[7l

Values of PNare shown in Figure 4 h m Loveday’s work in which he established a good correlation for power numbers calculated from carefully run pilot plant testwork on Palabora carbonatite ore and surveys at mill operations while varying the volumetric filling of the autogenousmill load.

766

0

WM pifot-nil tests on ctuomium ore MlY pilot-mid fosts

+ PMC. X

I

F.M.C. ore

prohction mi#

Primary outoqenour ni(l at Aitik ( S r r b n )

h a for bol-cnill 1y4.35)

-Thooreticol (0

on

1

I

I

20

3a

40

.

1

50 60 vbkrmr of miti load, per cent of total d u m a

1

to

Figure 4 Correlation of power numbers with total mill volumetric load at 73% C.S. PaIabora carbonatite ore, etc., pilot plant and milloperations (Loveday 1978)

767

In Figure 5, the author' has utilized this principle of power number, beginning with plant surveys in 1986, to generate a family of curves, similar in profile to Loveday's in Figure 4, for Werent mill speeds and the same range of total mill volumetric fiUing. Figure 5 illustrates power

number c w e s for conicalended mills and three selected mill speeds. Power numbers for flatended mills are usually up to 5% lower.

20

Figure 5 Power numbers vs total millvolumetric load for mill design (copyright Fluor 'GRINDPOWER")

Whereas this data is proprietary to Fluor as an integral part of the "GRINDPOWER" mill sizing program (Matthews and Barratt 1991), it has continued to benefit h m surveys with variable speed mills, particularly with gearless drives, over the last 15 years. Once the net grinding power requirement is set, "GRINDPOWER" requires input regarding: 0 0 0 0

0

Type of drive; single pinion, dual pinion, or gearless The desired number of mills to process the required throughput Anticipated ball charge volume, % vlv Total mill volumetric filling, % vlv Mill%cliticalspeed Mill pulp density, 'Yo solids wlw.

768

The program calculates a trial mill diameter based on an assigned power number (it automatically “looks up” the required power number) and a D:L ratio of 3:l in the case of a high aspect mill. This trial diameter is adjusted to the closest standard nominal diameter, then the program re-calculates the effective grinding length (EGL). A standard mill EGL is entered and minor adjustments are made to the operating conditions; e.g., mill speed, so that the calculated mill EGL is in agreement with the selected standard EGL. The process can be repeated for other operating conditions and the required net mill power draw is checked to ensure that it is higher than the power r e q W for grinding. In order to set the motor power, the engineer is prompted to input the maximum anticipated ball charge, and if necessary to re-iterate at a higher mill speed (in anticipation of the next stage of mill sizing investigation: the “tent” diagram). Motor power is then calculated and a standard motor horsepower is selected. This method can be used to size both high and low aspect autogenousand semi-autogenousmills, and pebble mills. It is important to note that power numbers can be calculated for any AG/SAG grindins operation based on the results of a detailed internal survey of a mill following a controlled “crash” stop. SAG mills that have been sized using the power number and ‘‘GRINDPOWER” approach include: Cadia Hill (study), Batu Hijau, Alumbrem, Collahuasi, El Teniente, Ernest Henry, Boddington (studies), La Candelaria, Fimiston, INCO Clarabelle, Freeport 95K Expansion, Lisheen, and Dreifontein. In-plant checks on mill performance have been made, largely in connection with optimization of shell 1inerKfter design, mill speed,and ball volumetric loading, for Los Pelambres, Batu =jay Alumbraxi, Troilus, and Lac Des Iles, as well as numerous studies and plant audits. Power numbers so used are believed to be accurate to within 2% to 3%.

Grate Open Area Much has been written about the inefficiencies of discharging pulp from SAG mills, and backwashing into the mill, with the associated detrimental effect on power draw due to slurry pooling and decrease in mill charge density. These phenomena are the subject of continuing studies; e.g., the AMIR4 P9 Project. Resolution of such a situation has two basic components: 0

Adequate pulp discharging capacity to accommodate new product plus the circulating load of pebbles and pulp. This is important for low aspect mills which are often singlestage to final product with high circulating loads, sometimes in excess of 500%. Such mills are either unidirectional with curved pulp lifters or have peripheral discharge. High a s p t mills have been commonly fitted with radial pulp lifters and are normally bidirectional, with the objective of opthizing shell liner/I&r life and mill operating availability. This design contributes to backwashing unless measures are taken to minimize this effect by cascading the length of alternate lifters (i.e., shortening the length of perhaps two out of every three), andor inamsing the depth of the septum during the design stage. If a mill is still bottlenecked, the only solution is to convert to some profile of curved pulp lifters, change to uni-directional operation,and benefit fiom the associated increased pumping action which contributes to what is called a ‘‘dry” mill. Curved pulp lifters are common in South Afiica and were a standard design for AG/SAG mills made by Dominion Engineering Co. of Montreal up until its absorption into the Svedala organization, as illusttated in Figure 6. The INCO Clarabelle SAG mill heads were drilled for both radial and curved pulp lifters. The operatodmaintenancehave elected to retainthe original curved pulp lifter design

769

MILL LINING SPlRAL PULP DISCHARGING SYSTEM

Figure 6 Typical curved pulp lifter design 0

Hydraulic gradient through the length of the AG/SAG mill.This should be steep enough to ensure that the mill charge is “dry” (i.e., does not slurry pool). In order to achieve this state, slurry (and pebbles) must discharge the mill as close to the mill periphery as possible and be subject to efficient discharge thereafter (preferably by curved pulp lifters). In following this philosophy, there is little point in installing grate sections in any position other than in the immediate peripheral row.

The ef€ective open area in this row is determined by mill diameter, slot width and length, desired pebble size, position of ball deflectinglpositioniuglclamping bars, and allowable casting integrity; e.g., whether the grate section covers one, two, or three pulp discharging septums, whether there are 18 sections in a 36 ft &a. mill or 36 for instance. It is common in high aspect mills with pebble crushing circuits to have a d o r m slot width which is commensurate with the desired pebble size. The reauired open area in this row is determined in part by either the pebble flow rate (tm) or the pulp flow (m’h). The parameta which the author’ has used, and which could be construed as b e i i conservative and subject to further studies, are based on Dominion Engineering Co.’s data (with curved pulp liftersas a basis) for either high aspect or low aspect ratio mills, namely: For pebbles: For pulp:

0.17742 m2(27.5 in?) per tph pebbles 366.12 m’h per mz (1.04 USGPM per in?) of open area,

It follows that, for a given mill diameter, if the required open area exceeds the available effective open area, then the engineer should consider basing the design on a larger mill diameter for efficient pulp and pebble discharge until a balance between the two categories of open area calculation is achieved.

770

Mill Shell LInerLifter Design Within the last five to six years, there has been a return toward Art MacPherson’s dictum that the ratio of shell liier height to spacing should be in the region of 4:l in high aspect mills for optimum power efficiency, kWh/t (Meaders and MacPherson 1964, Bigg and Raabe 1996). Dynamic simulationsdeveloped by Kajamani (Rajamani and Mishra 1996)have demonstratedthat the design of AG/SAG mill shell lkmhfbm should be based on mill operation at the highest planned mill speed to ensure that ball and rock trajectories concentrate at the toe of the charge for maximum impact and power efficiency at that speed, especially for harder ore types. This was proven at Alumbrera where a duction, within six months of commissioning, in the number of rows of shell lifters from 72 to 48 of Hi-Hiand a change in relief angle from 7” to 30” improved power draw and throughput (Shermau 1999). In Figure 7,the operating window for October and November, 1997 at Alumbrera is typical of the power draw and mill speed during the period with 72 rows of shell lifters from start-up in August/September, 1997 to March, 1998 for each of two 36 ft dia. x 17.25 fL EGL SAG mills, each powered by a 18,000 hp (13428 kw) gearless drive rated at 10.30 rpm for rated power. During this period, mill speed was restricted to the range 8.9rpm and 9.1 rpm (69% C.S. to 71% C.S.) to limit the impact of errant ball trajectories and consequent premature damage to the shell liners and lifters. Packing between lifters resulted in late release of balls and errant ball trajectories striking at between 8 and 9 o’clock. Failure to key in the mill charge resulted in insutficient l i i imparted to rocks and balls, particularly the larger racks, and consequent power inefficiency (high k W t ) and lower feed rates which averaged 1,500th per line in the window. Once the change to 48 rows of shell lifters was made in late Marchlearly April, 1998,power draw increased, power efficiency improved (lower k W t ) , and throughput increased to the design rates of 1,852 t/h per line on average. A survey (see Figure 7)in SAG 1 on April 13 recorded 1 2 151 kW vs 1 2087 kW predicted from the “CHARGEPOWER” section of ”GFUNDPOWER” for the measured charge volume of 12% v/v balls and 24% v/v total. In SAG 2 with a coarser feed, 13% v/v balls and 28% v/v total charge volume were measured and predicted power was being drawn over the full operating speed range up to 10.3rpm. Note that in SAG 1, the rate at which power can be dram diminishes above 75% C.S. at the lower total mill charge volume (24% v/v) with the finer feed. This trend is indicative of errant ball trajectories hitting the mill shell at the higher mill speeds and their failure to do useful work. In SAG 2 with 28% vlv total volume, mill power demand is increasing at a much faster rate; i.e., more power can be used for effective breakage of larger rocks on rocks and balls with a higher toe at higher mill speeds. Comparison of these operating conditions, SAG 1 vs SAG 2,confirms the selection of 10.30 rpm (80% C.S.) as the rated speed of the motor/mill to effectively grind harder, more competent, and coarser mill feed, in combination with the improved shell 1inerAifter design. With the softer and/or finer mill feed in SAG 1, the mill is slowed down and the operating conditions changed to stay within the capability of the motor; e.g., a slight reduction in the volumes of balls and ore with a small increase in the mill charge density: 3.924t/m3 vs 3.816 t/m3 in SAG 2.

771

PREDICTED MOTOR POWER OUTPUT AT MILL SHELL K W

Figure 7 is also a “tent” diagram (Barratt and Brodie 2001) in which motor Capability, as expressed by power and torque (current draw) outputs, is compared to the mill power demand required by the mill charge. The “tent” is described by the increase in motor power output at constant rated torque output (4190 A) as mill speed increases up to the rated design speed of the motor, 10.30 rpm (80% C.S.), and the rated motor power output of 13428 kW is reached, beyond which power output is constant and torque output (current) decreases. Mill power demand and operating conditions have to be within the “tent” in order to avoid exceeding the design limit for motor current (4190 A) at mill speeds lower than 10.30 rpm, and to be within the available motor current at mill speeds higher than 10.30rpm; e.g., available motor current at 10.30rpm +5% is 4190A x 95%= 3980A. Any condition “outside the tent” would result in overheating of the motor; i.e., the DCS current value display would turn h m “blue” to ‘W as a warning. In Figure 7, a “red‘‘ condition would be expected between 9.3 rpm (73% C.S.) and 9.8 rpm (76%C.S.) in SAG 2 with 13% v/v balls and 28% v/v total, particularly if the mill is being slowed down under load. Should the ball volume increase to 15% v/v, motor torque rating would be increased by 6% to 4444 A as shown, which would “raise the tent peak” to 14242 kW (19,090 hp) motor rating. Such an increase in operating flexibility was designed into the cooling system for the thyristors in the “E” house at Alumbrera, thereby providing a modest contingency on the initial motor rating. This enabled the operators to maximize feed rates on harder ore by raising the ball charge volume to 15% v/v and operating at medium mill speeds (e.g., 9.5 rpm to 9.8 rpm) and 24% v/v to 25% v/v total mill charge volumetric loading for optimum power efficiency. The design of fbture SAG milling circuits for harder ores should exploit this operating philosophy. Incorporation of power draws at higher mill charge densities (e.g., 15% v/v balls and 25% v/v total) into the mill design boosts confidence in scale-up for harder ores. For many years, the “1 ,OOO hp per ft of EGL” rule applied in assessing rated motor power for a 36 ft dia SAG mill processing porphyry copper ore. The SAG mills at Alumbrera have demonstrated that this rule should be increased by lo%, especially for harder ores, and that by doing so, the mill and motor are capable of operating at a higher torque rating and power output at medium mill speeds and a higher ball charge volume (e.g., 15% v/v) for power efficiency, as well as giving the operators the capability to ~n the speed up when necessary with a higher total mill charge volumetric loading (e.g., 30% dv). This operating philosophy leads into one of the most important factors in the selection of the motor for variable speed AG/SAG mills which is, in the authors’ opinion, definition of the design or rated speed of the motor or, in other words, definition of the “peak” of the “tent”; i.e., the power output capability at that motor speed, or in the case of gearless drives, mill speed. This is very o h an iterative process (Barrattand Brodie 2001), especially when bids from mill suppliers are being technically reviewed. Figure Nos. 8 and 9 illustrate the difference between selecting a lower mill speed at 72%C.S. and a higher mill speed at 76%C.S. for a 40 ft dia. x 21 it EGL mill with a 21000 kW gearless drive and fitted with 52 or 39 rows of Hi-Hi lifters at 30” relief angle. Figure Nos. 10 and 11 take the issue to the next step of ensuring that the motor power output and torque output are suf€icient to accommodate a 15% v/v ball charge at 30% v/v total mill charge volume with worn shell liners and raising the rated power to 22500 kW.This arrangement will allow the operators more flexibility in dealing with a range of softer to harder ores and also in optimizing shell linerfifter design to assist in defining the range of “sweet points” for optimum power efficiency at higher mill speeds. It is important to involve the mill vendor, liner supplier and liner handler supplier early in mill design process for optimization of liner design and liner replacement time.

773

VARIATION OF MILL POWER DEMAND WITH MILL SPEED AT DIFFERENT BALYTOTACMILL CHARGE RATIOS NORMALIZED ON A PER UNIT BASIS

I

MOTOR SPEED, PER UNlT u

P

1

1

nY a w n n 1

1

1

1

1

1

w a SCX

i 1

1

1

Figure 8 Variation of mill power demand with millspeed and comparisonwith 21000 k W motor power and torque capability at 72% C.S. rated motor speed (Barratt and Brodie 2001)

774

VARIATION OF MILL POWER DEMAND Wl'IB MILL SPEED AT DIFFERENT B U O T A L MILL CHARGE RATIOS NORMALIZED ON A PER UNITBASIS

GA9 C M Y S L C T m K.

'igure 9 Variation of mill power demand with mill speed and comparison with 21000 k W motor power and torque capability at 76% C.S. rated motor speed (Barratt and Brodie 2001)

775

VARIATION OF MILL POWER DEMAND WITH MILL SPEED AT DIFFERENT BALLITOTAL MILL CHARGE RATIOS NORMALIZEDON A PER UNIT BASIS

I

MOTOR SPEED, PER UNIT

lUt3MSUTAIVIJ;M

igure 10 Variation of millpower demand (new shell liners) with mill speed and comparison with 22500 k W motor power and torque capability at 76% C.S. rated motor speed (Barratt and Brodie 2001)

776

VARIATION OF MILL POWER DEMAND WITH MILL SPEED AT DEFERENT B m O T A L . MILL CHARGE RATIOS NORMALIZED ON A PER UNIT BASIS:

S

49

47

MOTOR SPEED, PER UNIT

l1lI) C - T m , M. 'igure 11 Variation of mill power demand (worn shell liners) with mill speed and comparison with 22500 k W motor power and torque capability at 76% C.S. rated motor speed (Barratt and Brodie 2001)

777

These same principles apply to lower speed ( I80% C.S.) low aspect mills. However, for higher speed mills in this category (most of which run at 91% C.S.) grid liners are used with hardly any lift except a rough surfhe. of pebbles that are packed into steel grids. Grinding by abrasion against this surface at near centrifugal mill speeds creates a wholly different dispersive action within the mill which allows the flow of pulp with high circulating loads through a discharge screen (grate) or, alternatively,a peripheral discharge.

Feed Chute Design Although details of the various types of feed chute designs are outside the scope of this paper, it is important to mention that maximum feed size (e.g., run-of-mine feed without primary crushing) and rock trajectories (with or without rock boxes) impact on the size of these feed chutes and also on the feed bearing trunnion liner diameter, and mill diameter: length ratio for optimum hydraulic gradient, and mill diameter. Diameter: Length Ratio The iterative process of sizing primary mills implies testing the sensitivity of mill length (EGL), motor rated power output, torque output, and rated speed (expressed as mill speed), and shell linerhifter wear so that the expected range of operating conditions can be accommodated once the mill diameter has been set. In this respect, the ratio of mill diameter to mill length (EGL) is important as it affects selection of the rated speed. In general for high aspect mills, higher ratios of D:L permit operation at higher mill speeds for a given rated power output. Lower ratios of D:L ate commonly rated at 72% C.S. or 74% C.S. but, as has been seen in the above example @:L = 1.88), there is an advantage to be gained in raising the rated power output and increasing the rated speed to 76% C.S. Both MacPherson and Turner predicted optimum ratios of 3:l for D:L, whether for wet or dry operation. At the time (1964), they were dealing principally with single-stagemills processing iron ore to finalproduct. For twostage circuits, D:L has been pmgressively moving toward 2:l or less as perceived and timedependent limitations on mill diameter have been exceeded. The D:L ratio for 32 ft dia mills decreased from 3.00:l to 2.17:l over 35 years until the advent of 36 ft dia mills in 1973.The D:L ratio for 36 ft dia mills decreased from 2.53:l to 2.06:l over 23 years until the first 38 ft dia mill was purchased in 1996.The 38 ft dia mills are 1.87:1,with special cases at Freeport (2.08:l)and Olympic dam (1.60:1), and the Cadia mill (40 ft dia) is 1.97:l and rated at 74% C.S. and 20000 kW (Jones 2001). For low aspect mills, in which D:L can vary &om 1:l to 1:2,the optimum ratio is governed by: 0 0 0 0 0 0

The selection of a single-stage or two-stage circuit Mill speed, 280% C.S., or 91% C.S. Linerhifter design, Hi-Hi or grid Pebble porting, survival rate of pebbles, breakage rates Slurry flowrate, including circulating load Discharge arrangement, grate or peripheral, open area, and effective hydraulic gradient.

For single-stage mills at lower mill speeds ( 2 80% C.S.) with conventional shell lher/lifkrs, D:L ranges from 1:1.5 to 1:2;at the higher mill speed (91% C.S.) with grid shell liners, D:L is generally 1:2.Lowerspeedprimarymillsarefoundtobewithin 1:l to 1:l.l.

778

SECONDARY MILL SIZING AND ASPECT RATIO The standard method for sizing secondary ball mills (usually the overflow type) in AG/SAG circuits follows published mill supplier information with aspect ratios in the range 1:1.5 to 1:2.0, depending upon product sizing and pulp flow; i.e., the highest expected circulating loads (Morrell 2001). The applied net power, calculated from testwork and power-based modelling, will have been based on mill speed and shell linerAih simulations and, particularly, ball charge volume and ball top size, which is dependent upon the contingency applied to the transfer size in the SAG circuit product. Iterations of these parameters can be performed in some models (e.g., "GRINDPOWR") to determine the desired motor power and drive rating, the maximum design mill speed,and the maximum design ball chatge volume (Barratt 1989). For single pinion and dual pinion drives, the possibility of pinion changes to a higher mill speed should always be considered when setting gear ratios, service factors, and motor power ratings. For gearless drives, motor torque output at the maximum desired operating mill speed and ball charge volume has to be determined. Pebble mills are sized in the same manner as autogenousmills.

SUMMARY AND CONCLUSIONS The art of selecting and sizing autogenous and semi-autogenous mills has evolved to include the largest mill, 12.192 m (40ft) in diameter and 20000 kW in rated motor power. The selection of an AG/SAG mill for a project is by no means a certainty, especially if the primary grindhg process is not power-efficient, in view of the larger and proven secondary/tertiarycrushers, screens, and ball mills that are now available. The technology of AG/SAG milling is better understood now after 20years and building/ o p t i n g some 200 mills larger than 8.535 m (22 ft) in diameter for ferrous, non-ferrous and industrial minerals (not including cement) projects world-wide. The process risk can be minimized by a thorough understanding of project geology and geotechnical parameters throughout the mineable resource, with comminutiontestwork by benchscale, pilot-scale, or a combination of both, coupled with an assessment of ore variability, methodologies for scaling up testwork results, and the application of known engineering and manufacturing procedures and practices with respect to ensuring the structural integrity and operability of both the mill and motor at the most economical cost. Success will depend on quantifying the design basis in terms of predicting the range of specific power consumption and power sharing between primary and secondary grinding for the foreseeable life-of-mine and potential planned expansions. Pilot testing is normally required to develop design criteria for AG/SAG circuits, but is not essential. In the context of engineering work preparatory to equipment purchase, power-based models that rely on bench-scale testwork can also be used, either in the absence of pilot plant testwork or to supplement it. This is particularly true for semi-autogenous grinding for which projects ranging in size from 500tpd to 120,OOOtpd have been designed without pilot plant testwork and operated successfully. For autogenous grinding and pebble milling, pilot plant testwork is preferred, but it can be sacrificed if project logistics dictate and a comprehensive database of geotechnicalparameters is available with a reference to existing autogenous operations and bench-scale ore competency testwork. It is now common place to assess the variability of grinding characteristics by testing a large number of ore samples that are related to geology, the geological model, and the mineable resource, as well as the results of pilot plant testwork or more comprehensive bench-scale testwork. The Minnovex SPI test is widely used in this regard.

779

The more precise power-based models (e.g., “GRlNDPOWER”) are used to scale-up pilotscale or bench-scale test results to the appropriate range of applied power draw, mill operating conditions, and to develop mill sizes. The underlying equation that is used for this purpose in this paper is based on the concept of power number, where:

Power numbers have been developed h m several mill surveys and are are related to a range of mill speeds and total mill charge volumetric l o a d i i . They are considered to be accurate to within 2%-3% and take into account motion within the mill charge and losses due to heat, sound, windage and no-load. This mill sizing methodology is used as a basic engineering tool and as a check on technical informationthat is provided by mill suppliers at the mill bidding stage. Selectionof a basic mill size is analyzedprior to equipment purchase for its integrity in terms of f d chute design, grate open area, shell liierAifter design, diameter: length ratio, rated motor power and rated speed, motor torque capability, operating speed range, and the range of mill charge operating conditions. All of these parameters are mounted into a series of ”tent” diagrams that are designed to assist in opthizing the capability of the motor in satisfying the power demands of the mill and its charge, as well as the production objectives. In conclusion, criteria for selectingthe diameter of the primary mill, its diameter: length ratio, and its applied power are influenced by the hardness, or competency of the ore with respect to impact, abrasion, or both, as well as the desired rated speed and the torque capability of the motor as challenged by power demands of the mill, especially with high ball charge volumes of 15% v/v or more. For the more competent ores, there should be a return toward a 3:l ratio for D:L &om the present 2: 1 for high aspect primary mills and, for higher mill throughputs,mill diameters in excess of 12.192 m (40 ft), both W i g necessary (together with optimized shell linerAifter design) for more efficient delivery of impact energy. For low aspect mills at higher mill speeds with grid shell liners,the technology is proven with D:L as 1:2. Use ofthis mill profile at lower millspeeds and D:L raaging h m 1:1.5 to 1:2 quires confiRnation of the optimum feed size, possibly with some form of pre-crushing, to achieve p r o d d o n objectives. The D:L ratio for lower speed primary mills is generally in the range 1 :1 to 1:l.l. In the area of pulp and pebble discharge, more industrial-scale surveys and engineering studies are necessary to establish design criteria for grate open area, ‘‘drf‘ mills, and efficient pulp transfer with curved pulp lifters. Whereas pebble mills are sized using the same principles and formula as is applied to the sizing of autogenous mills, secondary ball mills are sized using established Bond formulae with contingenciesfor maximum transfer size, maximum ball charge volume, and the highest operating mill speed in setting the rated motor power.

ACKNOWLEDGEMENT The authors wish to acknowledge material that originated from their work on projects associated with the following companies: Newcrest Mining Limited, North Limited, RTZ, Newmont Mining Corporation, Cia. Minera D o h Inks de Collahuasi SCM, Minera Alumbrera, Boddington Gold Mine, MIM, INCO, KCGM, Anglo American Corporation, Goldfields of South Africa,Phelps Dodge Corporation, Codelco-Division El Teniente, Cia. Minera Los Pelambres, INMET-Division Troilus, Lac Des Iles Mine, TeckCominco, JK Tech, Amdel, Lakefield Research of Canada, A.R MacPherson Consultants Ltd., Metso, FFE Minerals, h p p Polysius, ANI, and Fluor Daniel Wright Ltd.

780

REFERENCES

Dor, A.A., and Bassarear, J.H. 1982. Primary Grinding Mills: Selection, Sizing and Current Practices. Design and Installation of Comminution Circuits, eds. A.L. Mular and G.V. Jergensen 11, Chapter 24. New Yo& SME-AIME. Turner, RR. 1982. Selection and Sizing of Primary Autogenous and Semi-Autogenous Grinding Mills Design and Operation. Design and Installation of Comminution Circuits, eds. A.L. Mular and G.V. Jergensen 11, Chapter 25. New York SME-AIME. Jones Jr., S.M. 2001. Appendix of AG-SAG Installations World-Wide. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, I :373.

Parker, B., Rowe, P., Lane, G., and Morrell, S. 2001. The Decision to Opt for High Pressure Grinding Rolls for the Boddington Expansion. Proceedings Intem‘onal Autogenous and SemiAutogenous Grinding Technology2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, HI :93.

Barratt, D.J., Basic, J., Dunlop, G.A., and Phillips, R 1999. Autogenous and Semi-Autogenous Grinding: Laboratory and Pilot Plant Studies. Mineral Processing and Hydrometallurgy Plant Design, World’s Best Practice. Australian Mineral Foundation. July.

Barratt, D.J. 1989. An Update on Testing, Scaleup and S

i Equipment for Autogenous and Semi-Autogenous Grinding Circuits. Proceedings Advances in Autogenous and Semi-Autogenous Grinding Technology,eds. A.L. Mular and G.E. Agar, I :25.

Rowland, C.A. 1982. Selection of Rod Mills, Ball Mills, Pebble Mills and Regrind Mills. Design and Installation of Comminution Circuits, eds. A.L. Mular and G.V. Jergensen 11, Chapter 23. New Yo& SME-AIME. Siddall, B., Henderson, G., and Putland, B. 1996. Factors Influenciug Sizing of SAG Mills from Drill Core Samples. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 1996, eds. A.L. Mular, D.J. Bax~att,and D.N. Knight, 11:463.

Powell, M. 2002. South African Progress on Closing the Design Gap between High- and LowAspect SAG Mills. Proceedings 3p Annual Meeting of Canadian Mineral Processors. 12 : 189. CM. Matthews, B.D., and Barratt, D.J. 1991. “GRINDPOWER’? A Computer-Based Program for the S i g and Selection of Grinding Circuits. Computer Applications in the Mineral Zndushy, Second Canadian Conference,eds. A.L. Mular, R. Poulin, and RC.T. Pakalnis. Vancouver, September. Barratt, D.J., Matthews, B.D., and DeMull, T. 1996. Projection of SAGIAG Mill Sizes, Mill Speeds, Ball Charges, and Throughput Variation from Bond Work Indices. Proceedings Intem.onal Autogenous and Semi-Autogenous Grinding Technology 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, II : 541.

MacLarq D., Mitchell, J., Seidel, J., and Lansdown, G. 2001. The Design, Start-up and Operation of the Batu Hijau Concentrator. Proceedings International Autogenous and SemiAutogenous Grinding Technology2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, IV : 316.

781

Kosick, G., and Bennett, C. 1999. The Value of Orebody POWER Requirement Profiles for SAG Circuit Design. Proceedings 31st Annual Meeting of Canadian Mineral Processors. 16 : 241. CIM. Kosick, G., Dobby, G., and Bennett, C. 2001. CEET. Comminution Economic Evaluation Tool for Comminution Circuit Design and Production Planning. SME Annual Meeting. Morrell, S. 1992. Prediction of Grinding MU Power. Transactions Institution of Mining and Metallurgy. Section C. 101. January - April. Momll, S., 1996. Power Draw of Wet Tumbling Mills and its Relationsbip to Charge Dynamics: Parts 1 and 2. TransactionsInstitution of Mining andMetallurgv. Section C. 105. January - April. Loveday, B.K. 1979. Prediction of Autogenous Milling from Pilot Plant Tests. Proceedings 11” CommonwealthMining and Metallurgical Congress. Hong Kong. May. Meaders, RC., and MacPherson, A.R 1964. Technical Design of Autogenous Mills. Mining Engineering. September. Bigg, A.C.T, and Raabe, H. 1996. Studies of Lifter Height and Spacing: Past and Present. Proceedings International Autogenm and Semi-Autogenous Grinding Technologv 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, 111 :999. Rajamaui, R.K., and Mishra, B.K. 1996. Dynamics of Ball and Rock Charge in SAG Mills. Proceedings International Autogenous and Semi-Autogenous Grinding Technologv 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, I1 : 700. Sherman, M. 1999. Alumbrera’s SAG Mill Design and Optimisation. Mineral Processing and Hy&ometallurgv Plant Design, World’sBest Practice. Australian Mineral Foundation. July. Barratt, D.J., and Brodie, M.N. 2001. The “Tent” Diagram, What it Means. Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, IV : 368. Morrell, S. 2001. Large Diameter SAG Mills Need Large Diameter Ball Mills - What are the Issues? Proceedings International Autogenous and Semi-Autogenous Grinding Technology 2001, eds. D.J. Barratt, M.J. Allan, and A.L. Mular, 111 : 179.

782

Selection and Sizing of Ultrafine and Stirred Grinding Mills Jens K. H. Lichter’ Graham Davey’

ABSRACT The selection and sizing of mills for regrind and ultrafine grinding applications does not lend itself to conventional methodologies. It requires a more holistic approach that considers not only the mill but also the application of the mill within the process. The selection of stirred mills for ultrafine milling requires unique approaches that are able to answer the questions related to the selection of the circuit configuration, type of mill, media and operating conditions. Further considerations required for the selection of mills are the inherent difficulties of particle size measurement and an accurate definition of the product size required. INTRODUCTION Minerals with fine particle intergrowth, either with other metallic minerals or gangue, are being increasingly mined. These ores have provided new challenges in concentrator design, requiring fine and ultrafine grinding in order to obtain acceptable grades and recoveries. The advancement of flotation technologies now permits effective flotation in the sub 10 micron size range making it possible to separate finely disseminated minerals from gangue. Similarly, the ability to produce ultrafine feed for various leaching processes, including bio-leach and low pressure oxidation, often requires fine or ultrafine grinding to improve reaction kinetics to the level at which these processes become commercially viable. Economic ultrafine grinding processes also make it feasible to direct leach refractory gold ores, rather than the more conventional roasting or high pressure autoclave routes. The relationship between energy required and product size is not a proportional one. Theoretically, the energy required (per unit mass) to break a particle 1/1OOth of its size, is 4000 times greater. As we strive for ever finer grinds, the need to optimise the comminution process becomes self-evident. In order to achieve the required improvements, changes in milling technologies are needed, as well as a better understanding of the variables that affect them. The media in a mill generates a particular energy spectrum which is best defined as a frequency / magnitude plot of the energy delivered by the mill. It is possible to substantially alter this relationship for a mill by changing the operating conditions. Different mill designs will however differ in the range of energy spectra they can generate. The better a mill’s energy spectra matches the breakage requirements of an application, the more efficient the system will be. This has been demonstrated by Jankovic (A. Jankovic 2001) where clear optimum operating points could be determined for a mill by changing the operating characteristics. 1. Metso Minerals Industries Inc. Grinding, York, Pennsylvania

783

THE SELECTION OF MILLS - EQUIPMENT OPTIONS Time and innovation have resulted in numerous different mill designs capable of producing fine and ultrafine products. This paper will concentrate on the few that have seen mainstream commercial application in the mineral processing industry. In many cases these mills are of a unique and proprietary design, and do not have generic names and are known only by their commercial trademarked names. Mills for fine and ultrafine grinding fall into four primary categories, these being;

0

0

Ball Mills Stirred Media Mills Centrifugal Mills Jet Mills

The two first categories make up the bulk of the fine and ultrafine grinding duty. While ball mills still see extensive use for the production of fine powders, these tend to be predominantly dry applications in specific industries. Industrial minerals applications make extensive use of dry ball mills, often using ceramic grinding media (to avoid metal contamination of the product) for the production of fine and ultrafine powders. Ball mills in these applications are typically operated in closed circuit with dynamic classifiers. Long tube mills (length to diameter ratio’s in excess of 3 to 1) as well as batch mills are also used. For wet minerals applications, the application of tumbling ball mills is declining and limited primarily to very large tonnage applications and relatively coarse grinds. The efficiency advantages of stirred media mills over ball mills have largely seen the fine and ultrafine applications move away from conventional tumbling ball mills. Figure 1 shows a typical relationship for specific energy versus grind for a ball mill and a stirred media mill. At coarser grinds the stirred media mill requires approximately 30% less energy than a ball mill. For ultrafine grinds, this advantage increases to more than 50%. The data shown in this example is comparative data from closed circuit pilot milling campaigns using a conventional ball mill and a Vertimill. Media size and feed were identical for both mills.

100

1

0

50

150

100

200

Specific Energy (kW .hr/tonne)

Figure 1. Relative performance of Tumbling Ball Mills vs. Stirred Media Mills

784

250

These relationships hold for full size mills and are typical of the relative performance of stirred media mills in general when compared to conventional ball mills. Stirred media mills can be applied to relatively coarse feeds and grinds, with feeds up to 6mm, and products as coarse as 300 micron possible from some of the available mills. These do however represent extremes in the range, and more typically feed sizes will range from 300 micron down to 50 micron. Products are typically from 50 micron down to 5 micron. The definition for ultra fine products is not an industry standard, but for the purposes of this paper, the author defines it as sub 15 micron. Stirred media mills can be classified into a number of different sub categories predominantly defined by the speed, geometry and orientation of the media agitator or stirrer. The mill orientation can be either horizontal or vertical. The media agitator can consist of a shaft fitted with a spiral screw, pins or discs, and the media can be either agitated or fluidised. Although the basic appearance of the mills is often similar, the operating regime and performance can be very different. There are two fundamentally different classes of stirred media mills. The first class includes the Vertimill or Tower Mill and conventional pin mills. In these mills the agitator “stirs” the media with the fluid having limited effect on the interaction of the media with itself. In the second class, typified by the Stirred Media Detritor (SMD) and the Netzsch/IsaMill, the media size is very fine, and the speed of the impellor high enough to effectively fluidise the media. The media becomes highly mobile and takes on the flow pattern of a viscous fluid. Stirred media mills such as pin mills or the Vertimill, which use larger media, are more efficient with coarse, hard feeds. Fluidised media mills using low-density silica or ceramic media have the advantage for ultrafine milling with fine feeds. One commonly quoted characteristic of stirred mills, the energy intensity, does not have a strong influence on the relative performance of the mills. Mills such as the Vertimill and Tower Mill with low energy intensities operate as efficiently as mills such as the SMD and the Netzsch / IsaMill which have very high energy intensities. The key is correct media selection. A brief description of the mills follows;

785

The Vertimill or Tower Mill This is a stirred media mill consisting of a vertical cylinder with a relatively slow speed screw media agitator. See Figure 2. The Vertimill / Tower Mill is most commonly used for concentrate regrind applications with a typical feed size of around 100 to 300 microns and typical products of 100 to 15 micron. Finer products are possible with the use of suitable media. These mills predominantly use steel ball media ranging in size from 40mm to 6mm and are also capable of using cast pebble media such as “Cylpebs” or “Power Pebs” which become economically attractive for media sizes less then 25mm. The low impellor speed aids in reducing component wear, but results in a large mill size and volume. These mills are predominantly wet mills. The finest grinds in commercial applications grind down to approximately 80% passing 12 micron, but with suitable media, sub 5 micron grinds have been obtained in pilot installations at comparable efficiencies to other stirred media mills.

Figure 2. General view of a Vertimill

786

Pin Mill Pin mills have a vertical shaft impellor fitted with pins. The mill body is filled with spherical media, typically steel or ceramic, in the size range of 3mm to 12mm. Pebble or autogenous media can also be used. These mills are capable of operating either wet or dry. The preferred feed size is less than 300 micron and sub 10 micron grinds are achievable. Relatively high impeller speeds often result in wear issues with these mills, and they are most suitable for non abrasive feeds. Figure 3 shows a cutaway of a Metprotech mill.

Figure 3. General view of a Metprotech Mill The Stirred Media Detritor or sand grinder. This mill utilizes a vertical shaft pin agitator. Media is typically high grade alluvial silica sand or ceramic spheres in the range of Imm to 3mm in diameter (See figure 4.) The agitator speed is high enough to fluidize the media. Screens fitted to the upper circumference of the mill body allow product to leave the mill while retaining the media. These screens do not act as a classifier, but function only to retain the media. Feed size is typically in the range of 100 micron down to 15 micron, and products as fine as 98% passing 2 micron are achievable. Typical applications in the metalliferous industry have products down to 80% passing 6 micron. Figure 4 shows the Metso Stirred Media Detritor (SMD)

787

Figure 4. General view of a Stirred Media Detritor (SMD) The Netzsch or IsaMill. This mill utilizes a horizontal shaft media agitator most commonly fitted with discs. Media utilized ranges from prepared fine slag media through sand media and ceramic media in the lmm to 3mm size range. In the case of the IsaMill, the mill is fitted with an internal centrihgal screen fitted to the impellor for media retention. The application range is similar to the Stirred Media Detritor. See figure 5.

Figure 5. General view of a Netzsch / IsaMill

Centrifugal Mills This category of mills generates a high energy intensity inside the mill by moving the mill body around a central axis at high speed. It is therefore possible to create forces well in excess of the 1g force available to tumbling ball mills. These mills can be operated with conventional media or autogenously, and will operate wet or dry. One example is the HiCom nutating mill. This mill swings the mill body in a nutating motion. These mills will accept coarse feeds (limited by the throat diameter) and are capable of grinding below 10 micron. Media retention can be an issue when small media is required. See figure 6 .

Figure 6. Cutaway view of the HiCom Nutating Mill

Jet Mills or Fluid Energy Mills This is a stationary mill that uses the energy contained in a fast moving fluid to produce particle size reduction by impact or abrasion of the particles. Two main types are in use either the parallel type where the air is introduced to a circular grinding chamber or the opposed jet were two opposing fluid streams are impacted. The fluid used to carry the feed solids is normally compressed air, an inert gas or steam. No media is used in fluid energy mills the feed material and fluid providing the breakage forces. PRODUCT DEFINITION The selection of type and size of mill for an ultrafine grinding application has to start with a thorough understanding of the duty. A mill is not there to provide a product of a particular size specification, but to provide a product with the desired liberation characteristics. To amplify on this statement; different mill characteristics and operating conditions will affect the type of comminution that takes place in a mill, and will vary the balance between the fracture, attrition and abrasion components. This in turn affects the size distribution, liberation and surface characteristics of the product. It is important to consider this when developing pilot plant flowsheets. It is not advisable to utilize a conveniently available mill for the generation of feed to a pilot concentrator plant. The mill product characteristics, and therefore the grade and recovery performance of the plant, are closely related to the type of mill used. It is important to at least stay within similar classes of mills. If different types of mills are to be considered, then due cognisance is required of the differences between the mill products and their effect on plant performance. At

789

the very least, whole size distributions should be compared and evaluated for their recovery characteristics.

Defining product size It is a common practice in the minerals processing industry, to define the product size of a slurry by the particle size at which 80% of the particles by mass are smaller than that particle size (the P80.) This does not give a true picture of the mill product size distribution. Many industrial minerals producers, e.g. the paper filler suppliers, have long moved away from such “loose” product definitions. Multipoint product size distributions with very tight specifications (sometimes defining the required 99.9% passing size) are commonplace. Such stringent restrictions are not necessary in most minerals applications, and are often not achievable for wet applications. It would certainly be advantageous to move away from the customary 80% passing size specification. In many mineral concentration systems, the P95 or the P10, rather than the P80, will more accurately define the grade and recovery possible with that product stream. As an example, consider the typical flotation grade / recovery characteristics. Recovery of particles below a threshold size are severely impacted by limitations in the physical chemistry of the system, (e.g. the ability to depress ultra fine gangue or collect mineralize particles). Similarly, coarse particles will result in poor liberation affecting both grade and recovery. Recovery losses in most leach processes e.g. a cyanide gold leach, are largely determined by the coarse tail. It is therefore important to consider the whole mill product distribution curve in relation to the optimum grade recovery requirements of the downstream concentration stage. Consider the decision to mill in open circuit or in closed circuit. Figure 7 shows the grind versus specific energy characteristics of a Vertimill application. The data depicts the relationship between the grind and the specific energy required for open and closed circuit configuration. The grind, energy relationships are shown as functions of the P80 and the P95. Assume that open circuit pilot tests preceded leaching tests and a P80 of 10 micron was defined as the correct product size for optimum grade / recovery economics. This would equate to a specific energy requirement of 54 kW.hr/tonne milled. In closed circuit, the specific energy required to the same grind would be 37 kW.hr/tonne. This is a reduction of approximately 30%. Considering the cost differences between an open and a closed milling circuit at these product sizes (both capital and operating), the additional specific energy required might be considered reasonable. If this were a milling circuit preparing feed for a leach circuit, then the key recovery criteria would probably be a P95 of 22 micron. Based on this assessment, an open milling circuit would require 54 kW.hr/tonne, but a closed milling circuit would require only 18 kW.hr/tonne. If the selection criterion was a P95 rather than a P80, the energy reduction from open to closed circuit milling would then be 67%. Reaction kinetics will reduce the difference by a margin, but the basis for a decision would still be substantially different.

790

.Open circuit p80 -Open Circuit P95 -Closed Circuit P80 .Closed Circuit P95

Figure 7. Grind vs. specific energy - Comparison of open and closed circuit performance An equally important criteria is the product size specified. In the ultrafine product range, the relationship between the specific energy required, and the product size, is very flat. Significant increases in specific energy are required to produce moderate improvements in the grind. Figure 8 shows a typical specific energy vs. grind relationship for a stirred media mill. A change in the product specification from 80% passing 7 micron to 80% passing 6 micron would require an additional 50% specific energy. In this environment it is important to be precise as to the required product specification.

100

1

0

10

20

30

40

50

60

70

80

90

Specific Energy (kWNtonne)

Figure 8. Typical specific energy vs. grind relationship for a Stirred Media Mill

791

Particle Size Measurement The definition of the product size also warrants consideration as ultrafine milling adds an additional level of complication. Unlike typical grinds down to 38 micron, where screening is commonly employed, there are no “absolute” standards with which to measure ultrafine particle size distributions. Current particle size measurement methods include laser diffraction, settling, cycloning, optical, etc. The most common units are the Malvern Mastersizer, Microtrac, Cyclosizer, Coulter Counter and the Sedigraph. Each of these methods has distinct characteristics and is affected differently by shape, density and translucence amongst other particle properties. For the industrial minerals and pharmaceutical industries that have always worked in these fine size ranges, common industry standards are employed. That is not the case in the metalliferous industry where many different methods are still used. Table 1 shows comparative results between different particle size measurement methods for eight samples over a range of product sizes. It should be noted that the relative trends (i.e. which machine reports a finer size) is not always consistent for all powders. The particle characteristics, predominantly shape, have a strong influence. The discrepancies tend to be more pronounced at the finer sizes, where machine limitations begin to be encountered. In addition to potentially large differences in the product sizes “measured” by the different methodologies, machines of the same type, and brand, can give significantly different results. Even if the type and model of particle size analyser has been standardized, machine setup, maintenance, standard operating procedures and operator skill all have significant effects. Agglomeration of particles during measurement is also a concern, and the use of dispersants adds an additional level to the potential error.

7 8

I

I

I

I

5.6

7.9

I

4.1 4.7

Table 1 Particle size distributions reported by different size measurement techniques

If different milling technologies are being compared, using pilot or batch milling tests in different locations, it is important to be aware that the particle size measurement will be different, and probably by more than the potential difference between the different milling technologies. This would be true even if the same model particle size measurement device were used. The only reliable comparisons can be made if the products from both tests are tested on the same particle size analyzer using the same technician. Failing that, normal measurement errors may be too high to make meaningful comparisons. Where the same particle size analyser cannot be used, it is essential to at least use similar model machines, and employ the same operating procedures. Comparisons made using dissimilar particle size measurement techniques are not very meaningfid. Similarly it must be remembered that the particle size measurement method must be common with that employed when originally determining the grind requirement for optimum liberation. There are no easy solutions. The key is to exercise caution, and rigorously work according to best known practice. Be realistic about the reliability and accuracy of the methods used for sizing and comparing equipment.

792

DETERMINATION OF THE SPECIFIC ENERGY REQUIRED. Stirred media mill designs are generally unique and mill selection is often based on manufacturers testing, or alternatively on tests run in third party laboratories using lab scale versions of the mills being considered. Empirical methods, such as the Bond method, are largely unsuccessful in determining power draw requirements for ultrafine grinds and are unsuitable for stirred media mills. The Bond method, for example, incorporates a correction factor (EF5) for fine product sizes. This correction factor was specifically intended to correct for the inefficiency of ball mills using conventional media sizes when producing very fine products. With stirred media mills, the media size limitation is largely overcome and milling efficiencies are dramatically improved. Specific energies derived from the Bond equations would be unacceptably conservative. It is not possible to determine Bond Work Indices for the majority of fine and ultrafine grinding applications, as the feed size would not meet the test requirements. There is also considerable activity in the development of population balance models (PBM) for stirred media mills. The primary challenge is to accurately define the breakage rates and the effects of operating parameters and media on the breakage parameters. It is unlikely that these techniques will be used for mill sizing in the foreseeable future, as the challenges are considerable. One benefit of the finer feed size typical of ultrafine grinding applications, is that an accurate laboratory test is possible. To accurately size a SAG mill a minimum sample size of lOOkg is required for a laboratory test, and 20-100 tonnes for a pilot test due to the large feed particle size. Statistical relevance requires significantly larger samples. For stirred mills the feed top size is generally less than 200 microns, and therefore a sample mass of lOOg is sufficient for statistical reproducibility. The only reliable method currently available for the selection of mills for ultra fine grinding is a well planned and executed lab or pilot scale test regime. The test program should include an evaluation of all the primary operating parameters, listed below. Small sample size and relative ease of testwork make the evaluation of multiple operating parameters feasible. The primary data generated with such a laboratory or pilot scale test is the relationship between specific energy and grind. Depending on the type of mill being evaluated, the nett specific energy generated by the test mill can either be used directly without any correction factors or will need to be adjusted. This task is currently best left to the supplier of the equipment being evaluated or to the laboratory where the tests are being executed. Media Considerations Use of the correct media is important for all grinding applications, but in the case of ultrafine grinding in stirred media mills it becomes the most important variable. Media parameters that need to be considered include;

0

Size Type Competency Hardness

Stirred media mills use a wide variety of media from 25mm to 6mm steel balls and cylpebs commonly used in mills such as the Vertimill and Tower Mill, to Imm to 5mm high grade alumina and sand media used in mills such as the SMD and the Netzsch / IsaMill. Media can affect the specific energy required for an application by in excess of 30%. Any laboratory scale testwork should therefore include a range of different medias. Media availability is often regional and transport costs are a significant contributor to the total cost. When undertaking laboratory or pilot milling regimes, both local and known media types should be tested. Cost and performance are the primary considerations. The cost of conventional steel media increases rapidly as the size decreases below 25mm. This has largely restricted the use of small steel media in mills such as the Vertimill, Tower mill and pin mills. These restrictions are now

793

largely eliminated due to the recent availability of cast media and steel shot. Size limitations no longer exist, but there are still questions regarding the media consumption, and influence of iron (from the media) on some concentration processes. Material and sphericity are the strongest influencing factors for the lmm to 3mm media commonly used in “fluidized “ media mills. Choice of media for laboratory and pilot milling tests should reflect the reasonable choices available for the proposed plant. Using 2 kg of high alumina ceramic media at US$70 per kg may be acceptable for a pilot mill, but would be wholly undesirable for a commercial installation requiring 10 tomes of media. Similarly, using low cost local media may provide acceptable specific energies for the application, but media breakage and wear of mill components [which are not readily measurable in batch laboratory tests], may make the media unacceptable for a commercial application. Higher quality media could also reduce equipment size. Comparisons between different types of mills should consider media costs, and where reasonable, use similar media. It is advisable to always include at least one commercially available media of known quality into a test regime. Media Size. Media size has a significant impact on the performance of mills in fine and ultra fine grinding application. It is often the primary limitation of the fineness of grind possible from a type of mill. As feed and product sizes decrease, the energy required to break a particle also decreases, and the frequency of the breakages per unit mass increases. Excess energy from breakage events is largely converted to heat and does not contribute to the grinding process. The most effective way to increase the frequency of the grinding events and decrease the energy per event is to reduce the media size. Consider the relationship between media size and the number of balls (or other media shapes). Table 2 shows the relationship between the media size and the number of balls per unit mass (or volume). The number of media per unit volume increases by the inverse ratio of the media size to the third power. As the number of breakage events in a mill is proportional to the number of media, dramatic improvements are possible through the selection of the correct media size.

NUMBER OF BALLS NUMBER OF BALLS

BALL SIZE

SURFACE AREA

(mm)

&/tonne

/mton

Normalized

20

83.3

663 15

1

15

111.1

157190

2.4

10

166.7

530516

8

5

333.3

4 144132

62

3

555.6

19648758

296

2

833.3

663 14560

1000

Table 2. Relationship between media size, and the number of balls per unit mass

Because of the need to focus energy on ever increasing numbers of smaller particles, media above a certain size may become completely ineffective in producing ultrafine particles. A typical media size, specific energy plot for a Vertimill is given in Figure 9. Significant reductions in the specific energy required to achieve a grind can be obtained from the use of smaller media. The specific energy require to achieve a grind of 15 micron is more than 50% greater with 1Omm media than it is with 5mm media. A 15 micron grind was not achieved with the 18mm media over the range of specific energy tested.

794

18mm steel media 10mmsteel media 5mm steel media

40

50

Specific Energy (kWtonne) _. _.

~

Figure 9. Effect of media size on mill efficiency. Typical results for a Vertimill Similar effects are found with the media used in "fluidised" media mills. Figure 10 shows the effect of media size on a SMD. In this particular case the trend between energy required and media size reflected in the Vertimill data is reversed. In this example, the mill requires over 50% more energy to the same grind with the Imm media than is required with the 2mm media. The Imm media is too small to effectively break down the larger feed particles. A finer, or softer, feed might see this trend reversed. Also evident from the data is that the use of seasoned media charge is a requirement when generating design data. Mono sized media will not give a true reflection of the eventual performance of an operation.

1

.

_

__

~

- -~ ~

-

~~~

100

18mm steel media

I I

i

1Ommsteel media 5mm steel media

0 ~

10 ~

20

30

Specific Energy (kWh/tonne) -- - ._ . __

50

40 ~

Figure 10. Effect of media size on mill efficiency. Typical results for a SMD

795

Media Type. Media type and shape also affects mill performance. Media can be either ferrous or non ferrous. Non-ferrous media includes high grade alumina balls and beads, lower grade mullite ceramic beads, and silica sand. There is a large variety of exotic medias available for highly specialized industrial applications, where contamination, and not cost are the primary consideration. These medias are not normally suitable for high tonnage applications. Table 3 shows some typical medias, as well as examples of cost and relative wear (where applicable).

I

Main Components

1

Price ($/kg)

1 Approx. Consumption1

I

Zirconia Oxide Table 3. Media characteristics

Steel media is most commonly used in ball mills, vertical mills with screw agitators and pin mills. Fluidised media mills most commonly use small ceramic beads or silica sand. One consideration in the selection of a suitable milling technology for an application, is the effect of steel media on flotation recovery characteristics. In ultrafine grinding applications, iron in solution from the media can contaminate sulphide mineral surfaces with iron oxide thereby affecting the grade and recovery characteristics of the flotation plant. Iron in solution will also consume oxygen and affect some down stream processes. Under these conditions, a non-ferrous media may be preferred. Ball mills and stirred mills draw less power with non-ferrous media of a lower density than steel, and the mill sizes must therefore be increased. Although fluidised media mills can be operated with ferrous media, they are generally designed to operate only with non-ferrous media. Design modifications would be required and as such, this is not typically an option. Mullite ceramic beads have supplemented high grade alumina medias. These beads are typically kaolin based and while being hard and having good media properties, are significantly more affordable than more traditional alumina media. This class of media is seeing increasing application for use in ultrafine grinding. Sand media should be near spherical, and aspect ratios better than 1:l.l can be obtained. With all stirred media mills, the way that media moves over itself affects the energy utilization. Spherical media moves over itself relatively easily, whereas non-spherical shapes will have increasing tendency to “lock up” thereby consuming energy. This effect is most readily seen with “fluidized media” mills. Figure 1 1 shows a specific energy relationship for an SMD using sand and ceramic media. The ceramic media has advantages in both the s.g. of the media, the hardness, and the sphericity. For this particular application, the use of ceramic media would reduce the energy required by almost 50% over the use of sand media. Use of ceramic media for this application would almost halve the milling capacity required, as well as the energy consumed. This would need to be factored against the media consumption and cost differences between the two media types. This level of improvement is not always found, and is dependant on the feed size, hardness and grind required.

796

70

P

60

Sand Media

50

Ceramic media

.-

-E

40

0

(D

n +a

0

30

a

‘0

g

20 10 0 0

10

20

30

50

40

60

70

80

90

Specific Energy (kwhltonne)

Figure 11. Effect of media type on performance of a SMD Media Competency. Of particular importance with sand media used in “fluidized media” mills is the competency. The typical commercial sand media used in “fluidised” media mills is normally used for filtration or other duties. Mechanical strength is not specified. Ideal sand media is alluvial sand with rounded edges, and free of flaws. A common problem associated with low grade sand media is that it has internal flaws, and is not able to withstand the forces associated with milling. Flawed media tends to break up rapidly, degrading the media shape and size. Breakage increases media consumption. Broken media also presents sharp edges, which increase the wear of the impeller and wear liners in a mill. These faults are not typically evident during batch laboratory tests. Care is therefore required when designing plant based on the use of such media in batch or pilot milling tests. Media Hardness. One aspect sometimes not considered is the media hardness. With steel media, hardness does not affect mill performance, only media consumption. This is not the case where non-ferrous media is used. Consider the environment inside a typical “fluidised media” mill using sand or other non-ferrous media. If the mineral being ground is harder than the media, then the media will in effect be subject to Comminution, and size reduction of the feed will be reduced (C. Krause 1998). This data depicts the product size distributions for a fluidized media mill grinding a quartz feed. These are batch grinds for the same duration. Quartz has a Moh hardness of approximately 7. The data shows dramatically different results for media that is softer or harder than the feed. There is also a density effect, but this is primarily due to higher absorbed powers with the higher s.g media. The primary variable is the media hardness. While sand media may be acceptable for soft minerals, when hard minerals are encountered, higher grade alumina media may be necessary. Figure 12 shows the product size distributions for the various runs.

797

100

3.56sg - 8+ Mohs 3.27 Sg - 8+ MOhS 2.71 sg - ?+ mohs 2.68 sg - 6+ mohs

10 10 Size (micron)

1

100

Figure 12. Effect of media hardness Operating parameters. Outside of the effect of media selection on the performance of stirred mills, the strongest influencing variable is the slurry YOsolids, or viscosity. Vertimills, Tower Mills and pin mills are less sensitive to slurry viscosity than “fluidised media” mills. The former category of mills is able to operate over a reasonably wide range of % solids. Fluidised media mills require more careful analysis of the effect of % solids. Two sets of trends are presented below showing typical relationships between the grind and specific energy at various % solids. Figures 13 and 14 show the effect of slurry percent solids. As can be seen, the effect of slurry percent solids on the performance of the mills is significant and trends are not always consistent. Lab or pilot milling tests over a range of operating conditions are essential in order to determine the optimum operating conditions of the selected mill, and therefore the correct mill selection. 100

30% Solids 40% Solids

50% Solids

1

0

10

20 30 40 50 Specific Energy (kW.h/tonne)

60

Figure 13. Effect of slurry YOsolids on fluidised media mill performance

798

70

100

n 10 er

3

0

e a

30% solids 40% solids

50% solids l i

0

10

20

30

40

50

Specific Energy (kW.hr/ton)

Figure 14. Effect of slurry YOsolids on fluidised media mill performance CIRCUIT CONSIDERATIONS Classification The primary options are open or closed circuit milling, with the possibility to pre-scalp the feed ahead of milling. Scalping can be either in dedicated classifiers or alternatively by introducing the new feed to the mill discharge sump. Fine and ultrafine grinding circuits benefit from classification in much the same way that conventional grinding circuits do. Use of a classifier will reduce fines generation, produce a tighter product specification and reduce the overall energy requirements to achieve a specific grind. These benefits have to be weighed against the cost, both capital and operating, of classification circuits. For dry grinding circuits, the most commonly used classifiers are dynamic “whizzer” type units. The primary type of wet classifier for fine and ultrafine classification is the hydrocyclone. As grinds become finer, cyclone sizes need to be reduced, and the percent solids in the cyclone feed reduced. For sub 10 micron grinds, cyclone sizes are around 25mm to 50mm. In addition the product is very dilute, typically no more than 15% solids w/w. The downstream circuit determines whether this dilution is prohibitive or not. As an example, where a mill is used to regrind a concentrate, if the concentrate is re-introduced to the bulk float cell feed, and then the dilution may not be an issue. If however, the mill product is treated in dedicated float cells, then the dilute feed would not be acceptable. Added dewatering costs need to be factored into the overall economic evaluation. Another consideration is the slurry percent solids requirement for the mill feed. A typical concentrate from flotation cells would be too dilute to be used as mill feed for an open circuit mill. As shown previously, good control of the feed density is essential to efficient operation. A scalping cyclone would be required for this duty. The cyclone would have the combined benefit of removing finished product in the feed, thereby improving milling efficiency and producing a tighter product specification, as well as increasing the density of the mill feed. As the milling circuit is open circuit, the total flow to the cyclones, and therefore the number is reasonable, and the cyclone overflow is recombined with the higher density mill discharge and further dilution can be limited.

799

Number of Mills Milling circuits are always designed for a maximum duty. These circuits rarely operate at these values. During start up of greenfield plants, extended operating periods with significantly reduced feedrates can last for up to one to two years. Also, where mills are included in concentration circuits, feedrate to the mill will be affected by both the normal total circuit feed fluctuations due to the hardness of the ore, as well as swings due to the grade of the ore. In tough applications, these tonnage swings can be of the order of 100 percent to 200 percent. Under these conditions, the t u n down ratio (the minimum power at which the mill can operate) of the selected mills must be considered. Mills installed in fine and ultrafine regrind duties are normally constant speed, and therefore constant power machines. Ball mills and some stirred mills such as the screw agitated mil, and the pin agitated mill can operate at reduced ball levels and therefore power draws. These changes in ball charge can be used to minimize overgrinding during known periods of reduced throughput such as startup periods, but typically cannot follow the normal capacity swings resulting from short term changes in the feedrate. Fluidised media mills have very limited turndown and should be operated at their design power if severe wear problems are to be avoided. What this means, is that if the feedrate should drop below the design values, the mills will continue to input the f i l l power, and the product will be ground significantly finer than required. This would be more of a problem for flotation circuits than leach circuits, but the effect on the filters and thickeners in the plant must still be considered. The most effective method to control overgrinding is with the use of multiple units fed from a common source. Individual mills can be brought into service, or taken out, as required. Caution should be exercised when specifying a single mil for an application, if feedrate variations are expected to be a problem.

Vertical mills with screw agitators do have a measure of control over this problem. The mills include an internal recycle loop that controls the uprising velocity of slurry in the mill. On specification product can be removed from the mill and presented to classification, thereby reducing overgrinding. In some rare cases, specifications on the allowed coarse material in the product require the installation of two mills in series (where open circuit milling is required). This is not normally a requirement of metalliferous plants. REFERENCES

C. Krause, M. Pickering. 1998. Evaluation of Ultrafine Wet Mineral Milling Using Carboceramics Proppant products for Attrition Grinding media. A. Jankovic, 2001. Scale-Up of tower Mill Performance using Modelling and Simulation. JKMRC / Amira P9M project, Third progress report.

800

Grinding Plant Design and Layout Considerations By M. Ian Callow and David G. Meadows

ABSTRACT The emphasis in the paper is on the design and layout of two-stage grinding circuits consisting of large-diameter, short-length SAG mills followed by ball mills. The paper describes and discusses alternative layouts of coarse ore stockpiles and ore retrieval systems followed by a discussion of alternative layouts of SAG and ball mills, pebble crushing and associates support processes and systems. The paper includes tabulations of equipment installed at selected concentrators and arrangement drawings and photographs of typical layouts.

INTRODUCTION The emphasis of t h s paper is on the design and layout of two-stage grinding circuits based on North American large diameter short length SAG mills followed by second stage ball mills. There are many variations on this basic concept, particularly outside North and South America and also in the design and layout of smaller circuits. However the two-stage circuit has evolved over the last thu-ty years as the most cost effective system for grinding circuits with a capacity greater than say 20,000 mtpd. These circuits are the new “conventional” circuits from the 1980s onwards, almost completely replacing the older three-stage crushing, rod and ball mill circuits. Very few operators regret the passing of the rod mill with associated h g h maintenance and sometime nightmare rod tangles, but the ball mill remains the old-faithful heart of most grinding circuits. T h s is true after more than a century or more of grinding circuit development and the objective of later technology has been mainly to develop alternative techniques to prepare ball mill feed. This paper describes industry experience in the design and layout of storage piles, reclaim systems, SAG and ball mills, sizing and classification systems and support systems. The paper includes examples from some modem concentrators.

FEED STOCKPILE A summary of recent selected stockpile and reclaim feeder installations is included in Table 1 . The primary crusher product, with a top size of up to about 300mm, is conveyed to a feed stockpile ahead of the SAG mill. There are two basic designs for the stockpile depending on the number of SAG mill grinding lines. A single line SAG circuit typically has a cone shaped stockpile resulting from a single discharge point from the feed conveyor. A multi-line SAG circuit will have an elongated or trapezoidal stockpile formed either from the movable discharge of a shuttle conveyor or a traveling head conveyor that is fed from the primary crusher discharge conveyor. A design example of a trapezoidal stockpile, together with associated reclaim system, is shown in Figures 1 and 2.

801

Table 1 Feed Stockpile and Reclaim Systems

It is very important to establish early in any project if another SAG mill line may be part of a future expansion. If so, then the design of the initial stockpile must allow for expansion. This typically means leaving sufficient headroom below the initial feed conveyor discharge point to allow for the required modification of the conveyor. It also means installing the future concrete pad and reclaim tunnel in the initial construction since the initial stockpile will cover the area needed later for the expansion. The objective of good plant operation is to operate the stockpile within the walls of the dead storage i.e. within the live cone of the pile. %s will minimize variations in feed sue to the SAG mill caused by segregation that results from coarser rock rolling to the outside of the pile. Some emergency bulldozing of dead storage is necessary but scheduled bulldozing is not good practice. Live storage capacity must therefore be sufficient to cover interruptions in ore supply from the mine. Even though in most large operations the mine also operates seven days per week, there will be interruptions in mine supply typically caused by, for example, scheduled maintenance of the primary crusher and feed conveyor. Table 1 illustrates a range of 17 to 24 hours live storage. Dust control at the stockpile is a major consideration. Dust at conveyor transfer points can be controlled by normal techniques but it is more difficult to control dusting from the discharge point of the feed conveyor onto the pile. Pant legs have been partly successful but they are vulnerable to damage with consequent high maintenance. “Fogging” systems also help. It is important to orient the stockpile and concentrator so that prevailing wind will carry dust away from the concentrator, support buildings, transformers and switchyard. A stockpile cover is expensive and only partly successful in controlling dust in a h g h crosswind. However the use of flexible curtains at all access points can improve this situation. The need to cover the stockpile should be evaluated in each case based on: environmental requirements and regulations the dusting characteristics of the material wind strength and direction rainfall and snowfall

802

803

[

Y -I-

I&-

lia ! E

804

RECLAIM FROM THE STOCKPILE SAG mill feed is withdrawn from the stockpile by multiple feeders. The feeders are typically either apron or belt feeders. Apron feeders are more common and have been the traditional feeder for this application but, in recent years, there have been successful installations of the lessexpensive belt feeder. The choice is a trade-off between higher-cost apron feeders and lower cost, but possibly higher-maintenance belt feeders. Feeder arrangement influences the live capacity of the stockpile. The choice is either in-line with the SAG mill feed conveyor, or at right angles, or at right angles discharging onto short cross-belts that feed onto the SAG feed conveyor. Each arrangement results in a different percentage of live capacity to total capacity; increasing with more feeders and as the base area of Withdrawal increases. Normally there are a minimum of three feeders per SAG mill line and, for maximum flexibility to withdraw from different areas of the stockpile, each of the three feeders will sufficient capacity to feed the full design tonnage rate of the SAG mill. A similar ratio of unit capacity will apply if there are more than three feeders per SAG line (e.g. two out of six). Good design features of reclaim feeder layout that have proven of value at a number of operations are as follows: 0

0

0

0

Hang the SAG mill feed conveyor from the roof of the reclaim tunnel. Spillage is a problem around feeders and conveyors. A clear area under the feeders and conveyors facilitates clean up including the opportunity to wash down with hoses. Dribble conveyors Separate dnbble conveyors under the feeders are not required if the feeders are installed in-line with the SAG feed conveyor. Short cross conveyors will also act as dribble conveyors if a more complex layout of feeders is justified. Escapetunnel An escape tunnel is mandatory in all installations. Tunnel floor slope The tunnel floor should preferably slope down to the tunnel outlet to allow drainage of wash-down slurry.

The reclaim feeders discharge onto the SAG mill feed conveyors. SAG mill control systems will signal for a change in feed rate. The feeders are variable speed but the lag time between a signal reaching the feeders and the effect being reflected at the SAG mill is often too long for good control. It is therefore common to fit a variable speed drive to the SAG mill feed conveyor to shorten the response time of a change in feed rate.

SAG AND BALL MILL LAYOUT RELATIONSHIPS A summary of selected grinding mills, classification and pebble crushing systems is included in Table 2. Ths table is based on the original installation and does not include later modifications. The main choice in mill layout in two-stage grinding circuits is between locating the mills side-byside in a single grinding bay or locating the ball mills in a second bay. The number of ball mills associated with each SAG mill will influence the layout. Initially it would appear that the single-bay arrangement is the most economical layout in every case because of savings in building cost and associated facilities such as overhead service cranes. Indeed, in lower capacity grinding circuits such as Century Zinc illustrated in Figure 3, in which only a single ball mill is required, a side-by-side arrangement is the layout of choice because of the cost savings with a single grinding bay. However the potential for future expansion must always be considered because, as stated in the previous section, the stockpile layout is influenced by the

805

distance between the center-lines of the SAG mills and hence the location of the feed conveyors and reclaim tunnels. The addition of another grinding line in a side-by-side arrangement is simple with the ball mills located outboard of the two inside SAG mills. However further expansion with a third line will be complicated and costly because the SAG mill feed conveyors will be spread fkrther apart. As ball mills continue to increase in size and power, so does the opportunity for increased application of the single SAG plus single ball mill circuit.

Table 2 Selected Grinding Mill, Sizing and Pebble Crushing Equipment

The layout will change if there are multiple ball mills for each SAG mill. This is the case for most of the recent high-capacity grinding circuits installed in the copper industry. The two-bay layout is preferred over the side-by-side layout for the following main reasons:

0

the ball mills can be located side-by-side in the second bay and, in a multi-line circuit, h s reduces the distance between the centerlines of the SAG mills. the height of the ball mill foundations will be less than a side-by-side arrangement because the distance from the ball mill and SAG mill discharges into the common pump sump is reduced. a lower ball mill elevation means, in turn, a lower ball mill cyclone elevation with consequent savings in pump power and pump maintenance.

The two-bay layout at Cadia is illustrated in Figure 4. This circuit includes a 40ft.dia.x 22ft. long SAG mill and two 22ft.dia.x 36%. long ball mills. The SAG mill is the largest installed to date. Accurate distribution of ball mill feed is a problem. The flow volume in some modem applications exceeds 5,000 m3/hr and a distributor or splitter box has to accurately divide mass, volume and size distribution to each ball mill. A common technique in a SAG/two ball mill circuit is to discharge both the SAG mill and ball mills into a common sump. This essentially relies on turbulence to maintain a good distribution of slurry characteristics and does allow a reasonable split to two operating ball mill cyclone feed pumps drawing from the common sump. Sampling at operating plants has shown that this split is never ideal but is generally “good

806

enough’. The discharge f?om the ball mills, operating at circulating loads typically exceeding 300%, is obviously the dominating flow but reversal of SAG mill direction can often be detected by a change in split to each ball mill. Copperton is different and has two sumps (one for each ball mill) fed kom a splitter on the SAG discharge.

Figure 3 Century Zinc - Single SAG Single Ball Mill - Side by Side Layout Some less common circuits have more than two ball mills per SAG mill. These circuits obviously have a different power split between SAG and ball mills caused by the ore being softer for SAG milling but still needing a higher ball mill grinding power. Examples are:

0 0

Freeport No.4 that has four ball mills following a single SAG mill. Antamina that has three gearless ball mills following a single SAG mill Escondida 3.5 Expansion and Escondida IV that both have three ball mills following a single SAG mill.

Accurate distribution to three or more ball mills is difficult and the distribution system at each property is different: Freeport No. 4. This circuit has the advantage of a significant difference in elevation between the SAG mill and the four ball mills that allows gravity flow of the SAG mill discharge via an Soft. radius curved launder to a four-way splitter box ahead of the ball mills. The original split was not accurate due in part to the high velocity in the curved launder. The imbalance caused an efficiency loss in the ball mill circuit. Later modification has considerably improved the accuracy of the split. This success was achieved with one of the first applications to slurry systems of computational fluid dynamic (CFD) modeling techniques. This is a promising new design tool that simulates the fluid and solid flow dynamics of slurries. ( l )

807

Antamina. The split to three ball mills was achieved by the use of a weir system inside a threeway splitter box. Each of the three streams cross a weir and the depth of each weir can be adjusted by a rotating lobe. The owner has filed a patent application for this splitter. Escondida IV (and Escondida 3.5). SAG mill trommel and screen undersize are pumped separately to a three-way pressure distributor ahead of three ball mill sumps. This system was selected as the best compromise in a difficult application because the additional pumping stage, plus the need to pump undiluted coarse SAG mill product, is less than ideal. Therefore a complete standby pump and distributor will be installed at Escondida IV. Normally the discharge from the SAG mill is “lubricated” by the substantially higher volume of much finer ball mill slurry. Tlus “lubrication” is lost when pumping minus %,, SAG mill product alone, resulting in increased pump wear.

SAG DISCHARGE SIZING AND OVERSIZE RECIRCULATION The two common SAG sizing systems are trommel screens attached to the SAG mill and separate vibrating screens. Trommel screens have high capacity and are significantly less complex and less expensive to install that vibrating screens. Some early installations, for example at the Copperton concentrator of Kennecott, have been very successful. Other experience has been mixed, particularly when screening larger material from larger grate openings and pebble ports. There have been some famous failures caused by insufficient structural strength. Recently trommels have been beefed up and have increased torsional stiffness and their operating record has improved. Some trommels were undersized resulting in poorly washed oversize pebbles that gave problems in the pebble crushers. An example of a grinding circuit layout based on trommels is shown in plan in Figure 5 and in section in Figure 6 . Screens have limited capacity that sometimes does not match the duty required. The current maximum conventional vibrating screen size is 12 ft, wide x 24ft.long and smaller SAG circuits that only require one operating screen normally incorporate a second screen as a standby unit. Both the operating and standby screens are mounted on common rails so that the screens can be quickly switched for maintenance. The change typically only requires uncoupling and re-coupling the wash water system and the motor wiring. An example of a grinding circuit layout based on screens is shown in plan in Figure 7 and in section in Figure 8. Some high-capacity SAG mills require multiple screens. Extra headroom is required for the splitter that will accurately divide the discharge of a bi-directional SAG mill. This in turn normally has the costly effect of raising the SAG mill. Freeport No.4 has three screens mounted on common rails of which two screens operate and one is a standby. Sufficient headroom was naturally available because of the ground contours. The Escondida IV SAG circuit will have an interesting combination of short trommel screen followed by a vibrating screen and the operating results of this circuit will be followed with interest because it presents the opportunity to have both high screening capacity and the ability to produce a clean washed pebble for the pebble crushing circuit.

Los Pelambres plan to add pebble crushing to an existing SAG circuit that has trommels on the SAG mills. They plan to add a separate washing screen to ensure that pebbles are clean before crushing.

808

Figure 4 Cadia - SAG Mill Plus Two Ball Mills - Two-bay Layout PEBBLE CRUSHING Pebble crushing is now a very common feature of SAG mill circuits. Pebble crushing has often been justified by “the need to break down the critical size” produced in a SAG mill. This justification is sometimes not easily quantifiable but there is no doubt that pebble crushing has a similar effect to an increase in ball charge or ball size in that it decreases specific power consumption, and increases throughput and product size. The very significant advantage of pebble crushing is that it permits some external operating control of the SAG mill. The value of a well-designed pebble crushing circuit is often underrated. Too many circuits have been under-designed to reduce cost. Pilot testing typically shows that softer ores produce few pebbles and a pebble crushing circuit is not justified. In such cases, it is normal to allow for future installation of a circuit if there is evidence that the ore will harden. Harder ores typically produce more pebbles, often as the open-pit deepens or, as in the case of Collahuasi, production moves to a new ore body.

809

Plant operators have been hghly critical of the false economy of “cheap” pebble circuit design and their preferences are very clear. They include: A preference for multiple smaller crushers instead of a single large crusher. For example a single Nordberg MPlOOO fed at less than about 350 mtph may be damaged because the bearings do not seat properly. The upper capacity is about 600 mtph. Two MP 800s have a capacity range from about 150 mtph up to a maximum of about 1100 mtph plus the significant advantage of increased overall pebble crusher availability. Adequate pebble surge and storage capacity. Pebble production rates can vary widely as the ore changes. Pebble storage absorbs fluctuations in pebble production and assists steady operation of the SAG mill by feeding the pebbles back, either crushed or uncrushed, at a controlled rate. This helps maintain a steady grinding charge in the SAG mill. Ideally pebbles are stored in a stockpile ahead of the crushers. Cadia, having operated a pebble crusher since start-up but have recently added a 5,000 ton pebble stockpile and report a significant improvement in SAG circuit control. Some pebble storage capacity is essential. If a pebble stockpile cannot be justified then bins ahead of the crushers must suffice. A minimum of one-hour storage is recommended. There are some unusual circuit variations. Andina has a particularly interesting and flexible pebble crushmg circuit. A reversible conveyor after the crushers either returns crushed pebbles to the SAG mill or advances them forward to the ball mill feed chutes. The advantage of this circuit is flexibility of operation that allows better balance of power between the SAG mill and ball mill depending on the hardness of the ore. Pilot testing of many ores has consistently shown, particularly with harder ores, that SAG capacity increases substantially if crushed pebbles are forwarded directly to the ball mills. This circuit has been evaluated and studied a number of times but is not common practice in the industry. Inadequate control in the ball mills may be an issue here plus complications in layout. It is simpler to return the crushed pebbles to the SAG mill where they occupy some space as they pass through the mill and grates and generally report to the screen undersize. Good grinding steel removal ahead of the pebble crushers is essential. Poor steel removal at some operations is the major cause of downtime because of damage to the pebble crushers. Typically a well-designed circuit will have a minimum of two stages of magnetic separation, plus a metal detector, ahead of the crushers. This is not an area to be slumped in design and the choice of magnetic separator should be based more on a successful record in the industry rather than price alone since some magnets have proven to be not sufficiently robust for this duty. Thls is again an example of “cheap” design. A combination of rectangular core cross-belt magnets and stationary magnets at conveyor discharges (to extract steel when airborne) are currently preferred. It is important that the pebbles and steel are presented to the cross-belt magnets as a monolayer on a flat section of conveyor. A tramp screen was installed at La Candelaria and Chino as additional protection in the event of a grate failure. The screen removes pieces of the grate and the additional grinding steel discharged through the opening in the grate. Chino and La Candelaria also have copper-bearing magnetite in the ore that is removed with the steel balls by the magnets. No totally successful separation technique has been developed. Chino minimizes copper loss by allowing the steel and magnetite circulating load to build up and then periodically purge the circuit of both steel and magnetite. La Candelaria has tested an inclined conveyor with some limited success.

810

Los Pelambres have tested ring magnets inside the SAG mill trommel. These have been useful as a first stage of steel removal but in their proposed pebble crushing addition they plan to install two extra stages of magnets plus a metal detector ahead of the crushers. High-angle conveyors have been installed and successfully operated in pebble crushing circuits at a number of circuits notably Chuquicamata, Collahuasi and Batu Hijau. This type of conveyor allows greater flexibility in circuit design. Pebble crushers have been located over the SAG mill feed conveyor in some circuits so that crushed pebbles fall directly onto the conveyor. This apparent advantage is severely compromised if the SAG mill feed conveyor must be stopped for crusher maintenance.

BALL MILL PUMPING AND CLASSIFICATION The SAG mill undersize is joined by the re-circulating ball mill discharge and is pumped to a cluster of ball mill cyclones. The most common design of ball mill cyclone feed sump for a SAG mill followed by one or two ball mills is a shared single sump that receives both the screen or trommel undersize from the SAG mill and the discharge from the ball mills. The sump can either be located in the SAG mill bay or the ball mill bay. Figure 7 shows the sump located in the ball mill bay for a SAG mill fitted with a trommel screen and Figure 8 shows the sump located in the SAG mill bay for a SAG mill fitted with a vibrating screen. The optimum location should be evaluated in each case and may be influenced by the relative elevations of the SAG and ball mills. The ball mill cyclone feed pump system must be carefully designed. Flow volumes in large concentrators are very high and the slurry is abrasive. Pumps are now available with dnves of up to 2,200 kW as will be installed at Escondida IV. Pump wear is a major issue. Th~sleads to the need to consider three major items:

0 0

careful selection of materials for casings and impellers a sound standby pump system good maintenance practice.

A preference has developed for metal pumps for tlus duty that are typically white cast iron or high chrome steel. A contributor to high pump wear is chip steel discharging from the ball mills. Chip steel re-circulates through the ball mill cyclones sometimes plugging the apexes. This results in coarse un-ground material being thrown into the downstream process. Some recent installations of circular magnets in ball mill trommels have been most effective in removing chip steel, resulting in reduced pump wear and less plugging of the cyclones. The standby pump system is a key issue. There are three main alternatives as follows: 0

0 0

an installed spare pump for each operating pump, fitted with either: o valves on the suction and discharge of each pump o or Tech-Taylor style ball valves on the discharge o or movable rubber hoses on the pump discharge that can be switched to either Pump an installed spare pump with a separate discharge line and cyclone cluster for each pump. an uninstalled spare pump, stored in the same area as the operating pump, and fitted with quick-disconnect flanges on the suction and discharge lines.

81 1

812

813

Pump sizes and line diameters have increased as flow volumes have increased. Large valves and heavy rubber hoses become more difficult to operate and manipulate. Duplicate pump and cyclone installations are expensive. For these reasons the uninstalled spare option has become a more popular choice for larger concentrators. Ball mill cyclone clusters typically consist of up to 20 units in a radial arrangement with common overflow and underflow collection boxes. The most common cyclone size to match flow and size split in copper flotation circuits has historically been 26-inch diameter for an overflow product in the size range of 80% passing 120 to 160 microns. However in some high capacity plants the flows have become too high for a single cluster of 26-inch cyclones and this has resulted in the increased application of 33-inch cyclones. These larger cyclones have higher unit capacity and can be configured to give a similar size split as 26-inch cyclones. Los Pelambres plan to change out existing 26 inch cyclones for 33-inch cyclones to better match the current flow volumes whilst still maintaining the present split size. Both Batu Hijau and Escondida IV have a coarser overflow of about 80% passing 210 microns and have 33-inch cyclones. Ball mill cyclones are installed above the feed end of the ball mills. The underflow discharges into the feed chute of the ball mills and overflow discharges to the downstream process.

GENERAL CONSIDERATION IN THE GRINDING AREA The grinding circuit can either be installed in a covered building or in the open. Climate normally controls this decision. Most grinding mills worldwide are considered to need some form of protection and are installed in covered buildings. There are however a number of circuits in Australia that are not covered, notably the installations at Ernest Henry, Olympic Dam and Century Zinc, where low rainfall and moderate climate allowed major cost savings by deleting the building. Spillage and ease of cleanup is a major design issue. Most plants have sump pumps to assist in cleanup but spills in large grinding circuits swamp these relatively small vertical sump pumps and their value is doubtful. Good design features for grinding bay cleanup include:

0

steep floor slopes trench across the floor accessible by front-end loader large diameter drainage lines from the grinding bays to emergency ponds outside the concentrator if the topography allows.

SUPPORT OPERATIONS Ball Loading A concentrator will have steel grinding balls delivered in three or four sizes, ranging typically fiom a minimum of about 1.25 to 1.50 inch diameter for regrind mills up to, more recently, 6 inch diameter for SAG mills. More commonly there are two sizes - 4inch diameter or larger for the primary mill and 2.5 inch diameter or larger for the secondary ball mills. Each size will be dumped into a separate bin that, depending on location, will hold anywhere between a week to a month’s supply.

814

815

816

Recent designs are based on ball counters for the coarser sizes that are fed to the SAG mills and belt scales for the smaller sizes fed to the ball mills. The bins are rubber lined and typically slope to an adjustable outlet gate from which the large balls for the primary mills migrate to a slowly rotating ball feeder fitted with pockets that feed the balls individually onto a transfer conveyor or directly onto the SAG mill feed conveyor. Smaller balls are typically fed to the regrind mills by more traditional ball buckets or electromagnets. Designing to Simplify Operations and Maintenance There are desirable design features that simplify operation and maintenance, some of which are:

0

leave sufficient space around the mills for fork-lift traffic during liner changes set the operating floor level to afford easy access to the mills the operating floor must be strong enough to support fork lifts and small mobile cranes the operating floor should be at one level

Cranes Grinding bay cranes are selected on the basis of lifting the heaviest component of the mill or motor. Each grinding bay has a dedicated crane. If a single crane is selected then typically there is a big hook for large lifts and a smaller high-speed hook for maintenance duties. An alternative is to install a separate high-speed crane. The cranes are normally used in the construction phase to install the grinding equipment. Antomina is an interesting exception because, in this case, the savings in schedule justified transporting a heavy-duty mobile crane to a remote site in Peru at an elevation of 4200m to allow separate installation of the mills and cranes. Liner Change Changing out worn mill liners is a major component of overall mill downtime. The mass of individual mill liners has increased along with the increase in mill sizes. The use of liner machines is now universal in larger plants. Liner machines are mounted either on rails or tires. The grinding mill layout must accommodate easy maneuvering and insertion of the liner machine into the mill and sufficient space must be available on the operating floor close to the mills to store the new liners. In this regard the design of the mill feed chutes is important. The chutes are typically rock boxes with chrome-moly liners and are mounted on embedded rails in line with the mill. They are drawn back to allow insertion of the liner machine. Escondida and Cadia have self-powered feed chutes for faster removal. Bolt breakers and bolt punchers are used to remove the old bolts and liners. This equipment is normally suspended from monorails above the mills. Launders The hydraulic design of slurry conveying systems can be complex due to the many variables in slurry composition, physical design parameters and the h g h volume flows encountered in today’s plants. Launder slopes are critical. If too shallow then the slurry will not flow. If too steep then wear rates will be unacceptable. Computer programs are typically used to calculate sanding limit velocities based on a normal depth of flow in free-surface slurry systems. However good

817

engineering judgment and experience in open channel flow is still essential for success in tlus critical area of plant design. Mill Lube System Mill lube systems are complex and typically consist of oil reservoirs, high and low pressure pumps and associated piping, valves, and control systems. The systems are normally housed in a dedicated room located on the basement floor of the grinding circuit and as close to the mills as feasible to reduce pipe runs. Mill Cooling Water It is essential to use good quality water in the mill lubrication and motor cooling water systems. It is normal to use a closed-circuit system to conserve h s water and hence coolers and chillers are an integral component of the systems. It is important to maintain a constant air gap between the rotor and stator of large modem gearless ring motors drives. This requires careful control of the motor cooling water temperature and for this reason chillers are nonnally installed in this circuit. The control of lube system water temperature is less critical and atmospheric cooling towers suffice. References: 1. J.M.Berkoe, K.K.Knight et al.,”CFD Modeling Of Slurry Conveyance and Distribution in c Concentrator Grinding Circuits” SME Annual Meeting,Denver, CO, Feb. 2001.

818

Selection and Evaluation of Grinding Mill Drives George A. Grandy’, Craig D. Danecki2,and Peter F. Thomas3

ABSTRACT When considering the mill drive system for a grinding application, a large number of questions must be answered. Depending on the mill size and type, the mill can be driven by several configurations of drive systems: a single low-speed motor connected to a pinion driving a ring gear, two low-speed motors connected to two pinions and driving a ring gear, or a gearless motor mounted directly. In addition, a gear drive system can have a reducer placed between a high-speed motor(s) and pinion(s) or a power path splitting pinion stand. In some designs the reducer is integral with the pinion stand driven by a high-speed motor. The technical and commercial implications of several mill drive options are presented. Initial cost, cost of installation, cost of commissioning, and cost of operation and maintenance are presented for each drive option considered. The impact of each drive system on the design, cost and operation of the overall plant electrical system is discussed. Also, consideration is given to mill drive availability, and the potential cost resulting from delays in plant production. INTRODUCTION Over the past 30 years a number of papers have appeared discussing mill drives (see references). The primary emphasis has been on Semi Autogenous Grinding (SAG) mills, particularly the variable speed options for larger mills. Recently, an interest in variable speed for large diameter ball mills has developed. The intent of this paper is to present the latest technology considered for SAG mills and ball mill drives, plus provide some guidelines on when to select fixed and variable speed drives. For the variable speed options, gear and gearless drives are included. DRIVE OPTIONS Single-pinion fixed-speed gear drives offer the simplest design and the lowest capital and operating costs. In the past these drives were restricted to smaller mills, however recent technological changes in gear manufacturing have extended the range of mill diameters. Similarly, dual-pinion fixed-speed drives offer reliable and simple designs with proven load sharing capabilities. These fixed-speed drive systems can be used on the largest ball mills considered to date. With variable speed, particularly for SAG mills, the drive system became more complex and expensive. In the not-too-distant past, the variable speed choices were few (namely dc motors), but improvements in solid state technology and controls has resulted in a number of higher-efficiency, variable-frequency alternatives that have pushed dc motors out of the picture. Ldcewise, modern slip-energy recovery systems for wound rotor motors have overcome high-energy losses generally associated with liquid rheostats. Taken as a whole, the engineer is faced with many choices for a drive system. Table I gives 16 different configurations for the range of mills covered in this paper. The induction motors, both squirrel cage and wound rotor type, will be high-speed design with 6 or 8 poles, depending on the ~~~

1 Kvaerner, Metals E&C Div., San Ramon, California. 2 The Falk Corporation, Milwaukee, Wisconsin. 3 General Electric, Peterborough, Canada.

819

Table 1 Basic Arrangements of Mill Power Transmissions ~

Drive

Speed

Type Type Fixed

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Driver Primary

No. of Driver Pinions Secondary

iing Gear

Single

Reducer Reducer n.a. Reducer n.a. Reducer Reducer n.a. n.a. n.a. Reduier Reducer

Dual Jariable Xing Gear

Single

Dual

I

Gearless

I

n.a.

n.a.

Motor Speed

Motor Type

Power Supply

Wound Rotor Direct Induction Connect Synchronous to Wound Rotor Supply Synchronous Quadramatic Wound Rotor SER Induction PWM Synchronous LCI Synchronous PWM Svnchronous CCV

900-1200 900- 1200 180-200 900-1200 180-200 900- 1200 900- 1200 180-200 180-200 180-200 900-1200 900-1200 180-200 180-200 180-200 I 9-15

1

Synchronous

1

ccv

choice of gears. The synchronous motors will be low-speed design with 30 to 40 poles depending on frequency and gear ratio. Synchronous motors used with cycloconverters on geared systems will have 8 - 10 poles. Figure 1 sums-up the options for all mill drives, and particularly allows the engineer to understand where the overlap between gearless and ring gear driven systems is found. For the comparisons, the SAG mill aspect ratio (effective D/L) varies from 2.43 for the 26-ft diameter mill to 1.83 for the 44-ft diameter mill. However, the ball mill aspect ratio (effective L/D) is approximately 1.6 for all mill diameters shown (16-ft to 28-ft).

30 I

, Gearless Range ,

25

20

6

I

8

Dual Pinion Limit

.

I

1;

....ii..:.. ......:.:.:.:.,.,.,..,.,..:..,...,., J....... ............... . . . . .. ......... . . . . ............. ....,.,.........:.:...............c:.:.:

: , / . : . . . j . . j . j . . :

,;f

I

15

B

B

, I

Single Pinion Limit 10

5

0

I 10

/Ball 15

G A G Mill

Mill 20

25

30

DIAMETER, FT

Figure 1 Mill Power vs Mill Diameter

820

35

40

45

50

Although not the subject of this presentation, the first step to determine the mill size and drive type is the calculation of the design process specific energy ( k W t ) , plant size (t/d) and total SAG and ball mill process (shell) power. The required process power is divided into circuits and numbers of mills within a circuit, followed by the selection of the mill sizes that meet the requirements. To select the drive type several factors must be considered before coming to a final decision. The selection process involves quantitative (economic) and qualitative (subjective) factors. Choosing the correct factors is difficult and for most projects one of the most important in selecting the proper system. Fortunately, the engineer has a number of analytical tools to resolve the economic issues. Given the correct (and sometimes incorrect) assumptions, some subjective factors can be transformed into economic factors. However, in the end it may be the unresolved subjective factors that will determine the selection, even after exhaustive analysis of the capital and operating costs. These factors are discussed below. PROCESS CONSIDERATIONS Why Variable Speed? For SAG mills the need for variable speed depends on a number of factors. With variable speed drives the operator (or control system) can rapidly react to changes in ore characteristics, be it ore hardness or feed size distribution. Soft andor fme ore can result in a low total charge leading to liner damage and accelerated ball and liner wear. This condition can be corrected by increasing the circuit feedrate only when downstream conditions permit such changes. Otherwise it must be corrected by reducing the speed of the mill to force the balls to impinge on the charge and not the liners. When grinding out a SAG mill, variable speed is valuable for the same reasons. Variable speed drives also provide the advantage of slow starts (stops) of the mill and, for some systems, inching of the mill. However, for most overflow ball mill circuits, variable speed is only valuable when circuit feedrate control is required downstream. Without downstream constraints, ball mills are typically operated with a maximum ball charge and a fixed-speed, thus negating any need for variable speed. SAG and ball mill circuits are designed such that the circuit capacity is ball mill limited over most of the range of ore hardness expected from the ore body. Most variations are compensated by the variable speed drive on the SAG mill. In rare cases, the ore may be so variable that the speed (or power) range for SAG mill cannot compensate for an extremely hard (or soft) component, resulting in a need to reduce the power (or speed) of the ball mill to avoid over grinding and affecting downstream processes. In a few circuits, the operator can balance the SAG mill and ball mill by directing a portion or all of the crushed SAG-mill pebbles to the ball mill. Other plants allow partial cyclone underflow recycle to the SAG mill to further improve the balance of power. It should be noted that mine planning, ore blending and blasting practices play an important role in the design of milling circuits. Modern plant designs recognize that feed characteristic control is not perfect and that variable speed SAG mills are necessary for any successful operation. In nearly all cases, regardless of the SAG mill diameter, variable speed is not an issue for evaluation and is accepted as the standard design. On the other hand, variable speed ball mills require careful evaluation and in most cases cannot be justified based on process considerations. It is only the cases where gearless drives are considered that trade-off evaluations are required between variable speed and fixed-speed drives for ball mills and between gearless and gear-driven variable speed drives for SAG mills. Plant Layout Plant layout is a factor for evaluation when comparing geared and gearless drives. Obviously, the footprint for gearless, single-pinion, dual-pinion gear drives is different, as are the installed costs. Single-pinion drives require the least amount of space, but only marginally less than gearless drives. Dual-pinion drives have the maximum space requirement. In a single SAG mill circuit with

82 1

two ball mills, the differences are generally very small. However, for multiple line plants, the differences can be significant and require evaluation. Another factor for consideration is the position of the drive. Gearless drives are normally mounted on the shell near the feedend of the mill, while gear drives are typically installed on the discharge end of the mill. Spillage and drive contamination become an important consideration, particularly if the dnve is mounted below the cyclone circuit. Reliability, Availability and the Cost of Downtime There are many misconceptions concerning the reliability of the various drive systems and their impact on overall plant availability. Gearless drives have been presented as the “perfect” system that does not have the “inherent” faults associated with gears (misalignment and wear). The fact is that the differences are difficult to measure and, in most cases, subjective. Failure rates for components are better known for gear%because of the long history, whereas for gearless, the history is relatively short and no installation has come close to the full life. Proponents of gearless motors fault gears as having an impact on the plant availability because of the need for periodic inspections and re-alignment, neither of which is true. With modem infrared heat sensors, vibration accelerometers, and automatic lubrication systems, gears and bearings can be monitored for alignment (increased temperature differential across the gear face width) and unusual vibration, negating the need for visual inspections. Modem gear sets have substantially longer life than equivalent sets manufactured just a few years ago. Pinion life can be accurately predicted, and maintenance and/or replacement can be scheduled to have no impact on the overall availability of the plant. Catastrophic failure of gears and motors (gearless and gear driven mills) can occur, but the likelihood of such failures is extremely small. Ring gear failures occur generally over a long period of time, allowing early detection and planning for replacement or adjustment. Gearless motors have failed, requiring pole replacement, which resulted in extended downtime. Overall, the differences in availability cannot be accurately measured and should be considered equal. Reliability (or the perception of) has an impact on the capital cost of the plant because of the investment in critical spares. This is a factor for evaluation that must always be addressed. Gearless critical spares are generally associated with the components in the e-house. Spare pole pieces for the motor are sometimes considered. Gear drives have a different set of components, including elements in the e-house, spare motors, pinions, clutches and in some cases a spare ring gear. The need for a spare ring gear requires careful evaluation. There are many gear sets that have been in operation for more than 25 years without failure, and these gears did not benefit fiom modem manufacturing technology. If a difference in availability can be proven, the cost of downtime should be honestly measured. Operating costs are made up of fixed and variable components. When a circuit is down, only the fixed costs should be considered. The other cost component is the profit lost by not producing product, i.e., the lost opportunity for profit. This varies substantially from plant to plant. A plant with a low-grade feed and higher operating costs will have a smaller profit margin, whereas as plant with high-grade feed and lower operating costs will have a higher profit margin. With this evaluation procedure, each case will have a different total cost for downtime. Simply taking the total lost product production times the refiied metal selling price, as is often done, is very misleading and will result in an incorrect conclusion. Table 2 demonstrates a method for determining downtime costs. The analysis must match the plant with the mills sizes that are being evaluated. The 36‘ SAG circuit, which includes two 22‘ ball mills, will have a capacity of 30,000 todday (tpd) to 40,000 tpd for medium to hard ores, but can have capacity of 70,000 tpd to 80,000 tpd with very soft ores. Care must be taken to use the right assumption for plant capacity, etc., when calculating the net present value (NPV). When comparing plants with the same mill size but different capacities, the concentrator unit costs will be different because the grinding energy costs are essentially fixed (same power and steel consumption), but the feedrate is different. Grinding can be between 50% and 80% of the concentrator cost, depending on the ore hardness.

822

Table 2 Cost of Downtime Spreadsheet

Sten 1

- Calculation of nlant feedrate Plant nominal capacity Plant schedule for operation Assumed availability Plant feedrate = (t/d) I [ (h/d)*(availability) ]

t/d h/d % tlh ore

40,000 24 93% 1,792

40,000 24 93% 1,792

%Cu Yo

2 .O% 92% 32.975 72,710

0.6% 92% 9.892 21,813

Sten 2 -Calculation of mnner nroduction from nlant Assumed ore grade Assumed copper recovery Copper production = (th ore) * (%Cu) * (Recovery) Copper production = (th Cu)*2205

m

Sten 3 -Calculation of cost to nroduce conner Assumed mining cost Assumed Concentrator costs Assumed other costs (administration, infrastructure, etc.) Assumed smelting costs including transport Total operating cost to produce copper metal Sten 4

th Ib/h

m m

3.10 3.25 1 .oo

0.08 0.08 0.02 0.27 0.45

3.10 3.25 1.00

20% 10% 100%

0.02 0.01 0.02 0.05 0.40

20% 10% 100%

- Partition copner cost into fixed and variable comnonents Fixed (factored) Mining (assumed fraction) Concentration (assumed fraction) Other (assumed fraction) Total fixed Variable (bv difference)

0.25 0.27 0.08 0.27 0.87

m 0.05 0.03 0.08 0.16 0.71

$& !Q

Sten 5 -Calculation of nrofit Assumed selling price Total operating cost to produce copper metal Net profit (selling price total cost)

-

Sten 6 -Calculation of Lost Onnortunilv Cost This is the cost for extra downtime above the assumed average plant availability. Lost profit from not producing copper Plant fixed cost (occurs even when the plant is down) Lost opportunity cost (lost profit + fixed cost)

0.90 0.45 0.45

0.90 0.87 0.03

$/lbCu

m

0.45 0.05 0.50

0.03 0.16 0.19

0.50 72,710 36,121

0.19 21,813 4,056

10.0 36,121 361,207

10.0 4,056 40,557

Step 7 -Calculation of $/h Cost of Downtime Lost opportunity cost Lost Copper production for downtime Cost of downtime (cost losses)

$/lb Cu Ib/hCu

Assumed extra downtime per year Downtime cost (Lost opportunity) Total annual downtime cost

MY

Uh

Sten 8 -Calculation of Extra Downtime Cost

Step 9 -Calculation of NPV over 20 years

$/h Sly

Plant A Discount Rate NPV

823

Plant B

For the calculation shown in Table 2, we are assuming the ore is the same hardness (medium) and the same capacity for a 36‘ SAG mill. The spreadsheet is keeping everything constant between the plants, A & B, in order to study the effect of feed grade.

ELECTRICAL CONSIDERATIONS The selection of motors for grinding mills is covered in Chapter 35, Selection of Motors and Drive Systems for Comminution Circuits, where the following topics are discussed in detail. Enclosure and Ventilation Fixed-speed Operation Adjustable-speed Operation Motor Construction Bearings Soleplates Drive Configurations and Motor types Fixed-speed -- Single Pinion and Twin-Pinion Wound Rotor Motor(s) Low-speed Synchronous Motor(s) Variable-Speed Drives Wound Rotor Motor(s) with Slip Energy Recovery (SER) Squirrel Cage Induction Motor with Pulse Width Modulated (PWM) Drive Low-speed Synchronous Motor(s) with Load Commutated Inverter (LCI) Drive Low-speed Synchronous Motor(s) with Cycloconverter (CCV) CCV Driven Gearless Drives Power Factor Power Distribution Considerations Altitude Considerations

System Requirements - Transformers, Capacitors and Filters Each of the drive systems has it own unique power supply requirements. This must be addressed when making evaluations and comparisons of the various drives, particularly the need for transformers. In addition to the power supply, the need for power factor correction with capacitors (or other means) must be addressed and the resulting costs included. System harmonics generated by variable speed drives must be included in any evaluation. PWM, LCI and CCV systems each have their own harmonic characteristics. The interaction of these drives with the overall system, especially other variable-frequency drives, must be addressed and the appropriate filters included in the design. Some remote plants are at the end of a fragile electrical system that can have an impact on the starting of large motors. Each motor type has a different requirement (i.e. starting inrush current) and resulting cost impact. For example, if an electrical “fly wheel” is required, the size and cost must be included in any evaluation Overload Capability Unless specified and included in the price of the equipment, no overload capability is available from any adjustable-speed drive system or from a fixed-speed induction motor drive. Fixed-speed synchronous motors, however, do have mherent overload capacity, since it is possible to exchange their reactive power capability for active power. Therefore a 0.8 leading power factor synchronous motor can supply a 25% increase in its kW output without exceeding its temperature guarantees. However, if the plant operators plan to use the built in capabilities of the motor, this must be addressed in the original design of the overall system, particularly the mechanical design of ring gears and gear reducers. In addition, the mill dimensions and operating design points (speed and charge level) must be specified for the initial and future operating conditions.

824

Clutches The air clutch is used extensively with single and twin-pinion low-speed drives using synchronous motors. This device permits the motor to be started uncoupled, thereby enabling the use of low inrush current design motors, and also avoiding any torque amplifications during the acceleration of the motor. With weak power systems, reduced voltage starting of the motors is possible, and on twin-pinion drives each motor can be started separately, thereby reducing the impact on the power system. Maintenance Considerations Electrical equipment today are designed and built to be as free from maintenance as possible. Electronic solid state equipment has virtually no wearing parts, and is not expected to fail or wear out in normal operation. However, it is customary to carry an inventory of components for an adjustable-speed power supply. For motors of a geared drive, spare bearings, brushes, brush holders, brushless exciter components, and often a spare wound stator and rotor are mandated. Where more than one motor of the same rating is installed, one set of spares will be shared between them. For a gearless drive, the cost of a spare stator would be prohibitive so this is not called for. Spare rotor poles can be carried, and with some designs stator winding components can also be carried. Generally speaking, routine maintenance on geared dnves can be done during normal planned mill outages. Unplanned outages due to component failure can usually be handled in a relatively short time. Back-up systems can provide for continued operation until a planned outage can be arranged. With gearless motors, a stator winding fault will cause an extended outage which can range from two to three to several weeks, depending on the type of construction used by the motor manufacturer. MECHANICAL CONSIDERATIONS Present Power Capabilities Just a few short years ago, the perceived maximum pinion power capacity was 4,476 kW (6,000 HP). This changed in 1994 when the first 6,714 kW (9,000 HP) drives were designed, manufactured, and installed. The installation consisted of one 13,428 kW (18,000 HP) SAG mill driven by two, 6,714 kW (9,000 HP) pinions and two 6,714 kW (9,000 HP) ball mills each driven by a single pinion. This increase represented a 50 % increase in power, but as mentioned in previous papers, (see references) power is not the correct criteria, but rather, torque should be used to assess a gear drive’s capabilities. However, to simplify our discussions, the process section of this paper addressed mills in terms of power and speed which are the two variables of torque. We will refer to power for the remainder of this paper to simplify and keep our discussion in line with the examples presented. Today the maximum power capacity per pinion is approximately 10,000 kW (13,404 HP). This could increase by increasing some of the variables that define the design of the ring gear set, i.e., ring gear hardness, face width, center distance, etc. In 1994, manufacturing capabilities limited the power capacity to approximately 7,460 kW (10,000 HP). The case hardened and ground pinions were the limitation in manufacturing. To obtain power capabilities above 7,460 kW (10,000 HP), a coarser pitch tooth was required for the gear set to meet the tooth strength criteria. A larger gear grinder was required if coarser teeth were to be obtained. Additionally a larger gear checker would then be required to verify the quality level of the pinion tooth form. Typically the American Gear Manufacturers Association (AGMA) quality level of Q- 12 is specified. Surfaces finishes have improved dramatically. As - cut, through-hardened pinions would obtain finishes in the range of 125 to 250 Ra, while the current generation of ground pinions are achieving 24 to 40 Ra finishes.

825

Reliability The reliability of the ring gear set is greater today than it has ever been. Much of this can be directly attributable to the engineering firms and the mill OEM’s that establish specifications in an attempt to place gear manufacturers on equal footing for quoting purposes while at the same time obtaining a quality product. Unfortunately, all gear manufacturers are not the same, and quality can very widely among them. Reliability starts with obtaining a quality product. Once the product is installed and good alignment is obtained, proper maintenance will ensure the reliability and availability of the mill for many years of successful operation. The life of the ring gear set is designed for a minimum of 25 years. Many gear sets have operated successfully for much longer. The typical life for a carburized and ground pinion is better than 12 years, and it may be closer to 18 years. The life that is experienced by the ring gear and pinion is directly attributable to the improved materials and overall quality of the components.

Ring Gear Quality The quality begins with the rough material. The gear casting represents approximately 60% of the total cost of the gear set. Material selection to obtain the proper hardness through the root of the tooth is critical. All of the rating formulas are based on the assumption that the hardness at the root of the tooth is within the minimum hardness range used to rate the gear set. Unfortunately, just measuring the outside of the rough, whether it is a casting, forging, or plate, does not guarantee that the root of the tooth will be within the specification. Significant testing of heavy sections, which represent the total thickness of the casting, needs to be performed. These heavy sections are used to verify that the hardness gradient from the surface to a depth just below the root will not fall outside of the specification. Once a ring gear design is complete, FEA (finite element analysis) can be performed if necessary to assure that the section sizing is correct for the deflections that occur under load in operation. Next, another form of FEA is used, only it analyzes the solidification of the gear structure over time to provide assurance that the casting will be sound and defect-fiee. In steel castings, this is guaranteed by using ring risers and inside wedges. These practices assure that good directional solidification will occur. Directional solidification is important because steel loses 6% of its volume as it is solidifying. The riser must continue to feed the ring gear while the ring gear sections solidify from the drag side of the casting (bottom), up to the cope (gear riser interface). If this is not performed correctly, the casting will not pass the ultrasonic inspection. Gear castings are an unusual shape compared to most castings that are poured in typical foundries. It takes a foundry skilled in the casting of ring gears to make certain that the highest level of quality can be obtained in the rim area. To maintain high reliability, the ring gear quality is verified by performing ultrasonic inspections on the rim area (tooth section), along the split flanges (the flanges used to bolt the gear sections together), and the mounting flange (the flange that mates the ring gear to the mill flange). These areas are the most critical, especially the rim area near the tooth, due to the stresses that occur in these sections under operation. Once the casting has been inspected and its quality has been confirmed, machining begins. Just as the quality of the pinion is important to the reliability of the gear set, the machining of the ring gear to high quality standards is equally important. Ring gears are manufactured to tight runout levels and the ring gear teeth, on high power mills, are typically manufactured to AGMA Q- 10 quality levels. High quality levels are important to assure smooth operation, which will extend gear, pinion, and bearing lives. It is important to understand that it is very easy to fall outside these high tolerance levels. It is necessary to verify these^ tight quality levels with state-ofthe-art inspection equipment. This will ensure that the gear set will function properly at startup. The only attribute that cannot be measured is the ring gear lead. This attribute can be verified by checking the pinion lead on the proper checking instrument and verifying the ring gear’s lead by performing a fixed center contact check.

826

Before 1994, gear manufacturers were unable to accurately generate tooth forms on gears larger than 12 meters. Form cutting machines existed, however their accuracy, or tooth quality levels, did not meet AGMA standards for quality 10 gearing on these larger gears. Presently the gear cutter of choice to obtain high accuracy on large diameter ring gears is a MAAG type shaper cutter. The largest MAAG shaper cutter that has been manufactured can cut gears up to 14 meters in diameter which is large enough for a 40-ft mill. Pinion Improvements Pinion forgings have improved using ladle treatments, vacuum degassed and vacuum arc melt materials. Today forgers follow ASTM standards and ultrasonically inspect pinion forgings. Another improvement that has occurred to improve a ring gear sets reliability has been the application of tooth modifications to the pinion (Figure 2). These modifications applied during the tooth grinding of the pinion can have as much as a 30 to 50% improvement in herzian contact stress under misaligned conditions. Pinion Tooth Modifications ~

Mis-Alignment v s Contact Stress

1,000T

900

--

600Ii

-

0

~

A-

0

--_ *--

-$

500 -.

~

Q

0

.

-4-

-m-No~IcATIoNs -f

0

Q

400

-----A

*----+----+-----

CROWN ONLY NNCR3W4,MNPRORE

+MUCROWN, M

X PROFL

7

- 152

-.lo2 Mtor h d @en

- 052

- 002 Pinion Mis-Alignment(mm)

048

098 Mil End Cpen

14

Figure 2. Pinion Tooth Modification Graph If separate suppliers are used for the pinion and the ring gear, it cannot be assured that the two items will mate properly when installed at the mine site. One way to address this situation would be to have the pinion forwarded to the ring gear manufacturer’s facility for a final contact test. Timing is critical and could result in an unacceptable schedule delay if one of the components is not completed in time for the final contact check. Maintenance Availability to grinding facilities is vital to their profitability. Much has been misstated about the availability of mechanical drive systems. The fact is, no other drive system has demonstrated a better availability than gear driven systems, simple maintenance program can alleviate downtime and boost availability of mechanical systems. These maintenance procedures can all be performed dynamically while the mills are in operation. Others can be performed during routine shutdowns for other reasons such as conveyor maintenance or during liner replacement.

827

Wash downs of a modem ring gear for inspection are not needed. The maintenance program should follow the dynamic and static methods listed below: Dynamic: Stroboscopic viewing - This provides a good picture of the condition of the gear and the lubricant coverage. Infrared temperature measurements These can be performed either manually or automatically with readouts transferred directly to the control room. This procedure monitors alignment, which directly affects load distribution across the face width. If a shift in alignment occurs, it will usually take place over an extended period of time. This provides the operator with time to establish a schedule to make any adjustments needed during regularly planned shutdowns. Vibration monitoring - This can be accomplished either manually or automatically by mounting transducers on the pinion pillow block bearings and feeding back the signal to the control room. Typically a spare pinion is purchased to facilitate the change-out when the bearings wear and require change-out. These spare pinions are usually reused among the various mills at a facility.

-

Static: Visual examination - This is performed when the mill is stopped, usually during a routine shutdown like a liner replacement. If anything unusual is seen on the pinion or gear teeth during any of the above inspections, this procedure can more accurately document any anomaly.

Alignment Much has been stated about alignment and how difficult it is to align a twin-pinion drive system. Today, ring gear users can contact the original ring gear manufacturer and /or the OEM that supplied the drive system for assistance in making any mill drive adjustment. In addition, with the assistance of infrared alignment techniques, even the twin-pinion driven ring gears, are relatively simple to align. Software exists to help the user establish the proper movements to make, or the OEM’s can be called upon to provide this simple service. DRIVE FIRST AND EVALUATED COSTS Drive first costs are by far the most important element in the selection of a drive system. These are generally the easiest costs to obtain because all mill/motor/electrical suppliers are eager to provide budget and final costs for their equipment. However, to make fair comparisons, a complete scope of supply must be defined, together with a good specification covering all elements in the system. Often it is necessary to request several quotes for each component in the system to be sure that a “fair” price is obtained. Also, pricing for electrical equipment should be obtained independent of the mechanical elements; however, coordination between the various suppliers is required to assure the integrity of the overall system. Start-up, first year and critical spare parts must also be priced and included in the comparisons. In order to fully evaluate the drives, the cost for electrical losses needs to be included. Each drive system has its own combination of losses including filters, transformers, variable frequency drives, slip resistors, motors, gearboxes and ring gear, as needed. Tables 3,4 and 5 indicate the overall efficiency expected for each drive system. These are typical values that need to be reviewed each time a system evaluation is made, especially for comparisons of two or more drive alternatives. Typical efficiencies can be obtained from suppliers when first costs are requested.

828

Table 3 Single Pinion Drive Efficiency

I

I

SINGLE PINION A.5

- 10 - - -mW __ ..

FIXED

Transformer Vari.Frequency Drive Slip Resistor Motor Gearbox Ring Gear

nla nla nla 0.985

0.985 0.985

WR PWM da da 0.995 0.975 0.985 0.990 0.983 0.988 0.975 0.978 0.990 0.990 nla da nla nla 0.960-0.970 0.970-0.980 0.955-0.965 0.970-0.980 da ' nla da 0.985 0.985 I 0.985 0.985 0.985

I

I

I

I 0.946-0.955 I 0.927-0.936 0.917-0.927 I 0.91 1-0.921 Key: LSS - Low Speed Synchron NUS LCI - Load ( WR - Wound Rotor

0.894-0.904 0.919-0.928

CCV - Cycloconverter PWM - Pulse Width Modulated

Table 4 Dual Pinion Drive Efficiency

Filter Transformer Van. Frequency Drive Slip Resistor Motor Gearbox Ring Gear

VARIABLE FIXED LCI ccv WR PWM LSS WR nla nla 0.995 0.975 da da n/a nh 0.985 0.988 0.990 0.983 da da 0.990 0.990 0.975 0.978 da da da nia 0.980 nla 0.960-0.970 0.955-0.965 0.960-0.970 0.970-0.980 0.955-0.965 0.970-0.980 da da 0.985 nia 0.985 nla 0.985 0.985 0.985 0.985 0.9S5 0.9S5

\System Efficiency

0.946-0.955 0.908-0.9 18 0.9 17-0.927 0.9 1 1-0.921 0.894-0.904 0.919-0.928

Table 5 Gearless Drive Efficiency GEARLESS 9-30mW VARIABLE ccv I PWM 0.975 I 1.000 0.988 0.983 0.990 0.978 da nla 0.950-0.970 0.950-0.970 nla nla nla nla

Filter Transformer Variable Frequency Drive Slip Resistor Motor Gearbox Ring Gear I

System Efficiency

I

I 0.906-0.925 I 0.913-0.933

829

Key: LSS - Low Speed Synchronous WR - Wound Rotor LCI - Load Cornrnutated Inverter CCV - Cycloconverter PWM - Pulse Width Modulated

Based on the power draw an NPV for power losses can be calculated at the expected longrange unit cost of power. For example, at a unit rate of $0.0587 per kWh, 10% discount rate, 10year period and 95% plant availability, the NPV power cost is approximately $3000 per kW. A lOMW system with an overall system efficiency of 0.93 has an NPV cost for losses equal to $2.1 million, This added to the first costs to obtain the evaluated cost basis. ERECTION AND COMMISSIONING COSTS The cost of erection and commissioning are major components often overlooked in the evaluation process. The additional time of erection for a gearless drive system is around 6 to 8 weeks greater than that of a ring gear driven system, which could translate into 6 to 8 weeks of lost revenue. Unfortunately, the additional costs associated with this increase in erection and commissioning time are difficult to determine because labor costs vary significantly from site to site. In addition, if the mill is on a critical path, lost opportunity revenue needs to be considered. However, the Engineering firm must include these additional costs if a correct analysis is to be performed. Commissioning costs are those costs associated with taking a mill through the load testing stage to full production. With a gear driven mill, these would be the costs to align the mill gear set and set up the motor controls. Typically these costs are insignificant since would normally not exceed 3 to 5 days. In the case of a gearless mill, commissioning costs would be the cost associated with adjusting the air gap, and debugging the control systems. This normally requires the manufacturers to supply field supervision, which needs to be incorporated into the costs in the evaluation.

COST ANALYSIS The cost analysis was performed on several mill sizes. Both the capital costs and operating costs were analyzed. Several mill selections have been analyzed for two different concentrator plants. These plants utilize both SAG mill and ball mill systems. It should be noted that since the grinding circuit will usually comprise one SAG mill and two ball mills, the price differences on ball mill installations will be twice the amounts identified on a “per mill” basis. All mills used in the examples assume a projected life of 25 years, and the electrical power cost is .035 US$/kW-hr. The capital cost for each mill is defined for our examples as follows: Concentrator A Table 6 Mill diameter (feet): Mill Type: Quantity: Mill Total Power:

36 SAG 1 12 MW (16,086 HP)

Table 7 & 8 Mill diameter (feet): Mill Type: Quantity: Mill Total Power:

22 Ball 2 9 MW (12,064 HP)

Concentrator B Table 9 Mill diameter (feet): Mill Type: Quantity: Mill Total Power:

40 SAG 1 20 MW (26,810 HP)

Table 10 Mill diameter (feet): Mill Type: Quantity: Mill Total Power:

24 Ball 2 12 MW (16,096 HP)

Most of the inspection requirements are performed either while running or during downtime that is not related to the mill drive system. Therefore, production availability is not addressed, since it is not a valid factor in the analysis of the gear driven verses gearless drive system costs.

830

The analysis covers reducer connected as well as direct connect mills. Mills that utilize pinion stand arrangements have not been addressed due to the limited information about their performance and small number operating in the field, in the high power range that we are addressing. Equipment Costs The capital costs are tabulated in the tables 6 through 10. Motor and mill manufacturers have provided the costs for the mills in our examples. Prices will vary from project to project and will depend on the market conditions when the project is analyzed. The costs tabulated are established for the example of mills at the time of the writing of this paper. The costs in tables 6 through 10 are broken down into capital costs, installation and commissioning supervision costs, and operating costs. The capital costs consist of the mill costs, all drive train costs, spares, and any special studies that need to be addressed. The installation and commissioning costs for supervision are strictly supervision costs. Any costs associated with the labor force will vary greatly from one site to another. Therefore these costs need to be addressed in detail for each project separately. Spares can have a large impact on the total cost analysis. Most grinding operations will order a spare pinion assembly for insurance. A pinion assembly consists of a pinion, pinion bearings, a clutch hub and an incher hub if applicable. This way the pinion assembly can be quickly interchanged with the existing assemblies. Typically the change-out is required after approximately 6 to 10 years of operation due to the pinion bearings requiring replacement. Today’s carburzied and ground pinions will normally not require change-out for 10 to 15 years. The only time this is not the case is when an operation needs to increase the speed of the mill or the maintenance has not been adequate to either keep contaminants out of the gear guard, or properly monitor lubricant application. Operating Costs Each motor option discussed under the electrical section has been addressed. Tables 6 through 10 show the overall system efficiency for each option followed by its corresponding evaluated operating cost. This is then added to the total installed cost to produce the total capital and energy cost over a 10-year horizon. The gearless drive system will be used as the baseline for all of the comparisons. Concentrator A 12 MW SAG Mill Only variable speed was addressed in this analysis. The gear driven mill utilized a dual pinion drive train (table 6). Capital costs for the wound rotor motor option was the lowest, however this soon disappeared when the loss evaluation was applied. Installation and commissioning costs of the gearless drive system were significantly higher than the other three systems listed due to the increased complexity of the electrical system. The operating costs were the lowest for the LCI and CCV gear driven mills. The total 10-year cost was highest for the gearless driven mill. This was followed by the wound rotor induction motor which was 9.8% below the base line, then the low-speed synchronous with CCV at 10.2% and lowest cost was the low-speed synchronous with LCI at 12%.

831

-

Table 6 SAG Mill 12 MW Cost Anaylsis Mill Type :SAG Mill Power : 12MW (16,086HP) : 36 ft. Size DRIVE TYPE GEARLESS FIXED OR VARIABLE SPEED VAR-CCV 0 NUMBER OF PINIONS

LSS VAR LCI

-

2

WR VAR-SER 2

LSS VAR-ccv 2

ITEM

MECHANICALEOUIPMENT

1

2 3

4 5

(1) SAG Mill (with gears as required)

BRAKES METAL WEAR LINERS

Harmonic Study Transient Torsional Analysis

3,614,000 375,000 843,000

4,3 10,000

4,406,000

4,268,000

843,000

843,000

843,000

27,000 35,000

27,000 35,000

27,000 35,000

27,000 35,000

v 6 7 8 9

(Motor I Drive / Trxf I Switchgear) Ring motor Twin LCI Twin SER Twin CCV

4,185,000 3,135,000 2,550,000 3,425,000

SPARES 10 11 12 13 14 15 16

-

& COMM. SUPERVISION 17 Installation cost - Supervision Ring Motor 18 Installation cost - Supervision Gear Driven

19

60,000 190,000 147,000 98,000

Mill 1yr, Wear Spares Trunnion Brg. & Lube Spares Electrical Spares ( motor / drive / controls) Brake System spares (1) pinion, Clutch Spares, Pinion Brgs. (1) pinion, Gearbox Spares, Pinion Brgs.

22 TOTAL CAPITAL & ENERGY COST

52,000 92,000 150,000 184,000

184,000 187,000 $

9,574,000

$

8,828,000 $

8,287,000 $

9,076,000

900,000 195,000

205,000

195,000

S10.474.000S9.023.000S8.492.000S9.271.000

20 System Efficiency

EVALUG COST 21 Energy losses 10 years @, $30OOkW

52,000 92,000 95,000

52,000 92,000 150,000

92.30%

92.70%

90.40%

92.70%

2,772,000

2,628,000

3,456,000

2,628,000

$ 13,246,000

$ 11,651,000

$ 11,948,000

$ 11,899,000

Note: Mill prices include, trommel, lifting cradle jacks, plus all gears, clutches, and inches if applicable. In addition, Mill costs include reducer assemblies for wound rotor motor, high systems.

832

-

Table 7 Ball Mill 9 MW Cost Anaylsis

- Dual Pinion

Mill Type: Ball Mill Power : 9MW (12,064HP) Sue

: 22 ft.

DRIVETYPE GEARLESS FIXED OR VAKlABLE SPEED VAR-CCV 0 NUMBER OF PINIONS

US

WR FXD 2

FXD

2

LSS VAR-LCI 2

U S

WR VARSER

VAR-CCV

2

2

ITEM 1 2 3

4

5

6

7

-

( I ) Ball Mill (with gears as required) BRAKES METAL WEAR LINERS

-

9

LCI (Adjustable Speed) SCR (Adjustable Speed) CCV (Adjustable Speed)

275,000

275.000

275.000

275,000

27,000 35,000

27,000 35,000

27,000 35,000

(Motor I Drive I Trxf / Switchgear) hng motor LSS (Fixed speed)

WR (Fixed speed)

11

330,000 275,000

Harmonic Study Transient Torsional Analysis

8 10

S 2,249.000 $ 3,186,000 $ 3,079,000 $ 3,021,000 $ 2,989,000 S 2,988,000

275.000

27,000 35,000

3,765,000 1,820,000 775,000 2,635.000 2.075.000 2,840.000

SEdaES 12

13 14

15 16

17 18

19

20 21

22

76,000 96.000 I47.000 98.0

Mill lyr. Wear Spares Trunnion Erg & Lube Spares Electrical Spares ( motor / drive ) Brake System spares (I) pinion, Clutch Spares. Pinion Ergs (I)pinion. Grdrbox Spares, Pinion Brgs

-

Installation cost - Supervision Ring Motor Installation cost Supervision Gear Driven

n

Energy losses 10 years @ S3000kW

24

TOTAL CAPITAL & ENERGY COST

76,000 75,000 130,000

76,000 75.000 I50,000 295,000

295,000

$

7.098.000 S 5.802.000

$

76,000 75.000 150.000 295,000

300.000

300,000 4.620.000 S 6.589.000

$

5,982.000 $ 6.761.000

900.000 55.oOO

~

System Efiiciency

76.000 75.000 40.000

76,000 75,000 75.000

$7.998.000

$5,857,000 $4,685,000

92 50%

1.025.000

65,ooO

95.50%

1.215.OOO

S 10.023.000 S 7.072.000

S

195,000 S6,784,000

195,000

$6.187,000

$6.956,000

91.80%

92.70%

90.40%

92.10%

2,214,000

1,971.000

2,592,000

2.133.000

6,899.000

S 8.755.000

Note Mill prices include. trammel. litting cradle jacks. plus all gears. clutches. and inches ifapplicable In additioii. Mill costs include reducer assemblies lor wound rotor motor. high systems

833

205,000

S 8,779.000 S 9.089.000

-

Table 8 Ball Mill 9 MW Cost Anaylsis

- Single Pinion

Mill Type: Ball Mill Power :9MW (12,064HP) Size : 22 ft. DRIVETYPE GEARLESS FIXED OR VARIABLE SFEED VAR-CCV 0 NUMBER OF PISIONS

WR

LSS

WR

FXD 1

FXD

LSS VAR-LCI

VARaER

LSS VAR-CCV

I

1

1

1

ITEM

v ( I ) Ball Mill (with gears as required) BRAKES 3 METAL WEAR LINERS 1

$

2

4

5

6

7 8 9

10 11

Harmonic Study Transient Torsional Analysis

2249.000 S 3,554,000 330,000 275.000 275,000

$

3,514,000

$

275.000

3,389,000 $ 3,424,000 $ 3,359,000 275,000

275,000

275.000

27,000 35,000

27.000 35,000

27.W 35,000

27,000 35,000

(Motor I Drive I Trxf I Switchgear) Ringmotor LSS (Fixed speed) WR (Fixed speed) LCI (Adjustable Speed) SER (Adjustable Speed) CCV (Adjustable Speed)

3,765,000 I.025,000 600.000 2,285,000 1,800,000 2,550.000

SEBBES Mill lyr, Wear Spares Trunnion Brg. & Lube Spares 14 Electrical Spares ( motor / drive ) 15 Brake System spares 16 ( I ) pinion, Clutch Spares, Pinion Brgs. 17 (1)pinion, Gearbox Spares, Pinion Brgs. 13

$8

19

20 21

22

-

24

TOTAL CAPITAL & EBERGY COST

40,000

52,000 92,000 130.000

$

52.000 92,000 150,000

295,000

295,000 297,000

297.000

7,098.000 5 5,350,000 S 4,870,000 $ 6.600.000 S 6.132,000

$

6,835,000

900.000

$7,998,000

System Efficiency

Energy losses 10 years @ %3000/kW

52,000 92,000 150.000

52.000 92.000

295.000

Installation cost -Supervision Ring Motor Installation cost - Supervision Gear Driven

23

76,000 75,000 50,000

76.000 96.000 147.000 98.000

12

$

37,000

43,000

162,000

$5.387.000

%4.913,000

$6,762,000

168,000

$6,300,000 $6.997.000

92.50%

95.5Wh

93.60%

92.70%

90.40Dh

2.025,000

1,215,000

1,728.000

1,971,000

2,592,000

10.023.000

16 6,602.000

$

6.641.000

$

8.733.000

Note: Mill prices include. aommel, lifting cradle jacks. plus all gears, clutches, and inches if applicable In addition, Mitt costs include reducer assemblies For wound rotor motor. high systems.

834

162,000

$

8.892.000

92.113%

2,133,000 $

9,130.000

Table 9 SAG Mill - 20 MW Cost Anaylsis

Mill Type : SAG Mill Power :20MW (26,810HP) Size : 40 ft. x 22 ft. DRIVE TYPE FIXED OR VARIABLE SPEED NUMBER OF PINIONS ITEM MECHANICAT EQUIPMENT 1 ( I ) SAG Mill (with gears as required) 3 BRAKES 3 METAL WEAR LINERS

-

GEARLESS VAR-ccv 0

LSS VAR - LCI 2

WR VAR-SER 2

LSS VAR-CCV 2

I

4 5

6 7 8 9

$

6,970,000 $ 8,470,000 $ 5 10,000 1,550,000 1,550,000

27,000 35,000

Harmonic Study Transient Torsional Analysis

ELECTRICAL EOUIPMENT (Motor / Drive / Trxf / Switchgear) Ringmotor TwinLCI TwinSER TwinCCV

27,000 35,000

8,510,000 §!

8,395,000

1,550,000

1,550,000

27,000 35,000

27,000 35,000

6,085,000 4,325,000 3,540,000 4,670,000

SPARES 10 11

12 13 14

15

Mill lyr, Wear Spares Trunnion Brg. & Lube Spares Electrical Spares ( motor / drive / controls) Brake System spares (1) pinion, Clutch Spares, Pinion Brgs. (1) pinion, Gearbox Spares, Pinion Brgs.

16 TOTAJ.CAPI TAL COST INSTALLATION & COMM. SUP17 Installation cost - Supervision Ring Motor 18 Installation cost - Supervision Gear Driven

320,000 190,000 147,000 98,000

22 TOTAL CAPITAL & ENERGY COST

185,000 92,000 130,000

185,000 92,000 150,000 295,000

295,000 298,000

$ 15,932,000 $ 15,129,000 $ 14,367,000 $ 15,399,000

900,000 195,000

205,000

195,000

$16,832,000

$15,324,000

$14,572,000

$15,594,000

92.50%

92.70%

90.40%

92.10%

20 System Efficiency EVALUATED OPERATING COST 21 Energy losses 10 years @, $3000/kW

185,000 92,000 150,000

4,500,000

4,380,000

5,760,000

4,740,000

$ 21,332,000

S 19,704,000

$ 20,332,000

$ 20,334,000

Note: Mill prices include, trommel, lifting cradle jacks, plus all gears, clutches, and inches if applicable. In addition, Mill costs include reducer assemblies for wound rotor motor, high systems'.

835

Table 10 Ball Mill - 12 MW Cost Anaylsis Mill Type: Ball Mill Power : 12MW (16,086HP) Size :24 ft. DRIVE TYPE GEARLESS FIXED OR VARIABLE SPEED VAR-CCV 0 NUMBER OF PINIONS

LSS FXD 2

WR FXD 2

LSS VAR-LCI 2

LSS

WR VAR-SER 2

VAR-CCV 2

ITEM

v I ( I ) Ball Mill (with gears as required) 2 BRAKES 3 METAL WEAR LINERS

$

2,497,500 $ 3,690,500 $ 3,728,500 $ 3,525,500 $ 3,638,500 S 252,000 323,000 323,000 323,000 360,000 323,000

3,490,500 323,000

ENGlNEERlNGSTUDIES 27,000 35,000

4 Harmonic Study 5 Transient Torsional Analysis

27,000 35,000

27,000 35,000

27,000 35,000

ELECTRlCALEOUlPMENT (Motor I Drive l Trxfl Switchgear) Ring motor LSS (Fixed speed) WR(Fixedspeed) LCI (Adjustable Speed) 10 SER (Adjustable Speed) I I CCV (Adjustable Speed)

4,185,000

6 7 8 9

2,175,000 875,000 3,135,000 2,550,000 3,425,000

l52PmEs 215,000 70,000 147,000 98,000

12 Mill lyr, Wear Spares 13 Trunnion Brg. & Lube Spares 14 Electrical Spares ( motor / drive )

15 Brake System spares 16 (1) pinion, Clutch Spares, Pinion Brgs. 17 (1) pinion, Gearbox Spares, Pinion Brgs

-

120,000 70,000 130,000

120.000 70,000 150,000 103,000

103,000 107,000

7,886,500 $ 6,556,500 $ 5,263,500 $ 7,488,500 $ 7,000,500 $ 7,743,500

900,000

-

55,000

22 System Efficiency

EVALUG COST 23 Energy losses 10 years @ $3000/kW 24 TOTAL CAPITAL & ENERGY COST

I20,Ooo 70,000 150,000

107,000

INSTALLATION & COMM. SUPERVIWN 19 Installation cost - Supervision Ring Motor 20 Installation cost Supervision Gear Driven 21

120,000 70,000 40,000

103,000

$

18

120,000 70,000 75,000

195,000

205,000

195.000

$8,786,500

$6.61 1,500

$5,328,500

$7,683,500

$7,205,500

$7,938,500

92.50%

95.50%

91.80%

92.70%

90.40%

92.10%

2,700,000 $

65,000

1,620,000

2,952,000

2,628,000

3,456,000

2,844,000

11,486,500 S 8,231,500 $ 8,280,500 $ 10,311,500

$ 10,661,500

S 10,782,500

Note: Mill prices include, trammel, lifting cradle jacks, plus all gears, clutches, and inches if applicable. In addition, Mill costs include reducer assemblies for wound rotor motor. high systems.

836

9 MW Ball Mill A combination of fixed and variable speed was addressed in this analysis. The gear driven mill utilized a dual pinion drive train (table 7). Capital costs for the fixed-speed systems were $1.3 M to $2.5 M lower than the baseline. Installation and commissioning costs of the gearless drive system were significantly higher than the other five systems. The operating costs were the lowest for the fixed-speed synchronous gear driven mills. The total 10-year cost was highest for the gearless mill. The most efficient drive was the fixed-speed low-speed synchronous which was $3 M or 29.4% lower followed by the variable speed synchronous and CCV systems which were $1.27 to 1.24 M respectively or 12.7 to 12.4% lower than the baseline. The wound rotor with SER was the least efficient. This concentrator requires two ball mills, therefore the cost difference will be approximately double, i.e., $3 M x 2 = $6 M. It is approximate because the spares would not be increased by a factor of two, but neither would the cost of installation be increased by two, therefore doubling the cost difference will result in a good estimation. 9 MW Ball Mill A combination of fixed and variable speed was addressed in this analysis. The gear driven mill utilized a single pinion drive train (table 8). Capital costs for the fixed-speed systems were $1.75 M to $2.2 M lower than the baseline. Installation and commissioning - same as previous ball mill. As with the dual pinion ball mill, the operating costs for the fixed-speed synchronous gear driven mill is the lowest. The total 10-year cost was highest for the gearless mill. This analysis demonstrated that the most efficient drive was the fixed-speed synchronous system that was $3.42 M or 34% followed by the fixed-speed wound rotor, which was $ 3.4 M or 34% below the baseline. The variable speed system with the least expensive 10-year costs was the low-speed synchronous with LCI with a savings of $1.29 M or 12.9% below the baseline, closely followed by the wound rotor with SER which demonstrated a savings of $1.13M or 11.3% below the baseline. The CCV provided a savings over the baseline of $.89 M or 8.9%. The single pinion fixed-speed system demonstrated significantly higher savings over the gearless system for a mill in t h s power range. This concentrator requires two ball mills, therefore the cost difference utilizing the lowest cost fixed speed option will be approximately double, i.e., $3.42 M x 2 = $6.84 M. Concentrator B 20 MW SAG Mill Only variable speed was addressed in this analysis. The gear driven mill utilized a dual pinion drive train (table 9). Capital costs for the gear driven wound rotor system option was the lowest. Installation and commissioning cost of the gearless drive system are higher as with the last analysis. The operating costs were the lowest for the LCI and CCV gear driven mills. The total 10-year cost was the lowest for the LCI system. The LCI was $1.63 M or 7.6% lower than the baseline. The CCV and the wound rotor SER system were tied with an approximate savings of $1 M or 4.6% below the baseline.

837

12 MW Ball Mill A combination of fixed and variable speed was addressed in this analysis. The gear driven mill utilized a dual pinion drive train (table 10). Capital costs for the fixed-speed systems were $1.3 M to $2.6 M lower than the baseline. Installation and commissioning cost difference remained basically the same. The operating costs were the lowest for the fixed-speed synchronous gear driven mills. The total 10-year cost was the lowest for the fixed-speed synchronous gear driven system with a difference of $3.26 M or 28% below the baseline. The fixed-speed wound rotor system is a close second with a difference of $3.21 M or 27.9% below baseline. The variable speed system with the least expensive 10-year cost was the low-speed synchronous with LCI with a saving of $1.18 M or 10.2% followed by the CCV with $.88 M or 7.2% and the wound rotor with SER with $.70 M or 6% below baseline. This concentrator requires two ball mills, therefore the cost difference will be approximately double, i.e., $3.26 M x 2 = $6.52 M. CONCLUSION The factors involved in selecting a mill drive system have been examined from the differing perspectives of process engineering, electrical equipment, and mechanical equipment currently. It has been shown that the opportunities for conservative plant designs using modern machinery are broad in scope and afford the owner a selection of drive methods. This paper has shown that for two large ball mills the difference in initial capital cost between the highest and least cost is in the order of $3 million per ball mill, or $6 million for the grinding circuit, without compromising availability or reliability. A sound method of calculating the real costs of downtime, based on a mine’s individual data has been given. Tabulations that allow the reader to examine the economic consequences of equipment selections have been provided based on fm pricing data for late 2001. The conclusion is that every drive selection and cost analysis is subjective to a particular mine, and that each case demands a careful study by a mine owner’s in-house engineers or by his consulting engineering company, and that all three disciplines, process, electrical, and mechanical should participate. There is much to be gained by involving the vendors in these discussions at an early stage before critical decisions are made. ACKNOWLEDGEMENTS The authors wish to thank Stuart Walters for his input and his diligence in keeping us all together and on track. In addition we thank Metso Industries Inc., especially Vytas Svalbonas and Subhas Hazra for their support and help in writing this paper. In conclusion, we wish to acknowledge Kvaerner, Metals E&C Div., General Electric, and The Falk Corporation for their support and assistance provided during the preparation of this paper. REFERENCES Antosiewicz, M., “The Use of Mill Gear Operating Temperatures for Alignment Evaluation,” IEEE Technical Conference, Tarpon Springs, Florida, May 1979, Section: Drives and Related Products, Paper No. 5. Barratt D., Brodie M., & Pfeifer M., 1996, “SAG Milling Design Trends, Comparative Economics, Mill Sizes and Drives,” SAG Conference 1996, Vancouver, B.C., 1228 pp. Bassarear, J.H., and Thomas, P.F., 1985, “Variable Speed Drives for Semiautogenous Mills,” Trans. SME-AIME, Pre-print No. 85-20. Danecki, C., 1997, “Workshop SAG 97: Girth Gears at 18000 HP - Operation and Performance, (Update April 1997),” Workshop SAG 97 Conference, Vina Del Mar, Chile, 1997.

Danecki, C., 1996,” SAG Mill Drives Girth Gears at 18000 HP - Operation and Performance,” SAG Conference 1996, Vancouver, B.C., 425 pp. Danecki, C. and Kress, D., 1995, “Grinding Mill Gear Drives for the Future,” IEEE Technical Conference, San Juan, Puerto Rico, June 1995,381 pp. Kress, D. and Hanson, D., 1989, “Girth Gear Design Concept Through Operating Criteria,” SAG Conference 1989, Vancouver, B.C., 609 pp.

839

The Design of Grinding Mills Vytas Svalbonas’

ABSTRACT Grinding mill design seems like a straightforward application of structural, mechanical, and electrical concepts. But, like many other items, it is actually based on a series of approximations, Prior to actual manufacture, the component sizes and properties are approximate, the loading is approximate, the future maintenance can only be approximated, the site soil characteristics are approximate, and even the future desires of the operators as to throughput are not well known, over the period of mill life. Thus, the mill designer does not have a single “bull’s-eye” to target, but must aim to cover a whole range of possible future desires. INTRODUCTION A good design for a grinding mill system will allow future operators a satisfactory flexibility in operations. To do this, the mill designer must consider all the variables that will affect his design, structurally and mechanically, and provide a safety margin for the variation of those variables considered relevant. At first glance this seems quite simple, since the mechanical requirements are controlled by the agreed power to the mill, and the structural requirements are controlled by mill loading. However, the designer must also consider what future operators may reasonably expect to do with the delivered mill, and he must be aware of the approximate nature of both loads and materials in his design. INTERACTION WITH PROCESS VARIABLES The process variables that will affect the mill structural design include: power versus mill grinding chamber shape and size; shape, weight, and size of lifters and liners; discharge systems like trommels; charge ball volumes. If one considers that these are only weightlload parameters, then the design seems straightforward. However, weight is not the only relevant variable. Grinding Chamber ShapdSize This is usually related to the mill configuration design, like trunnion mounted or shell mounted mills, as shown in Table 1. Choosing the process variable as grinding chamber volume, some relations between volume and effective grinding length (EGL) are given in Table 1. It can be seen that the conical configurations, be they externally shaped or internally shaped, will change the grinding volume, and thus the EGL, when compared with a purely cylindrical mill with end plates. It may be argued that the conical volume is a less efficient grinding volume. But the same argument can be used for the first few cylinder feet in the feed end of a flat-ended mill with higher

1 Metso Minerals Industries, Inc.; York, Pennsylvania, USA.

840

Table 1 Comparison of Mill Shapes for 26 ft x 37.5 ft (7.92 m x 11.43 m) EGL Gearless Ball Mills

EQUAL VOLUME

EQUAL EGL

MILL TYPES

EGL

INSIDE INSIDE PER- NEW SHELL VOLUME CENT EGL

ADD’L NEW NEW SHELL INSIDE VOLUME SHELL

Shell Supported: A - Square with flat ends

37.50 ft 38.00 ft 18 897 f? 88.8% 42.23 ft 4.73 ft 42.73 ft 21 282 ft3

B - Square with

37.50 ft 38.00 ft 19 357 ft3 91% 41.32 ft 3.82 ft 41.82 ft 21 282 ft3

l5 degree ends

11.43 m 11.58 m 11.43 m 11.58 m

535 m3 548 m3

12.87 m 1.44m 13.02m 12.59 m 1.16 m 12.75 m

602m3 602 m3

C - 15 degree ends 37.50 ft 39.92 ft 20 003 ft3 94% 40.04 ft 2.04 ft 42.45 ft 21 282 ft3 protected) 11.43 m 12.16 m 566 m3 12.20 m 0.62 m 12.94 m 602 m3 Trunnion Supported:

D - 30 degree ends

37.50ft 38.00ft 21 282f? loo% 37.50 ft

---

.11.43 m. 11.58 m- 602 m3.

--- .11.58 m- 602 m3

11.43m

38.00 ft 21 282 ft3

volume flow. Also, recent Discrete Element Modeling (DEM) work indicates that the contribution by the conical volume exceeds traditional allowances, as the head lifters have a significant impact on the lifting of the charge. In any case, the process arguments are outside of the scope of this study. What must be considered by the mill designer is that it is a certainty that future operators will want more power out of any delivered mill. Thus, to supply a mill with insufficient grinding volume based on an artificial EGL is a design flaw. Designing a mill configuration by grinding chamber volume will require the flat-ended mills to have a longer cylindrical length for direct comparison. This will have a direct effect on mill design loads, and all other parameters. Note also that one configuration (‘C’) of the shell mounted designs in Table 1 has a conical volume due to a protective liner configuration directly under the riding rings (requiring a longer inside shell dimension). This is one way liner design may affect mill structure. Other effects are considered in the next section.

841

Lining Configurations Liners are considered “dead weight” in mill structural analysis. Thus, the total lining “load” has a limit set in the design. Some operating plants, in the interest of production optimization, try to add “extra” balls to the mill charge at the rate of liner wear. This can be done, keeping the dead weight and the live charge weight to a total limit, although the effects of dead weight and live charge weight on mill stresses are slightly different. One must remember, of course, to make a ball adjustment when the mill is relined. Liners, however, have other effects on mill design than merely dead weight. Today it is common to study liner configurations as a variable for throughput and power optimization [References 1, 2, 31. Figure 1 shows a mill power variation, Figure 2 comes from a typical ball trajectory study, and Table 2 has a listing of mines doing such studies. As long as liner configurations are bounded by overall total liner design weight constraints, one might expect they do not interact with mill structural design. The new lifter angles, however, are typically accompanied by reduced power draw and increased liner wear. That, in turn, leads users to heavier liners and more charge or higher mill speeds. Liner configuration also can lead to different possible accident loads. Figure 3 shows a liner portion that has been extruded significantly by direct ball impact. The upper 25 mm of the liner plate has been cold worked to elongate approximately 20+ mm. This was caused by running a 30 ft (9.1 m) SAG mill with an improper mix of balls and ore for the lifter spacing and configuration chosen. Within a few months of operation, the liners had been cold worked so that all liner gaps were closed, and liners were buckling and forcing the mill shell apart at the longitudinal joints, as shown in Figure 4. Segmented shell-riding ring configuration mills will be weaker against accidents causing “internal pressure” overloads, in the unbolted sections of the riding rings. This contributed to the speed of the failure and mill shutdown (see later sections). Earlier liner configuration designs, with closer lifter spacing, would have taken the ball impacts on the lifters, and the cold work elongation of the liner plates would not have occurred.

Discharge Systems As mill diameters increase, mills are required to handle ever increasing volumes per unit discharge 2.5 area. Capacity and flow typically increase as D . The part of the discharge system that is of most concern to structural designers is the trommel. While the trommel usually has little effect on the mill design, even its adjacent trunnion bearing, its own structural design is complicated by the possible very large fluctuations of its live load, compounded by accident conditions such as panel blinding or mill grate breakage. During its operational life, the trommel may see different dam heights and positions, which change live load volumes and distributions, different conditions in the wear of the rubber covering of its main members, which may change the fatigue criteria to include corrosion, and unforeseen dynamic loads if large balls enter a trommel with buckets. In order to accurately analyze trommel life, two load histories are necessary: 0

A worst case operating load, defined by the mill operators.

0

A worst case accident load (such as the SAG mill operating with a broken grate), and its tolerated duration and frequency.

This will allow design for long term fatigue, with consideration of a short term pattern of higher accident loads. Luckily, ball mill trommels are usually not subject to such complications.

842

1.OE+07

i.OE+06

'.

/

P

1.0E+03 i 1.OE+02 L

1

0.1

10

Mill Diameter [m] Figure 1 Comparisons of Power Draw for Ball Mills of Different Diameters (Ref. [2])

Figure 2 Typical DEM Study (3-D Model)

843

-

Table 2 BHP Survey of Large SAG Mill Shell Liners Release Angle and Lifter Spacing Evolution

I

I

40. of Lifters/ Release No. of Lifters/[ Release Property/ Location Alumbrera (Argentina) Cadia (Australia) Candelaria (Chile) Escondida (Chile)

Style 48

(2) @ 36

{iT

(2) @ 3 6 (1) @ 36'

7.5

I

Rail

I

8.5

I

20

I

1 1 I I

(1) @ 26'

(2) @ 34.5' (2) @ 32' (2) @ 32'

I

HAT Hi-Lo 26 Hi-Lo HAT 48 Rail 64 Hi - Hi 64 Lift

I I

18

I

I

20 25

(2 Tests) T i 3 0 WedgeBad 18.5 ClampBar 17.5 1 2 2 1 & Plate Plate

T

I Hi - Lo

Hi - Lo

(1) @ 3 6

8.5

HAT -LoPlat 36 Hi-Lo Plate w/lifter 69 HAT

22 12

Pending

18 11.5 6

I

~

(also current)

Plate

(2) @ 36'

[iT

25

69 HAT

Freeport (Indonesia) Freeport (Indonesia) Inmel - Troilus (Canada) Kemess Mine (Canada) Kennecott Copperton ,(USA) K.I.O.C.L. (India) PT Nusa Tenggara (Indonesia) Pasmin co Century Zinc (Australia) Rand Gold/ Morila (Mali) SNIM (Mauritania) Collahuasi (Chile) Chuquicamata (Chile)

Rail 72 HAT

1 1 1 :l

(degrees) Style (degrees) 25 & 30 36 ?A JU Tested IHAT-LoPlatd

I

6

Los Pelambres (Chile) BHP - OK Tedi

Hi - Hi

(Papua New Guinea)

TOP HAT

I

Hi-Hi Hi - Hi

I

25

Figure 3 Liner Extruded by Ball Impact

Figure 4 Mill Shell Separation

845

Maximum Charge versus Ball Volumes For SAG mills, the power draw is a function of both the charge density and the total charge volume. For example, a SAG mill could draw similar power with a 12% ball, 20% total charge level, or an 8% ball, 35% total charge level. However, the charge mass of the two cases would be different, as well as its possible distribution in the mill shell. Thus, the mill needs to be structurally designed for a maximum charge mass condition, rather than power. Typical ore specific gravity may also be deceptive, since mills tend to concentrate the heavier, harder to grind portions in the grinding chamber. Thus, grinding chamber contents may be heavier than the average ore being fed to the mill. ACCIDENTS AND ACCIDENTAL LOADS A liner accident caused by ball impact was described in a previous section and in Figures 3 and 4. These types of accidents have been relatively rare, but might increase as SAG mills get larger and liners are varied before deeper experience with individual mill operation is obtained. With increasing mill diameters, the designer’s initial fear is of high ball trajectory cracking a liner and causing additional damage to the internals. Thus, the initial preference is to start up a large diameter SAG mill with slightly softer liners and go to harder configurations as the mill operator’s expertise increases. The occurrence of the cold work accident, described earlier, therefore has to be prevented during mill start-ups, with the softer liners. Another form of accident, more prevalent in big ball mills, is the drop charge, described in Ref. [4]. Damage from such an accident is shown in Figures 5 and 6 for a 24 ft (7.3 m) diameter ball mill. Reference [4] analyzed this drop charge, and concluded that an 8% volume “frozen” charge (18% of the design 45% volume) dropping the distance of half the diameter, would result in 5.5 times the magnitude of the static design loads in the 24 ft (7.3 m) diameter ball mill. This will conclusively result in head casting cracking and cylinder shell yielding under such accidents. Thus, all possible care needs to be taken in the operation of large ball mills to prevent such accidental occurrences. Some “static snapshots” of a DEM study of a large ball mill, in Figures 7 and 8, show another phenomenon to consider. The start time of many large gear driven ball mills is determined by a clutch setting of approximately 6 seconds. This is a relatively fast start, especially compared to gearless motors. The speed of such a start initiates a different, more violent starting ball cascade (where a lot of balls overshoot the charge toe; see Figure 7) than that in the steady running speed of the ball mill (see Figure 8). Figure 7 shows a greater tendency for balls to be “launched” before the charge has a chance to start shearing and cascading, as in a steady state (Figure 8). Thus, there is a start-up load, even in a liquid charge, which is of higher energy than the steady state. If the slurry is allowed to dewater, allowing solids to pack between balls, the balls will remain bound together for a longer time before the solids fluidize, thus increasing the risk of a partial drop charge. In fact, a mini-drop charge can occur on each start-up, with possible accumulation of damage such as mill bolt loosening, which will eventually result in flange bolting problems if not diagnosed early and remedied. Thus, for large ball mills, some thought must be given to slower start drive trains and the addition of drop charge prevention devices. There are other overload conditions to consider, such as an operational feed overload, and the earthquake condition. Normal operation overload does not impact mill design. The overload will eventually back up the SAG feed chute and stop the mill. However, even if the mill is 50% full, this rarely occurring condition will not affect long term mill stress conditions because the mill overload stresses will still be significantly below any limits such as yield stress.

846

Figure 5 Cracked Head Flange

Figure 6 Paint Cracks on Distorted Shell Flange

a47

Figure 7 First DropKascade on Ball Mill Start

Figure 8 Steady State Ball Mill Cascade

848

The earthquake condition, however, can result in more significant stresses in certain mill components, and deserves more consideration in mill specifications. In the power industry, there are several different earthquake conditions defined, such as the “operating base earthquake” (OBE) and the “safe shutdown earthquake” (SSE). The names essentially define how the equipment is supposed to behave under these conditions; it is expected to operate right through an OBE, and be capable of a safe shutdown under an SSE, although some repairs (or component replacements) may be expected afterward. In mill specifications, no such requirements are given. Thus, the mill designer is at liberty to make his own criteria. For the overall grinding mill structure, the earthquake conditions are not the determining design criteria. However, components such as bearings see very significant stresses during an earthquake. So, the designer needs to decide at least two things: 1.

When can an earthquake occur? This will determine the load condition for the earthquake acceleration (g) coefficients. It might seem that a full power operating condition would be the most significant. But what if an earthquake occurs for a pad bearing supported mill when one of the pads has been removed for maintenance or replacement, and the mill remains fully loaded?

2.

What is the stress criterion for the earthquake? Again we can consider a pad bearing structure. Some pad bearings are designed so all loads are carried through machined areas that define the corresponding stress. Some other designs depend upon “point” contact, like a ball bearing on a plate structure. The stresses at these concentrated contact points are already very high under maximum design, and will go into general yield under an earthquake. The more critical areas, in this consideration, are the mill thrust bearings and their attached structure. This is due to the fact that under normal loads this is a lightly loaded area and under earthquake conditions the stress change (from normal operation) is the highest. Under present mill specifications, the mine operator does not know what he is buying in these areas, as the stress criterion (how much yielding) is optional to the designer.

A final accident condition to be discussed is grinding mill structure erosion, and any allowances for this and other environmental degradation in the designs. This is usually not considered in mill specifications. However mill designs usually reach critical design stresses at only a few points, with the rest of the mill structure at significantly lower stress. Therefore, there is inherent allowance in those parts of the structure. The structure should be protected by rubber backing no matter what lining system is used. The rubber backing sheets serve primarily a mill protection function, and only peripherally deal with bedding of liners. Critical areas may deserve some extra protection. For example, configuration ‘C’ in Table 1 is designed with special corner protection under the riding ring bearing. Trunnion bearings usually have dual protection, in the form of a trunnion liner, and sealing and rubber covering of the area beneath. In many designs, trunnion structures have experienced erosion and have continued to operate satisfactorily for many years. Shell mounted riding ring designs have to be looked at for the same relative safety. Corrosion has to be considered, especially for trommel structures that are operated without or with worn rubber or urethane protection. It should be realized that the welds are normally designed for a corrosion free environment, and corrosion attack will reduce the allowable fatigue stresses by up to a factor of 2, with a corresponding exponential life decrease. This relationship has to be considered in cases where a “corrosion allowance” is sought. The best protection is to prevent corrosion attack.

849

UNCERTAINTIES IN DESIGN VARIABLES The uncertainties in live operating loads arise from inaccuracies in the initial estimates of average ore densities, percent solids, etc. These are expected to exist in all cases, and should not influence final designs by significant amounts. There are some characteristics, however, which do have significant effects and have to be considered in detail. Soil Characteristics Soil characteristics, if they are known at the time, are usually not given to the grinding mill designer with the bid specification. In the past, most worries about soil characteristics involved differential settling of the bearing piers. With good foundation design this is not a problem, although in one case in India there was continuous differential settlement between the piers, as the mill was being erected, in the neighborhood of 50 mm. This kind of problem, however, is purely foundation engineering, and is not covered in the mill design. A more insidious problem exists if the overall soil spring constant is relatively soft and a mill dynamics problem can be initiated. A recent, well-known and documented [4, 5 , 61 mill system dynamics problem involved the 40 ft (12.2 m) gearless Cadia SAG mill stator. Therein, on a “typical” soil and foundation, the gearless drive stator structure was found to be too flexible. In some upper mill speed ranges, the nonlinear, steady state dynamic response of the stator was found to violate the motor air gap set limits and thus be in an “unsafe mode” [4, 5 , 61. Thus, the upper portion of the stator structure was suitably stiffened to reduce this steady state response. In the initial stages of the El Teniente 38 ft (1 1.6 m) SAG mill system design, both the stator structure and the foundation structure were somewhat stiffer than that of Cadia. However, the overall soil stiffness was so low as to overcome both of these improvements. This did not become evident until the design was well along, and since the soil stiffness could not be easily remedied, additional adjustments in the stator and foundations had to be sought. This demonstrates the possible existence of a variable which may not even be known in the preliminary design stage, yet may significantly affect the final structures. Load Paths Another “indefinite” design variable that affected structural design was discussed in References [7] and [8]. This was the amount of friction a pad bearing design could possibly contain, which, in turn, would modify the pad bearing load into the mill structure from pure radial pressure to pressure plus moment. This “unknown” depends on the possible friction coefficient that can be generated in different pad bearing designs, and their operating environment. Also some mill configuration designs, such as the shell supported riding ring, would be more sensitive to this, as well as bearing load position change due to overall thermal expansion, than trunnion mounted designs. A similar statement could be made comparing possible pad bearing load distributions among the pads versus sleeve bearing designs. Thus, it is noted that different uncertainties exist for similar phenomena, depending upon the structural or mechanical configuration chosen.

Acceptance Criteria A final design uncertainty is concerned with the choice and application of the fatigue acceptance criteria for materials used in grinding mill construction. The acceptance criteria are chosen depending upon what is absolutely known about the material and what is only statistically estimated. Examples can be taken either from welded fabrication or cast fabrication. Let us start with a cast head example. A cast head will have one surface machined to a specified finish, and one side as-cast. The as-cast surface may have varying surface roughness, embedded sand granules, notches due to mold making methods, etc. The overall casting may have distorted in the mold, and thus, with only one side machined, may have a varying thickness not

850

considered in the drawing. The machining may have been such that some of the stock allowance is purposely left on the produced casting for a variety of reasons. The end result is that the manufactured mill head is only an approximation of the drawn one, and more importantly of the analyzed one. The approximation must be considered in the original choice of fatigue design acceptance criteria. If we look at an experimental specimen fatigue value, we know that cannot be used directly for design. The test specimen is smooth machined, and the mill casting will contain notches. Therefore, we have to start with the obtained values from notched fatigue specimens, which automatically incorporate the material fatigue notch sensitivity. This will take care of the surface discontinuities, but does not consider the thickness variations mentioned earlier or where the physical notches may occur. Suppose, for example, the head casting contains a mold surface notch in a locally as-cast, radiused area. Thus, we could have an average stress concentration in the radiused area, and simultaneously a point stress concentration due to the surface notch. This combination is not considered in the notched fatigue specimen, and thus the fatigue acceptance criteria must have a second reduction factor for this. Alternately, we must have analyzed every geometric radius in detail on the cast head, which is not done. Even at this point we are not finished, for the casting may have subsurface flaws such as shrink, or some dross areas on the surface. Since the notched fatigue specimen has specifically defined geometric discontinuities, it must be established that these other surface and subsurface discontinuities do not produce worse fatigue reduction factors. This can be done experimentally for the multiple types of flaws in question [9]. Only upon the cumulative consideration of all of these approximationscan the fatigue acceptance criteria be chosen. A similar example can be given for the weld fabrications. Again, geometric approximations are always present from the fact that the large diameter mill shells are made from as-rolled plate (with its thickness and radius variations), with hand grinding of weld contours. Welding flaws are also present, although they may be size limited by weld inspection and repairs. An additional variable, however, is residual stress [ 101. The significance of this practically undefined magnitude variable is discussed in Reference [ 101, and the following example from Reference [ 111: Two coupons of AISI 8629, one atmosphere carburized and one vacuum carburized, are ground under controlled conditions. Grinding is performed using a wheel with coolant flow, and removing no more than 0.001 inch (0.025 mm) of stock per pass, for a total of 0.004 inch (0.10 mm) stock removal. X ray diffraction then showed the tensile residual stress of the vacuum carburized coupon to be 37 ksi (255 MPa) higher. Rechecking provided the cause as the resulting higher grinding heat (friction) on this coupon caused by it being 1 HRC point (15 HBN) harder on the beginning surface, and 4 HRC points (60 HBN) harder on the final surface 0.004 inch (0.10 mm) below. From this example, one can draw conclusions how uncontrolled hand grinding of welds can vary residual stresses on these fatigue parts. Therefore, similar to castings, we cannot work with a weld metal fatigue allowable found in laboratory specimens. The fatigue criteria we choose must consider variations in geometry, unknown levels of residual stress, variations in surface finish notches, and metallurgical notches and allowed subsurface flaws due to the different welding processes. This is the way welding codes [12] choose their fatigue criteria, from experimental evidence. There is one more simplification used by welding codes for setting the fatigue criteria for their joint classes, and that is stress (or load) pattern. Since the choices come from a series of experimental studies, usually the loadings are simplified by the laboratory apparatus. Therefore, a direct finite element analysis (FEA) stress, for a weld joint in a complicated structure under a complex loading pattern, is not always compatible with the code allowable, and some modifications [ 121 are needed.

851

Codes such as BS7608 [I21 base their allowables on the statistical analysis of their experimental data. Thus, their statistical safety factors can be changed by designers using 8 deterministic load factors, or requiring the same accuracy for extra long lives (say, 2 x 10 cycles) 7 where the code experimental data is much sparser, than the code “cutoff” of 2 x 10 cycles, or by adding accident conditions of indefinite magnitude or time using Miner’s Rule. The limits of code data have to be considered for any accuracy in such modifications. Thus, these codes are an excellent guide, but not a substitute, for carefully collected data on long term (20-30 years) grinding mill life history from a designer’s own experience.

GRINDING MILL DESIGN The design of grinding mills today is being driven by three main factors: direct purchase cost, advertising of a category the author will classify as “techno-novels,” and reliability. The factors are listed in the order of priority that seems prevalent today; although, in some few, distinct cases, reliability leads the groupings. There is no denying that costs are driving forces, and will always be very important. However, it is to be hoped that the costs would be evaluated on a dollars per total purchased value rather than absolute dollar magnitude basis. This, however, puts heavier burdens on the evaluators, especially in sorting out the prevalent techno-novels that have arisen. Techno-novels The author uses this phrase to contain the many categories of dubious advertising statements and claims that are being made today about various technical aspects of grinding mill designs. The best way to explain this is to look at some category examples. 1.

The half-true statement. This approach starts off factual and ends with an assumption. For example: “Fabricated mill construction is characterized by the absence of cast trunnions or end plates.(True) This results in a lower mill mass.(True) The fabricated end plate design is associated with absolute certainty in respect to material quality.” (Assumption) Can you have absolute certainty of the material quality of steel plates? Two experiences can answer this. A manufacturer buys steel plate from a warehouse that sources the structural grade plate from Eastern Europe. The plate, as delivered, meets all the requirements of the structural grade (ASTM A36), but also has microalloy elements that are not required to be measured by A36 (expected to be negligible). The end result is that this plate cannot be cold-rolled without breaking, and needs a special welding procedure, all of which is found out “after the fact.” A second experience concerns flange plate which again meets all the inspection requirements, but, due to a slight oversight of cropping of the ingot, also contains some elongated, semi-fused porosity. This only shows up as magnetic particle indications on the final machined surface. Finally, let us not forget that welding itself is a variable, flaw-inducing process for plates.

2.

Presenting half of a prolcon argument. This approach presents half of the facts. For example: “The shell (supported) design allows for straight end walls which results in less liner wear due to abrasion and impact.” This may be perfectly true, but it also neglects to mention that what this means is there is less charge mixing, agitation, and grinding in this area.

3. Misapplying engineering theory. This approach takes a well-known engineering theory and applies it out of context, or outside of its limitations. For example: “The shell supported mill design results in significantly lower bending moments than the trunnion supported mill design. The design is inherently stiffer and has less deflection resulting in

852

a more favorable ring gear to pinion contact (or air-gap in gearless applications) through a range of load conditions.” This statement is based on engineering beam theory which assumes a rigid cross-section, and calculates the overall displacement of that crosssection center for a beam, which is defined as having a length over 5 times the magnitude of other dimensions. This may have been true for the long tube mills built 30 years ago, before mill structure analysis was done by FEA. But, the theory does not apply to today’s SAG mills with diameters almost twice the length, or even to large ball mills. The details of the statement are even more interesting, in that, when the theory is applicable and the cross-section remains rigidly circular, that is the point where a gear-topinion, or an air gap, is the easiest to set up, for any design. It is the out-of-circularity displacements that create the difficulties. 4.

Unproven or irrelevant statements. These statements have no general applicability. For example: “The ring gear is positioned adjacent to one of the shell supported bearings which allows the ring gear to be exposed to lower dynamic forces when compared to a conventional mill.” Ring gears have been placed on the corner between the cylindrical shell and the conical head of SAG and ball mills, in the middle of older SAG mill shells up to 32 ft (9.75 m) diameter, attached to trunnions of the same size SAG mills, on the cylindrical shell (as described in the quotation), and sometimes intermediate places, all without any detrimental effects. The dynamic forces in each case (that the gear experiences), are a function of charge dynamics and the various thicknesses of the structural components, and thus vary for each individual application, never having much significance in any of the designs known to the author.

Choosing a Mill Configuration It is unfortunate that mill evaluators have to cut through the above advertising techno-novels to get to the basic reliability comparisons of different grinding mill configurations. Basically, what drives a mill configuration choice, a choice of appurtenant structures, and a choice of options is usually the same: price. The challenge is to make this value. Other than that, the engineer has complete freedom in choosing a configuration, as long as customer requirements are met. At this stage, the configuration is not governed by segmentation, or abilities of multiple shops to make pieces as opposed to a sole source, etc. These are secondary constraints to be applied later in reaching an optimum design. Thus, some relevant customer variables at this point are: 1. The overall size of the mill to match the customer’s processing (kilowatts, tons per hour) requirements. 2. Basic openings to match customer flow and velocity requirements. 3.

Estimates of weight for preliminary bearing configurations.

For example, at this stage, based on the accuracy of the process flow requirements, the positioning of the mill bearings may be configured as in Figure 9a. This will progress to the eventual decision as to the use of pad or sleeve bearings. Sleeve bearings will require the use of “smaller” diameter trunnions, while pad bearings can fit equally well on smaller and larger openings. It is important to note that construction materials are chosen at a later stage, and the term “trunnion” need not denote a casting nor “riding ring” a fabrication. Similarly, the bearing lubrication type can be chosen later with reference to details such as bearing local geometry and pressure.

853

a) Basic Configurations

-

b) Adding Segmentation

-

-

c) Introducing Materials

D d) Introducing Details Figure 9 From Configuration to Materials Processing

854

-

-

The next important variable in configuring the mill will be segmentation requirements, as noted in Figure 9b. This may come from fabrication constraints, but will most likely be driven by economic shipping and handling limits. In fabrication, with vertical boring mill (VBM) cutting capacities up to 20 meter diameters, we are not likely to reach dimensional limits soon. However, the cut diameter has to be considered in conjunction with VBM table weight capacity and crane capacities. The inconsistent combination of these three variables (as far as mill pieces are concerned), at most shops, will produce the fabrication limits requiring segmentation. One final area of configuration is concerned with details, and comes u . e r material choices are made (see Figure 9c and 9d). This is so interrelated with material procurement that it may be misrepresented. Nominally, Figure 9d shows the choice of a “contoured” flange, which may be said to have some advantage over a plate flange. (Note that it is the engineer’s job to design both configurations with equal safety so that the advantages, if any, are reduced to commercial.) However, the advantages are dependent on the dimension ‘a’, and its choice will be dependent on raw material price, the cost of machining to net section, and the accuracy of the FEA model for stress analysis. Thus, some contour flanges are “more contour than others,” and the only technical advantages tend to quickly disappear with decreasing ‘a’. With the configuration chosen, the typical characteristics of the two classes are shown in Table 3. This is not a “pro/con” table but merely a description of characteristics. Erection sequences of the configurations are shown in Figure 10 (shell supported) and Figure 11 (trunnion supported).

Table 3 Configuration Characteristics

SHELL SUPPORTED DESIGN

rRUNNION SUPPORTED DESIGN

- Can be used with all bearings

- Cannot use sleeve bearings

- Can use all pad bearings

- Can use all lube systems

- Can use all lube systems

- Can use all drive systems

- Can use all drive systems - Polysius Combiflex can use common housing for bearings and gear - Hofmann SMTP can use common housing for

bearings and gear - Uses castings in main components

- Stiffer bearing trunnion - Smaller diameter bearing seals - Longer mill - Simpler lining assembly - Access - one floor - Faster erection of bearings

- Uses weldments in main components - Can use castings for end closures - Riding ring deflections should be checked - Large diameter bearing seals - Shorter mill (need equivalent volume for power) --------- Access - possible two stories - Lesser concrete foundations - Faster erection of rotating structure (no trunnions)

855

Figure 10 Syama - Erection

Figure 11 Cadia - Erection

856

Since the FEA will be assumed to be carried out accurately without errors [lo], and since the stress limits will be chosen appropriately, the pro and con comparison of these designs will most likely be in the structurallmechanical details, such as: Elimination of the segmented riding ring weaknesses to leakage and contamination, while keeping bolting in a reasonable location for inspection and correction (see Figure 12). Older designs allowed mill connection bolts to protrude into the mill interior, the nut or head to be covered up by mill liners. What this approach did not consider may be described by any mill operator faced with a decision of running or shutting down if one of those bolts loosened, elongated, or broke. 0

Designing equal ease of access to the mill interior through the feed opening, and to the mill bearings through the housing openings wherever the bearings may be located. If the bearings are low on the mill, the access for liner handlers, feed chute, etc will be from a flooring close to the mill centerline, while the bearings are substantially below this level. Thus, access and approach must be designed from a two levelltwo story scenario. Eliminating local structural details that are designed as “erosion traps”; see example in Figure 13. The area ‘A’ is a perfect location for fine slurry collection, and continuous erosion circumferentially, until eventual failure of the flange. Such erosion has been experienced by various older SAG and ball mills in Canada, Sweden, and elsewhere. Designing the chosen bearing lubrication system to be compatible with mine personnel experience. While it is true that the simplest will be the easiest to install, replace parts for, and be the cheapest, it is not always the easiest for trouble-free operation. One mine operator learned that buying a pad bearing lubrication system with no automatic flow control to each component required an extra full time operator for continuous adjustment of manual flow valves for the system to operate properly. Choosing which accidents and maintenance occurrences are important for design, and which can be neglected. For example:

-

In case of an accidental power outage, some pad bearing systems must have extra oil from accumulators, and some can be marginally forgiving without the use of accumulators. The need for these devices is not solely the difference between hydrodynamic and hydrostatic pad geometry and sliding area. It is based on details such as: how long the mill will rotate or rock; how much residual oil adheres to a large riding area; was the price of the lube system further cheapened by neglecting a cooling circuit, thus lowering the viscosity of the residual oil; how free are the pad bearing pivots; etc.

-

In case the mill design is based upon equal pad loads, what happens if the pad loads vary? How long does it take to balance the loads each time, and what is the access for this operation?

-

In the case of large diameter seals, what are the variables for good operation, and what is the access to adjust these variables? If the seal depends on tensioning and sliding, what is the wear life, how does it perform when worn, and how fast can it be replaced and readjusted?

In the case of all the above examples, the question to be answered is whether the cheapest solution is also the optimum, from the point of view of trouble-free “up time” for the specific site conditions, and the operation’s personnel preferences/skills.

857

riding ring

t"1

segmented riding ring, or offset ring design

modification to avoid segmenting riding ring

Figure 12 Segmenting Riding Ring Designs

858

\

Figure 13 Local Connection Detail with Erosion Trap (Ref. [13])

Figure 14 Projected 2-D Mesh of 3-D Element Idealization (Ref. [13])

859

The Analysis Steps We have considered the process conditions, defined the loads, carefully considered acceptance criteria, picked our mill configuration, segmentation, and materials; so we are now ready for the stress analysis [lo]. This is actually a three step process: pose the “question” correctly to the FEA computer program (the computer program chosen must be capable and accurate enough to provide a good “answer”); interpret the “answer”; and assess its meaning for the mill system. The second step is perhaps the easiest, as most of the FEA programs in commercial use today have been validated with reference to some kind of standard benchmarks. There are international organizations that can provide a set of standards. The potential program user only needs to check that his chosen program has been so validated, and that the elements he has chosen to use have also been part of that validation, in a way consistent with his application. The other two steps are harder to establish. The user could have had FEA course work in the process of obtaining his accredited engineering degree, or completed introductory FEA vendor training from the program developer. Unfortunately, the advances of FEA software in ease of use, and the corresponding increase in users, far outstrip the user’s efforts to obtain the learning required for the techniques, engineering, and discipline required to successfully use these programs. Ongoing education is required in FEA basics, engineering mechanics, and failure theory, to keep up with advancing software application. Once we have qualified users, the analysis is actually quite straightforward and only needs a few decisions: 0

How many specific load conditions need to be individually consideredhnalyzed?

0

Are there portions of the mill structure which need deflection checks as well as stress checks?

0

What will be the FEA progradelements used? What will be the idealization type (2-D or 3-D)?

The pitfalls of FEA modeling are mentioned in Reference [lo], with some demonstrations. It should be noted that expertise, in the mining industry, in the use of this method, has progressed very little in over 20 years of application. For example, in 1979 [14], the author presented an example of several 3-D elements used for a sample plate bending problem, wherein the solution accuracy was shown to “collapse” as the element aspect ratio increased from a little over 3, to 10. Yet it is easy to find examples today, where one of the same elements is being used to analyze mill shells, with aspect ratios in the 60 to 100 range. Worse yet, finite element shape regularity conditions, as discussed and demonstrated in [lo], are also violated in the same models, as shown in locations ‘A’ and ‘Byof Figure 14. Thus, the answers from such models must be seriously suspect as to accuracy. The above should serve as a warning, that requirements of specific FEA configurations in mill specifications, by laymen, are very serious lapses in judgment, and should never be used. Let us characterize, quickly, the two types of models termed 2-D and 3-D. The 2-D model uses elastic ring-type finite elements, so that the grid idealization is only two-dimensional. The variations in the third direction (hoop direction) are usually expressed in a Fourier series. The variation of the non-axisymmetric load usually determines the number of Fourier terms required. The 3-D model uses “brick” elements, and thus has a grid pattern in all three dimensions, including the hoop.

860

Either modeling technique can produce good results. The 2-D model gives a refined grid through the thickness (see Figure IS), but is incapable of analyzing around discrete geometry such as axial flanges. The 3-D model, if used for the entire mill structure at once (instead of substructuring), needs to use isoparametric elements and a properly configured grid pattern. By comparison, the 2-D model can use about 6000 degrees of freedom, and will need 25 to 30 Fourier terms. The 3-D model with isoparametric “brick” elements will need between 300 000 and 500 000 degrees of freedom for a quarter symmetry model, with good accuracy. Using substructuring techniques this can be reduced in 3-D, but will then need both overall and local area models. Analysis can be enormously enhanced by obtaining and keeping detailed historical records of mills. These records would contain correlation between mill analyses and strain gauge results in local areas, a history of mill local cracking for different workmanship conditions, long term survival of mills which are then reexamined and reanalyzed at decommissioning,etc. This kind of data allows for analytical modifications with some degree of confidence, both in the areas of stress analysis and acceptance criteria. For example, while a 2-D FEiA model does not give data around axial flanges, several strain gauging programs in this area can provide such data and refer it back to areas away from the flange. Thus, the axial flanges no longer need modeling, and the local stress can be obtained by using a multiplier on the “remote” stress, provided strain gauge programs were used to obtain such a multiplier.

I

Figure 15 Advantages of 2-D Meshing

861

Similarly, it should be recognized that the welding codes [ 121 are neither directly created for mill structures nor do we use them exactly as intended for bridge structures [9]. Thus, any deviations should be backed by engineering logic and historical data from surviving mills. This will give a realistic assessment of weld finish improvement, effects of component thickness, effects of large magnitudes of compressive stress range on fatigue of oven stress-relieved structures, mill flange bolting, etc, all within the “average” environment actually experienced by grinding mills over 25+ years of life.

-

QA The Link Between Design and Fabrication One other piece of data provided by a study of surviving mills is the sensitivity of specific component designs to fabrication flaws. This of course, has to be supplemented by laboratory testing of specimens with typical flaws [9]. This kind of data can be converted to “fitness-forpurpose” information, since 25 year old mills were made to various, different QA requirements than those built in the last 10 years. In setting up a QA program, the fitness-for-purpose aspect of flaws has to be known in order to set up the inspection limits. Equally important is the knowledge of inspection reliability, along with the “typical” flaw characteristics of various individual fabricators. The buyer must realize that the whole QA program is a matter of statistics. The I S 0 QA programs are aimed at insuring that the inherent statistics are not further compromised by bad record keeping, but they do not eliminate the statistical nature of QA. In advertising literature one can often find the statement that fabrication (of something) can be assured with “a proven NDT program.” Let us see what steps need to be taken to create such a program: 1. The inspector’s reliability depends upon: 0

0

0

His overall NDT (nondestructive testing) education and experience. The reliability of the NDT instrument in finding the specific flaw being sought, His familiarity with this NDT instrument use, and the inspection specification. Whether he is having a good or bad day.

2. The inspection reliability depends upon: 0 How well the specification is written. 0 How accurate is the match between the capability of the equipment used and the sensitivity required by the inspection. 0 The calibration of the equipment.

3. The inspection suitability depends upon: 0 Whether the inspection is being performed at the right step in the fabrication. 0 The engineer’s knowledge that this is the critical flaw and of its characteristics. 0 The engineer’s knowledge of the behavior of the flaw through various fabrication stages. 0 The engineer’s accuracy in assessing inspection reliability; i.e., what is the largest flaw that can be missed, not the smallest flaw that can be found. 0 The engineer’s knowledge of which typical flaws are expected from this particular shop. Note that the last few points are dependent upon the engineer (or specification writer) being familiar with the specific shop doing the fabrication work. Cost pressures can make mill designers use one shop in the USA for one mill, and make the next mill in China or South Africa.

862

Knowledge is required of each shop’s procedures in welding, casting, heat treating, machining, and NDT personnel capabilities and equipment, to make sure that all of the written specifications or procedures apply equally. It is not unusual to find one shop using different manufacturing steps or having different machinery from another, requiring different inspection points or techniques, or one NDT group being more capable or more familiar with a particular procedure than another. Knowledge of only the shop’s fabrication costs is grossly insufficient. Specifications are usually written for the most familiar shop. Thus, it is important to realize that a single specification or NDT program is only “proven” for that one shop. Every time fabrication is sourced from a different shop, the required procedures must be reviewed for full applicability to different methods of fabrication, which may have different “most probable” flaws, and possibly the different expertise of the shop inspectors. Therefore, to have a “proven NDT program” or specification, the mill designer must also have full time NDT inspectors, failure engineers, and QA technicians on its staff. Otherwise, sourcing fabrications at different shops, by cost, is just a “luck of the draw” process, and will produce a corresponding fabrication.

SUMMARY While the methods of mill design may be relatively straightforward, the choices of all the variables determining the final design are not. The choices should be governed by a conscious effort to produce the most suitable mill design for the specific characteristics of the mine site (including maintenance), but often can be overwhelmed only by first cost considerations. In any case, the design should cover all of the following:

1. The process variables affecting accidental and normal load conditions should be considered. 2.

Some accidents made possible by operation cannot be covered in the limits of the structural design. In those cases, mechanical designs should be available to prevent, minimize, or warn against such accidents.

3.

Reaction of the structures to possible earthquakes should be explicitly defined.

4.

For large mills, the dynamic interaction of soil and foundation, and their effects on the mill and related structures, should be considered.

5.

Design acceptance criteria should be defined for the practical (actual) mill structures, not for theoretical structures.

6 . Design engineers must take responsibility and be accountable for the tools and techniques they use for analysis. Following the above considerations, mill configurations can be compared using equal reliability guidelines, and economics can be used for final decisions. With any configuration, the analysis and fabrication steps are linked by the QA program, which should be based on “fitnessfor-purpose’’considerations defined by the analysis. Thus, the accuracy and applicability of both the analysis and historical performance records in finding such limits cannot be overemphasized, The QA program should not be viewed as a theoretically perfect paper study, but should take into account all the reliability variations possible. Doing this will require QA program modifications as different fabrication shops are sourced.

863

REFERENCES [I] J.A.Herbst, and L.K.Nordel1, “Optimization of the Design of SAG Mill Internals Using High Fidelity Simulation,” Vol. IV,p. 150-164, proceedings SAG 2001, Oct. 2001. [2] R.K.Rajamani, and B.K.Mishra, “Three Dimensional Simulation of Charge Motion in Plant Size SAG Mills,” Vol. IV, p. 48-57, proceedings SAG 2001, Oct. 2001. [3] D.Rogston, “Interpretation of Charge Throw and Impact Using Multiple Trajectory Models,” Vol. IV,p. 115-123, proceedings SAG 2001, Oct. 2001. [4] VSvalbonas, “Updates in Design for Large SAG and Ball Mills,” Workshop de Molienda SAG 2001, Chile, May 2001. [5] C. Meimaris, B.Lai, and L.Cox, “Remedial Design of the World’s Largest SAG Mill Gearless Drive,” Vol. 11, p. 74-83, proceedings SAG 2001, Oct. 2001.

[6] H.Kummlee, and P.Meinke, “A Mechatronic Solution: Design and Experience with Large Gearless Mill Drives,” Vol. 11, p. 11-24, proceedings SAG 2001, Oct. 2001. [7] V.Svalbonas, “Difficulties in Mill Comparisons - Shell Supported vs Trunnion,” Vol. 11, p. 128-141,proceedings SAG 2001, Oct. 2001. [8] V.Svalbonas, “Mechanical Design of Large Grinding Mills - AGISAG,” in Mineral Processing and Hydrometallurgy Plant Design - World’s Best Practice, Australia Mineral Foundation, Oct. 1999. [9] V.Svalbonas, and P.Ulrich, “Fitness for Purpose: Should You Buy One Grinding Mill for the Price of Two?,” Paper 01-25, presented SME Annual Meeting, Feb. 2001. [ 101V.Svalbonas, and M.Fresko, “How Safe are Your Recent Mills? The Compatibility Between

FEA and Welding Codes,” Paper 02-40, presented S M E Annual Meeting, Feb. 2002. [ 111Heat Treating Process Magazine, p. 39-40, Nov. 2001. [ 121British Standard BS7608, “Code of Practice for Fatigue Design and Assessment of Steel

Structures,” 2/95 Revision. [ 131J.Hadaway, and E.Hecht, “Extended Detailed Finite Element Analysis of a 9.6 Metre

Autogenous Sag Mill,” AN1 Internet web page, 2001. [ 141V.Svalbonas, “Some Considerations in Computer Structural Analysis of Large Grinding Mill

Shells,” AIME Preprint 79-17, Feb. 1979.

864

Sizing and Application of Gravity Classifiers W. Michael Reed

ABSTRACT T h s chapter discusses the key sizing parameters used for selection of gravity classifiers. The three basic elements used for calculating the operating size requirements of mechanical gravity classifiers: settling rate, overflow capacity, and raking capacity are presented. Typical applications with examples are described. Finally, adjustments that can be made during operation to alter operating performance are reviewed. INTRODUCTION Before the advent and widespread use of hydrocyclones, mechanical gravity classifiers, primarily the spiral type, were commonly used in closed circuit grinding applications. Due to the smaller ball mill sizes then in use, a spiral or rake classifier could be conveniently located adjacent to the ball mill, receiving ball mill discharge and returning oversize back to the feed end of the mill. With the ever-increasing diameter and length of ball mills, this convenient fit disappeared so that in modem applications, gravity classifiers are seldom if ever used in closed circuit grinding applications. However, there remain several applications in minerals processing where spiral classifiers are typically used, such as sizing, washing, and dewatering of ores, sands, and salts. Applications for closed circuit grinding and some common modem applications are presented. Although t h s chapter discusses spiral classifiers specifically, the principles also apply to rake and drag classifiers. EQUIPMENT DESCRIPTION A spiral classifier consists of an inclined elongated tank of fabricated steel plate and structural steel in which one spiral assembly (simplex classifier) or two spiral assemblies (duplex classifier) mounted parallel to the tank bottom revolve slowly without touching the sides or bottom of the tank. Classification is accomplished according to size or specific gravity through the difference in settling rates of the particles in the tank. The feed normally enters at the pool level through a rectangular feed opening on one side of the tank. The tank is equipped with an adjustable overflow weir to maintain pool level and an overflow box for collecting the overflow product, which generally consists of fine solids and water. The coarser, or heavier, settled solids, commonly called rake, are conveyed along the drainage deck by the revolving spiral(s) and discharged at the upper end of the tank. Depending on the desired size of separation, the tank depth or spiral submergence at the overflow end generally ranges among manufacturers from about 90% to 150% of the spiral diameter. Lower submergence is used for coarse separations and higher submergence is used for fine separations. To increase settling area, tanks can be flared, which reduces overflow velocity allowing a finer separation. Side overflow weirs may be added to W h e r reduce overflow velocity. The slope of the tank is normally 14" to 18". Generally, hgher slope (smaller settling pool area) is used for coarse separations and lower slope (larger pool area) is used for fine separations. The raking mechanism consists of a single spiral (single pitch) or double spirals (double pitch) mounted on a shaft. The spirals are continuous from the overflow weir to a point above the pool

867

level. Manufacturers in the United States have generally built production size classifiers with outside spiral diameters ranging from 610 mm (24 in.) to 1981 mm (78 in.) in 152 mm (6 in.) increments. In size tabulations, the diameter in mm is typically taken as 25 times the diameter in inches. The pitch of the spiral (advance per revolution) is generally 50% to 75% of the spiral outside diameter. Double pitch spirals are constructed with the second spiral ribbon 180" apart from the first. Triple pitch spirals are constructed with the spiral ribbons 120" apart. At the same rotating speed, a double pitch spiral will have double the raking capacity of a single pitch spiral and a triple pitch spiral will have triple the raking capacity of a single pitch spiral. When raking capacity requirements exceed the capability of a single spiral assembly (simplex classifier), a second spiral assembly can be added. A classifier with two spiral assemblies is called a duplex classifier. A spiral classifier is generally identified by two dimensions, spiral diameter and inside tank length. Further identification includes the quantity of spiral assemblies, simplex (one spiral assembly) or duplex (two spiral assemblies in one tank); the quantity of spirals per assembly (single pitch, double pitch, or triple pitch); and submergence. A further modifier is flaring of the tank to increase the overflow length. The tanks will be either straight (no flare), medium flare (partial flaring), or full flare. The key component of a spiral classifier is the main shaf, which is generally large diameter seamless steel tubing of sufficient wall thickness to prevent deflection due to the spiral weight and the thrust load imposed by the spiral. Cast iron or steel arms are clamped to the main shaft. The spiral consists of sectionalized circular steel flights bolted to the arms to form the continuous spiral. Replaceable abrasion resistant wear shoes are fastened to the lead edge of the flights with countersunk bolts. Shoe materials are polyurethane, Ni-hard, abrasion resistant steel, or ceramic. The shaft is designed with a pivot at the upper end. The lower submerged bearing consisting of sealed roller bearings is bolted to the lower end of the shaft, and is easily raised by the spiral lifting device for inspection or maintenance. The submerged bearing is lubricated from a grease pipe extending along the center of the shaft to a grease fitting at the upper end. The spiral drive is connected to the upper end of the shaft beyond the discharge point. Modem drives consist of shaft mounted gearmotors, or shaft mounted reducers driven by V-belts and electric motors. DEFINITIONS

Classification is the separation of a stream of particles of different sizes in a fluid medium into two streams. Gravity classijkation is a process using the force of gravity to effect a classification. Settling rate is the terminal settling velocity of the smallest particle that will overflow during gravity classification. Overflow is the fluid stream carrying the finer or undersize fraction during classification. Underflow (or rake) is the fluid stream carrying the coarser or oversize fraction during classification. SIZING A GRAVITY CLASSIFIER Regardless of application, there are three basic steps used in the sizing of a spiral classifier. These are determination of settling rate, determination of required overflow capacity, and determination of raking (or underflow) capacity. Each of these basic sizing elements is reviewed here. A machine is then selected which dimensionally satisfies the sizing parameters.

868

The primary consideration in classification is a determination of what s u e of particles to pass (overflow) or to retain for W e r processing. In the minerals processing industry, classifiers are generally used to make a separation at a specific particle size in a slurry with solid particles of varying sizes in water. For gravity classifier sizing, it is necessary to determine the settling rate of the smallest particle that will overflow. Particles larger than this size will settle and be carried out by the rake(s) and particles smaller than this size will also overflow. Some basic rules of classification that apply are: 1. Coarse particles have a higher settling rate than fine particles of the same specific gravity. 2. Irregularly shaped particles generally settle more slowly than spherical particles. 3. The settling rate of solid particles becomes progressively slower as the density or viscosity of the fluid increases. This is commonly known as hindered settling.

To make a split at a given particle size, a complete mass balance around the classifier is needed. At minimum, this will consist of the following information: 1. Feed solids size distribution 2. Feed slurry solids content 3. Feed slurry volumetric flow rate 4. Specific gravities of solids and liquids 5. Volumetric flow rate of dilution liquid 6 . Required underflow solids content 7. Required underflow slurry volumetric flow rate

From mass balance calculations using the required operating conditions, the required flow rate and solids content for the overflow are determined. Through the following calculation procedure, a classifier is then selected that will have the required pool area (for overflow) and raking capacity (for underflow). Settling Rate The particle settling rate is dependent on particle shape, slurry density, solids specific gravity, and liquid density and viscosity, which all affect the Reynolds number and particle drag coefficient in a complicated way. In many cases, the hindered settling rate V, m y already be known from similar applications or from laboratory testing. When the settling rate for a given application is not known, it can be estimated from empirical relationshps. The settling rate for the smallest particle to overflow must be determined. For a standard condition of water at 20°C, the terminal settling velocity (or settling rate) for a spherical particle of diameter d mm is as follows: V, = 8.39 ([l

+ 0.0103 (1 - Cv)4.66 g d3 ( S - l)]0.5- l}/d

Where V, = settling rate, m d s d = particle diameter, mm g = gravitational constant = 9,810 &s2 C, = solids volume fraction or percent solids by volume S = ratio of solids to fluid specific gravity Refer to Appendix B for the derivation of this formula, for necessary adjustments when water temperature is different than 20°C, or when fluids of different density than water are involved in the process.

069

It should be noted that the settling rate derived above is for particles of spherical shape, whch are not normally encountered in minerals processing. The experimental factors for particles of different shapes vary with Reynolds number in a complicated and obscure way so that any tabulated values are only approximations. Generally accepted values for the shape factor for typical minerals are between 0.90 and 1.00. The settling rate determined above should be multiplied by the shape factor. However, due to the dependence of the settling rate on many variables, and its greater dependence on machme dimensions, Equation 1 without correction would be valid within an acceptable engineering accuracy. Once the settling rate of the finest particle to overflow is established, it will be used to calculate the required overflow capacity for the classifier. Overflow Capacity Sizing the classifier from the overflow standpoint is a matter of calculating the required pool area needed to allow settling. For most cases we have determined the flow rate from the mass balance calculations and the settling rate fiom test work or calculations, so we solve for the product of classifier dimensions E and Win accordance with

Where E = the distance along the classifier pool between the feed opening and the weir, m W = effective weir length, m, and Q = theoretical overflow, m3/h *Thederivation for this formula can be found in Appendix C. Most manufacturers’ literature includes pool area tabulations for their standard spiral classifier sizes, types, and slopes. Raking Capacity From material balance considerations, we know the solids flow rate required to be raked. Generally, a machme of standard size is selected that will provide this raking capacity. Raking capacity is a complicated function of spiral dimensions, tank slope, and rotational speed. One formula for raking capacity is as follows (Hill 1982):

T = w P p (D- 0.75 ~ ) / 2 8 . 3

(Eq. 3)

Where T = dry t/h per spiral rpm w = flight width, m P = flight pitch, m p = (rho) dry bulk density of material transported, kg/m3 D = outside flight diameter (spiral diameter), m, and 28.3 = combination of conversion factors and an arbitrarily assumed efficiency factor It should be noted that this formula does not consider classifier slope or physical characteristics of the material being transported. However, some general observations can be made. At a given spiral diameter and tank slope:

1. Raking capacity is directly proportional to spiral speed 2. Raking capacity is directly proportional to dry material bulk density

870

3. Raking capacity is directly proportion to spiral flight pitch Most manufacturers have developed proprietary raking capacity calculations, and their literature includes ralung capacity tables for standard spiral classifier sizes and slopes. Machine Selection Tables A.l, A.2, and A.3 located in Appendix A provide standard overflow capacity and raking capacity information for a major manufacturer. Tables A.4 and A S provide fine sand washing capacity information for a major manufacturer. EXAMPLE 1 CLOSED CIRCUIT GRINDING In this example we assume we are grinding copper ore, producing flotation feed of 95% passing 212 pm (70 U.S. mesh) and 30% solids. Ball mill discharge is 200 t/h of solids at 70% solids by weight, and solids specific gravity is 2.7. The ball mill discharge sieve analysis is given in Table 1. Because the overflow is to contain very little plus 212 pm material, size the classifier with overflow capacity to theoretically retain and rake nominally 150 pm (100 U.S. mesh) particles. With the above information we can develop a material balance as shown in Table 2. From Equation 1, the settling rate is calculated as: V, = 8.39 {[1+0.0103 (1 - 0.137)4.66(9810) (0.150)3 (1.7)}0.5- 1}/0.150=7.64 m m / s

From Equation 2, the required pool dimensions E Ware calculated as:

E W = 143.3/[( 1.8) (7.64)] = 10.42 m2 Table 1 Ball mill discharge sieve analysis Size Cumulative % Retained 9.5 mm (318 in.) 3.0 10.0 6.3 mm (1/4 in.) 425 pm (40 U.S.mesh) 41.0 300 pm (50 U.S. mesh) 55.2 150 pm (100 U.S. mesh) 73.5 106 pm (140 U.S. mesh) 77.1 75 pm (200 U.S. mesh) 81.0 -75 pm (200 U.S.mesh) 19.0 Table 2 Closed circuit grinding classifier mass balance Feed Rake Makeup Water 200.0 Solids, t/h 147.0 Water, t/h 85.7 36.8 74.8 285.7 183.8 74.8 Total, t/h % solids (weight) 70.0 80.0 Sp. Gr. Solids 2.70 2.70 Solids, m3/h 74.1 54.4 Water, m3/h 85.7 36.8 74.8 159.8 91.2 74.8 Total, m3/h % Solids (volume) 46.4 59.7

871

Overflow 53.0 123.7 176.7 30.0 2.70 19.6 123.7 143.3 13.7

Thus, a machme with raking capacity of 91.2 m3/h and pool area of 10.4 m2 is required. Referring to Table A.1, we select a 1,500 mm (60 in.) spiral diameter Model 125 simplex medium flare design, which has 12.4 m2 pool area. From Table A.3, the longest tank length of 9.73 m (3 1 ft - 11 in.) is selected. From Table A.2, spiral speed of 5.0 rpm is selected, which provides 97.5 m3/h raking capacity with a double pitch spiral. The drive motor is 15 kW.

EXAMPLE 2 SAND WASHING Special types of spiral classifiers are specifically designed for “washing” fine sand, crushed aggregate, and other materials. Washing refers to removal of slimes, silt, and clay to the overflow water. These machines generally feature fully submerged spirals and fully flared tanks with side weirs added to minimize overflow velocity. Due to the side weirs, feed is generally introduced through a center feed box and a baffle plate is included to reduce the velocity of the incoming slurry and prevent turbulence. Extra long dry decks and 18” slope provide maximum dewatering of the fine sands. The washers are usually selected based on published water capacities for retaining 150 pm (100 U.S. mesh), 106 pm (140 U.S. mesh) or 75 pm (200 U.S. mesh) particles, and on published raking capacities at standard spiral speeds. For example it is desired to wash 200 t/h of specification sand of 1.60 t/m3 (100 lb/ft3) bulk density containing 10 t/h of minus 106 pm (140 U.S. mesh) material. 150 m3/h of water is introduced with the feed. Again, we can develop a material balance as given in Table 3. The rake material is assumed to be conveyable at 85% solids by weight. Thus, a machne with water capacity of 150 m3/h and raking capacity of 190 t/h of solids is required. Referring to Table A.4, we select a single screw washer of 1,350 mm (54 in.) spiral diameter x 10.67 m (35 ft) length, which has 227 t/h raking capacity and 301 m3/h maximum water capacity. The drive motor is 30 kW.

EXAMPLE 3 SEPARATION OF INSOLUBLES IN TRONA PROCESSING Following calcining of crushed trona ore, the dry crude soda ash material is dissolved with weak liquor in agitated tanks. Temperature exiting the dissolving step is about 90°C. Typically, the insoluble portion of the dissolver feed solids is about 20% by weight. From the dissolvers, the hot slurry flows to a spiral classifier at about 6% to 7% solids by weight. The coarse plus 150 pm insolubles settle out in the classifier and are removed by raking to the tailings handling system. The finer minus 150 pm insolubles in the overflow are advanced to thickeners for solids removal. Table 3 Sand washing mass balance Feed Solids, t/h 200.0 Water, t/h 150.0 350.0 Total, t/h 57.1 % Solids (weight) 2.70 Sp. Gr. Solids 76.9 Solids, m3/h Water, m3/h 150.0 Total, m3 h 226.9 % Solids (volume) 33.9

872

Rake 190.0 33.5 223.5 85.0 2.70 73.1 33.5 106.6 68.6

Overflow 20.0 118.2 138.2 14.5 2.70 7.7 118.2 125.9 6.1

A typical classifier mass balance is given in Table 4. Classifier feed is 25 t/h of solids at 6.5% solids by weight. The size analysis indicates 32% passing 150 pm. The liquid is soda ash solution at 1.30 specific gravity. Viscosity of the 90°C solution is 1.4 mm21s. From Equation B.2, the settling Archimedes coefficient is calculated as: A l = {[0.0103 (1 - 0.0109)4.66(9810)]/(1.4)2}=48.9857 From Equation B. 1, the settling rate is calculated as: V, = 8.39 {[1+48.3957 (0.150)3(1.0769)}0'5- l} (1.410.150) = 6.69 mm/s

From Equation 2, the required pool dimensions E Ware calculated as:

E W = 275.2/[(1.8) (6.69)] = 22.85 m2 Thus, a machme with raking capacity of 106.6 m3/h and pool area of 22.85 m2 is required. Referring to Table A.1, we select an 1,800 mm (72 in.) spiral diameter Model 150 simplex full flare design, which has 27.8 m2 pool area. From Table A.3, the average tank length of 11.15 m (36 ft - 7 in.) is selected. From Table A.2, spiral speed of 4.0 rpm is selected, which provides 116.4 m3/h raking capacity with a double pitch spiral. The drive motor is 22 kW. OPERATIONAL ADJUSTMENTS AND MAINTENANCE FEATURES There are two ways to adjust the classifier cut point during operation. The first is water flow or dilution water addition. The separation size is extremely sensitive to changes in water flow or water addition. If water flow in the classifier feed is increased, the overflow rate will increase, therefore coarsening the particle size split to the overflow. Conversely, reducing the dilution water addition rate will reduce the overflow rate, therefore causing a finer particle size split to the overflow.

The second method for adjusting cut point during operation is adjusting weir height. Increasing weir height at constant feed flow will increase the E dimension, therefore increasing settling time, which causes a finer particle size split to the overflow. Conversely, lowering weir height will coarsen the particle size split to the overflow. The lower end of the spiral assembly is equipped with a manually, electromechanically or hydraulically operated lifting mechanism to remove the spiral fiom settled solids for ease of starting and maintenance. The spiral drive is arranged so that the spiral will rotate and be lifted or lowered simultaneously. Table 4 Insolubles separation in trona processing mass balance Feed Rake Overflow Solids, t/h 25.0 17.0 8.0 Liquid, t/h 359.6 5.7 353.9 384.6 22.7 361.9 Total, t/h % Solids (weight) 6.5 75.0 2.21 Sp. Gr. Solids 2.70 2.70 2.70 1.30 1.30 1.30 Sp. Gr. Liquid 3 .O Solids, m31h 9.3 6.3 272.2 Liquid, m3/h 276.6 4.4 285.9 10.7 275.2 Total, m3/h 3.25 58.9 1.09 % Solids (volume)

873

Appendix A Manufacturers Capacity Information Table A.l Metso Minerals spiral classifier pool areas Spiral diameter, Tank Type mm design Model 100 Simplex 600 S 1.40 (24 in.) M 1.55 F 1.72 Simplex 750 S 2.12 (30 in.) M 2.37 F 2.65 Simplex 900 S 3.01 (36 in.) M 3.36 F 3.77 Simplex 1,000 S 4.12 M 4.61 (42 in.) F 5.17 Simplex 1,200 S 5.30 M 5.96 (48 in.) F 6.71 S 8.23 Simplex 1,500 M 9.25 (60 in.) F 10.44 S 11.8 Simplex 1,800 M 13.3 (72 in.) F 15.0 Simplex 2,000 S 13.7 M 15.5 (78 in.) F 17.5 Duplex 1,800 S 22.6 M 25.6 (72 in.) F 29.1 Duplex 2,000 S 26.5 (78 in.) M 30.0 F 34.1 Metso Minerals Pumps and Process Colorado Springs, Colorado Model 100 = 100% spiral submergence Model 125 = 125% spiral submergence Model 150 = 150% spiral submergence S - Straight Tank M - Modified Flare Tank F - Full Flare Tank

874

pool area, m2 Model 125 1.79 2.08 2.41 2.70 3.21 3.72 3.86 4.53 5.30 5.25 6.17 7.25 6.77 7.99 9.40 10.5 12.4 14.7 15.0 17.7 20.9 17.5 20.8 24.6 26.1 30.5 35.3 30.3 35.4 42.1

Model 150 2.17 2.64 3.19 3.28 3.97 4.89 4.72 5.81 7.06 6.39 7.89 9.62 8.2 10.2 12.4 12.8 15.9 19.4 18.1 22.6 27.8 21.2 26.5 32.6 29.3 35.2 42.1 34.0 41.1 49.2

Table A.2 Metso Minerals spiral classifier raking capacities Spiral Tank Spiral Capacity, diameter, m3k/rpm type design Type mm 0.34 S SP Simplex 600 M DP 0.68 (24 in.) F TP 1.02 750 S SP 0.58 Simplex (30 in.) M DP 1.16 F TP 1.74 Simplex 900 S SP 1.19 (36 in.) M DP 2.38 F TP 3.57 Simplex 1,000 S SP 1.64 (42 in.) M DP 3.27 F TP 4.92 1,200 S SP 2.96 Simplex M DP 5.92 (48 in.) F TP 8.88 Simplex 1,500 S SP 5.89 (60 in.) M DP 11.78 F TP 17.67 1,800 S SP 9.47 Simplex M DP 18.94 (72 in.) F TP 28.41 2,000 S SP 10.73 Simplex M DP 21.46 (78 in.) F TP 32.19 Duplex 1,800 S SP 18.95 M DP 37.90 (72 in.) F TP 56.85 2 1.46 Duplex 2,000 S SP M DP 42.92 (78 in.) F TP 64.38

Speed range, rpm 1 to 16

Metso Minerals Pumps and Process Colorado Springs, Colorado SP - Single Pitch DP - Double Pitch TP - Triple Pitch *Note: For specific gravity of mineral = 2.7 and assuming 100% raking efficiency

075

1

to 13 1 to 11 1 to 10 1 to 8 1 to 6 1 to 5 1 to 5 1 to

Motor ICW 1.1 to 2.2 1.1 to 2.2 1.5 to 3.7 1.5 to 5.5 2.2 to 7.5 5.5 to 15 7.5

to 22 7.5

to 22 7.5

to

5 1

22 7.5

to 5

22

to

Table A.3 Metso Minerals spiral classifier inside tank lengths and weights Model 100 Model 125 Model Empty Empty Spiral dia, Tank weight, Tank weight, Tank mm kg kg length, m length, m length, m 600 2.90 1,150 1,460 3.96 3.43 (24 in.) 3.43 1,400 3.96 1,650 4.50 Simplex 3.96 1,590 4.50 1,840 5.03 750 3.23 1,470 3.76 1,710 4.55 3.76 1,690 4.55 (30 in.) 2,060 5.33 Simplex 4.55 2,000 5.33 2,370 6.12 900 3.76 1,990 5.28 2,770 6.04 (36 in.) 4.52 2,350 6.04 3,130 6.81 Simplex 5.28 2,710 6.81 3,500 7.57 1,000 4.98 3,120 5.74 3,560 6.50 (42 in.) 5.74 3,520 6.50 3,970 7.26 Simplex 6.50 3,930 7.26 4,400 8.03 1,200 5.38 4,580 6.10 5,190 7.16 6.45 5,380 7.16 6,010 8.23 (48 in.) Simplex 7.16 5,910 8.23 6,460 9.30 1,500 5.77 7,960 7.09 9,920 8.41 7.09 10,500 8.41 11,200 9.73 (60 in.) Simplex 8.41 10,900 9.73 12,700 11.05 1,800 7.09 12,400 9.12 15,600 10.14 8.10 14,000 10.14 17,200 11.15 (72 in.) Simplex 9.12 l5;500 11.15 181800 12.17 2,000 8.13 14,400 10.16 17,600 11.18 (78 in.) 9.14 16,000 11.18 19,300 12.19 Simplex 10.16 17,600 12.19 20,800 13.21 1,800 10.14 (72 in.) 11.15 Duplex 12.17 2,000 11.18 (78 in.) 12.19 Duplex 13.21 Metso Minerals Pumps and Process Colorado Springs, Colorado Note that the duplex spiral classifier is available only in Model 150.

876

150 Empty Weight, kg 1,790 1,990 2,190 2,190 2,560 2,890 3,350 3,920 4,120 4,260 4,690 5,250 6,070 6,9 10 7,760 11,600 13,300 15,000 18,600 20,300 22;OOO 20,800 22,600 24,400 33,200 36,300 39,400 37,200 40,500 43,700

Table A.4 McLanahan Single Screw Washer capacities Maximum water, m3/h Screw dia. & Solids 150 106 75 length, cap., micron micron micron mmxm tih product product product 5 0 0 ~ 7 . 6 2 32 114 75 42 (20 in. x 25 ft) 6 0 0 ~ 7 . 6 2 45 148 97 55 (24 in. x 25 ft) 7 5 0 ~ 7 . 6 2 73 179 117 66 (30 in. x 25 ft) 9 0 0 ~ 7 . 6 2 95 207 136 76 (36 in. x 25 ft) 9 0 0 ~ 8 . 2 3 95 207 136 76 (36 in. x 27 ft) 1,100 x 10.06 159 402 263 148 (44 in. x 33 ft) 1,350 x 10.67 227 459 301 169 (54 in. x 35 ft) 1,650 x 10.97 363 516 338 190 (66 in. x 36 ft) 1,800 x 11.58 431 604 395 223 (72 in. x 38 ft)

Nom. screw rpm 38

Motor kW (1,500 rpm) 3.7

Weight with motor, kg 2,630

32

7.5

3,170

26

11

4,060

21

11

4,420

21

11

4,840

17

18.5

8,410

14

30

11,800

11

45

19,500

10

55

22,700

Nom screw rpm 21

Motor kW (1,500 rpm) 2-1 1

Weight with motors, kg 7,690

21

2-1 1

8,430

17

2-18.5

15,100

14

2-30

23,100

11

2-45

33,800

10

2-55

43,900

McLanahan Corporation Hollidaysburg, Pennsylvania Table A.5 McLanahan Double Screw Washer capacities Maximum water, m3/h Screw dia. & Solids 150 106 75 length, cap., micron micron micron mmxm till product product product 900 x 7.62 190 3 12 204 115 (36 in. x 25 ft) 900 x 8.23 190 3 12 204 115 (36 in. x 27 ft) 1,100 x 10.06 317 597 391 220 (44 in. x 33 ft) 1,350 x 10.67 454 704 46 1 259 (54 in. x 35 ft) 1,650 x 10.97 726 780 51 1 287 (66 in. x 36 ft) 1,800x 11.58 862 915 599 337 McLanahan Corporation Hollidaysburg, Pennsylvania The capacities indicated in Tables A.4 and A S are based on minus 8.5 mm (3/8 in.) x 0 solids weighing 1.60 kg/m3. For finer products, the screw speed is reduced to allow proper dewatering. At slower speeds, capacity rates are reduced accordingly.

877

Appendix B Terminal Settling Velocity Calculations The terminal settling velocity V, for particles of diameter d mm in water with kinematic viscosity v is given by (Menne 2001): V,

= 8.39

{[l + ( A 0 d3 ( S - 1)]0‘5-l}(v/d)

(Eq. B.l)

Where V, = settling rate, m m l s Al is the solids-correctedparticle Archimedes coefficient for the system: AZ

=

{[0.0103 (1 - C,)4.66g]/v2}

(Eq. B.2)

g = gravitational constant, 9,810 mmls2 v = (nu) kinematic viscosity of water, about 1 m 2 / s C, = solids volume fraction or percent solids by volume, and S = ratio of solids to liquid density (equals solids specific gravity if liquid is water)

Using the appropriate units, this settling velocity correlation applies to any fluid. Note that temperature has a very significant effect on water viscosity. Kinematic viscosity of water at various temperatures is given below. Kinematic viscosity of water, Y Temperature,OC 0 10 Viscosity,mm2/s 1.8 1.3

30 0.8

20 1.0

40 0.66

50 0.55

60 0.47

70 0.41

80 0.36

90 0.32

100 0.28

For a “standard” condition of water at 20°C, Equation B.2 can be substituted into Equation B. 1, then simplified and rewritten as follows:

V,=8.39 {[l +0.0103 (1 - C,)4.66gd(S-1)]0‘5-l}/d

(Eq. B.3)

As an example, the terminal settling rate of a 212 micron (0.212 mm or 70 U.S. mesh) solid particle with a specific gravity of 2.7 in a slurry of 20% solids by volume at a temperature of 20°C is : V,

= 8.39

{[l

+ 0.0103 (1 - 0.20)4.66(9,810) (0.212)3 (2.7 - 1)]O.’ - 1}/(0.212) V, = 10.15 mmls

Appendix C Overflow Capacity Derivation Velocity of flow from the feed entry to the weir is calculated from flow volume divided by overflow area: V,

=

1,000 Q/60 W h = 16.667 Q/W h

(Eq. C.l)

Where V, is flow velocity, mlmin Q = theoretical overflow, m3/h W = effective weir length, m h = overflow weir crest, mm 1,000 = conversion from m to mm, and 60 = conversion from h to min

878

The retention time of the particle is then: t = ElV,

(Eq. C.2)

Where t = the retention time, min E = the distance along the classifier pool between the feed opening and the weir, m Experience has proven that if the particle settles a distance of 2 h mm before it gets to the overflow weir, it will not overflow; if the particle hasn’t settled a distance of 2 h, it will overflow. So then, 2 h = V, t = V, (60 EIV,) 2 h = V, (60 ql(16.667 QlWh) = 3.6 V, E WhlQ

(Eq. C.3)

Solving for Q, we get Q = 1.8 V,E W

(Eq. C.4)

For most cases we know the desired flow rate and settling rate, so we solve for classifier dimensions E W:

E W = Ql(l.8 V,)

(Eq. C.5)

REFERENCES Hill, R.B. 1982. Selection and Sizing of Gravity Classifiers. Design and Installation of Comminution Circuits, ed. A.L. Mular and G.V. Jergensen, 11, Chapter 33. New York SMEAIME. Menne, D. Slurry and Free Solids Discharge Calculations [online]. Perth (Western Australia, Australia). Date of publication unknown [cited 13 February 20011. Article 6. Particle Hindered Settling Velocities [ d s ] : [and Free Settling Velocities if C, -> 01. Available from the Internet: http://members.iinet.net.au/-menne/sluny.htm Riethmann, R.E., and B.M. Bunnell. 1980. Application and Selection of Spiral Classifiers. Mineral Processing Plant Design, ed. A.L. Mular and R.B. Bhappu, 2nd ed., Chapter 16. New York: SME-AIME.

879

Hydrocyclone Selection for Plant Design Timothy J Olson and Patrick A . Turner

ABSTRACT Hydrocyclones are used in many and various duties in mineral processing flowsheets. There is a wide range of sizes, styles and fittings to select from, however, and the focus of this paper is to provide a basis to specify a hydrocyclone for a given application. A general description of how a hydrocyclone works is included to provide background to the discussion of process and hydrocyclone geometry variables. The mechanism for selecting a hydrocyclone for an application includes the use of the corrected D50 as the key separation parameter. Important manifold design options for new projects and hydrocyclone maintenance and materials considerations are identified. Included for reference are typical mineral processing hydrocyclone applications. INTRODUCTION “It speaks highly of the versatility of the hydrocyclone that notwithstanding our lack of knowledge of its basic principles, it has proved satisfactory in so many varied applications” (Bradley, 1965). The hydrocyclone is used in various applications in many industries, from degritting sewage sludge to removing oil droplets from produced water. The governing principles are difficult to quantify because of the complexity of the fluid dynamics with multiple phases in highly swirling flows. The majority of applications are in the processing of mineral ores however, and experience has helped develop a basis for predicting the hydrocyclone classification performance in these duties. The factors that affect performance, both process and hydrocyclone design, will be covered in this paper. The focus will be on providing information that an engineer who is designing a hydrocylone system will find useful. GENERAL DESCRIPTION A cutaway of a hydrocyclone is shown in Figure 1. The slurry enters the area of the hydrocyclone called the inlet head from the inlet feed pipe. The slurry is introduced next to the wall of the

Figure 2 Hydrocyclone, Axial Velocity Profile

880

cylindrical inlet, which induces a swirling action. This action helps develop the inertial forces that enable the classification of particles within the hydrocyclone. The slurry is further accelerated in the conical sections of the separator. The swirling action produces a low-pressure vortex in the center of the hydrocyclone where the finer, lower-mass particles migrate. The relatively light particles are removed with the overflow stream by an upward swirling flow through the vortex finder. The heavier particles are removed with an underflow stream by a downward swirling flow through the a ex region of the hydrocyclone classifier.

Figure 4 Hydrocyclone, Pressure Distribution

Figure 3 Hydrocyclone, Tangential Velocity Distribution

Figures 2 and 3 show the mean axial and tangential components of the velocity at different crosssections in the upper portion of a 250-mm diameter hydrocyclone (Petty et al., 2002). These single-phase numerical calculations were developed using the Reynolds averaged Navier-Stokes (RANS) equation, and standard transport equations for the Reynolds stress (RSM model) and the turbulence dissipation. The simulation imposes a backpressure on the overflow and underflow streams to avoid the air core. The Reynolds number based on the effective diameter of the feed entry and the volumetric flow rate of the feed stream is about 200,000. Figure 4 shows the pressure distribution predicted by the simulation. The results, which are qualitatively similar to experiments by Kelsall (1952) and to multiphase flow calculations reported by Devulapalli and Rajamani (1994), predict a Rankine vortex flow with a maximum tangential velocity near the radius of the vortex finder (see Figure 3). This feature distinguishes hydrocyclone flows from other swirling flows encountered in centrifugal separators. As illustrated by Figure 4, the swirling action of the flow field causes a lower pressure to develop in the core of the hydrocyclone. It is noteworthy that the Computational Fluid Dynamics (CFD) simulation captures the important qualitative flow features of a hydrocyclone classifier illustrated by Figure 1. SIZING AND SELECTION OF A HYDROCYCLONE The factors involved in sizing a hydrocyclone for a specific application also provide information on the important variables related to how a hydrocyclone works. To select the appropriate hydrocyclone, the engineer must know the solids concentration and size distribution plus particle and liquid specific gravities along with the solids tonnage and slurry flowrate.

88 1

Cyclone Diameter VS. D50 (For “Typica I” Cyc l a n e s )

D

M)

Parue* Dianarr (mcrbns)

1

2

3

4

5 678910

15

2 0 2 5 3 0 40M60

Cyclone Diameter (Inches)

Figure 5 Recovery Curves

Figure 6 D50 For Typical Hydrocyclones

It would also be helpful to have information on the liquid and slurry viscosity and particle shape. For design of a new plant, this information is often estimated, based on experience with similar applications. In many requirements, an estimate of the feed characterization is known. In these instances the selection is completed by matching the estimated performance of a specific hydrocyclone and the required separation. Hydrocyclone performance is often evaluated using a graph of particle size versus percentage of the particles recovered to the underflow. An example of a recovery curve is shown in Figure 5. The actual recovery curve shown does not reach zero, which is typical. It has been shown that this offset is due to particle entrainment caused by the watersplit to the hydrocyclone underflow (Lynch and Rao 1975). This curve is corrected by assuming entrainment in every size in proportion to the feed concentration. After the curve has been adjusted for the water-split, this Corrected Recovery Curve can be plotted as shown in Figure 5. The characteristics of this curve are often used to describe the hydrocyclone performance, most notably the D50. The D50 is the particle diameter with a 50% recovery on the corrected recovery curve. The D50 is shown on the corrected recovery curve in Figure 5. The slope of this curve (alpha) is a measure of the sharpness of separation, which is also an important parameter. Hydrocyclones come in a variety of sizes or diameters. The larger the hydrocyclone diameter, the coarser the separation. Each size hydrocyclone has a base DSousing standard operating conditions and a “typical” geometry (Arterburn1976). The D50 (base) shown in Figure 6 is valid with the following conditions: 1. 2. 3. 4. 5.

Feed Liquid - Water at 20 degrees C (viscosity 1 cp) Feed Solids - 2.65 SG spheres Feed Concentration - < 1% solids (wt%) Pressure Drop - 10 psi Hydrocyclone Geometry - Standardized hydrocyclone with; vortex finder 30% of hydrocyclone diameter, feed orifice 7% of feed chamber area, 20-degree cone for larger hydrocyclones, cylinder section included, vertical mount.

882

The actual D5o in a process is determined by adjusting the base Dsousing a set of correction factors related to process and equipment variables.

where D50(base)is the base separation for the specified hydrocyclone diameter (Figure 6 ) and CP are correction factors for process variables and CD are corrections factors for hydrocyclone design variables Process variables include feed % solids, particle specific gravity, feed pressure, slurry viscosity, among others. Hydrocyclone design variables, within a given hydrocyclone diameter, include vortex finder size, inlet area, length of cylinder, cone angle, and mounting angle.

PROCESS VARIABLES Feed Concentration The most dominant process variable affecting hydrocyclone performance is the concentration of solids in the feed. This is also the most important variable because the operator can normally vary this with water addition or other means. Most of the other important variables cannot be easily changed. In grinding circuits, it is common to monitor and adjust hydrocyclone overflow size distribution with changes in the feed rate to the mill or water addition depending on system or pumping limitations. The correction for this variable is shown in Figure 7 for three separate materials with different size and viscosity characteristics.The first line represents a tailings application with a very fine feed and a high concentration of clays. The curve on the right represents a closed circuit grinding application producing a coarse product. Based on the solids concentration, the hydrocyclone feed has a high amount of very coarse, well de-slimed solids at a relatively low slurry viscosity. The curve in the middle is appropriate for most applications with broad size ranges.

I

Correction For Feed Concentration

I

0

10

20

30

40

50

60

70

~

I

Feed Concentration (%Solids, Volume)

~

Characteristics on Feed Solids Correction Factor

Although most mining applications are generally known and the proper adjustments can be made in the selection formulas, in some cases it is recommended that a sample be tested. This is often critical for proper hydrocyclone selection. In a test loop if all the other process and hydrocyclone variables are known, the appropriate correction for solids concentration can be determined.

Feed Pressure In many applications this parameter is the only other variable that can easily be adjusted. This is normally done by opening or closing feed valves to change the number of operating hydrocyclones for a given flow rate. In some cases, variable speed pumps can also effect pressure changes. Because changing the feed pressure does not make a major change in the hydrocyclone

883

performance the hydrocyclone pressure is often controlled over a broad range. For about a 50% increase in pressure drop the expected change in D50 would be about 10%. However, pressure does have a large effect on component wear rates within a hydrocyclone. The effect of pressure is as follows: CP(pressure)= 3.27 x P4.'' Where P = Pressure drop, kPa An estimate of the pressure drop across a hydrocyclone can be determined by referring to manufacturer capacity curves which will show the hydrocyclone flow-rate as a function of inlet pressure for each vortex finder size for a selected hydrocyclone size. These capacity curves are for water, and a correction is required for slurries. Slurry Viscosity In practice the slurry viscosity is rarely known at the time of selecting the proper hydrocyclone. The slurry viscosity is directly related to the solids density by volume and to the total surface area of the solids. Thus relationships are developed to correct the D50 point for slurry viscosity by relating a specific application or size distribution of the feed solids with the slurry density. An example is shown in Figure 7. The effect of slurry viscosity is approximated by a combination of the feed solids concentration correction, discussed above, and the liquid viscosity in most hydrocyclone sizing simulations. This is because of the difficulty of measuring the slurry viscosity with coarse solids. Some work has been performed to measure this variable with special devices to keep the slurry moving (Kawatra 1996). Another approach is a methodology that allows a traditional viscosity measurement of only the fine segment of a given feed because this is the source of the material that will have a significant effect on the viscosity. In addition, with applications that have a significant amount of material that exhibits non-Newtonian flow characteristics, the apparent viscosity will vary with shear rate. Temperature also has an important but often overlooked effect on the fluid viscosity. This often can explain seasonal variations in hydrocyclone performance. For example, with a plant water temperature variation of 10 degrees from 10-20 degrees F, the fluid viscosity will change by 30% from 1.3 to 1.O. Specific Gravity Because hydrocyclones are classification devices that separate particles based on mass, the particle and liquid specific gravities are important variables. This has long been one of the important features of hydrocyclones in a grinding circuit because of the preferential way the heavier metalbearing mineral particles report to the hydrocyclone underflow and are ground to a relatively finer size than the lighter gangue minerals. Figure 8 shows the hydrocyclone recovery curves for the lighter gangue particles and gold particles. Notice the D50 for the gold particle is 57 microns with a 160-micron D50 for the bulk solids. Corrected Recovery Curve Gold Velaus Tola1 SollQ Rsrmsry

FEE0 CONDlllONS 1 0 4 4 % Solido. 36.41% -Yo Mesh. 18 PSI

100

g

90

:

80

U

5 s

10

f f

50

Ip

40

g

30

L

u

2o 10

0 10

1DO

Partlclr Slzr, mlcrons

loOD

Figure 8 Gold Versus Total Solids Recovery in a Grinding Circuit

The specific gravity of the fluid also is important in applications with high concentrations of dissolved salts like potash where processing is done in a concentrated brine. Also, in some locations, the clarified water will have a build up of salts causing an increase in the liquid specific gravity as well as the viscosity. CP(specific gravity) = (1.65/(solids sg-liquid sg))O.j DESIGN VARIABLES, HYDROCYCLONE GEOMETRY Hydrocyclone Inlet Design Hydrocyclones designed prior to 1950 featured outer wall tangential feed entry and 12- 15 mm thick rubber liners. This design was not adequate for tine separations or for abrasive slurry applications. Most hydrocyclone manufacturers have redesigned their inlets to include some form of involute, ramped or scrolled feed style and all of these provide a measured advantage in hydrocyclone performance compared to earlier tangential designs. Figure 9 illustrates the various types of hydrocyclone feed entries. The inlet opening or cross-sectional area of the orifice feeding into the cylindrical section of the inlet has an effect on capacity as well as D50, and most hydrocyclone models have several options to increase or decrease this area based on the desired flowrates and cutpoint. In general, the larger inlet area, the higher the hydrocyclone capacity and the larger the predicted D50.

4 14 OUTER

WALL TANGENTIAL

1 4 INVOLUTE

4 ICl R4YPED ENTRY OR

PI INVOLUTE W P

4

SCROLLED EVOLUTE

Figure 10 Hydrocyclone Cylinder Length

Figure 9 Hydrocyclone Inlet Styles

Cylinder Section Typically hydrocyclones have a cylinder section length equal to the hydrocyclone diameter. This can be a separate section or integral to the inlet head. Figure 10 illustrates a hydrocyclone without a cylinder section plus hydrocyclones with a single and double cylinder. While the longer cylinder section provided greater residence time and thus more capacity, it also reduces the tangential velocity. The added cylinder length results in minimal improvement in hydrocyclone separation and will increase hydrocyclone capacity at the same pressure by 8-10%. Larger 660-840mm diameter hydrocyclones typically have shorter cylinder sections. Cone Section Figure 11 illustrates the different hydrocyclone cone angles that are used in different applications. The 20-degree cones have been a standard in the minerals industry. The flat bottom hydrocyclone has been used to make very coarse separations with characteristic D50’s 2-3 times that of the standard hydrocyclone. The longer 10-degree cone will produce a finer and sharper separation at a

885

higher unit capacity compared with the 20-degree hydrocyclone. Use of this longer cone angle can change the predicted D50 by 15-20%. A hydrocyclone that has multiple cone angles is also shown in Figure 11. Apex Capacity

SI

Figure 11 Various Hydrocyclone Cone Angles

..

3 .

.*

I, ,I I.*.

I

1

1

Apex Diameter (inch’es)

5

*

,,,n

Figure 12 Apex Capacity

Vortex Finder A range of vortex finder (diameter) sizes are normally available for each model hydrocyclone.

The vortex finder sizes can range in sizes from 20-45% of the hydrocyclone diameter. Larger vortex finders will increase the hydrocyclone capacity but provide a relatively coarser separation. Smaller vortex finders will have the reverse affect. The correction for vortex finder size is: CD(vortex finder) = (Dv/(0.3 x Dc)’.~ Apex Design The apex angle and design also have a large effect on performance. Proper selection of both the apex opening and the angle will allow a high underflow % solids as well as maintain the intended hydrocyclone separation. This design incorporates an optimal apex angle in combination with a straight section to maintain the finest possible separation with maximum underflow solids concentration. The proper apex size must be selected to insure the maximum underflow density and limit the fines that the additional water will entrain and carry to the underflow. Some hydrocyclone sizing methodologies have used apex opening as a factor in determining the expected hydrocyclone performance. Others maintain that apex size is a matter of determining the correct apex size to handle the estimated solids flow at the highest density. However, closer inspection shows that changing apex opening, in most cases, also means changing the apex angle and, in that context, the apex change does have an effect on the expected D50 produced. Figure 12 illustrates the capacity of different size apexes. The selection of apex or spigot size is performed after the basic hydrocyclone for the application and the expected material balance has been performed. The apex is then selected by knowing the expected tonnage and flow that will report to the underflow and selecting the correct opening from the attached chart. For existing installations, the discharge pattern of an apex can provide information about the required size of apex opening. A wide spray pattern is indicative of a dilute underflow and an apex that is too large for the application. If an apex is too small, the underflow will “rope” with a very tight narrow discharge. Sampling the underflow to determine % solids for different openings is recommended. Underflow solids concentrations of 50% by volume are normally a good target.

886

Mounting Angle Hydrocyclones can be mounted at angles ranging from vertical to nearly horizontal. The effect of lowering the mounting angle will increase the expected D50 for a given hydrocyclone by 20-40% depending on the angle. This has been a popular tool to increase concentrator tonnage by effectively producing a coarser product. However, mounting at angles less than 45 degrees from horizontal have resulted in some maintenance problems, most notably with inlet head wear life.

Vertical Cyclones vs. Horizontal Cyclones

Figure 13 Mounting Angle

Most large mill circuits have 660 - 840 mm diameter hydrocyclones. These large diameter hydrocyclones are normally 2,500 to 3,000 mm tall. These tall hydrocyclones provide a substantial head on the underflow discharge of the hydrocyclone. In order to maintain a high underflow density, the apex size must be closely monitored and maintained. Installing the hydrocyclones at 45 degrees from horizontal greatly reduces the head on the underflow discharge. A consistently high underflow density can be achieved because apex diameter is not as critical as in vertical installations. In addition, the lower head results in reduced velocity of the slurry spraying out of the hydrocyclone. This increases the component life of the apex by about 100% compared to vertically mounted hydrocyclones. Approximately 50% of the hydrocyclones installed in the past 10 years on large SAG circuits have been installed at 45 degrees from horizontal. The reason for this trend is to maintain a higher average underflow density compared to vertically mounted hydrocyclones. This also allows the operator greater flexibility to change tonnage or density because the oversized apex will not easily plug and misplace coarse material to the hydrocyclone overflow. Drawbacks include longer overflow pipes, shorter inlet headliner life, and difficulty accessing the lower part of the hydrocyclone. For hydrocyclones installed at 45 degrees, most operations must completely remove the hydrocyclone in order to work on the apex or lower cone. SAMPLE CALCULATION The discussion above is intended to serve as a guide to the relative importance of hydrocyclone variables on performance. To select a hydrocyclone for a specific application there are several simulation programs that use similar methods to calculate the expected performance of a given size hydrocyclone in a selected application. To complete our general example: Where: D50(actual)=D50(base)X CP1 X CP2 x CP3 ... x CD1 x CD2 x CD3 etc ... If we start with the following hydrocyclone: 250mm hvdrocvclone with Standard Configuration: 20 degree cone angle, 250 mm cylinder, 75mm vortex finder, vertical mount, 123 mm2 inlet area, (CD1, CD2, CD3, etc., all 1.0 for standard hydrocyclone)

DSo(base)

= 24

microns

887

At the following feed conditions for a silica sand classification process: 20% Feed Concentration, Solids Liquid SG 2.65 and 1.O, 69 KPA Pressure, 20 Degrees C The CP for feed solids is 2.0 from the middle curve and the others calculate to 1 for the standard specific gravity water viscosity and pressure. D50(actual) = 24 microns x 1 .O (all CD factors) x 2.0 (solids factor) x 1.O (all other CP factors) microns

= 48

TYPICAL MINERAL PROCESSING HYDROCYCLONE APPLICATIONS Closed Circuit Grinding One of the most prevalent hydrocyclone applications in a concentrator is to classify grinding mill discharge. This can be from a SAG, primary, secondary, regrind, or other auxiliary mill. Depending on the application and mineral liberation of the ore, the hydrocyclone will typically achieve an overflow product size ranging from a P80 of 300 micron to a P95 of 25 microns in closed circuit grinding duties. Typical hydrocyclone performance is shown for the large 840mm hydrocyclones in a coarse copper circuit and 5 lOmm hydrocyclone data from a gold circuit producing a P80 of about 75 microns in the following table. Table 1 also shows hydrocyclone data from primary grinding circuits in two different iron ore installations. The first example is from a concentrator producing a very fine product and, for contrast, the second is a much coarser grind in another location. In most plant design situations where a grinding circuit is involved, the hydrocyclone feed conditions are not known, and the selection is based only on specified conditions for the overflow product. Hydrocyclone selection requires an experienced-based correlation that links the required overflow size with the separation size required In grinding circuit applications, the separation achieved by the hydrocyclones is normally defined by a percent passing point in the hydrocyclone overflow rather than the D50 point or "mesh of separation". The most common definition of the separation is the "PSO" point or particle size that is 80% finer in the overflow. The PSO is different from D50 because the P80 point is dependent upon both the D50 separations achieved by the hydrocyclone and the size distribution of the hydrocyclone feed. For example, if the hydrocyclone feed size distribution is 50% +lo00 microns and 50% -1 micron, the P80 achieved by the hydrocyclone will be less than 1 micron regardless of the size of hydrocyclone installed. Thus, it is not a simple matter to go from D50 separation to the P80 of the hydrocyclone overflow. In some applications, this relationship is one to one, but it varies dramatically from application to application. The D50 separation is normally not an important data point for the hydrocyclone operator. They normally will want to know the "mesh of separation" or the particle size that has a 95-99% (D95 - D99) chance of reporting to the underflow. This is often called the cut or separation achieved by the hydrocyclone and is typically two times the D50 point. The recovery curve shown in Figure 5 illustrates the relationship of the various recovery points versus the D50 point. In some instances, the hydrocyclone and the mill are modeled as a circuit and the complete system performance is evaluated as a function of hydrocyclone selection variables. This method is often helpful in an existing circuit where most parameters are known. Conditions in a grinding circuit are not constant, and the hydrocyclone feed solids concentration and particle size distribution change continuously. The changing parameters and the self-regulating nature of the hydrocyclone in a grinding circuit tend to mask many of the effects of hydrocyclone changes.

888

Table 1 Primary Grinding Circuit, Typical Hydrocyclone Performance ron Ore :oarse

,on Ore CoDDer

Gold

~

Over- Under-

Cyclone Product

Mesh

Feed Cum Wt% Finer

Microns 4800 10 1700 20 850 28 600 35 417 48 300 65 212 100 150 150 106 200 75 270 53 325 45 400 37 4

% Solids Circulating Load Pressure, PSI Cvclone Diameter

flow

flow

Cum Cum Wt% Wt% Finer Finer

93 75 65 54 43 32 24 17 13 10 8

100 100

7

25

99 98 93 82 70 57 46 37 30

90 73 59 46 35 23 15

10 7 5 4 3

65 36 81 430 I1 840mm 133")

Feed Cum Wt% Finer

Overflow Cum Wt% Finer

Cum Wt% Finer

100

100

100

100

97

100

97

88 82 75 53 50 34 24 18 16

100 100 100 100

86 78 68 35 32 17 12 8 7 6

99 97 95 92 90 87 84 81 78 72 68 62

15

99 93 81 68 63 60

Underflow

Over- Under'eed flow flow :urn Cum Cum Vt% Wt% Wt% 'iner Finer Finer

eed 'um Vt% iner

100 100 100 100 100 100 100 100 100 100 100

99 97

100

99 96 92 89 85 82 78 74 69 61 55 47

99 97 94 92 87 81 71 56 44 34 28 24

Over- Under flow

flow

Cum Cum Wt% Wt% Finer Finer 100 100 100 100 100

99 96 86 72 59 49 41

98 94 89 86 80 74 58 40 27 18 12 II

u 43 21 6' 305 16 510mm 120")

14 5 240 30 38Omm 115")

71

66 54 130 11 660mm 126")

8

It should be noted that the hydrocyclone is often limited in closed circuit applications by the capacity of the grinding mill. Changes made to affect the hydrocyclone classification will result in a change in the circulating load and the amount of material returning to the mill feed. A finer separation will result in a higher circulating load and a higher tonnage feeding the mill at a constant new feed rate. This will increase the feed YOsolids feeding the hydrocyclone and this will coarsen the separation. Therefore changes to the hydrocyclone must be evaluated in the context of the overall grinding system. Tailings Hydrocyclones Construction of tailings dams is a very important consideration in a hard rock mining project. Hydrocyclones are often used to produce a low cost source of sand from mill tailings. The required characteristics of this dam building material are that it must: 1. Provide suitable drainage characteristics to allow materials to consolidate adequately and display fill-strength. 2. Minimize the risk of failure by liquefaction under dynamic loading conditions. 3. Provide good percolation rates of water through the coarser solids. A different but very closely related application is for the production of paste back-fill in underground mines. The amount of available sand will depend on the grind, and as a result, tailings applications are much more typical in copper operations which have a comparatively coarse tailings. The traditional "Rule of Thumb" for high quality sand is a particle size less than 20% finer than 74 microns (Turner 1997). Recent experience has shown that this is not a reliable measurement in that it does not quantify the amount of silts and clays. The amount of minus 20 micron or the amount of 5 micron or finer material is a much better measure of the sand quality. Hydrocyclone size (diameter) can vary from 250-660mm for tailings applications. Hydrocyclones can either be mounted along the crest of the dam with the underflow discharging directly or on a hydrocyclone manifold at a central station where the underflow is pumped to the dam location. The hydrocyclones are often mounted horizontally in this application to decrease the

889

amount of water and fines that are carried out the apex. In some cases, a two-stage hydrocyclone system is required to recover an adequate amount of sand at an acceptable quality. This involves another set of hydrocyclones to process the first stage underflow. A low cost alternative to the two-stage process is to introduce an attachment to the lower portion of the hydrocyclone where fresh water is introduced to wash entrained slimes to the overflow. Because of the importance of achieving a good quality dam building material, hydrocyclone testing is recommended. The test-work can be performed on actual tailings samples collected at the concentrator and sent to a pilot plant. On large operations, predicting the amount sand that can be recovered at an acceptable quality often will determine the viability of a project. Dewatering and Desliming Dewatering and desliming applications are very common in non-metallic operations where the amount of very fine material, including clays, has a large impact on reagent usage as well as product grade and recoveries in the subsequent process. In some phosphate operations, for example, the ore is deslimed with hydrocyclones prior to feeding a hydrosizer. The products from the hydrosizer are also dewatered in hydrocyclones to increase the feed solid concentration for adequate conditioning prior to flotation. The rougher flotation products are dewatered again in hydrocyclones prior to conditioning for cleaner flotation. Similar applications are found in coal and industrial sand processing. In many dewatering applications, where hydrocyclone underflow density is important, special attachments can be provided to the hydrocyclone that control or restrict the flow of material out of the apex. This is done with a duck bill shaped attachment added below the apex and an elongated pipe on the hydrocyclone overflow, specially designed to create a siphon. A valve regulates this siphon, and the amount of the siphon effects the force required to open the duckbill valve. This design - is used in low solids applications with variable feed conditions, which makes proper apex sizing impossible. This is shown in Figure 14.

__

Figure 14 Apex Attachment, Controlled with Siphon

Ultra-fine Particle Separations In some applications, very fine particle separations are required. It is often necessary to remove ultra fines to enable the separation and recovery of minerals like cassiterite that occur in these very fine sizes. Typically 50mm (2”) diameter hydrocyclones are the preferred size for these fine high capacity separations. The design of the 50mm hydrocyclones has been refined to allow separations in the 5-10 micron D50 range. Smaller diameter 13-25mm hydrocyclones are available but because of the inherent lower capacities and smaller orifices of these very small hydrocyclones the operational problems are not insignificant with any amount of tramp oversize. Small diameter hydrocyclones are used in high flow-rate applications by incorporating them in a pod or group of hydrocyclones, with a common feed and overflow compartment and accessible individual underflows where any plugging of the apex can be seen and addressed. Many pods can be fed though a radial distributor as if they were individual hydrocyclones, to process these larger flows. An example of the installation of these is shown in Figure 15.

890

Figure 15 Canister or Pod Design, Small Hydrocyclones

Figure 16 Radial Manifold or Cyclopac, Large Hydrocyclones

Figure 17 Spider Manifold

MANIFOLD SYSTEMS The specified design for hydrocyclone manifold systems are also an important consideration. In most mineral processing plants, this consists of a radial feed distributor with isolation valves on the lines feeding the individual hydrocyclones (figure 16). The valves are normally a heavy-duty knife gate style that is either actuated manually with a hand wheel or automatically with a pneumatic cylinder. Most modem plants will have the capability to remotely open and close hydrocyclones based on changes in hydrocyclone feed pressures. Pressure transducers are also included on the feed header so that the hydrocyclone feed pressure can be monitored remotely. A diaphragm is recommended to keep the pressure gauge or transducer from plugging with solids. In coarse grinding circuits the feed distributor and the launder collecting the hydrocyclone underflow are lined with at least 12mm (l/2”) rubber. The wear is not as concentrated in the hydrocyclone overflow launder and the pipes feeding each hydrocyclone, and, as a result, these will be lined with 6mm (1/4”) rubber in most cases. In lighter duty applications, the rubber lining is reduced in the feed and underflow launders and eliminated in the feed pipes and overflow launder. An alternate material that has been used recently in hydrocyclone overflow piping is HDPE. This is more wear resistant than unlined steel pipe and also fairly inert to chemical attack. Another important consideration is the design of the feed distributor. It is important to have the feed evenly distributed to each of the operating hydrocyclones. It is in this area that system designers can be tempted into considering a less expensive in-line feed distribution design. This always leads to an unequal distribution in both the feed solids concentration as well as size distribution. This in turn leads to different hydrocyclone performance within the same bank, often requiring different apex sizes. The metallurgist’ task in hydrocyclone performance evaluation or optimization is made much more difficult as a result. A radial design feed distributor, with the feed coming from either the top or bottom, and into a header with multiple nozzles allowing the feed slurry to be distributed to each hydrocyclone feed line, is the preferred design. This is shown in figures 15-17 for different style manifolds. The header should have a removable top for maintenance. It is also a good idea to allow extra nozzles for possible future capacity or as a location to sample hydrocyclone feed. MATERIALS OF CONSTRUCTION AND MAINTENANCE CONSIDERATIONS Hydrocyclones used in the mining industry normally have steel or fiberglass housings with replaceable liners. In some less wear sensitive applications, molded urethane hydrocyclones are used. In metallurgical hydrocyclones the most prevalent lining is gum rubber. These liners are 89 1

normally available in 12mm (l/2”) or 25mm (1”) thickness. Because ofthe amount of coarse solids feeding a hydrocyclone in many of the mining applications the more wear resistant hydrocyclone design will include the thicker rubber liners in the upper portion of the hydrocyclone and utilize a combination of ceramic liners for the lower cones and apex. In large hydrocyclones (660 and 840 mm), the wear is highest in the apex followed by the lower cone and then the middle cone. The next highest wearing part is the inlet headliner. Thus, in order to keep all of the parts on the same cycle, premium ceramics are required in the apex and lower cone to bring the life of these parts up to the inlet headliner life. It is common commercial practice to use one type of ceramic in the apex, another type in the lower cone liner, and a third type in the middle cone liner (figure 17). The objective is to increase the life of the lower liners to be consistent with the inlet head liner life. The ideal situation is to have all the liners wear out at the same time. Hydrocyclone maintenance practice can also affect performance. In addition to worn apexes causing low underflow solids, worn cone liners can have grooves or edges that will cause misplaced coarse solids to the overflow. If a worn lower cone is replaced but the cone right above is also worn and not replaced, a negative edge may result which will adversely affect performance. In most cases at the plant design stage, because the actual hydrocyclone feed conditions and wear requirements are unknown, the hydrocyclones are provided with all rubber liners and rubber apexes with alternate sizes. After the optimum apex size has been determined at the commissioning stage a proper size ceramic apex is provided.

Figure 18 Hydrocyclone Cross-section

Figure 19 Various Hydrocyclone Liners and Materials

While the rubber and silicon carbide liners are the primary wear material used in closed circuit grinding applications many mining hydrocyclone installations use different wear materials. For example, many phosphate and iron ore installations utilize polyurethane liners in the hydrocyclones. In coal, hydrocyclones are either lined with an alumina or silicon carbide ceramic or polyurethane. Different hydrocyclone liner materials are illustrated in Figure 19.

892

ACKNOWLEDGMENTS We are grateful to Dr. Charles Petty, Professor Department of Chemical Engineering, Michigan State University and Dr. Steve Parks also of Michigan State University for their assistance and review of this work. REFERENCES Agar, G.E., and J. A. Herbst “The Effect of Fluid Viscosity on Cyclone Classification”, Transactions SME, December 1966. Arterburn, R.A., “The Sizing of Hydrocyclones”, Krebs Engineers, 1976. Arterburn, R.A., “The Sizing and Selection of Hydrocyclones”, Design and Installation of Comminution Circuits, Volume 1, Chapter 32, 597-607. Bradley,D., “The Hydrocyclone” ,Pergamon Press, 1965. Devulapalli,B. and R. K. Rajamani., “Hydrocyclone Modeling of Swirling Flow and Particle Classification in Large Scale Hydrocyclones,” KONA Powder and Particle Journal, No. 12, pp. 95-104,1994. Dorr, J.V.N., and A. Anable, “Fine Grinding and Classification”, Trans. AIME, Vol. 112, pp. 161177, 1934. Hill, L.N., “Installation of 0.84 M (33 IN) Cyclones on the Primary Grinding Circuit at Cyprus Sierrita Corporation”, SME Annual Meeting, February 14-17, 1994. Hochscheid, R.E., “Horizontal Cyclone in Closed-Circuit Grinding”, SME Annual Meeting, February 1985. Hukki, R.T., and K. Heisdanen, “Two-Stage Hydraulic Classification - A Report on an Industrial Application”, AIME Annual Meeting, Chicago, IL, February 1981. Kawatra, S.K., A.K. Bakshi and M.T. Rusesky, “ Effect of Viscosity on the Cut (D50) Size of Hydrocyclone Classifiers”, 1996, Elsevier, Great Britain, Minerals Engineering, Vol. 9, No. 8, pp 881-891. Kelsall, D.F. “A Study of the Motion of Solid Particles in a Hydraulic Cyclone”, Trans. Inst. Chem Eng., Volume 30, 1952,87-104. Kelsall, D.F. “A Further Study of the Hydraulic Cyclone”, Chem. Eng. Sci.. 2,1953,254-270. Lynch, A. J. and T.C. Rao, “Modeling and Scale-up of Hydrocyclone Classifiers”, 11’ International Mineral Processing Congress, 245-269, 1975. Mular, A. L. and N.A. Jull, “The Selection of Cyclone Classifiers, Pumps and Pump Boxes for Grinding Circuits, Mineral Processing Plant Design”, SME, Port City Press, Baltimore, MD, Chapter 17, pp. 376-403, 1978. Olson, T. J., “Hydrocyclone Design for Fine Separations at High Capacities”, Symposium on Recent Advances in Cyclones and Hydrocyclones, AIChE Annual Meeting, Nov 12-17,2000. Petty, C.A., S.M. Parks, and T.J. Olson, “Flow Simulations within Hydrocyclone Separators”, Symposium on Centrifugal Separation, Minerals Engineering Conference on Solid-Liquid Separation, June 18-20,2002, Falmouth, UK. Petty. C.A. and S.M. Parks, “The Influence of Hydrocyclone Geometry on Separation Performance”, Symposium on Particulates and Multiphase Flows, Annual AIChE Meeting, November 4-9,200 1, Reno, NV. Rodgers, R.S.C., A.M. Hukki, G.J. Steiner, and R.A. Arterburn, “An Evaluation of The Use of Two Versus One Stage of Hydrocyclones in a Pilot Scale Ball Mill Circuit”, Cyclone Symposium 1 10” Annual AIME Meeting, Chicago, IL, February 1981. Schlepp, D.D and P.A. Turner, “Influence of Circulating Load and Classification Efficiency on Mill Throughput”, SME Annual Meeting, February, 1990. Slack, M.D., R.O.Prasad, A. Bakker and F. Boysan, “ Advances in Cyclone Modeling using Unstructured Grids”, Fluent Europe LTD, 200 1. Turner, P.A.and M. E. Hoyack, “The use of Hydrocyclones in Gold Mills without Thickners: Design Considerations”, SME Annual Meeting, 1993, Reno NV. Turner, P.A., W. van Ommen, J. Zutman, “Application of Hydrocyclones for Producing Sand From Mill Tailings - Design and Operating Considerations”, XX International Mineral Processing Congress, Aachen, Germany, September 2 1-26 1997.

893

Coarse Screening Mark A. Bothwell’ and Andrew L. Mula? ABSTRACT Of the coarse screens employed by the mineral processing industry, vibrating screens are very common. A variety of screen media have been used including woven wirekloth, perforated screen plate and profile wire (rodshars). Typical screen media are made from high carbon steels, rubber and synthetics such as reinforced polyurethane (tyrethane). Typical screen performance measures, such as the efficiency of undersize removal from the oversize stream, depend upon design variables (e.g., screen area, open area, aperture size and shape, slope of screen deck and deck motion) and operating variables (e.g., particle size, shape and distribution, solids feed rate, bed depth and feed moisture content). Two traditional screen sizing methods are still in use, namely, the feed rate method (screen area proportional to the screen feed rate and the throughput method (screen area proportional to the flow rate of screen undersize in the feed). These are briefly reviewed. Advances in vibrating screen selection and sizing methodology are described for a SAG mill discharge screen example. The current approach to screen design and corresponding features of sizing, mechanical limitations, increased availability and media selection are discussed. INTRODUCTION Industrial screening is a size separation strategy where particulate material is separated on the basis of size and is one of the oldest of unit operations in world-wide use by industry. Size separation permits the transfer of oversize and undersize streams to suitable processing steps (e.g., SAG mill discharge oversize being recycled to a SAG mill and the undersize being transferred to a grinding circuit). Many screening devices are available in the marketplace with a variety of screening surfaces to choose from. In general, screens involve the passage of particles through screen apertures (openings) and when size separations are made at 4 mesh (4.75 mm) or larger, screening is arbitrarily “coarse” (Matthews, 1985). Screen feeds may be wet or dry. Typical applications of coarse screens are: to separate undersize from oversize fed to a primary crusher, to separate oversize from crusher discharge in crushing plants, to separate oversize from SAG mill discharge and to dewater screen oversize being transferred to a conveyor belt. Typical coarse screens (see Matthews, 1985) are grizzlies (stationary or moving), vibrating screens (horizontal or inclined including multi-slope screens), shaking screens (oscillating or reciprocating) and revolving screens (trammels and barrel screens). Commonly encountered in mineral processing are grizzlies, trommels and vibrating screens. The latter are discussed in some detail below. COARSE VIBRATING SCREENS Major components of vibrating screens are the screening surface, the vibrator assembly, the base frame, the support frame, the vibrating frame, the motoddrive assembly and the feed boddistributor. Auxiliary components may include feed chutes, dust enclosures, conveyor belts and dust collection systems. Figure 1 is a schematic of a typical 3-deck vibrating screen. Screening Media Matthews (1985) stresses that the most important element of any screen is the screening media surface, where stratification and separation take place. Screening media is typically the most expensive maintenance item on screens. Screening surfaces are placed into one of three general categories, namely, woven wire screedcloth, perforated screen plate and profile wirehar. These ~~

1 Product Support Manager, Screens, Svedala, Appleton, Wl, USA ’ProfessorEmen’tus, UBC, Vancouver, BC, Canada

894

various designations refer to the sub-components, which comprise the media panel. A wire mesh is a series of individual wires which are woven together to form a mesh panel. A perforated plate consists of a flat piece of plate, which has a series of holes torch cut, molded or punched through it. One of the biggest advances in screening media has occurred over the last 20 years. This has been the introduction of synthetic materials as a screen mesh. These media systems are molded or punched, and are similar in form to a perforated plate. They are typically made from reinforced polyurethane (tyrethane), natural rubber, or synthetic rubber. There are many sizes and types available that offer various advantages involving the methods in which they are fixed to the screen, as well as their material characteristics which make them advantageous for specific applications. Woven WirdCloth. This material has been widely used an has accounted for upwards of 70% of sales in the early eighties. With the introduction of synthetic media this percentage has dwindled to around 50% of sales. For severe applications and coarse sizes, steel or rubber plate is often employed. When finer sizing is desired, profile wire or synthetic membrane media are selected. A variety of materials have been used to manufacture wire mesh. These include alloy steels that are abrasion resistant, high carbon steels, and stainless steels.

Figure 1: Schematic of a 3-deck Vibrating Screen (Courtesy of METSO) A screen surface must withstand the stressesAoads applied to it and maintain a high degree of resistance to abrasion and corrosion. Once the opening size and capacity characteristics are arrived at and a screen is fully operational, the “best’ screen surface is one which never has to be replaced. Realistically, this means that the replacement cost per unit of throughput is the minimum. For example, carbon steel screen is consumed at a rate as measured by a replacement cost of C dollars/section/year plus cost of replacement. A synthetic material does not wear out as fast (perhaps lasting longer by a factor of 5 or more). It is more expensive to purchase, but is usually easier to replace due to snap-in components. The biggest downside to synthetic media is typically the loss in open area which results in less screening efficiency, or the requirement of larger screens, If C + labor cost for the synthetic is greater, then carbon steel will continue to be the material of choice due to the higher amount of open area.

895

Square mesh surfaces are often selected for coarse applications when accurate sizing is necessary andor particles are slabby. However, on an incline the effective square mesh aperture and capacity may be adversely influenced. On the other hand, rectangular mesh surfaces of comparable sizing will exhibit a higher capacity, because the proportion of open area is greater, and the probability of a particle falling through an elongated aperture is greater. Moreover, rectangular surfaces are not as susceptible to blinding (i.e., openings become plugged with near mesh particles or wet, sticky ones; the latter also cling to decrease the effective aperture) and are suited for needle-like particles, for high moisture ores and for ores with a high clay content. When blinding is a severe problem, special screening media and decks should be considered. Screening media with flexible characteristics like rubber or thin polyurethane membranes (see Figure 2) can work well in very severe applications.

In moderately sticky applications a flexible Z-Slot stainless steel wire or long slot wire with polyurethane alignment strips can work well while not sacrificing open area.

Figure 3: Stainless steel Hi-T and Z-Slot wire

The screen decks can also be modified to allow standard wire meshes to work in sticky applications. A heated deck is useful for fine ore of high moisture content. Ball decks rely on rubber balls bouncing against a screen bottom to loosen material. As a last resort, water sprays are recommended. The flow of feed material can be either parallel or perpendicular to the longer dimension. In parallel, a higher capacity is observed and blinding is reduced for high moisture andor clayey ores. In perpendicular flow, a screen blinds less for dry ore, a longer life is observed and the efficiency is higher. Perpendicular slots allow fewer elongated particles to pass to the undersize stream since elongated particles tend to align with the flow of material as they move along the length of the screen.

896

Perforated Screen Plate. These are employed for coarse separations and are useful as an upper deck screen to reduce wear and damage to a lower deck screen of smaller opening. Plates are more expensive, but they have exhibited high resistance to wear, long life, less blinding, higher efficiency and a high degree of accuracy in sizing when used with larger aperture sizes. For screen openings less than about % inch, percentage open area is less. Perforated plate is often covered with thick rubber to increase the wear life. Thicker media mounted at an incline is detrimental to effective aperture size. Profile Wire (RodsBars). These have been used for coarse and fine screening, for dewatering applications and for special screen assemblies (cone shapes, etc.). Wire in parallel with flow is most common, but transversely placed wire is effective for wet screening in finer size ranges. Profile wire surfaces are not widely used in crushing circuits. VIBRATING SCREEN APPLICATIONS Since it is not feasible or practical to perform laboratory tests on the capacity and efficiency of every screening application, a number of models for predicting screen performance have been devised over the years. The results from these various models can vary depending on the assumptions made, but they all tend to rely on two basic principles of screening. These include “stratification” and “probability of separation”. To better understand these principles, refer to Figure 4.

Stratification: process of filtering small rocks to the bottom of the material bed

Separation: process of small rocks falling through the apertures in the media

Figure 4: The difference between stratification and separation Figure 4 shows that stratification is the process of getting the undersize material to the screen surface, and the probability of separation is the process which takes place once the particles reach the screen surface. A given particle will often require multiple attempts before it finally misses a wire and hits an aperture with the proper alignment to fall through. All of the screening methods described in this section will define how one of these two principles is affected by the various screening factors.

897

Common Measures of Screen Performance The performance of a screen may be neglected in an operating environment, until there is an incentive to increase productivity. Changes in screen aperture size, feed rate (anaiogous to circulating load in many cases), feed size distribution, frequency and amplitude of vibration, and other factors will then be studied to determine their influence on throughput and operating costs. At the design stage, screen performance must likewise be considered, so that measures of screen performance become equally important. There are a variety of screen performance measures in use. Three common ones are Nominal Capacity, Efficiency of Undersize Removal From Oversize Stream and Efficiency of Undersize Recovery. These are defined with reference to Figure 5 which shows the variation in mass of undersize particles falling through an inclined vibrating screen along its length. In Figure 5, F is the stph of dry feed solids, 0 is the stph of the screen oversize stream, U is the stph of the screen undersize stream and f, and ox are, respectively, cumulative weight fraction finer than screen separation size in streams F and 0 respectively. Nominal Capacity. Nominal capacity is expressed as the tons of solids passed through the screen per square foot of screen area per hour per millimeter of screen aperture. Nominal capacity depends on the type of material being screened (e.g., coal versus crushed ore) and other factors. For a given material, the nominal capacity permits a rough comparison between various screens by accounting for the influence of differences in area and screen opening. Efficiency, E,, of Undersize Removal From Oversize Stream. For a screen operating at steady state, the oversize stream should contain almost all (>95%), particles coarser than the size at which the screen makes a separation (called cut size when measured with square mesh sieves) along with at least some particles that are finer. If the oversize stream is the important one, then some measure of how well a screen diverts fines from the oversize to the undersize stream is useful. Equation 1 shows that:

F, fx A-B: Feed enters. Vibration causes particles to stratify.

Feed

B: Maximum stratification. B-C: Maximum particle removal due to high % fines. C-D: Particle size to opening ratio close to 1; separation is by repeated trials.

Figure 5: Stratification and Separation of Particles on a Continuous Vibrating Screen

898

E,

=I*

F(1-f ) ]=(1-0,)

where ox is the cumulative weight fraction finer than cut size x in the oversize stream and E, is expressed as a fraction less than or equal to 1 (multiply by 100 to convert to percentage). Efficiency of Undersize Recovery, Ru-. When the undersize stream is of main interest, a measure of the effectiveness of the screen to transfer feed particles finer than the cut size to the undersize stream is the recovery of feed fines in the undersize stream. Thus:

where Ru. is expressed as a fraction less than or equal to 1 Note that E, and R,. are related by:

Screen Efficiency Versus Capacity. Screen efficiency for a given screen varies with capacity. This is illustrated in Figure 6, where Percent of Rated Capacity is plotted versus Percent Screening Efficiency, E,. The Rated Capacity is based on a probability function of the undersize particles falling through the apertures before they reach the end of the screen. It does not describe the bed depth as it relates to stratification. Note that the maximum efficiency is 95% and occurs at 80% of rated capacity. Below 70% capacity, the efficiency decreases rapidly, because the light load permits particles to bounce away from apertures. It should also be noted that the dashed line should be referenced when the Rated Capacity is below 70% and a bed of material to aperture ratio of 1:1 or higher is still maintained. These conditions that commonly occur with lower open area synthetic media, will actually allow efficiency to be increased above the 95% mark.

EFFICIENCY (OPERATING AT 80% RATED CAPACTPI), INCREASE CALCULATED AREA BY 20%.

899

The curve in Figure 6 was developed for screens operating at about 90% efficiency. At this efficiency, the capacity was defined as 100% of rated capacity. Thereafter, changes in feed rate produced efficiency variations as shown for a given screen. Remember that efficiency, h,is related to R,. by Equation (3). NOTE: When a new screen is being sized, some applications will call for an efficiency, E,, of 95%. To ensure of this, increase the calculated screen area by 20% if the Feed Rate Method (Equation (6)) is used. Influence of Traditional Variables on Performance Traditional variables which influence performance include design variables such as: Screen Area and Open Area, Aperture Size and Shape, Slope of Screen Deck, Deck Motion, and operating variables such as: Particle Size, Shape and Distribution Solids Feed Rate and Bed Depth Feed Moisture Content. Screen Area and Open Area. Other things being equal, the capacity of a vibrating screen varies directly with screen area. For a given area, the capacity is proportional to screen width, while the efficiency is proportional to screen length. An overall compromise seems to be a length that is about two or three times the width, since at some point an increase in length has minimal influence on efficiency. For a given screen area, the “best” capacity and efficiency occurs when the material at the discharge end is all oversize and one layer in depth. In practice, screens operate at lower efficiency to obtain higher capacity. The area of a screen is limited by the strength of the screen deck which must be able to handle heavy loads in motion. In turn, deck strength depends upon the proportion of the screen area that is open to particle passage. Open area can vary between 6 and 90 percent, depending upon the characteristics of the screen and projected usage. Capacity is directly proportional to open area, while efficiency is expected to increase likewise. Open area is chosen to minimize the possibility of screen ruptureldamage. Thus, when wire is employed, wire diameter should provide maximum open area consistent with the strength required for the application based on past experience. The possibility of blinding enters into it, since larger diameter wire is less flexible, and may induce moist or elongated particles to plug openings. The effective area of a screen may be less than the actual area, because space is necessary for tension rails, center clamps and blockage of openings by support bars. Unless manufacturer tables list effective areas, a safety factor of 10% should be sufficient. Some sizing methods already incorporate an adjustment for this loss in open area, and will require a relative open area to be used when sizing screens utilizing synthetic media that does not require support bars, center clamps or tension rails. Aperture Size and Shape. The capacity of a screen will decrease with a decrease in size of the opening. At a fixed capacity, screen efficiency likewise will decrease as aperture size decreases. If aperture size decreases, wire diameter tends to decrease so as to maintain open area percentage. When strength must be retained, wire diameters may not change severely so that open area must decrease and capacitylefficiency is reduced. Blinding becomes more of a problem as aperture size decreases, particularly with dry feeds that have an abundance of near size particles or with feeds of high moisture content, Dry screening below about 6 mesh is typically not a commercial success, because capacities are too low for mineral processing plants. There are however specialized screens and media that allow this to be done when moisture and clay content is very low (less than

900

1% moisture). In most cases that fall outside this range, wet screening is the only feasible alternative. The shape of the aperture, i.e., square or rectangular or round or slotted, will be important to screen performance. Rectangular or slotted openings offer more open area and less blinding for most ores, thereby increasing capacity and efficiency. However, square and round openings permit of a more accurate split at the cut size of interest. In general, apertures are staggered to prevent particles from riding on screen material too long before encountering an aperture. In most cases, industrial screens do not split particles in accord with their aperture size, e.g., a screen with 2-inch square openings will not split at a cut size of 2 inches as measured with laboratory sieves. This is due to factors such as aperture shape, deck slope, bed fluidization and rate of travel, particle shape and distribution and screen motion. For square openings, the aperture size in inches is X= .0238 + 1.155X,, where X, is the desired cut size in inches (after Matthews, 1985). Relationships between X and X, for other than square openings are often available in tables and/or graphs. Slope of Screen Deck. When the discharge end of a screen deck is inclined downwardly from the horizontal, material cascades more rapidly down the slope and either passes through an opening or over the screen surface in accord with some probability. Hence the capacity of a screen must increase as deck slope increases. Efficiency will increase up to a critical slope. Thereafter, it decreases rapidly. Coincidently, an increase in deck slope will decrease the effective aperture size at a given feed rate. In crushing plants, screen decks will be installed at angles of 0" to 30" from the horizontal with 20" being most common in stationary crushing plants and 0" in portable crushing operations. Multi-slope screens that vary the slope along the length of the screen from 30" to 0" are a relatively new concept in the industry. They allow increases in capacity for a given size of screen of up to 70%. These screens work on the concept of quick stratification on the feed end due to the high angle of inclination and high G-Force, and better separation of near size at the discharge end to lower travel rates. Rates of material travel on circle throw screens with counter flow rotation are approximately proportional to inclination. This at 20°, flow rate is 80 ft/min, while at 22' it is 100 fdmin. Horizontal screens can utilize either a straight-line motion, or elliptical motion, which is aligned at 45' to horizontal. A screen at 0' with straight line motion will typically have a travel rate of 50 fdmin, versus 65 ft/min for an elliptical throw screen. This characteristic allows an elliptical throw screen to achieve higher capacity while still achieving a good efficiency. The downside of an elliptical throw screen is the added cost and maintenance of the more complicated elliptical throw mechanism. Caution must be exercised, because flow rates vary with speed, throw characteristics, type of screen surface and aperture size. Deck Motion. Vibration of screen decks is produced either by circular or straight-line or elliptical motion, where the vibrator rotates in a flow or counter flow direction at amplitudes of 3 to 20 mm and shaft speeds of 700 to 1200 rpm. Frequencies of 700 to 1000 cycles/min are normal, while high speed devices exhibit frequencies of up to 3600 cycledmin. Trommels do not vibrate-they rotate, whereas sifter screens can exhibit vibratory and rotary motion combined. In mineral processing, coarse vibrating screens are popular, where the vibration lifts the material to cause stratification of particles and conveys particles on an incline or at horizontal. Speed, throw, slope and direction of rotation affect blinding of screens by near-size particles. The throw has the greatest influence on blinding, however. Too small a throw will permit near-size particles to plug openings; too large a throw will keep them away from the screen surface and reduce the efficiency. Large throws at high speeds, reduce bearing life which varies inversely with the 10/3 power of the bearing load. The combination of throw amplitude and speed, can be used to measure the G-Force of a screen, as represented in the following formula:

G-Force = (RPM)' x (Throw in Inches) / 70,400

901

(4)

The G-Force of a screen is a measure of energy at which a screen is operating. The G-Force should be higher on heavily loaded screens, or when screening sticky material. Horizontal screens also tend to run at a higher G-Force as they do not have gravity helping move the material along the deck as on an incline screen. Rotation in the flow direction increases capacity by increasing the rate of flow of material, while efficiency may be reduced. Counter flow rotation (reversing the motor) will tend to retard the rate of flow on the screen, which increases the efficiency but decreases the capacity. Counter flow rotation may create blinding, although the effect may be compensated for by increasing deck slope. Speed (frequency of vibration) produces a lifting component for stratification and a conveying component in which the screen pulls back at the end of a cycle. High speeds go with small throws, while low speeds go with large throws. This compromise reduces bearing wear as well. Generally, large throws are required for screening coarse material or when bed load is substantial. For screening fine material, small throws with higher speeds are best. Crushed material calls for larger throws than required for rounded material as the irregular shape of the particles tends to cause them to stick in the apertures. Particle Size, Shape and Distribution. For a given screen aperture, both screening rate and passage probability (hence capacity) will increase as particle size decreases, although particle shape will modify the effect. With design variables constant, screen capacity will be less for acicular particles in comparison to more rounded ones. For a given material, the size distribution defines the proportion of fines, near-size particles and oversize particles present on the screen in relation to the aperture size. Near-Size particles are those whose sizes are between Y, to 1% times the size of the aperture, so that a small particle for one size of screen will be a near-size particle for a small size of screen. Oversize particles never pass through a screen; small particles (fines) pass through rapidly. Thus, near-size particle passage is the rate-determining step in screening, since, if poorly aligned, near-size particles will not pass through. To obtain higher screening rates (higher capacities), exposure of fines and near-size particles must be maximized and the quantity of oversize must be reduced such as by employing an upper screen deck of coarser size. This deck is commonly referred to as a relief deck. Large, rapid variations in the proportion of oversize and near-size particles in the feed stream will influence the load on the screen. Efficiency may suffer accordingly, so it is desirable that the feed size distribution be stabilized if possible. Solids Feed Rate and Bed Depth. For a screen of fixed throw, speed and aperture, bed depth depends upon factors such as feed rate, deck slope, feed size distribution and direction of rotation (either concurrent or counter to the flow direction). At steady state, large particles on top of the bed prevent finer ones from bouncing around, thereby keeping them close to the screen surface. Likewise, oversize helps to push near-size particles along or through the screen to reduce blinding. Since bed thickness increases with feed rate, there is an optimum thickness because efficiency increases with feed rate, passes through a maximum and then decreases. For a given feed rate, the screen width is selected to maintain the bed depth at the discharge end, so that screen width determines capacity. Nichols (1982) recommends that bed depth for dry ores at the discharge end of the screen should not exceed n times the screen opening, where n = 2 + 0.02p . For a given rate of flow of oversize material, F (stph), the bed depth, D (inch), may be estimated from:

where vt (ft/min) is the rate of travel of the bed material and W (ft) is the effective width of the screen and p is bulk density (lbs/ft3). At the feed end, D may be estimated by letting F represent the feed rate to the screen. For other materials, Bothwell (2002) recommends:

902

D = No more than 4 times the opening for Dry Aggregate D = No more than 3 times the opening for Dry Raw Coal D = No more than 8 times the opening for Wet Mill Discharge D = No more than 3 inches for Wet Drain & Rinse or Desliming Matthews (1985) provides an alternative to Equation ( 5 ) :

I

Screen Length, ft 6 to 10 12 to 16 20 to24

I

Bed Depth, Dry Rock 1.5 to 2.0 times avg. particle size 2.0 to 2.5 times ave. Darticle size 2.5 to 3.0 times avg. particle size ]

Bed retention time, t (min), is estimated from t = L,/v,, where L, (ft) is the effective screen length. The rate of travel, v,, varies with deck slope and motion characteristics. On inclined circle throw screens in counter flow rotation, between angles a, of 18" and 25", v, is approximated by v, = -120 + 10a. Thus at a = 20", v, = 80 ft/min. For flow rotation v, is somewhat higher. In production environments, bed depth is adjusted by manipulation of the feed rate where possible. If capacity cannot be sacrificed, then deck slope should be modified to obtain the desired depth. To stabilize bed depth as much as possible, the feed size distribution should be as constant as possible. Feed Moisture Content (Dry Screening). Moisture content in excess of a few percent and/or a high clay content in screen feed may lead to blinding of screens andor to a reduction in efficiency and/or capacity. In the absence of blinding, moisture causes fine particles to stick to oversize. Moreover, fines may tend to agglomerate in the presence of clay which acts as a binder. Small particles thus become larger ones. When such problems are severe, efficiency is lowered significantly. Agglomerates which reach a size equivalent to 95 that of apertures may cause blockage and reduce capacity. Depending upon aperture size and feed size distribution, moist fines may adhere to screen mesh and decrease the effective screen opening. Efficiency is lowered as a result. In exceptional cases, apertures may be totally closed by adhesive fines (clays or near clays). When problems are severe, special screen cloth may be used, or heated decks and rubber ball trays can be used, (as described previously). Alternatively, wet screening should be considered.

TRADITIONAL SIZING METHODS Nichols (1982) traces briefly the history of vibrating screen development, where box and mechanism designs were improved sufficiently to make them an important part of most processing plants. By 1940, two empirical methods to estimate the size of a vibrating screen for a particular task had been developed. These incorporated experimentally-determined base capacities (still in use today with modifications) where modifying factors were necessary to force a match with actual operation. One method has screen area proportional to the mass flow rate of dry feed to the screen, while the other has screen area proportional to the mass flow rate of dry undersize fed to the screen. For both methods, empirical capacities were developed under the following criteria (Nichols, 1982): 1) 2) 3) 4)

Material being fed to the screen contains 25% oversize. Feed to the screen contains 40% half size material. Material has a bulk density of 100 lbs/ft3(1.6 tons/m3). Screening surface has an open area of 50%. 5 ) Material is dry, free flowing and relatively cubical. 6) Maximum screening efficiency required is between 90 and 95%.

903

The above conditions may not occur for natural or crushed size distributions, so that experimental modifying factors were introduced. Involved is the calculation of an effective screen area from input data that permit the selection of a basic capacity and corresponding modifying factors. Once effective screen area is determined, consideration is given to width, length, severity of duty, support structures, feeding arrangements and screen enclosures for dust collection.

Sizing Vibrating Screens By the Feed Rate Method This method has screen area, A (ft2) proportional to the screen feed rate, F (stph), at about 90% efficiency where:

C = base capacity, stph/ft2of screen when other factors are 1. Varies with aperture size. M = oversize factor for stratification difficulty when % of oversize in feed changes. K = half-size factor accounts for % of feed passing Yi of aperture size to not plug holes. Q1 = bulk density factor = p I 100 ,where p is bulk density of ore in lb/ft3. Q2 = aperture shape factor accounts for effect of round vs square vs rectangular openings. Q3 = particle shape factor accounts for effect of cubical vs slabby elongated shapes. Q4 = open area factor = %OA/50 for bulk densities of 5 1 to 100 lb/ft3. Q4 = open area factor = %OA/60 for bulk densities less than 50 lb/ft3. Qs = wet screening factor increases with opening size. Qs = surface moisture factor increases with decreasing moisture. Q7 = deck location factor Qmf= multiflow factor accommodates for special inclinations in screen deck. Factors have been presented either as tables (Gluck, 1965) or as graphs with tables (Nichols, 1982).

Sizing Vibrating Screens By the Throughput Method This method has screen area, A (ft'), proportional to the mass flow rate, U (stph), of screen undersize in the feed at steady state. Matthews (1985) shows that:

A=[

" I

(7)

Ff Fo Fe Fd Foa Fs Fw

C = base capacity, stph/ft2,at 95% efficiency R".. Ff = fines factor. Accounts for difficulty of screening the % passing ?4of the aperture size. F, = oversize factor. Accounts for stratification difficulty if % finer than aperture size changes. F, = efficiency factor to account for desired efficiency. Fd = deck factor to allow for area lost on lower deck. F,,= open area factor = ratio of %OA used to a standard %OA. F, = slot factor. Accounts for effect of shape with long dimension parallel to flow direction. F, = wet screening factor. An efficiency factor is incorporated, such that screen area will be reduced to maintain the same capacity at any desired lower efficiency. An efficiency greater than 95% is not considered practical. A somewhat similar approach was introduced by Karra (1979) who showed that:

904

Screen Area (m2) =

" I

U = mtph of undersize in screen feed = Ff, (see Figure 2) A = base capacity for crushed stone B = oversize factor C = half size factor D = deck location factor W = wet screening factor F = bulk density factor = U/1602 G = near size factor R = efficiency factor related to R,. Kana (1979) provides regression equations that permit the rapid calculation of the above factors from the usual input information for sizing a vibrating screen.

Estimation of Screen Width, Length and Deck Angle Once effective screen area, A, has been estimated the screen width, W (ft), and screen length, L (ft), must be determined. Remember that a length to width ratio of 2 to 3 is not uncommon. Recall that the effective area of a screen can be found in tables. Otherwise, increase the calculated area by about 10 %. W is chosen to maximize capacity; length is chosen to maximize efficiency. W can be estimated from Equation (3, when bed depth, D, is pre-selected. Bulk density of the ore must be known and v,, the rate of bed travel, must be estimated. For inclined circle throw machines in counter flow rotation, v, = -120 + 10a, where a is deck angle with values between about 16 to 28 degrees. In flow rotation, v, will be about 25% higher. Starting deck angles at various ideal oversize flow rates, F (stph), and various screen widths, W (inch), are estimated from

a =1

5

4

(9)

for angles between 10 to 28 degrees. Standard widths are, in inches, 24,36,48,72,84 and 96 and a should be rounded to the nearest whole number. Length, L (ft), is found from L = A/W, where A is the effective screen area. Values for L and W should be matched as well as possible with off-the-shelf machines to keep costs down.

Traditional Input Data For Calculation of Screen Area Minimum data necessary to calculate screen area are: Feed size distribution Feed rate Screening Efficiency Required Separation size Vibrating screen type Screening media type Type of material being screened Oversize and Undersize limits ADVANCES IN SELECTION AND SIZING METHODOLOGY Over the years, the various sizing methods have been refined, and factors adjusted to compensate for changes in the industry. One of the areas that has seen a considerable change is in applications that have a high percentage of fine material in the feed, combined with a large amount of water. Application engineers discovered that when using traditional methods to size these screens, the

905

result was a very lightly loaded screen. This resulted in the theory that slurries of this type act more like a liquid than a typical agglomeration of particles. By applying fluid dynamic principals and fluid flow curves, a new method of sizing screens was developed. This sizing method, known as the Pulp Sizing method, is discussed in the following section. The methodology is typically used on FAG and SAG mill discharge screens. Also discussed in this section are techniquess used to design custom screens to meet special operating parameters.

Screen Design Because vibrating screens are relatively easy to manufacture, a “made to order” approach is realistic and operating personnel can have influence at the design stage. The latest design techniques, such as a finite element analysis (FEA) to analyze operating stress distributions in screen components and structures under vibration and material load, are employed successfully. The idea is illustrated in Figure 7. In Figure 7, the application factor, which is expressed as % of total fatigue life, involves screen design to ensure of structural integrity under stress and varies with type and/or tonnage of feed material and with location inside the screen structure. The factor can be anywhere between 10 to 95% of the total screen fatigue life and is limited by carrying capacity, “G-Force” (see under Mechanical Limits) and availability. Selection of Mill Discharge Screens Factors that should be taken into consideration are sizing, mechanical limitations, increasing availability and media selection. Sizing. Mill discharge screens must perform two functions. The first is to achieve an efficient separation and the second is to de-water the oversize material to the point at which it can be conveyed. Separation: Mill discharge screens are normally set-up to make one or two separations with a single or double deck machine, respectively. Screens performing two separations are required in mill circuits with FAG (Fully Autogenous Grinding) mills, which utilize pebbles from the primary mill as a grinding media in the secondary mills. Depending on the loading and gradation of the ore, it is also common to use the top deck of a mill screen, as a relief deck. The apertures in the screening media generally range from 0.5 mm to 20 mm for the bottom deck, and 6 mm to 40 mm for the top deck. In mill discharge applications it is common for the majority (90% or more) of the feed to pass through the opening in the screen deck. A “Pub Sizing” computer-modeling program, which predicts the size of screen required for this type of application, is employed. It is based upon empirical data collected from free flowing slurries which were presented to a mesh with a given open area. When a bed depth of material inhibits the free flow of the fluid, the screen must be over-sized to allow for efficient screening and good drainage. The bed depth correction factor, C, for bed depth to aperture ratios, (DK,) of 4 to 8 inclusive is estimated as (note: set C =1 for a ratio c 4):

c = 0.89 + o . o 4 [ 4 Figure 8 shows how deck % open area depends upon feed rate (gpm) from zero inclination to 20 degrees.

906

_I+

Process Assessment Sizing

e-

DESIGN

-

Process Factors Finite Element Analysis Fatigue Life

OPERATION

I

-II

Figure 5 Current Approach to Vibrating Screen Design

907

65 60 55 50 45

40 35 30 25 20 15

0

20

40

60

80

100

Gallons per minute Figure 8: Deck % Open Area versus Screen Capacity (gpm)at Various Inclinations The parameters necessary to predict the size of a mill discharge screen are: 0

0 0 0 0 0

Feed rate in stph (short tons/hour) of solids or gpm (gallons/minute)of slurry Percent solids of the slurry feed Percent of solids passing the aperture in the screen media Specific gravity of the solids Separation size Percent of open area in the screen media

The expression applies to bottom deck openings less than 12 mm. For apertures greater than 12 mm, the bed depth to opening ratio should be reduced to 4: 1 as aperture size approaches 25 mm. There are several methods of reducing the bed of material on the deck of a screen. The first is to change the angle of inclination. Although this may increase screening efficiency and carrying capacity, it will have a negative effect on the de-watering capability of the screen. A second method of decreasing bed depth is to add a relief deck above the separation deck. The media on the relief deck should be selected to keep the percentage of feed passing the opening above 90%. The bed depth should also be balanced to allow the bottom deck to perform a good separation within the parameters outlined above. In some applications the amount of mill screen feed which passes the opening will be less than 90% and the bed depth will be within the range specified. For these situations, traditional sizing methods (e.g., see Nichols, J., Vibrating Screen Theory and Selection, METSO Minerals, formerly Allis Chalmers-Svedala) are employable such as by using Equation (5). Table 1 on the next page shows computer printout of a portion of pulp screen sizing output based on input data as shown at the top of the table. De-Watering: A second function of a mill discharge screen is to reduce the amount of moisture in screen oversize. Since most mill screens discharge onto a closed circuit belt conveyor, it is necessary to reduce the surface moisture of the screen oversize, to a level which can be easily

908

handled. This moisture level is normally in the 5 1 5 % range depending on the characteristics of the conveyor belt and material being conveyed. When moisture levels are higher than this amount, “belt slides” and other related problems can occur. The surface moisture level of the discharge material depends on the following factors:

Table 1: Computer Printout of Pulp Screen Sizing Output

PULP SCREEN SIZING FOR SLURRY SCREENING customer:

[ ~ g ~ r e e n ~

I

lO.OR-xrx!%

Low Head Screen a

q r d

(Does mtircludebbrkmpactafea)

The Bed DepthDpenmg Ratio k 3.22

0 0 0 0 0 0

Percent solids of feed Particle density Retention time on the deck Bed depth of material Open area of the media G-Force of the screen

Under normal conditions, the first 1/4 to 1/3 of a screen is used to remove thefur-size material. The remainder of the deck removes near-size material and drains the water from the screen. The percent solids coming onto a mill discharge screen is normally in the range of 40% to 75%. As more water is fed to the screen, a longer drainage section is required to compensate for the larger amount of area used to pass the initial surge of slurry. As the percent solids increases above 65%, it is often necessary to add spray bars to the screen. The spray bars should be located a minimum of 1 to 2 meters from the discharge end of the screen to allow some drainage. Normal mounting of the spray bars is at an angle of 75 degrees to the deck, against the flow of material. The amount of spray water should be added to the total capacity of the screen feed, and the percent solids should be adjusted accordingly. Particle density is measured in terms of MPS, “Mean Particle Size”. Most slurry feeds from grinding mills have an MPS of 3mm or more. In these cases the MPS does not have a major effect on the de-watering capabilities of a mill discharge screen. There will be a much larger effect on moisture levels in cases where the MPS is less than 1.5mm, and the bed depth to opening ratio is near or above the maximum recommended. When this occurs, the surface tension bond of the

909

water will reduce the inter-particle movement of the material. The result is an oversize product which has a much higher inherent moisture. When sizing screens for these types of applications, the DSM, “Dutch State Mines”, screen sizing method should be considered. Since this type of application is a relatively rare occurrence in mill discharge screens, it is not considered herein. The retention time of the material on the screen deck has a big effect on the moisture level at the discharge end of the screen. Most mill discharge screens will retain the material for a minimum of 15 seconds after the initial surge of slurry feed is removed. This is apparent from Figure 9 shown below.

I-.i-Coarse

a

+Fine

I

20

25

70

4 60 50

40 c,

a

30 20 10

0 0

5

10

15

30

Retention Time (Seconds) Figure 9: Oversize Discharge % Solids versus Retention Time The retention time must be longer in applications with smaller apertures, which tend to have a smaller MPS in the oversize product. In the “Pub Sizing” computer program, a correction is made as the angle of incline increases. This is due to the sheeting action of the water at steeper inclines. For applications requiring a lower moisture discharge product, the surge factor in the “Pulu Sizing” program should be increased to allow more retention time. As the bed depth of material increases, the ability of a screen to stratify the material before it reaches the end of the screen, is decreased. A higher bed depth also tends to compress the bottom of the material bed, thus making it more difficult for water and fine particles to work their way down to the media where they can be removed. The open area of the media directly effects how quickly the initial surge of slurry can be removed. A screening media with a higher percent of open area can reduce the size of the screen required, but at the sacrifice of wear life. Selecting the proper media is discussed in more detail later in this paper. The G-force (energy level) that a mill discharge screen is operating at can greatly effect its de-watering capability (see under Mechanical Limits). A higher G-force is more effective in

910

breaking the surface tension bond between water droplets and ore particles. Higher G-force is also instrumental in achieving a higher carrying capacity over a screen. Mechanical Limits. There are also mechanical aspects of a screen, which must be taken into consideration. The first of these mechanical factors is the G-Force (see Equation (4)), a parameter for measuring the energy level of a screen. In horizontal de-watering screens it is important to maintain a G-Force of 4.5 to 6.0. This value compares to G-Forces of 3 to 4 for a typical incline screen. As aperture sizes decrease and bed depths increase, screens should be operated near the top end of this range for maximum performance. Tests have shown that even higher G-Forces are advantageous for these types of applications, but the exponential cost increase of building screens to run at higher G-Forces has limited the market to the levels currently available. The second mechanical limit, which must be considered in a screening application, is the “Carrying Capacity” of a screen. The Carrying Capacity is defined as the maximum amount of material, which can be carried over the decks before the momentum of the screen body is overcome by the weight of the material. An empirical expression has been developed for calculating the value of this mechanical limit, where the carrying capacity is a function of the following parameters: 0 0 0

0 0 0

Screen weight (W) Screen speed (N) Screen throw (T) Screen length (L) Material travel rate over decks (vt) Empirical constant (K) Carrying Capacity, C,, is calculated as follows:

”=[

]

WV,T~N~ KL

The signs of exceeding carrying capacity are easily identified. These include: 0 0 0 0

Irregular material distribution andlor travel rate along the screen decks Differences in throw when screen is empty or loaded Bottoming of screen isolation springs Failure of cross tubes in the screen body

It should be noted that other factors could cause these same types of problems, but usually not all at the same time. Increasing Availability. Due to the large investment involved in setting up a milling circuit, it is the goal of most mines to have their equipment available for operation over 90% of the time. Many mines have been able to increase their availability to over 95% by utilizing state of the art preventive maintenance techniques and other innovative ideas. In order to reach these high levels of availability, the maintenance cycle must be optimized. The length of this maintenance cycle varies depending on ore characteristics and site conditions. Most mines have a scheduled maintenance cycle in the range of 1 to 6 months. It is important that any equipment installed in these types of operations be designed to meet these maintenance cycles. In the past, many mines were set-up with a single piece of equipment to perform each operation. Today the industry has realized that the equipment in the circuit with the lowest availability governs the availability of the entire milling circuit. Due to this, equipment requiring higher maintenance has been identified, and many mill circuits are designed with parallel lines of equipment. With this type of set-up the material flow can easily be by-passed to one line, while the other receives maintenance.

911

Since space is rather limited in the mill discharge area, in most cases it is not possible to easily divert material flow from one screen to another. A solution to this problem is to mount the mill screens on a “roll-away cart” as illustrated in Figure 10. When a screen is mounted in this manner, it can easily be rolled to the side for maintenance, while an identical screen is rolled into position for the next maintenance cycle. This entire process can be completed in a few minutes. There are four major areas, which affect the availability of a screen. In order of importance, these include: 1. 2. 3. 4.

Screen Media and liners Drive components Structural Components Screen Mechanisms

The search for innovative components and predictive maintenance methods to increase the cycle life in each of these areas is never ending. Some of these include:

Figure 10: Movable Mill Discharge Screen

0 0

0 0

Specialized screening media Special electrical motors and motor mounts Special bearings in vibrator mechanisms Modal analyses to prevent critical frequency problems State of the art testing procedures

912

0 0

Special screen isolation springs Special surface coatings for corrosive applications

The features listed above are important to consider when comparing mill screens, but will not be discussed in detail as they are beyond the scope of this paper. Screen Media. One of the most expensive maintenance items on a mill screen is the screening media. Since media is continually exposed to the abrasive ore and scouring action of large amounts of water, it tends to wear out at a relatively high rate. It is important to look at not only wear rate, but also open area and change-out time required, because these factors can greatly effect the production and availability. Some factors to consider for mill discharge screens include: 0 0 0 0 0 0 0

Rubber, polyurethane or steel media Open cast or injection molded polyurethane “Effective Open Area” of the media Thickness of the media near the aperture Relief angle of the apertures Aperture shape Method of fixing the media to the screen

Type of Screen Media Material: Rubber media is generally used in applications where high impact and large material top size is encountered. Due to this, the top deck of a mill discharge screen is normally a good application for rubber. Polyurethane is used in applications, which require a higher open area while still providing good wear life. There are polyurethane media systems available which will accept the feed sizes seen on the top deck of a mill screen. If there are petroleum products or other chemicals in the ore, polyurethane is less likely to have a chemical reaction as compared to rubber. Due to the highly abrasive characteristics of mill screen applications, steel wires are only used in cases where higher open area is required. Types of Polyurethane Media. When choosing a polyurethane, it is important to determine if the media is injection molded or open cast molded. The injection molding process is done in a hot mold, which requires panels to cure very quickly. The open cast molding process is done in an open mold at room temperature. In this process the curing time is much longer. Since the curing time for the open cast system is longer, molds are usually setup to make multiple panels at once. This results in a higher tooling cost than the injection molding system which requires only one small mold for each aperture size. The longer curing process of the open cast molding system results in an interlaced molecular mesh, while the injection molding process tends to create a polyurethane with a linear grain structure. The two different processes also require different types of urethane mixes. The injection molding system normally uses a polyether type urethane, while open casting uses polyester. The result of this comparison is that open cast screen media panels normally last 30%to 50% longer in a direct comparison. Injection molded panels do have a higher accuracy which is important in aperture sizes lower than 1.5 mm. When screening below this size range, the acceptable level of product contamination needs to be considered. If open cast panels can not meet the accuracy required, it becomes necessary to sacrifice wear life by using an injection-molded media. Open Area: The methods used to compare the open area of a media panel are often misunderstood. Most manufacturers of polyurethane and rubber media compare open area by using ROA, “Relative Open Area”. ROA measures the opening to urethane ratio in the mesh area only and does not consider the thicker side frames of the panels. In order to get a true comparison of the open area, the EOA, “Effective Open Area” must be considered. The EOA is calculated using the following formula: EOA = [(Number of Apertures)x(Areaper Aperture)]/[Area of Entire Media Panel]

913

The EOA is normally considered when calculating the volume of slurry a given screen can pass. Thickness and Relief Angle: When choosing a screening media, thicker is not always better. Reasons for this anomaly can be explained in terms of the relief angle. The apertures in polyurethane and rubber screening media are molded to be smaller at the top, and larger at the bottom. This tapered design prevents pegging of material, and allows ore particles to fall freely through the panel once they pass the narrowest point of the aperture. As media wears, it tends to wear in the narrowest section of the aperture. At some point during the life of the panel the walls of the aperture will become parallel, and pegging will begin to occur. Since pegging reduces the effective open area, the production will be reduced. The only way to increase the production back to normal levels is to change the panels, which results in a larger percentage of the panel being thrown away. Small apertures will normally have larger relief angles than large apertures. This is due to the fact that larger particles retain more energy and can be thrown out of the aperture easier than smaller particles. Media with square apertures, which are very rigid, require larger relief angles than those with slotted apertures or flexible meshes. Aperture Shape: Slotted apertures can increase screening efficiency by 25% as compared to square apertures. This is due to a higher probability of a particle hitting an aperture, rather than the mesh surrounding the apertures. Slotted openings will allow more slabby material through the screen deck. This is normally not a problem in mill applications. Round apertures can reduce pegging and provide longer media life; the trade off is a reduced open area. Round apertures must also be sized larger than square apertures to make the same separation. Fixation to screen: There are many types of media systems available. Each type uses a different method of fixing the media to the screen. In applications with large feed size or high bed depths, it is important to choose a system with good support below the panels. Bolt-down systems are often used in this type of application. In applications with small feed sizes and lower bed depths, snap in media is normally used due to ease of maintenance.

OPTIMIZING PERFORMANCE AND WEAR COSTS Optimizing performance means that you seek to operate screens at their “highest” capacity and efficiency consistent with production targets at the “lowest” possible wear costs. Optimizing Performance. A higher screen feed rate is obtained by increasing the screen feed rate of solids (or mill discharge) in some way. However, in the absence of additional changes, screen efficiency will likely be sacrificed (see Figure 6 ) . Efficiency is associated with stratification and the probability of particle passage through screen apertures. Stratification: When efficiency decreases due to poor stratification, the operator should consider the following if possible: Decrease feed rate Increase travel rate (produce a lower bed depth) Increase the screen energy (G-Force) which depends on throw, speed and rotation direction Change to wet screening if screening is dry Change the screen media (e.g., switch to multi-flow screen decks if possible) Ensure that screen feed uniformly distributes across the width of the screen at the feed end

Probability of Passage: When probability of passage causes poor efficiency, the operator should consider the following if possible: Decrease the travel rate (increase retention time) Increase angle of incline while running at counter flow Optimize G-Force (speed and throw) for the screen opening size Change to wet screening Change screen media (such as by increasing open area) Convert circle throw machine to counter flow if with flow conditions in use

914

Ensure that screen feed uniformly distributes across the width of the screen at the feed end

Optimizing Wear Costs. Wear costs depend upon the following wear items: Screening media Clamp bars and side liners Feed box and discharge spouts Motors and drive mechanismshelts Springs Bearings and mechanism supports Support frames and side plates To lower wear costs, the operator should consider the following if possible: Use synthetic media and liners Reduce frictional force between media and material by lowering bed depth using a smaller particle size to opening size ratio Decrease the energy level Replace belt drives with direct drives Use straight line motion instead of circular motion

SUMMARY Screening principles have not changed over the years, but a wide variety of new materials and modeling tools are available to today's screen engineers. By utilizing these innovations modern screens are able to achieve higher production in the same space with higher availability than in the past. As this trend continues screen manufacturers will be able to build larger, more cost effective designs that keep up with the demand for lower cost screening. Cost reduction in the screening process will become more critical as ore reserves are depleted and lower grade ores are mined. ACKNOWLEDGEMENTS We wish to thank METSO for permission to publish this paper REFERENCES Matthews, C. W., (1985), General Classes of Screens, Ch 1, Section 3E in SME Mineral Processing Handbook, Ed. N. Weiss, SME, Littleton, CO. Matthews, C. W., (1985), Screening Media, Ch 4, Section 3E in SME Mineral Processing Handbook, Ed. N. Weiss, SME, Littleton, CO. Colman, K. G., (1980), Selection Guidelines For Size and Ty e of Vibrating Screens in Ore Crushing Plants, Ch. 15 in Mineral Processing Plant Design, 2 Edition, Eds. A. Mular and R. Bhappu, SME, Littleton, CO. Colman, K. G., (1985), Selection Guidelines For Vibrating Screens, Ch. 2, Section 3E in SME Mineral Processing Handbook, Ed. N. Weiss, SME, Littleton, CO. Nichols, J. P., (1982), Selection and Sizing of Screens, Ch. 27 in Design and Installation of Comminution Circuits, Eds. A. Mular and G. Jergensen, SME, Littleton, CO. Gluck, Samual E., (1963, Vibrating Screens, Chem. Eng., McGraw-Hill, NY. Gluck, Samual E., (1969, Vibrating Screens: Surface Selection and Capacity Calculation, Chem. Eng., McGraw-Hill, NY. Bothwell, Mark, (1999), Selection of Mill Discharge Screens, Svedala Screening USA (currently METSO), Appleton, WI. Karra, V. K., (1978), Efficiency Sizing of a Vibrating Screen, SME-AIME Preprint No. 79-23, AIME General Meeting, New Orleans, Feb. 18, 1979.

"B

915

Karra, V. K., (1979), Development of a Model For Predicting the Screening Performance of A Vibrating Screen, CIM Bulletin, 72, #804. Nichols, J. P., (1982), Vibrating Screen Theory and Selection, Allis Chalmers Corp (currently METSO), Appleton, WI. Grant, Douglas C., Richard J. Hornung, Keith E. Rouch, Pradip N. Sheth, (1980), Predicting Screening Performance, Allis Chalmers Advanced Technology Center, Appleton, WI Moses, John and Lyle Gray, (1 992), DSM and Pulp Sizing Methods, Svedala Australia Limited. (1980), Vibrating Screen Manufacturers Association of America Handbook.

916

Fine Screening in Mineral Processing Operations Steven B. Valine’ and James E. Wennen2

ABSTRACT Technical advances in screening machine design and the development of long-wearing, nonblinding screen surfaces have made fine screening in mineral processing plants a practical consideration. Factors that affect the selection and performance of fine screening operations will be reviewed. Benefits associated with the use of fine screens in several operations will be presented. INTRODUCTION Screening is the process of classifying particles according to size. While factors such as particle shape and specific gravity may have an effect, the separation is largely dependent upon particle size. In general, fine screening applies to particle separations ranging fiom 10 mm (3/8 inch) to 38 microns (400 mesh). Fine screening is normally accomplished with high frequency, low amplitude vibrating screens employing either elliptical or straight-line motion. These types of screening machines with the advantage of high unit capacity and high efficiency are the subjects of this paper. Stationary screens, such as sieve bends, are generally less expensive, but have lower capacities and usually require multiple stages to achieve acceptable efficiencies. In cases where very close tolerances are required, sifting screens with flat, circular motions are often used.

Fine screening can apply to wet or dry separations although the mechanisms involved are quite different. In wet screening, particles are fed to the screen as a slurry. Particles small enough to pass through the openings are carried through with the fluid and the process is completed in a relatively short screen length. Once most of the liquid has been removed, the screen simply acts as a vibrating conveyer until more water is added to facilitate further removal of additional fine particles. In most cases, equipment requirements can be determined on the basis of unit feed rate per unit width of screen, for example, t/h/m. Dry screening is more of a statistical process where particles are presented to the screen surface multiple times as they roll, bounce, or move down the length of the screen. To pass though the screen, an undersize particle must be presented to an opening in just the right way and probability plays a role. Dry screens require a certain length for the process to occur efficiently and screen area becomes an important design parameter. For design purposes, dry fine screens are typically sized on the basis of unit feed rate per unit area, for example, t/h/m2. The size distribution of one or more products from a fine screen is an important performance criteria for most screening applications. However, screening efficiency is just as significant in evaluating screen performance. From a practical standpoint, screening efficiency is analogous to recovery while the size analysis of a particular stream is similar to grade. Screening efficiency can I

Derrick Corporation. Buffalo, New York.

2

Consulting Engineer, Grand Rapids, Minnesota.

917

be thought of as that fraction or percentage of particles that were correctly placed by the screen. At a desired particle size, three efficiency values can be calculated from the size analysis of the screen feed, oversize, and undersize streams. Oversize eficiency is the fraction of oversize particles in the feed that were recovered to the oversize product screen. Similarly, undersize eficiency is the fraction of undersize particles in the feed that were recovered to the undersize product stream. OveraN eficiency is the total fraction of particles that were correctly placed. At the desired separation size, the following data is required for the calculation: A - percent of oversize in the feed; B - percent of undersize in the feed (1 00 - A); C - percent of oversize (coarse) in the oversize product; D - percent of undersize (fines) in the undersize product. The weight splits and efficiency values are calculated as follows. U = Undersize weight (percent)=

IOO(C-A) C+D-100

0 = Oversize weight (percent)= 100 - U

E"

UD

= Undersizeeficiency = -

B

oc

E~ = Oversize eficiency = A

E = Overall efjlciency =

UD+OC 100

For example, a 140 mesh separation is required for a feed material analyzing 95.9 percent passing 140 mesh. Sieve analysis of the screen products shows the oversize to be 58.4 percent passing 140 mesh while the undersize product is 98.7 percent passing 140 mesh. Therefore, A=4.1, B = 95.9, C=41.6, and D = 98.7. Using the above equations, U=93.1, 0 = 6.9, EU= 95.8, Eo = 70.5, and E = 94.7. In this example, the screening machine correctly placed about 95 percent of the material. It has been said that screening is more of an art than a science (Matthews 1985). Unlike coarse screening and despite thousands of data points, the development of accurate mathematical models and/or handbook-style guidelines for fine screening has not been possible or practical. Accurate fine screening machine requirements, operating conditions, and performance are best determined by conducting full-scale screening tests with representative samples. Screening machine manufacturers are usually in the best pqsition to make preliminary recommendations and conduct full-scale tests. WET FINE SCREENING The selection and operation of the appropriate screening machine for a wet separation depends upon the process objective. For example, a process may require maximum oversize efficiency (the correct placement of oversize), such as the use of fine screens in a classification circuit. All coarse,

918

non-liberated particles should be recovered by the screen and sent back to the grinding mill. Undersize efficiency is equally important, but the process can tolerate some fines going back to the mill with oversize. The recovery of media with a fine wet screen would be a second example where maximum recovery of oversize (high oversize efficiency) would be critical. If a product specification called for a minimal amount of fines, then undersize efficiency (the correct placement of undersize) would be important; the screen must remove almost all the fine particles in the feed. Finally, in a dewatering application, poor undersize efficiency would be desired; the objective would be to capture and dewater as many particles as possible. In each of these cases, a different type of screen would be selected and the operating variables would be different. Factors Affecting Wet Screening Feed Rate. The capacity of a screening machine is defined as the optimal feed rate to meet the desired product specifications. Feed rate, usually expressed as dry mass flow (t/h), is one of the more critical factors affecting screen performance. The capacity of the screen will determine the number of screening machines required. Exceeding capacity (or over feeding a screen) will result in the misdirection of undersize particles and fluid to the oversize stream as well as a reduction in screen surface life. Depending upon other factors, the optimal feed rate can be exceeded to some extent without a significant decrease in efficiency. The capacity of a wet fine screen is best determined by full-scale testing to optimize all factors affecting screen performance. Feed Density. As explained above, undersize particles are transported through the screen openings by the fluid and therefore, the volume fraction of fluid will affect screen efficiency. Screening efficiency will increase with decreasing feed density. From a practical standpoint, a screen feed density of roughly 20% solids by volume has been found to be a reasonable compromise, independent of dry solids specific gravity. For example, high screening efficiency could be obtained with the screen feed at 45% solids by weight for a silica sand slurry with a dry solids specific gravity of 2.6. For a mineral with a specific gravity of 5.0, the screen feed should be about 55% solids by weight to obtain reasonable screening efficiency. To maximize undersize efficiency (the correct placement of undersize), the screen feed slurry could be even lower, perhaps as low as 10 to 15% solids by volume. It has also been shown that it is usually more beneficial to add water to the screen feed slurry than to add the same amount of water directly to the screen surfaces with spray nozzles. For a dewatering screen application, the screen should be fed at the highest percent solids obtainable, since the goal is to retain the maximum amount of particles on the screen.

Feed Size Distribution. The size distribution of the material fed to a screen is one of the more important factors affecting both capacity and performance of a wet screening machine. The oversize particles must be conveyed off the screen and capacity usually decreases as the amount of oversize increases. Another important factor is the amount of near-size material in the screen feed. Near-size material is defined as the particles that are 2 mesh-size equivalents larger and smaller than the screen opening. Near-size, oversize material inhibits the ability of the undersize material to get through the screen openings and, in some cases, can cause some plugging problems. Selection of screen media is quite important when dealing with significant amounts of near-size material. Screen Opening and Open Area. The larger the opening, the greater the machine capacity. Conversely, as the desired separation size decreases, so does machine capacity. For example, say that full-scale tests determine that machine capacity is 100 t/h with a 250 micron (60 mesh) screen opening. Machine capacity could drop to 20 to 40 percent with 150 micron (100 mesh) openings. At a given opening size, the open area of the particular screen surface also affects capacity. To increase screen panel life, it may be desirable to use a more robust screen cloth with lower open area. However, doing so will result in a lower machine capacity.

919

Types of Wet Screens Single Feed Screen with Optional Water Addition. Single feed wet screens, especially when designed with spray water addition, are typically used to achieve maximum removal of undersize particles from the oversize. An example is the Repulp screen illustrated in Figures 1 and 2. The Repulp screen is equipped with spray nozzle directed towards lined wash troughs located between one or more screen sections. As discussed previously, water passing through the screen openings is the primary means by which undersize particles are transported through the openings. By reslurrying the oversize one or more times, additional fines or chemicah can be removed. The wash troughs are important since spray water directed towards fine mesh screen surfaces can dramatically reduce screen panel life.

Figure 1 Repulping screen (photo courtesy of Derrick Corporation)

Figure 2 Repulping trough (photo courtesy of Derrick Corporation)

920

Multiple Feed Point Screens. As mentioned previously, the process of fine wet screening can be completed in a relatively short screen length. Therefore, the ideal type of machine geometry would be a short, wide screening machine. Since this concept is not yet practical, several manufacturers have designed screening machines with multiple feed points. As illustrated in Figure 3, this concept is actually 2 or 3 short screens in parallel and accomplishes the same thing as a single short, wide screen. Multiple feed point machines have been shown to have 50% to 125% more capacity than an equivalent single feed machine. Feed

1

Feed

Figure 3 Multipfe feed point screen A multiple feed point screen is preferred when the objective is to produce an undersize with minimal amount of oversize (high oversize efficiency) and some misplaced undersize particles can be tolerated in the oversize. Derrick Corporation has expanded on this concept with the introduction of their Stack SizerTMscreen (patent pending). As shown in Figure 4, the Stack Sizer is actually 5 short screens in parallel, stacked one above the other.

Figure 4 Stack SizerTM (photo courtesy of Derrick Corporation)

921

Dewatering Screens. There are two general types of dewatering screens. The first type, and most common, is a horizontal, linear motion screen as shown in Figure 5. This type of machine is generally fed at high percent solids to minimize the amount of free water. Minimizing the amount of water in the feed will maximize oversize recovery. The screen simply acts as a vibrating conveyer, shaking water from the solids as it conveys the material to the discharge chute. The oversize moisture content decreases as the acceleration or g-force produced by the screen increases.

Figure 5 Dewatering screen fed by hydrocyclones (photo courtesy of Derrick Corporation)

The second type of dewatering screen is an inclined machine with some type of vacuum feature to assist in removing water from the oversize. As illustrated in Figure 6 , the oversize discharge section of the machine offered by Derrick is connected to the suction side of a small blower. Using this type of screen can generally result in lower product moisture levels than similar horizontal, linear motion machines and is best suited for particles larger than 75 microns (200 mesh).

Figure 6 Vacuum-assist dewatering screen (photo courtesy of Derrick Corporation)

922

New Type of Fine Wet Screening Surfaces - Polyurethane The development of high open area, long wearing, fine polyurethane screen surfaces is perhaps one the greatest advances in fine wet screening technology. Derrick Corporation offers panels with openings as fine as 100 microns (140 mesh) and open areas similar to fine, woven wire panels. For example, a 0.15 mm (100 mesh) polyurethane panel has about 35% open area. Fine polyurethane panels can last 10 to 20 times longer than woven wire and are resistant to blinding. Before the advent of fine polyurethane panels, plant designers would avoid using fine screens due to high consumption rate of fine, woven wire screen panels. With the long life typical of polyurethane screen panels, applications once considered impractical are now economically feasible. The last section of this paper sites several examples of the long screen cloth life typical of fme polyurethane screen panels. DRY FINE SCREENING Industrial minerals and sands are typically fine screened to obtain products that meet required size specifications. Fine screens are normally installed to operate after size reduction equipment or after drying operations. Multiple stages of sizing with screening machines operating in series on different floors are commonly required. Mineral products produced using dry fine screening equipment include glass sand, nepheline syenite, olivine sands, limestone, graphite, polypropylene and polyethylene pellets, (others).

Dry screens, properly sized for the intended application, are capable of producing materials meeting stringent size specifications. Examples include undersize material with little, if any, oversize particles and oversize products with very low amounts of undersize. Carefid testing is usually required to select the appropriate type of screening machine, the machine angle, and the optimum screening media for a given application. Types of Dry Screening Machines High frequency vibrating screening machines for fine dry screening applications are normally supplied as single and double-deck machines. Triple deck units are available, but are not normally considered cost effective. It is possible to make two undersize products and one oversize product with a single deck screen. This requires a double compartment undersize hopper; the coarser panel is in the first screen position. Similarly, a double deck unit can be configured to make four products. An example of a single deck machine is shown in Figure 7.

Vibrating screens for fine dry sizing usually operate at 1500 to 3600 cycles per minute with particle acceleration of 3 to 5 G’s. Coarser separations can be made at the lower end of the range and finer separations usually require higher frequency. The angle of screen panels for fine, dry separations can range from 25 to 45 degrees with some applications (fine graphite, for example) requiring even steeper angles. Generally, material is screened at or near its angle of repose. Screens with multiple angles, where the angle decreases down the length of the screen, are commonly used where the presence of near-size material would otherwise cause inefficient screening. Factors Affecting Dry Screening Several factors affect the selection and performance of dry screening machines. Feed characteristics, as well as machine design and operating parameters, will affect the performance of a dry screen. Moisture. The moisture content of the feed to a dry screen can be the most troublesome single factor affecting performance. For fine separations, even 0.5% moisture can cause a screen to either plug or blind. This is due to the fact that the finest particles, due to their high surface area, will

923

Figure 7 Single deck dry screen with double undersize hopper (photo courtesy of Derrick Corporation).

contain more moisture than that of the bulk of the material. The finest particles with higher moisture will agglomerate with other fine particles or stick to larger particles and cause screen openings to plug. This eventually reduces the effective screen area and prevents fines from being able to go through the screen openings. Therefore, feed to a fine, dry screen should be as dry as possible; “bone dry” is not an unreasonable specification. Angle of Repose. Dry, fine screening is generally accomplished with the screening machine positioned at or near the angle of repose of the material being screened. The angle of repose is measured from the horizontal and is one of the important parameters to be determined with starting the selection process for an inclined, vibrating screen. Some materials must be screened at an angle just slightly greater than the angle of repose to obtain the necessary shear and permit fine particles to make their way to the screen surface. Bulk Density and Specific Gravity. Another factor to consider is the loose bulk density of the feed material. Heavier material will fall through a screen more rapidly than lighter material. Particle density, or more specifically, the difference in density between the particle and the surrounding fluid (air) affects the velocity of a particle through the fluid. The surrounding fluid (air) has a lower relative density and restraining force due to drag. Size Distribution. Similar to wet screening, the size distribution of the material fed to a screen, as well as the amount of near size particles, affects both capacity and performance of a dry screening machine. For a given material, as the amount of undersize in the feed increases, more unit area will be required for the fine particles to find their way through the screen openings. Near-size, oversize particles have a tendency to stay close to the screen surface and may inhibit the ability of an undersize particles to pass through an opening. Near-size, undersize particles move through screen openings at a slower rate than finer particles; therefore, more unit area will be required. Particle Shape. In some cases, particle shape is another factor that must be considered. For example, some natural silica sand deposits consist of well-rounded grains. If the diameter of a

924

significant portion o f these grains is close to the opening size of the screen cloth, screen capacity and performance can be affected and special consideration must be give to the screen panel design. Flat particles, such as mica and natural graphite, can also present some special problems. This can usually be overcome by a combination of screen panel design and increasing the machine angle. Feed Box Feed Design And Selection The design of the screening machine feed box is also important. The objective of the feed box is to achieve an even distribution of feed across the width of the screening machine and minimize wear on the screening media. Depending upon the particle size and wear characteristics of the material screened, feed boxes can be simple chutes or more complex assemblies designed to reduce the velocity of the feed stream before it is distributed to the screen cloth. DESIGN CRITERIA Some basic information is required to determine preliminary equipment requirements, operating conditions, and estimated screen performance. For fine wet screening, the following data is required. to the screen, including circulating load, if applicable 1. Total dry mass flow (th) 2. Dry solids specific gravity 3. Screen feed slurry density 4. Anticipated minimum and maximum range of flow and density 5 . Feed particle size distribution 6. Desired separation size 7. Required product specifications, including any downstream constraints on slurry density 8. Any machine layout requirements, including headroom constraints The following data should be known for dry screening applications.

I. 2. 3. 4. 5.

6. 7. 8.

Total dry mass flow (th)to the screen, including circulating load, if applicable Loose bulk density Feed moisture content Angle of repose of material screened Desired separation size and product specifications Temperature of material to be screened Ambient temperature range Any machine layout requirements, including headroom constraints

In addition, it is useful to have a process flow diagram to gain a better understanding of how the screen will be used. EXAMPLES AND BENEFITS OF FINE SCREENING Classification The use of fine screens in grinding circuits has the greatest benefit when there is a significant difference in specific gravity between the valuable and waste minerals. Conventional classification methods, such as a hydrocyclone or screw classifier, separate particles by differences in settling velocity. However, when there is a difference in mineral specific gravities, large gangue or middling particles and finer concentrate particles have similar settling velocities. This results in the misplacement of particles; gangue and middling particles end up in the circuit product and fine, liberated valuable minerals are sent back to the mill and are over ground. The use of fine screens, which separate particles only on the basis of size, can result is significant benefits, both in terms of mill capacity and power consumption per ton.

925

For example, the grinding circuit at an ilmenite producer was originally closed with hydrocyclones. To improve classification efficiency, the cyclones were replaced with fine screens. The circulating load dropped from 300 to 350% to less than 100% and circuit capacity increased approximately 30%. The National Steel Pellet Company (NSPC), an iron ore company in Minnesota, needed to increase capacity and improve product quality. Rather than spend significant capital and time on the installation of additional grinding capacity, NSPC instead decided to modify its secondary grinding circuit to improve classification (Wennen, Nordstrom, and Murr 1997). In the original circuit, mill discharge was pumped to a hydrocyclone and the underflow fed back the mill. Pilot plant tests and computer simulation studies led to the development and installation of a circuit where new feed is pumped to the cyclone and the cyclone underflow directed to the mill. Mill discharge is pumped to vibrating screens equipped with fine polyurethane screen panels. Screen undersize joins the cyclone underflow as circuit product. Screen oversize is fed to magnetic separators and only the concentrate sent back to the mill. The benefits of this new circuit were quite significant. The production rate increased by 30 to 34%. Power consumption per ton decreased by 24%. Furthermore, by minimizing the production of middling particles with improved classification (Bleifuss 1968), the same concentrate grade was produced at coarser grind. Esan’s two feldspar processing plants in Milas, Turkey benefited from the use of high frequency screens and 0.23 to 0.50 mm polyurethane screen surfaces (Guven and Bozdogan 1998). The screens are used to close ball mill circuits and are fed at 48 to 55% solids. This relatively high feed density was required to maintain the screen undersize at 45% prior to froth flotation without the need for an additional dewatering step. Even at this relatively high screen feed density, overall efficiency ranged from 94 to 99%. The use of screens was desired over other classification methods to minimize the production of fines which have a detrimental effect in flotation. The overall economics were further aided with the use of fine mesh polyurethane panels. Typical life of 0.43 mm polyurethane panels at Esan is over 7,200 hours.

Improvement in Concentrate Grade Fine screens are used in numerous iron ore plants to reduce the silica level in final concentrate. By removing particles coarser than 53 to 75 microns (200 to 325 mesh) from final magnetic concentrate with fine wet screens, silica levels are reduced by 1.0 to 1.5 percentage points (Weinert and Salmi 1984; Derrick, Wennen, and Nordstrom 1989). The cost of using fine screens is far less than finer grinding or the use of froth flotation. Originally, sieve bend screens were the industry standard. In the early 1980’s, high frequency vibrating screens with multiple feed points became more popular. Improved Recovery in Gravity Separation The Iron Ore Company of Canada installed single feed point fine wet screens fitted with 0.35 mm polyurethane screen surfaces ahead of an existing spiral plant to recover fine iron from magnetic tailings (Penny 1996). By removing the coarse silica from spiral feed, the spirals could be adjusted to recover more particles in the 425 to 75 micron (40 to 200 mesh), as well as fine hematite that is usually lost. Screening before the spirals was shown to have the potential to increase overall plant iron recovery by up to 2.5%, accounting for production of an additional 500,000 t/y. Recovery and Dewatering of Fine Particles Systems consisting of small diameter hydrocyclones feeding vibrating screens have been developed to recover and dewater 150 to 38 micron (100 to 400 mesh) particles from dilute waste streams at quarries and aggregate facilities (Guven and Kelley 1996). The waste stream, typically the overflow from a screw classifier, is pumped to multiple 102 mm (4 inch) diameter hydrocyclones fed from a circular manifold mounted above a horizontal, high g-force, linear

926

motion vibrating screen. The size separation is actually accomplished by the hydrocyclone and the high density hydrocyclone underflow is directed to the screen for dewatering. Since there is minimal free water in the viscous, 60 to 70% solids hydrocyclone underflow stream, screening efficiency is poor, which is desired. The high g-force screens, equipped with 0.5 mm polyurethane screen surfaces, transform the viscous cyclone underflow into a conveyable or stackable material. To completely eliminate the waste pond, hydrocyclone underflow can be further thickened and dewatered. Interstage Screen for Gold Processing The development of an efficient and reliable screening system has been an essential component of the CIP/CIL process for gold extraction (Reinhofer 1988). Known as the interstage screen, this linear motion machine fitted with fine polyurethane screen panels retains the coarser, activated carbon while allowing large volumes of finely ground ore to pass to the next tank in a cascading series. Since carbon has a relatively low specific gravity, it was critical that the separation be based solely on particle size rather than hydraulic classification. Improved Flotation Performance Samitri originally produced 3.0 to 3.5% silica iron ore pellet feed at its Alegria Mine in Brazil. To meet new customer requirements for concentrate containing less that 1.O% silica, Samitri developed additional process steps consisting of fine screening followed by flotation (Valine et a1 1996). Flotation alone was not successful since the feed contained about 35% plus 0.15 mm (100 mesh). Laboratory and pilot testing determined that flotation feed must contain less than 5 to 10% plus 0.15 mm (100 mesh) to meet concentrate grade requirements. In 1992, Samitri therefore installed 24 multiple feed point fine screens to remove plus 0.15 mm (100 mesh) particles from flotation feed. For the first year of operation, fine woven wire screen surfaces were used at Samitri. In 1994, polyurethane screen surfaces with 0.18 mm (80 mesh) openings were installed in place of the fine wire panels and screen operating costs dropped significantly. While the woven wire screen surfaces lasted an average of 250 hours, the polyurethane screens had an average life of over 7,000 hours or almost 1 year of operation. CONCLUSION Recent technical advances in screening machine design and the development of polyurethane screen surfaces as fine as 100 microns (140 mesh) have made fine screening a more practical option for mineral processing operations. The metallurgical benefits that result from true particle sizing can now be obtained with high capacity machines equipped with long life, non-blinding, polyurethane screen surfaces. REFERENCES Matthews, C.W. 1985. General Classes of Screens. In SME Mineral Processing Handbook, ed. N.L. Weiss. 3E-1. Wennen, J.E., W.J. Nordstrom, and D.L. Murr. 1997. National Steel Pellet Company’s Secondary Grinding Circuit Modifications. Comminution Practices. 19-25. Bleifuss, R.L. 1968. The Mineralogy of Taconite Products as Related to the Augmentation of Magnetic Middlings. Proceedings of the 41”’ Annual Meeting of the Minnesota Section AIME and the 29lhAnnual Mining Symposium. Guven, O.N. and I. Bozdogan 1998. High Speed Screening Technology in a Grinding and Flotation Circuit. Innovations in Mineral and Coal Processing - Proceedings of the 7’h International Mineral Processing Symposium. Istanbul, Turkey. Weinert, J.D. and R.W. Salmi. 1984. Recent Applications of Fine Screens at Minntac. Skillings Mining Review. 23-Jun-1984.

927

Derrick, H.W., J.E. Wennen, and W.J.Nordstrom. 1989. Proceedings of the 50thAnnual Universiv of Minnesota Mining Symposium. Penny, B. 1996. An Integrated Approach to Iron Ore Recovery at the Iron Ore Company of Canada. 281hAnnual Operator’s Conference of the Canadian Mineral Processors. 2 13-225. Guven, O.N. and C.P. Kelley. 1996. Dewatering of Fine Particle Waste Using the Derrick Dewatering System. Changing Scopes in Mineral Processing - Proceedings of the 6lh International Mineral Processing Symposium. Kusadasi, Turkey. Reinhofer, R. 1988. The Design and Development of the Derrick CIP/CIL Interstage Screen. Intermountain Mining and Processing Operators Symposium. Elko, Nevada. Valine, S.B., J.R.V. Futado, G. Martins, D.L.V. Policarpo, F.C. da Silva Quintao. 1997. Process Improvements at Samitri, Brazil. Mining Engineering. 49:4.

928

THE USE OF' HINDERED SETTLERS TO IMPROVE IRON ORE GRAVITY CONCENTRATION CIRCUITS Steve Hearn'

ABSTRACT Hindered settlers or hydraulic classifiers have been used for many years for the classification of industrial sands and other industrial minerals to produce closely sized products for market needs. These units varied from multi compartment free settling units like the Eagle Iron classifier, to perforated plate classifiers. Such devices used rising currents of water with limited control techniques, which caused varied efficiency in size separation. Later generations of classifiers utilized the principle of hindered settling to achieve more closely sized separations. In recent years, this technology has been successfully applied to augment gravity concentration circuits so that the use of technologies such as spiral concentrators would be more effective. It is important to note that when treating heavy minerals using a hindered settler the device not only classifies but also performs density separation, upgrading according to specific gravity. Hindered settlers are therefore often referred to as a density separator. INTRODUCTION The use of a rudimentary type of hindered settler at Quebec Cartier Mine's pellet plant in Sept Iles, Quebec, generated early interest in the device as a concentrate cleaner. The units were installed to upgrade pelletiser feed by reducing silica in spiral concentrate that had been shipped some 250 km from the Mount Wright mine. This application of hindered settling as a process complemented gravity spirals and allowed the operator to achieve separations that spirals alone could not. This paper will discuss spiral operation and the limits of such devices, and describe the principle of hindered settling. I will show how these methods can be retro-fitted into existing operations to result in practical flowsheets that are easier to operate and cheaper to install. To illustrate these techniques, this paper will cover the industrial application of such methods using recent examples from the iron ore industry in Sweden and Canada. The improvements are illustrated and each example is given as a case history. PRINCIPLES OF OPERATION OF GRAVITY SEPARATION DEVICES Spiral concentrators The lineage of the spiral can be traced back, beyond the original cast iron Humphreys spiral, to the mining industry folklore of cut-up and stretched car tires. Spirals are probably the lowest capital cost mineral concentrator available today. Combined with their low operating cost, they have continued widespread use. The leading suppliers of spirals are Outokumpu with the Carpco@ spiral, Mineral Technologies from Australia and Multotech from South Africa. All three companies agree that urethane lined helices make the best surfaces for a separating spiral, as it provides wear life together with enough friction to aid separation of the different mineral components. The fiberglass backing provides structural support, being in turn wound around a center support column. Center columns have been used for carrying concentrates and, at times, wash water. Manufactures are constantly vying with each other to offer improvements that claim to squeeze out an extra point or

' OUTOKUMPU TECHNOLOGY INC. Physical Separation Division, Jacksonville, Florida USA

929

two of mineral recovery from an ore. It is quite possible that all variations of helix pitch, profile and diameter have been tested over the years. Some went to extremes, such as the large 2-m diameter units made by the Russians and Chinese a few years ago and the Budin spiral where the spiral diameter increased as the slurry descended. Modem conventional spirals generally range from 20 - 40 ins. (500 - 100 mm) in diameter, 13 - 18 inch (330 - 460 mm) pitch, and up to 7 turns. Given the consensus on overall geometry, the search to differentiate this product leads to improvements such as re-pulpers (Figure 1) that aid mixing of the slurry, particularly after concentrate removal or intermediate high-grade concentrate take-offs, or splitters that aid quick removal of separated minerals. (Figure 2) In essence, spiral designers strive to make up for the inherent shortcomings of these devices. The spiral is overall effective for mineral recovery, but with limitations regarding particle size distribution of feed and its effect on feed concentration efficiency. One area of improvement exploration is in reducing losses of fines (generally defined as minus 75-micron material).

REPULPFA

Figure 1 Repulper To Direct Water From Outer Spiral Wall To Concentrate Collection Area

930

Figure 2 Intermediate High-grade Concentrate Splitter The spiral operates as a flowing film separator, which is usually loaded well beyond the appearance of a film separator. Valuable fines are trapped either in regions of high velocity flow (at the spiral helix outer wall) or beneath coarse material. Coarse gangue cannot easily be released from between layers of heavy mineral, nor can these gangue particles be persuaded to travel to the outer wall where the excess water and other gangue or light particles are transported (Figure 3). Spiral design improvements described above all attempt to redress these operational shortcomings. In summary, spiral performance is best, in what can be described as a rougher application, where the feed concentration of heavy minerals is low (for example less than lo%.) The main objective of such processing is recovery, rather than achieving a high concentrate-upgrading ratio. Spiral tailings may well be final tailings such as in stage I primary concentration of ilmenite from mineral sands. Spiral operation under these circumstances can be relatively hands-off from the operators’ point of view and, therefore, represents a truly low cost processing step. Conversely, where spirals are treating high concentrations of heavy mineral, recovery is often poor, as mineral entrapment can occur leading to the need for additional stages of spirals to clean concentrates or scavenge tailings. High loadings of middling also characterize these conditions often leading to high re-circulating loads. Additional wash water may be needed to aid flow in these high-density situations. In any case, the net result is generally a high level of required operator attention with associated costs and the possibility of errors in judgment. Hence, any complementary process, which can encourage the use of spirals in their area of comparative advantage, must be an improvement.

931

Figure 3 Separation Regions Across a Spiral Profile Density Separators It is generally accepted that when a rising current of water is introduced across the bottom of a classifying vessel, a sand or other mineral can be expanded into a state of teeter. The water introduced at the bottom of the column has the greatest velocity. When the falling particles achieve the same velocity as the upward current, they will fall no more. The continuous motion of particles following paths of less resistance to lower pressure regions (in the upper part of the column) and being replaced by other particles is said to be a state of teeter. In this teetered state, mineral grains will classify themselves so that the coarse grains report to the bottom of the column, where high velocities of water flow through the interstices, staying relatively close to each other. The finer particles disperse to the higher levels of the column where they stay more openly suspended(Figure 4).

932

Figure 4 Principle of Hindered Settling This teetered suspension is, in effect, a dense media of solids with an apparent high specific gravity (s.g.) and can be used to float off other solids, such as silica, that are light enough to be supported by the suspension regardless of particle size. (Figure 5 )

933

Hindered settling classifiers are invariably provided with an overflow for the discharge of the finer or lighter material from the classification, which helps to regulate the overall depth of the teetering suspension. A pressure sensitive device may then be inserted into the teetering pulp to give an indication of its specific gravity, which can be used to control the separation taking place in the device. For any upward current of water, the specific gravity is indicative of the average particle size of the sands. A pressure sensitive device may be used to provide a variable signal to operate a valve that controls the discharge of the coarseheavy material that has accumulated at the bottom of the column. If insufficient material is being discharged, coarser particles will accumulate above the water spray pipes. As they will be lying closer together, these coarse particles will be at a higher pulp gravity and will give a higher pressure reading on the sensor. The control loop, in an effort to prevent this, will increase the opening of the discharge valve. If, on the other hand, too much material is being discharged only the finer particles will be left above the sensing position and, as these will be more dispersed, they will give a reducing pressure as the pulp gravity falls, thus closing the discharge valve. Hindered settling classifier tanks have been made in many shapes and sizes. The majority of developments have been variations of the multi-spigot hydrosizer where the prime function was the preparation of mineral sands for treatment on shaking tables. Single spigot machines have been developed by a number of mineral processing companies, originally intended for the production of industrial sand. These devices usually followed the principles used in the multi-spigot units with no new developments that would allow larger tonnage to be handled efficiently. The main failings of the single spigot, hindered settling classifiers, were as follows: 1. The method of distributing the upward current water through a perforated plate from a water box from beneath the teetering column was effective while the teetering column was relatively static, but suffered once the discharge valve opened to let out a high tonnage of sand. 2. When the coarse product was discharged directly from the teetering sands, a high proportion of water was discharged as well. As this water came from the pre-set supply for the upward current, fluctuations in discharge rate greatly affected the quantity of remaining water, which caused variations in sands dispersion. 3. 3. When the coarse product was discharged directly from the teetering sands, anything other than very small tonnages created a short circuit from the feed to the discharge. (Figure 6) I y

Normal Operation

Separator filling with Settled solid5

Figure 6

934

Separator Flushing

To overcome this limitation machines were designed with multiple discharge valves to reduce the volume being discharged by any one position and thus reduce the limitation of discharging from the teetered state. The Density Separator is a hindered settling classifier designed specifically to overcome the limitations referred to above and intended to bring the advantages of hindered settling into the realm of high tonnage process plants. The remedy is distributing the upward current of water through an arrangement of perforated spray pipes spaced across the tank at the bottom of the sorting column. This allows the coarse, classified material to collapse through the spaces between the spray pipes over the whole area of the tank, thus achieving an infinite number of discharge positions from the teetering sands. Beneath the level of the spray pipes, the tank is closed by a dewatering cone of sufficient slope to allow the settled sands to freely flow to the discharge valve at the apex. The settled sands are discharged with a minimum of entrained water and, therefore, variations in the tonnage of material discharged do not greatly affect the true upward current. (Figure 7)

PERIPHERAL

-0vERFLow LAUNER

FEEDWELL

CAPACITATIVE PRESSURE SENSOR

Figure 7 Detail Of Typical Density Separator The Density Separator is designed to produce an extremely sharp partition curve when classifying particles between 20 mesh (840 microns) and 200 mesh (74microns). During the classifying process, (which is the hindered settling of larger, denser particles and the floating of lighter, finer particles) the underflow, consisting of faster settling particles, is effectively dewatered. By measuring the density in the hindered settling region and comparing this density with known values, discharge of the underflow is accordingly regulated.

935

This effect can be used to: 1. Classify particles by size. 2. Upgrade a feed consisting of two or more groups of particles (often minerals) of differing specific gravity. Heavier faster settling material builds up a heavy media in the vessel and, in effect, allows the lighter material to float on top, eventually overflowing. This effect can be used to upgrade minerals containing lighter fraction gangue minerals.

The achieved product is dependent on the composition of the feed material and the density of the individual particles of the feed. The separator can be fed with dry or wet material. Empirical models have been developed to accurately predict classifier operation with a given a feed size distribution. This type of model has been extended, using the specific gravity of minerals in a slurry feed, to predict concentration in a Density Separator. For ease of manufacturing the water distribution system, Density Separators are usually made in square or rectangular cross section. The added advantage is a greater classifying area per unit floor area in a process plant than that of circular tanks. Units larger than 2.44m x 2.44m (8’x8’) are divided up to avoid excessively steep discharge cones. Hence a 12’x12’ (3.6m x 3.6m) unit for example, would be, in effect, four 6x6 (1.8m x 1.8m) units grouped together. The controller will synchronize the operation of the four discharge valves. (Figure 8)

Single Valve

Multiple Valvcs

Figure 8 Controller Operating Single or Multiple Discharge Valves CASE STUDIES

LKAB Iron Ore Mine - Sweden When LKAB decided to re-investigate the exploitation of hematite ores (it was until that time only mining the magnetite ores), the company surveyed international mineral processing equipment suppliers for available solutions. Many techniques were considered and tested, from gravity, including centrifugal concentrators, to flotation and magnetic separation. Spirals were reviewed, but suffered in Scandinavia from “bad press.” The perception, perhaps from the Canadian operations in Labrador and Quebec, was that many units would be needed, resulting in a very labor-intensive operation, equaling a high cost recovery system given labor costs in Scandinavia.

936

The company had heard of the successful employment of Density Separation in iron ore at an installation in India (at Kudremukh Iron Ore Company, Ltd) and decided the technology was worthy of further investigation. Samples were initially sent to Jacksonville, Florida, for testing in a laboratory-scale Floatex@unit. Initial results lead to the installation of a model 460 (about 0.5m x 0.5m at the overflow lip) at the company's Malmberget facility in Sweden (Figure 9). The unit was installed in the company's pilot plant operating at a feed rate of about 3 tph. This produced quantities of upgraded material used for other tests including grinding.

Figure 9 LKAB Floatex"Density Separator Installed at Malmberget This ore liberated at around minus I mm making it a very suitable feed material for the hindered settler. The silica content of the coarser underflow product was reduced to almost the target level of around I%. The overflow product contained the majority of the gangue together with finer hematite values. A screening process was an option for the over flow but not favorably regarded by the company, being perceived as a high maintenance step, so spirals were, somewhat reluctantly, considered. The now leaner and well prepared spiral feed proved an excellent candidate for upgrading. In addition, because the feed tonnage to the spirals was reduced, and hence less spirals were required, plant management concluded that their reduced operator staff could manage a spiral circuit operation. During testing, visual inspection of the spiral concentrate product, still containing 7%SiO,, showed that the spiral had to a considerable degree, size sorted the feed, thus making it again a good candidate for further upgrading in a hindered settler. Testing this material in a hindered settler, demonstrated that by closely controlling the teeter zone, virtually all the remaining silica could be rejected to the underflow product. The resultant overflow, now free of coarse silica, could be blended with the coarser primary settler concentrate (underflow) to make acceptable product. (Figure 10)

937

.....................................................................

..........................................................................................................................

Floatex Prediction 0 1 1 Sire Separation For LKAB Hematite Ore .....

..........................

. . . . . . . . . . . . . . . .

.................................

Figure 10

As testing continued, the density separator proved additionally beneficial for retreating primary separator underflow in a cleaning stage. The retreated underflow product, whilst only displaying a minor reduction in SiO,, showed appreciable lowering of the P,O, content. Acceptable P20,content is essential when selling this ore to steel makers. While the overflow contained even coarser hematite, it was still easily upgraded in combination with the primary overflow using the existing spirals. The eventual gravity circuit operated at 216 tph feed, achieved the performance requirements of > 80% iron recovery, and an overall SiO, reduction to < 1% (Table 1). Operation was easy: primary hindered settler underflow feeds directly into the secondary hindered settler; the underflow density providing an ideal feed density for the process whilst eliminating a pumping stage. Spiral operation is simplified (although a wash water spiral had to be added) because the function of the spiral is to produce a throw away tail at < 20% Fe. Concentrate upgrading was achieved with the tertiary Floatex. (Figure 11)

Capacity (tlh) ”

hgree of enrichment

First s&ge separation Noni. 216 Max.246 Min. 186 > 64.5% Fe

Second b w e separation

Third stage separation

- 132

- 52

2 68%Fe

2 68% Fe

Or 68% in total over the two stager, 5 0.15% P < 1.0% Si02

-___________

Iron recovery (each stngc) Iron recovery (compared to incoming Fe-content) Solid9 concentration in underflow (% by

> ail%

> 81%

;t 61%

> 83% > 79

Average 2 76

938

Feed 48% Fe 15% SiO, I

I

i

Cyclone Overflow

I

PRIMARY FLOATEX' Density

, Carpco" Spiral Concentrators

Separator

SECONDARY FLOATEC Density Separator

P

Tailings

I L < 2 0 % Fe

p" ~

i.. ...

..

loss

TERTIARY FLOATEX' Density Separator

I

i

Final Gravity Concentrate 68% Fe 1% sio,

Figure 11 Gravimetric Separation Flowsheet Swedish Iron Ore Plant Using the automatic density control capabilities of these hindered settlers means that a consistent density underflow product is achievable, which in turn defines the silica content in the final product. Feed grade variability is not translated into product variability - the primary hindered settler produces a consistent underflow and excess gangue is sent to the spiral circuit where it is easily rejected. LKAB determined that spirals, augmented with hindered settlers in an innovative circuit design were a cost effective process approach.

Iron Ore Company of Canada, Labrador When the Iron Ore Company (IOC) was acquired by North Mining of Australia in 1998, it embarked on an extensive audit of their concentrator. IOC had, until that time, relied on the exclusive use of thousands of wash water spirals for the recovery of their hematitic ores. Except for the original Humphreys cast iron models being supplemented with Mineral Deposits fiberglass model WW spirals, there had been few fundamental changes to the circuit in nearly 40 years of life. Several stages of spirals together with a variable feed ore led to a complex circuit making it difficult to control and ensure consistent product quality. Practically speaking, controlling this quantity of spirals is almost impossible. A simplified circuit was deemed highly desirable. (Figure 12)

939

Crude Feed

(Wet Mill Screen Undersix)

Cyclone U/F

Figure 12 Current Concentrator Spiral Plant Flowsheet During their investigation of available alternate methods, the concentrator staff looked extensively using Density Separators in different locations in the circuit, to find the most beneficial position of such technology. A pilot model Floatex 460 was acquired for the concentrator at IOC's pilot plant-Carol Lake and was tested in conjunction with a number of spiral stages. They investigated hindered settling to treat cleaner spiral concentrates (which is in affect at the adjoining Wabush Mine and the aforementioned Kudremukh Iron Ore Company, Ltd in India) but found it unsatisfactory Investigations moved to using hindered settling as a primary classifier to provide a sized feed for subsequent gravity treatment and allow a more "customized " approach to the task of selecting spirals for further silica reduction. Nine different flow sheet variations were investigated before the so-called CUP (Concentrator Upgrade Program) was selected. A nominal separation size of 1 15 microns was selected, with the criterion being to maintain a 35/65 weight split to underflow to create controllability of the subsequent spiral circuits. (Figure 13) Hence, control relied heavily on upward current water addition and little on actual density sensing. Independent process controllers for each hindered settler unit were replaced by a DCS system, monitoring all machines in the control room.

940

CUP Hindered Settling Performance Curve 4.5 tonne& vs. 5.5 tonne& Loading M

a

Za

P

d=

EEI

j:

m

w 0

m

I

m

1D

--(-I

Figure 13 The "ww" wash water spirals, being fairly new, were retained for duty treating. The coarse fraction (the settler underflow) and two stages of spirals were required to maintain optimum product grade of 65.6% Fe. Losses are identified from this coarse circuit, as the spirals do not recover the coarsest particles. As these are generally unliberated middlings, their recovery would dilute the final grade to below an acceptable level. This apparent inefficiency is allowed to occur as the net gain, in fines re covey, with the new circuit, more than offsets the loss. The classifier overflow presented a new challenge as the feed characteristic was completely changed. The impact lay in the material, at approximately 36% Fe and averaging 90% minus 150 microns, being an ideal candidate for feeding wash water-less spirals. One major benefits for IOC, apart from easier operation, was water reduction. Using a wash water-less design on just one stage was predicted to save IOC some 20 000 gpm water. CUP Fine Rougher Spiral Performanee Curves Weigbt Recovery vs. Fe Recovery & Weight Recovery vs. Corn. Fe Grade

ioI

mn

.((Lo

mn

IDd

M

WIsbtmc-UYfx)

Figure 14 Again, the on-site pilot plant was key in evaluating various spiral models. One feature demanded by IOC operating personnel, was a splitter design that would retain its position even when the spiral was subjected to the frequent and necessary wash down. A unique patented design

94 1

using a locking "outboard" adjusting handle was developed by Outokumpu specifically for IOC's needs and found ready acceptance by their technicians. The primary fines spiral tailings at 18% Fe exit the gravity circuit for scavenging (using magnetic separation) in a separate magnetite circuit. The concentrates, not yet sufficiently upgraded, were treated with secondary wash water spirals. Investigation on further wash water reduction, and possible elimination was halted since a number of the newer spirals had been released from duty when the circuit changed. These were re-configured as secondary spirals or cleaners in the fines circuit. Another major effect was eliminating the need to deslime gravity circuit feed. Mill discharge, which is classifying screen underflow at nominal minus 1.4 mm, is pumped directly to the Floatex Density Separators. This eliminates the cyclones previously used to prepare feed to the spiral circuit. IOC also determined they could feed directly from the classifier overflow to the spiral circuit, again obviating densifying cyclones, this time ahead of the fines spirals. (Figure 15)

,

c-

Figure 15 IOC's New Iron Ore Recovery Circuit

942

The concentrator will be refurbished i n three mill lines with incoming feed at a maximum 6000 tph. The primary Floatex units will be installed as six 8 ft x 8ft (2.44m x 2.44m) units arranged as twins. The distribution layout for the classifiers as well as the spirals was arranged for maximum flexibility, so that units can easily be taken off line as feed grades or product requirements change. (Figure 16)

Figure 16 Density Separator Similar to the Primary Classifiers at IOC's Carol lake Concentrator

ACKNOWLEDGEMENTS The author would like to acknowledge misty Trull and Erik Spiller, also at Outokumpu, for their patient editing and constructive contributions. REFERENCES Elder, J., S.Hearn, P.Vekatraman. Application of Floatex Density Separator for the heavy minerals sands industry. 1999 SAIMM Heavy Minerals Conference. Johannesburg RSA, Holland-Batt, A.B. Some Design Considerations for Spiral Separators. 1995 Mineral Technologies. Littler, A. Trans. Inst. Min Metall. (Sect.C. Min Process.Extr.Met.) 1986, vo1.95, pp CI 33- 138. McKnight,K., N.Stouffer, J. Domenico and M.Mankosa , Recovery of Zircon and other economic minerals from wet gravity tailings using Floatex Density Separator. 1996 SME Annual Meeting, Phoenix Az. Nevens, M. Concentrator upgrade program at the Iron Ore Company of Canada. Proceedings 2002 , Canadian Mineral Processors meeting, Ottawa. Wills, B.A. Introduction to the Practical Aspects of Ore Treatment and Mineral Recovery. Sixth Edition, 1997, Butterworth and Heinemann

943

Types and Characteristics of Gravity Separation and Flowsheets Richard 0 Burt’

ABSTRACT Gravity separation has been an established method of concentration for well over two thousand years. Old as the process is, it is far from obsolete, and continues to be used in the processing of a wide variety of minerals. This paper introduces several gravity concentration papers by discussing basic technology, the range of equipment types available, and their main operational parameters. It includes some typical flowsheets; from simple to complex, and from those that use only gravity to those where it is but a part. INTRODUCTION The principles of gravity concentration have been well known for well over two thousand years: for most of that time it was, with hand picking, the only processes available for the separation of minerals. While alternative processes now predominate in some sectors of mineral processing, gravity concentration remains a vibrant alternative, one that always warrants consideration. Indeed, as much - if not more - tonnage continues to be treated by gravity than other processes combined. Why is this? Why use ‘old’ technology, when more modem technologies appear more efficient both in terms of recovery and throughput? However, maximising recovery is not the primary goal of any venture: that should be maximising the economic performance, which often is not the same thing. As this and succeeding papers will highlight, many of the recent developments in gravity concentration equipment have been related to increasing unit capacity and efficiency. Consequently, gravity concentration generally has a lower installed cost per tonne than competing technologies. Furthermore, the lack of requirement for reagents not only ensures a low operating cost of the process, but also minimises the environmental impact of the operation, with the absence of organic chemicals and their reaction products. Gravity tailings are also, generally, coarser, with the subsequent advantages to tailing dam operation. However, one of the great advantages of gravity concentration, in terms of its economic benefit, is its ability to effectively treat coarse materials, thereby allowing for the rejection of barren waste at a coarse a size as possible. This can have a major impact on the size of the subsequent plant. Furthermore, as the rejected waste can also be the harder portion of the ores, the actual reduction to the grinding plant alone in terms of k W t of ore can be substantial. The resultant capital and operating cost savings will generally far outweigh the often marginal loss of values in the coarse waste. In other cases, while rejection of a coarse, barren, tail is not practical, it can be shown that the production of a coarse concentrate, with subsequent processes acting as ‘scavengers’, can have a positive impact on overall recovery and economics. (Bath Duncan and Rudolph 1973). Coarser concentrates are also easier to filter. A later section will highlight the main areas of application for gravity concentration: suffice here to note that the simplicity of the gravity is such that it is the only logical technology for many of the small, unsophisticated artisanal type plants in the developing world.

’ Cabot Corporation, Elora, Canada 947

Notwithstanding the many advances in equipment design, the chief problem with gravity concentration remains the recovery of the finest fractions. Below about 50 microns, recovery decreases with decreasing particle size, and the practical size limit for gravity remains about 10 micrometers, even for the separation of particles with a substantially different density. Consequently, for many complex ores, where the liberation size is such that rejection of a barren tail is impractical, and grinding to below 100 microns is mandatory, the applicability of gravity concentration is limited to areas where other processes are even less efficient. Gravity concentration suffers the same disadvantages as most other, competing, processes the need for removal of slimes prior to treatment, stage concentration. While some devices (cones, spirals for example) dynamic loads are less than competing processes, with others - especially the shaking table -the dynamic loads make structural integrity paramount.

SOME BASICS Gravity concentration can be defined as the separation of two or more minerals of different specific gravity by their relative movement in response to the force of gravity and one or more other forces, one of which is generally the resistance to motion by a viscous fluid. Factors that are important in determining the relative movement of a particle in the fluid include the specific gravity, weight, size and particle shape, not only in absolute terms but also relative to all other particles in the system. It is a characteristic of gravity separation devices that, in order for the particles to move relative to one another that they must, at some stage of the process, be held slightly apart from each other, or ‘dilated’ by a superimposed force. A rough guide of the potential efficiency of any separation can be gained from a simple relationship, known as the concentration criterion

where

A,, Specific gravity of the heavy mineral A, Specific gravity of the light mineral Af Fluid density

The finer the particle size, the more difficult the separation: consequently the higher the required concentration criterion value. The impact of particle size is understandable when one considers that particle mass decreases, and the number of particles per unit of mass correspondingly increases, at the cube of the decrease in particle size, thereby rapidly increasing the number of individual particles of rapidly decreasing mass that have to be separated. Four mechanisms essentially explain the operation of gravity concentration devices: density; stratification; flowing film; and horizontal shear. Increasing the gravitational force enhances the intensity of the above mechanisms; it is not a different mechanism in itself. Density: The suspending fluid has an apparent (or actual) density in between that of the minerals to be separated, such that one fraction will have a net negative buoyancy and will sink while the other will have a net positive buoyancy and will float. Heavy medium separation is the obvious example: this will be dealt with elsewhere. It is arguably the most efficient of all processes. Strattfication: The various mineral constituents are stratified by being subjected to an intermittent fluidisation in a vertical plane. Jigs are the prime example of a stratification device, although the action of a shaking table behind the riffles is another. Flowing film: The various constituents are separated by their relative movement through a stream, which is flowing down an inclined plane under the influence of gravity. Flowing film concentration is probably the oldest mechanism applied. Sluices, cones and even the spiral are examples.

948

Horizontal Shear: The various constituents are stratified by superimposing on the flowing film a horizontal shear force, which may be continuous or discontinuous. Shaking tables, the Crossbelt Concentrator and some centrifugal separators are examples. While these mechanisms are common to apparently diverse devices, no device can be fully described by any single mechanism: a combination of two or more mechanisms is generally required. In recent years, there has been considerable activity with regard to the mathematical modelling of both these general concentration mechanisms and of the specific equipment involved: citations in the technical literature are extensive. These models have undoubtedly led to a better understanding of the underlying principles, and of the ability to predetermine equipment performance; specific examples will be discussed in the forthcoming papers. In addition, circuit modelling software allows for circuit design and performance to be carried out with a good degree of confidence. Notwithstanding these advances, as with other mineral processing options, except in the very simplest of separations, effective mineral characterisation and other basic testwork is a prerequisite for successful circuit design. Where head grade warrants, mineralogical examination can be carried out on samples of the ore; however, in other cases it is often advisable to upgrade the ore to the point where reasonable characterisation can be made on small quantities of material. Heavy liquid analysis is one of the most powerful tools for gravity concentration testwork. Sequential separations of a sample of the ore ground to liberation, carried out on a range of closely spaced increasing (or decreasing) fluid densities will permit full characterisation of a separation. Alternatively, separation on sized fractions of the ore, at two, or maybe three fluid densities, will often provide sufficient information to enable not only accurate prediction of liberation size, but also in many cases for circuit design. (Burt 1984) For higher density separations (>3.3) it is now often necessary to use a unit such as the Magstream Separator, as the health risks related to the higher density fluids are generally no longer acceptable. The results from one such, hypothetical, analysis is shown in figure 1. 100 v)

L

80

Q)

r u0

60

.-ws 3 a

40

si

20

L

6

0 0

500

1000

1500

2000

Size, microns

Figure 1: example of heavy liquid separation for liberation study In this example, the liberation size at which a rejectable tailing (floats at 2.9) and that at which As, apart fiom the the values are optimally liberated (sinks at 4) are significantly different. obvious potential for cost savings, the efficiency of gravity concentration decreases with

949

decreasing particle size, optimal flowsheet design would incorporate stage grinding, producing a rejectable tailing at about 1 mm, as well as a rougher concentrate. This latter would require stage re-grinding to about 0.3 mm to optimise final concentrate grade. The 10% of values reporting to the 4 + 2 . 9 s.g. fraction even at the finest sizes could suggest some degree of fine locking with an mineral of intermediate specific gravity, or even a lower spedic gravity value. Mineralogical examination would provide the answer. UNIT PROCESSES The effectiveness by which process engineers utilise the interactions between the various mechanisms involved in gravity concentration, and the particular mineral suite at hand, determines the effectiveness of the separation. The overall size range that can be treated by gravity concentration devices is larger than with any other process. The top size is theoretically that at which the different minerals are sufficiently liberated from each other for there to be a density difference; however, the practical top size is probably 500 mm. The practical lower cut-off is about 10 micrometers. Likewise, unit capacity of the equipment suitable for recovery at different size ranges also decreases with decreasing particle size. No one device developed to date is capable of effectively treating the fill size range, and some form of feed preparation is, therefore, normally required. Units that are essentially mass classification devises, such as Jigs, Reichert Cones, the Knelson and Falcon centrifugal units are capable of treating a fairly wide range of sizes, but some loss of efficiency does occur at the top, middle and bottom of the size range. Serial gravity concentration is practical with these units. Other types, relying primarily on reverse classification or flowing film, such as tables, spirals, Multi-Gravity Separator (MGS) operate significantly better on closely size or classified feed, requiring parallel circuits (figure 2) Feed

Feed

I Sizing

Figure 2: Series gravity and parallel circuits

There are several factors that may determine the choice of the correct device for a specific separation. These include: the duty required; the size range and shape of the particles to be separated; the throughput required; the efficiency expected; and, all other things being equal, the unit cost - both capital and operating - of the device. In terms of duty, some devices, especially higher tonnage units and those with an inherently lower enrichment ratio are ideally suited to roughing or scavenging duty, while others, with inherently lower throughput, may be suitable primarily as cleaning devices. For example it would be more common to install rougher/scavenger/cleaner spiral circuits, with final concentrate cleanup with shaking tables, than to install all-spiral or all-table plants. All concentrating units, be they gravity or other types, require: a means of introducing the feed stream to the unit; a means of segregating the mineral constituents; and a means thereafter of separating and separately discharging the discrete streams. Whilst apparently a comparatively simple requirement, the efficiency with which different units, and different manufactures, address all three requirements determines which units succeed and which eventually fail.

950

Jigs For coarse concentration devices, especially as roughers and scavengers, the Jig remains of major importance, particularly in coal preparation. Jigs sort individual particles into layers of increasing density within a relatively thick ‘bed’, by intermittent fluidisation in a vertical plane. All machines consist essentially of a tank filled with a fluid, generally water, with a horizontal, or slightly inclined jig screen, upon which the jig ‘bed’ a ‘ragging’ of intermediate density particles is supported. Fluidisation of the bed, and ragging, is effected by pulsation of the jig screen itself, or by the cyclical upward and/or downward flow of the fluid caused by an external mechanism, including mechanical or hydraulic, or by pulsation by air or water. The characteristic of the cyclical pulsation is one of the key elements of effective jigging: the actual pattern used is at least partially dependent upon which of the several, sometimes mutually exclusive, theories of the mechanism of separation the jig manufacturer has determined to be of greatest impact. The heavies fractions are removed from the bed either ‘through the screen’, or ‘over the screen’ by appropriate cutter. The majority of jigs, many of which are reviewed by Cope (2000), are rectangular; however some units are radial, such as the Cleveland Jig, first used in Brazil for the recovery of diamonds,, or trapezoidal, such as the IHC Jig. One novel approach is the Australian invented InLine Pressure Jig (IPG). Not only does this radial unit have a moving screen, it is totally enclosed, allowing it to installed within circulating load circuits without additional pumping. (Gray 1997) Sluices and Pinched Sluices: Notwithstanding the efficiency of jigs, sluice boxes, in use for over a millennium, are still widely used particularly for gold recovery, both in the USA and Russia, where many mechanised configurations have been developed, as well in some of South-east Asia’s older tin plants, where they are also known as palongs. They also remain the unit of choice for artisanal miners, where the sluice ‘box’ is often nothing more than a long, winding, channel hewn through the rock, or it is a relatively short unit built from last month’s tailings. (figure 3).

Figure 3: The simplest of them all - a ground sluice in Nigeria. While such sluices are generally incapable of recovering fine material (less than 150 micron) in the majority of applications such fines generally account for a minor to insignificant portion of the mineralization. Settled heavies need to be manually removed from the unit.

951

More advanced are pinched sluices, which includes the Reichert Cone. Essentially flowing film devices designed originally for beach sands treatment in Australia, they utilise large surface area to achieve high capacity: in the case of the Reichert Cone, up to 300 tph (Holland-Batt 1998). Essentially, as the film passes through the ‘pinch’ or in the case of the cone, flows inward across the cone surface, the stratified film thickens to the point where it is practical to peel the lower, heavies enhanced, layer from the upper, heavies depleted, layer. Any single cut is relatively inefficient and the key to the success of these units is the ability to multiplex the process, often in quite complex formats, in one machine. For example, at the Kemi chromite mine in Finland, roughing units incorporate twelve concentrating segments and the cleaner units another sixteen. (Ruokonen et al 1998). Cones, and sluices, require a high density, and deslimed, feed slurry for optimum performance: however, they are ideally suited to modelling (e.g. King 2000), and as such are often highly automated. Spiral Concentrators Spirals have evolved dramatically since the first experimental units made in the 1940’s by Humphreys out of reject automobile tyres; modelling now allows manufacturers to predetermine which of the many model types available is best suited to a specific application. Nevertheless, the principles of design are the generally same: a multi-turn smooth bottomed helical sluice of modified semi-circular cross section, with or without an additional wash-water channel, and with either a series of concentrate ports down the sluice or splitters at the end. Modem variants include a range of flat-bottomed sluices for fines recovery, the Chinese ‘rotating spiral’, which both rotates and has grooves cut into the bottom of the sluice to direct the settled material toward the centre, and Carpco’s stationary unit with a similar groove pattern. These newer units have reduced the bottom size of effective separation to at least 30 micrometres (Burt 1999, Richards et a1 2000) Relatively inexpensive, with small footprint, and requiring little in the way of structural steel, spirals are ideal units for all plants of all sizes and sophistication, from the largest iron-ore plants to individual units installed in the outback. Shaking tables While, in recent years, high throughput centrifugal separators have replaced most of the largescale applications for shaking tables, these units have been the workhorses of the industry for well over a century, and they still play an important role. The high recovery/enrichment ratio that can be achieved still provides a cost effective and user-friendly solution for the concentration of heavy minerals in lower tonnage streams, where the key remains the need for effective feed preparation. The table combines several complex mechanisms, including, stratification, hindered settling, and flowing film concentration. The end-mounted, reciprocating head motion provides the motive force for not only the former two, but also, in most cases, the means of separating the stratified layers, by the lateral movement of the concentrates away from the pulp stream. The Holman table’s unique vertical component, of approximately 1 mm., partially overcomes this, by imparting a jigging type action, which both enhances stratification within the riffled part of the deck, and also lateral movement in the smooth part. (Burt 1984). Riffle pattern design, including grooved rather than elevated riffles, as well as the diagonal deck design has improved the performance of standard shaped decks. However, some new concepts in table shape include the double-sided, trapezoidal decked Gemini table, as well as the Chinese circular deck concept (Hehrong 1991). Fine particle separators The physical constraint of settling time of ultra-fine particles continued to be the problem with all devices, and recovery decreased rapidly finer than about 15 micrometers. Much of the emphasis of equipment innovation has been aimed at improving separation of the finest particles. Richard Mozley’s several successful attempts to push the envelope of gravity concentration made him

952

arguably the most successful of the innovators of fine gravity equipment. His Mozley Frame (later becoming the Battles-Mozley separator) was a direct outgrowth of the shaken helicoid, a device he developed with Prof. Burch at Bristol University: it overcome the lack of capacity of that machine by incorporating the multi-deck approach of the Denver Buckman table. It was a semi-batch unit, in that the heavy fractions settled to the deck surface, and had to be removed while the feed was arrested. Nevertheless, it, and its later companion, the Crossbelt Concentrator, with a means of continuous discharge of concentrate, allowed the recovery by gravity of finer particle sizes than ever before. Significant numbers of these units were installed in the period 1966- 1990, but the pursuit of capacity caught up with them, and they have been superseded in recent years by other devices. Centrifugal separators: The technology that has probably caught the most attention in recent years has been the development of a range of centrifugal - or elevated gravity - concentrators. While an early attempt in the 1950’s, Ferrara’s Tube, had technical merit, its inherent lack of capacity meant that it was nothing more than a research device. (Ferrara 1960) Some concentration units, such as the Gilkey Bowl, Knudsen Bowl and YT separator incorporate centrifugal force have been in operation, in China, Russia and in Alaska for close to thirty years, primarily treating gold. The centrifugal force simply enhanced gravity, with rotational speed of the drum chosen such that the heavier particles settled to the surface, while lighter particles remained in suspension. The two units that have the most success to date, and both invented in Canada, are the Knelson Separator and the Falcon Concentrator. (Casteel 2001) Both manufacturers started making only semi-batch machines, but both have now also developed continuous units to complement the batch units, as a result of an iterative process, with collaboration between Government, McGill University (Laplante 1993) and the mining industry. The Knelson consists essentially of a conical drum with a series of parallel ‘Vee’ shaped riffles. Gravitational force which force particles into the riffles is counteracted by hydrostatic water injected into the bedding to form a fluidised bed, and the concentration mechanism can best be likened to a hindered settling classifier. (Laplante 1993). The Knelson has become the unit of choice in many free milling gold circuits but is finding applications elsewhere. The basic concentrating element of the Falcon concentrator is also a conical, vertically spinning, but smooth walled drum, with slurry moving upward from the central feed point. The main separation mechanism is percolation trickling, enhanced by the centrifugal force (up to 300g). In the continuous model. concentrate is withdrawn from the bottom of the bed through a series of ports that either open intermittently, or remain open continually, producing a relatively low enrichment ratio product. The Falcon found earlier application in mid-density separations, such as tantalum (Deveau 2000). The Mozley Multi-gravity separator (MGS), which essentially wraps a shaking table into a cylinder, and rotates it, operates at significantly lower ‘g’ force than either of the Canadian units. Unlike them, also, the MGS, was developed originally for the tin industry, treating the finest fractions - the MGS has truly pushed the envelope in terms of ultrafine particle recovery, reducing the bottom size of effective recovery to as low as 3-4 microns. Wheal Jane pioneered Mozley’s MGS as a final cleaner of tin flotation concentrates, due to it’s ability to achieve higher grade than straight flotation, without sacrificing overall circuit recovery (Turner and Hallewell 1993). As an answer to it’s relatively low capacity 4-8 tph, the company developed the higher capacity MeGaSep, rated at 30 tph for minerals and up to 50 tph for coal: the prototype was successfully tested in a UK colliery in 1999 (figure 4).

953

Figure 4: The Prototype MeGaSep at a UK colliery The Kelsey Jig, developed in Australia (Clifford 1999) has, as the name implies has combined a Harz type jig with an intermediate centrifugal force (approx. 60g). In addition, the fine-grained rigging may act partially as a heavy medium (Malvik, Sandvik and Rein 1998). The Altair jig is another centrifugal unit: while the earliest models were tested for gold, more recent development has been aimed at fine coal cleaning (Mohanty and Honaker 1998). Considering the high rotational speeds and consequent centrifugal forces, all these types of separators require a high level of engineering, including balancing of the rotational unit, and careful selection of materials of construction. Tanco carried out extensive work on the Falcon Concentrator to improve wear characteristics of the concentrate ports (Deveau 2000) while Kelsey has answered the conundrum by biannual shop rebuilds. Nevertheless, properly maintained and properly fed machines can provide good long-term reliability and performance. Dry separation: Developments in dry gravity concentration are more modest, at least in the mineral processing field. Nevertheless there is a wide range of equipment available, much of which could be regarded as the dry equivalent to its more commonly recognised counterparts, such as, dry jigs, sluices and tables. In the right place, dry gravity concentration can be very effective, and it will have an ongoing role in the processing engineer’s arsenal. The water balance While not specifically a ‘unit process’ - at least in terms of gravity equipment - the water balance in gravity plants is of special importance, that often requires additional, ancillary, unit processes. Different devices have different optimum operating solids content, such as cones (55-60%), shaking tables (generally 20-30%) and the old Bartles-Mozley (6- 10%). Unfortunately these optimum pulp densities are not always the same as the preceding equipment’s product and dilution, or thickening, is often required. In areas where water is at a premium, it may indeed be necessary to reduce circuit complexity at some expense of recovery to maintain an adequate water balance.

954

While gravity devices often have a much greater latitude to the inclusion of ‘slimes’ in the feed than flotation, excess slimes do impact on fluid viscosity and ultimately performance. There is some latitude, therefore, in using ‘dirty’ water to dilute feed streams. However, in general, there is no such latitude in the wash-water additions required by many devices: in such cases only ‘clean’ water will do. APPLICATIONS OF GRAVITY CONCENTRATION Gravity concentration, ancient process that it may be, remains vibrant. Applications in the mining industry run, literally from A to Z, with gravity concentration being used in the concentration of minerals from andelusite to zircon. Furthermore, it is arguably the only process that is suitable for the whole range, from the largest and most sophisticated operations to the smallest and most basic of plants, where instrumentation - and even electricity - is an unheard of luxury. Many of the largest plants are in the iron ore industry, where cones and spirals have typically been the main unit processes employed. One large, highly automated plant, is the Kemi chromite concentrator in Finland, producing in excess of % million tonnes of product per year. The plant incorporates various gravity circuits, including heavy medium separation, a Reichert cone plant, and spirals, as indicated in the simplified schematic, figure 5 (Ruokonen et al, 1998) South-east Asian tin plants (figure 6 ) are also examples of high throughput plants, but ones where the very low grade nature of the feed, and the premium of space, demands simplicity, both of circuitry, and of equipment both in terms of operation and maintenance.

Crushed ore

I

Conc

Tail

1-~

I Cone plant I YT

Foundary sand

Met grade chrornite

Figure 5: Simplified schematic showing the gravity sections at Kemi concentrator

t

I olsize

J i g 1 2

I

I T 1

- Jig

conc to tin shed

Figure 6: Typical tin dredge jig plant

955

For many years, gravity concentration has been the process of choice for the concentration of essentially all mineral sands throughout the world, with spirals, and to a lesser extent Cones, being the main unit processes in use. Fairly complex rougher/scavenger/cleaner circuitry is often employed, where a spiral plant as shown in figure 7 may well result. Feed

T

I

C

Figure 7: An example of a complex spiral heavy minerals concentrating scheme A typical, mid-sized sophisticated all-gravity plant for the concentration of metal oxides is that of Tantalum Mining Corporation of Canada, Manitoba Canada. (Ferguson et a1 2000). The flowsheet (figure 8) demonstrates several key factors for optimum gravity concentration: stage grinding and concentration; feed preparation; stage upgrading, as well as a willingness to use diverse equipment at different stages of the flowsheet. At the bottom end of the spectrum is the artisanal miner. Each operation may only produce a minuscule amount of product, but such is the number their cumulative output accounts for a significant proportion of production, especially in developing nations, of such minerals as gold and tantalum. Generally far from infrastructure, and with minimal power, a typical gravity concentration plant will consist of little more than a ground sluice followed by an upgrading sluice - all hand fed and operated - as well as the final winnowing of concentrate. For duties such as these, gravity concentration is not just the best solution - it is the oJry solution. Gravity concentration is often used in conjunction with other processes, in roughing or scavenging applications as well as final cleaning. Recovery of free milling gold prior to cyanidation or CIP circuits has a long history: jigs were most commonly used in North America, while in South Africa Johnson barrels and Plane Tables were more common. However, today, the Knelson concentrator, and to a lesser extent the Falcon and the IPJ, have become the units of choice, with or without different units for further upgrading. At Avocet’s Penjom mine in Malaysia, all three units are employed (figure 9). (MJ 2001)

956

Closed circuit

n I

I_

I

I

Spirals

I

I I

Screen at 2mm

1 I

I

I

I

I Deslime cyclones I

I U

I

I

[Stokes1

L___t

I

1-GiG-j

I Hydrosizer I

I ' IX-Beltl I I FalconC20

1

CONC

Figure 8: Schematic flowsheet at Tantalum Mining Corporation of Canada Ltd.

I

Ulf

I

I

J*

1

"i S irals

To CIP plant

tables

To smelting

Figure 9: schematic of Penjom gravity gold concentration section

957

Further processing

There are many examples of the potential for gravity concentration in the treatment of minerals not normally regarded as typical applications. Recently cited examples include the proposed Mt Weld, Australia, flowsheet which incorporates gravity concentration of coarse rare earth oxides prior to flotation of gravity tails, reducing both grinding and reagent costs, (Guy et a1 2000), and the use of Mozley MGS units for celestite concentration (Oliazadeh and Mohammedi 1998). Similarly, scavenging of molybenite flotation tails for the recovery of heavy minerals in Rumania (Zlagnean et al., 2000), PGM recovery from nickel-copper flotation tailings in Russia (Petrov et al. 2000) have been proposed. Conventional wisdom suggests that, with base metal ores, where the efficiency of flotation is generally excellent, there is no role for gravity, where recovery would generally be lower. Nevertheless, it could, and should, have a much greater role than it does, primarily as a preconcentrator. The rejection of a reasonable proportion of the barren (and generally harder) feed, prior to grinding of the softer sulphides, would have a major impact on costs, with only a minor impact on recovery. The most likely process to be successful would be heavy medium separation. Notwithstanding such diverse applications in the metal, and industrial, minerals industry, the greatest tonnages treated are probably still in the coal industry, where, in general, the most important unit processes are heavy medium separation and jigging. Even here, however, the newer generation of centrifugal concentrators, especially the MGS and the Falcon are finding some applications. (Abela 1997) Elsewhere, gravity concentration is employed in such diverse areas as the food and the oil industries In the former, air tables and its equivalents are being used for such diverse applications as: removal of stones, metal, wood and glass from grain and from peppercorns; the separation of shelled from unshelled peanuts; the grading of peas and corn and the removal of doughballs from baby cereal. In the latter, tables have been used in refinery thermal catalytic cracking units, to maintain the balance and activity level of the catalyst, while the potential application of the MGS for the separation of oil shales has been studied. Similarly spirals would perform a key role in a potential heavy minerals co-production facility treating Alberta oil-sand tailings (Owen and Tipman 1999) - thereby not only recovering an economic by-product, but also reducing the quantity of tailings to be stored. Ever increasing emphasis is being placed on environmental protection, and the need for a full life-cycle approach to the treatment of minerals and the use of their products and to the land from which they are extracted. Simply extracting minerals is not enough any more: recycling and reusing are equally important, and these present a new set of challenges to the mineral processing engineer. Gravity concentration - often dry - is used for such applications as the recovery of brass from foundry sand (Mankosa and Venatraman 1997) and the recovery of precious metals from computer boards. As many plants run many very different materials in relatively small batches through their plant, visibility and adaptability of the equipment is important, suggesting that such units as the shaking table are ideally suited. (Burt 1999). As many recycling plants are in, or close to, cities, and effluent disposal is problematical, often requiring almost complete recycling of water, ‘benign’ processes such as gravity, which do not require reagents or other chemicals, are the obvious answer. The growing awareness of the environmental cost of landfill sites is spurring the development of profitably operating secondary recovery plants. Not only do such plants recover valuable products, but they also reduce the volume of the sites (and hence reduce dumping costs), and enable operators to comply with ever more stringent and costly regulations that are being placed upon them. Likewise, rather than simply starting new landfills as existing ones become full, consideration is now being given to “mining” old fill sites to recover values and to reduce volumes. A variety of mineral processing technologies are being applied including gravity concentration (Finch 1998, Amaratunga 2001). Truly full-cycle, back to the days of Mr Boffin, albeit with more sophistication!

958

CONCLUDING COMMENT One paper cannot hope to do more than briefly introduce the subject of gravity concentration, as well as to indicate some of its advantages and potential applications. The succeeding papers in this section will describe unit processes and practices in more detail. If all this paper has done is to remind process engineers of the potential for gravity, giving them pause to at least fdly consider such potential in their next flowsheet design, it has succeeded. ACKNOLWEDGEMENTS I wish to thank Cabot Corporation for their permission to present this paper. Special thanks also to Chris Bailey, Nigel Grigg and Mike McGarr. I also thank the various operators and manufacturers who once again have allowed me to make free with photos of their plants and equipment. REFERENCES Abela, R.L. (1997). Centrifugal concentrators in gold recovery and coal preparation. Extraction Metallurgy Africa, 2526 June Amaratunga, L.M. (2001). The mineral processor and the environment: waste to resources. CIM Bulletin 94 Nov/Dec. pp 62-70. Bath, M.D, A.J. Duncan, and E.R. Rudolph. (1973) Some factors influencing gold recovery by gravity concentration. J. S. Afr. Inst Min and Metal1 73 (1 1) June pp.363-378. Burt, R.O. (1984). Gravity Concentration Technology. Amsterdam: Elsevier Science Publishers BV. 607pp. Burt, R.O. (1999). The role of gravity concentration in modem processing plants. Minerals Engineering. 12:11 pp.1291-1300. Casteel, K. (200 1). The process story. World Mining Equipment October pp.20-28. Clifford, D. (1999) Gravity Concentration. Mining Magazine March, pp. 136-148. Cope, L.W. (2000). Jigs: the forgotten machine. E&MJ, August pp.ww.30-34. Deveau, C. (2000). The evolution of Falcon Continuous Concentrators at Tantalum Mining Corporation of Canada. Canadian Miner. Processors Ann. Mtg, Ottawa. 17pp. Ferguson, W., S. Young, C. Deveau, W. Hazel, A. Lussier and R. Copeland. (2000). Cabot Corp. Mining Corporation of Canada Ltd. Canadian Milling Practice. CIM Special Vol. 49.

- Tantalum

Ferrara, G. (1960). “A process of centrifugal separation using a rotating tube” Proceedings 5th Int. Miner. Process. Cong. London: IMM. pp. 173-184. Finch, J.A. (1998). And where are we going? Comminution, flotation and gravity separations, CIM Bulletin, 91 Nov/Dec pp. 68-72. Gray, A.H. (1997). InLine Pressure Jig - An Exciting, Low Cost Technology with Significant Operational Benefits in Gravity Separation of Minerals, The AusIMM Annual Conference, Ballarat, March, pp. 259-265.

959

Guy P.J., W.J Bruckard and M.J. Vaisey. (2000). Beneficiation of Mt. Weld rare earth oxides by gravity concentration, flotation and magnetic separation In: Seventh mill operators' conference, held in Kalgoorlie, Western Australia, 12-14 October 2000. Carlton, Victoria: Australasian Institute of Mining and Metallurgy. (Australasian IMM publication series, no. 6/2000.) p. 197206, Liu Hehrong. (1991). Short concentrating path shaking table. In: XVIIth International Mineral Processing Congress, Dresden, FRG, 23-28 September 1991. Preprints. 111, 199 1. pp.67-76. Holland-Ban, A.B. (1 998). Gravity Separation: A Revitalised Technology. SME Annual Meeting, Orlando Florida, March, SME Inc., Littleton, preprint 98-45. 5pp. Laplante, A.R. (1993). A Comparison of two Centrifugal Concentrators. Annual Canadian Miner. Proc. Conf CIM, Ottawa. 20 pages. King R.P. (2000). Flowsheet optimisation using simulation: a gravity concentrator using Reichert cones. In: Proceedings of the X Y I International mineral processing congress held in Rome, 23-27 July 2000. Massacci P., ed. Amsterdam: Elsevier, 2000. B, p.B9,1-9 Mohanty, M.K., and R.Q. Honaker. (1 998). Evaluation of the Altair Centrifugal Jig for fine particle separations. SME Annual Meeting, Orlando, FA, March. SME Inc., Littleton, Preprint 98-169 llpp. Malvik, T, K.L. Sandvik. and A. Rein. (1998). Scandinavian experiences with the Kelsey centrifugal jig. Innovation in physical separation technologies, the Richard Mozley Symposium Volume. Papers presented at a conference held in Falmouth, UK, 4-5 June 1997. IMM, London, pp.113-122. Mankosa, M.J., and P.Venkatraman. (1 997). Application of mineral processing techniques for secondary materials recovery, SME Annual Meeting, Denver CO, SME Inc., Littleton, Feb. 15pp. MJ (2001) Penjom foils the preg-robbers. Mining Journal, London. Sept 28,2001. pp.238-9. Oliazadeh, M. and H. Mohammedi. (1998). Beneficiation studies of Zagras celestite. In: Innovations in mineral and coal processing. Proceedings of the 7IhInternational mineral processing symposium held in Istanbul, Turkey 15-1 7 Sept. Atak, S., Onal, G. and Celik, M.S. Eds. Rotterdam: A.A Balkema. pp271-3. Owen, M and R. Tipman. (1999). Co-production of heavy minerals from oil sand tailings. CIM Bulletin 92 No 1030 May. pp.65-73. Petrov G.V., T.N. Greiver, L.I. Gyrskia and Y.V. Andreev. (2000). A process mineralogical study of platinum-bearing chromite ores from the polar Urals., Prepr. SOC.Min. Metall. Explor., no.00-30, 4pp. Richards, R.G, D.M. MacHunter, P.J. Gates and M.K. Palmer. (2000). Gravity separation of ultra-fine (-0.1 mm) minerals using spiral separators. Miner. Engng. 13:1 Jan. pp65-77. Ruokonen E., H. Pekkarinen, R. Bergstrom and E. Puukko. (1998). Recent developments in controlling complex cone concentrator at Kemi concentrating plant. Innovation in physical separation technologies, the Richard Mozley Symposium Volume. Papers presented at a conference held in Falmouth, UK, 4-5 June 1997. IMM, London, pp.101-110.

960

Turner, J.W.G. and M.P. Hallewell. (1993) Process improvements for fine cassiterite recovery at Wheal Jane. Minerals Engineering. 6 : 8-10 pp.817-829. Zlagnean M., N.Tomus and C.Vasile. (2000). Processing of valuable vein minerals, as molybdenite, monazite, magnetite, pyrite and ilmenite In: Proceedings of the X X I International mineral processing congress held in Rome, 23-27 July 2000. Massacci P., ed. Amsterdam: Elsevier, 2000. C, p.Cl0 5-9..

961

Types and Characteristics of Heavy-Media Separators and Flowsheets Robert A. Reeves, P.E.’

ABSTRACT Heavy-media separators, and cyclones in particular, are widely used by the coal, mineral, and chemical industries to separate lower-density materials from higher-density materials. Heavymedia processes are typically more expensive and complex than wet or dry gravity concentration processes such as jigs and hydrocyclones. However, heavy-media processes recover more highgrade products and have greater flexibility and tolerance to variable feed conditions. In selecting a separation process, one must consider the washability characteristics of the raw feed. A heavy-media process includes subsystems to prepare the feed, process the prepared feed, and recover the media from the products. Factors in selecting the specific types of equipment for each subsystem include feed rate, type and size of raw material, and product quality requirements.

The feed preparation step deslimes the feed to reduce the amount of fine particles that may contaminate the media, and provides closely sized screen fractions that can be efficiently treated by the heavy-media separators. Baths, drag tanks, and drums typically process particle sizes greater than 12 mm. Cyclones typically process particle sizes between 25 mm and 0.5 mm. Special separators have recently been developed to process a wider range of particle sizes, including fine particles, in a single step. The media circuit typically uses finely ground magnetite and ferrosilicon to prepare media of the desired density. Other materials, such as halogenated hydrocarbons, have been proposed, but are not commonly used due to environmental concerns or high cost. The media are recovered from the low- and high-density products and concentrated for reuse. New magnetite or ferrosilicon is added to make up for losses.

INTRODUCTION The heavy-media separator, and the heavy-media cyclone (HMC) in particular, is a widely used piece of process equipment that separates particles based on their specific gravities. Heavy-media baths, vessels, and cyclones use precisely controlled heavy-media pulp densities to clean thermal and coking coals, recover diamonds, separate scrap materials, and pre-concentrate ores. Cyclones have the advantage of using centrifugal forces (between 30 and 60 times the force of gravity) to aid the separation process. Coal is the primary application for the heavy-media processes and will be used for many of the examples presented in this paper. The HMC will be discussed in detail because of its popularity. Baths and vessels normally process particles ranging from 10 mm to 150 mm. Cyclones of various diameters now can process particles ranging from 0.15 mm to 100 mm. Two other items of equipment, the Large Coal Heavy-media Separator (LARCODEMS) and the Dynawhirlpool, have been widely used to process larger particle sizes than thought practical for

1

Hazen Research, Inc., Golden, Colorado.

962

the cyclone. References to the LARCODEMS, Dynawhirlpool, and other matters relating to the HMC are provided in the reference list. Dutch State Mines developed the HMC in the 1940s to wash coal. Since then, the cyclone has been installed in over one-quarter of the coal preparation plants worldwide. The HMC is most widely used in Australia (used in over half of the coal preparation plants), followed by Canada, Republic of South Africa, Indonesia, India, China, and the United States, Coal users are demanding higher quality products that satisfy environmental regulations and reduce utilization costs. The HMC is ideally suited to produce these high quality products because they cannot be as efficiently produced by other separation processes. Certain toxic elements, i.e., mercury, are partitioned by gravity-based coal preparation equipment. Cyclones may be used to reduce the concentration of these elements in the cleaned product.

HMC CIRCUIT DESCRIPTION The HMC or vessel is not a stand-alone piece of equipment, and it should be examined as part of a circuit that includes ancillary unit operations critical to efficient cyclone operation. A typical HMC installation, shown in Figure 1 includes five basic functions that: Prepare the raw feed material Separate the feed into low- and high-density products Recover the media from the products Concentrate media solids Provide circuit control

Feed Preparation The feed preparation process sizes the raw feed, providing a range of particle sizes specified for the application. In standard practice, large vibrating screens fitted with various wire, punch plate, cast urethane, and profile wire decks are used to provide a sized, deslimed feed with a predictable surface moisture content. Limiting the feed top size protects the cyclone from blockage and provides the best possible performance for a particular cyclone diameter. Table 1 lists the range of sizes normally processed by various diameter cyclones. Desliming the feed removes fines, typically smaller than 0.5 mm, which would adversely affect heavy-media quality and decrease cyclone performance. Cleaning feed that has not been deslimed has been discussed, but most commercial plants built today deslime the raw feed. Heavy-media baths require screening the feed to a typical bottom size of 10 mm. Table 1 Range of particle sizes processed by heavy-media cyclones Source: Multotec Process Equipment (Pty) LTD Cyclone Diameter

Bottom Size

Top Size

Inches

Meters

Tyler mesh

mm

Inches

mm

20.1

0.5 1

14

1.4

1.4

35

24.0

0.61

2

1.7

42

26.0

0.66

2.4

1.7

42

28.0

0.71

3

1.7

42

31.5

0.80

6

3.4

3.3

85

39.4

1.oo

4

4.8

3.9

100

8

963

Raw Feed

Sizing Screen

ieavy Yedia

b

Cocrse Process

Process Water

Process Water

I LOW-SGProduct Drain & Rinse Screen

4

HighSG Product Drain & Rinse

Heavy-rneda Process

Smell

Oiaini

I

I Final Discard

Heavy-media Circuit

d

Diluterneda Circuit Nonmagnetic Tails Fines Process

Magebc Concentrate Process Water

Media Circuit

Make-up Media Supply

Dewat ering

Final Product

Figure 1 Typical heavy-media circuit block diagram

964

Standard plant practice has relied on inclined and horizontal vibrating screens to prepare the feed. More recently, multiple-angle screens (the so-called banana screen, named after the characteristic its curved profile) have become popular with large operations because of their high capacity and low unit operating costs. The traditional horizontal vibrating screen in shown in Figure 2, and the multiple-angle inclined screen is shown in Figure 3.

Figure 2 Horizontal vibrating screen sizes and deslimes feed Source: Tabor Machine Co. Mixing Prepared Feed With Heavy Media Sized and deslimed feed is mixed with heavy media, which is a suspension of magnetic, highdensity, powdered solids and water. Finely ground magnetite is preferred for coal applications (specific gravity of separation less than 2.2) because of its relatively low cost and ability to be recovered and concentrated by conventional wet-drum magnetic separators. For higher-gravity separation applications, i.e., ores and diamonds, ferrosilicon is preferred because it has a higher density than magnetite. Two methods of mixing are typically used depending upon how the HMC is fed. Two feed methods, gravity or slurry pumps, provide the feed at the cyclone inlet at the required pressure. Feed pressures vary by application and cyclone diameter. Typical feed heads (feet or meters of head) ranges between 9 and 15 times the diameter of the cyclone. A specified volume of heavy media is required to process one ton of feed. The ratio is usually expressed on a volume of media to volume of solids basis; typical coal applications use a 4:l ratio. Higher ratios can be used to improve separation efficiency depending on the washabilty of the feed and other considerations.

965

Figure 3 Multiple-angle inclined screens have high sizing and desliming capacity Source: Tabor Machine Co.

Gravity-fed systems use a pulping box located at the discharge of the desliming screen, which has been elevated to the required distance above the HMC to provide the necessary hydrostatic head. Deslimed feed falls into a pool of heavy media, and the mixture flows downward by gravity to the cyclone inlet manifold. Since the pulping box and cyclone are secured at fixed locations, the feed pressure cannot be easily varied. Pump-fed systems use a sump in which deslimed feed and heavy-media are mixed. The sump is designed to allow the feed to become completely engaged with the heavy media and not segregate. The mixture is pumped and delivered to the cyclone inlet manifold at the desired pressure. Pumpfed systems are popular because of lower headroom requirements and greater operational flexibility. Product Separation The HMC consists of four basic elements, as shown in Figure 4: cylinder feed section, cylinder section, conical section, apex (spigot), and vortex finder. The cyclone manufacturer specifies these components to provide the desired capacity and separation characteristics.

966

Feed

cticrn

Figure 4 Cyclone components Source: Krebs Engineers

967

The cylindrical feed section directs the incoming slurry into the cylinder section. Careful design minimizes turbulence and provides a smooth transition from a linear flow path to a circular flow path. The circular flow imposed by the cylinder section subjects the slurry to high centrifugal forces. These forces propel the feed particles, depending on their specific gravities relative to the pulp density of the media, to some distance out from the center of the cylinder section. The length of the cylinder section is selected to provide sufficient residence time for separation to take place. The entlre mixture flows from the cylinder section to a cone section, typically at a 20-degree included angle. Density zones are created in this section where feed particles more heavy than the media move against the wall of the cone section, and feed particles less heavy than the media move to at the inner core. The high-density particles are forced down the cone section by incoming feed and are discharged though the apex. The low-density particles move into an upward-moving stream and pass out the cyclone through the vortex finder located along the centerline of the cyclone. Splash or "kill" boxes located at the discharge points of the cyclone absorb kinetic energy and distribute the material across the width of the drain and rinse screens. Cyclones have increased in size, up to 1.45 meters in diameter. Figures 5 and 6 show typical views of larger cyclones now available.

Figure 5 1-meter (40-inch) diameter cyclone Source: Krebs Engineers

968

Figure 6 1.45-meter (57-inch)diameter cyclone Source: Multotec Process Equipment (Pty) LTD The heavy-media vessel is shown in Figure 7.

Figure 7 Heavy-media vessel Source: Hull Bulk Handling Limited

969

Drain and Rinse Screens High- and low-density products discharged from the HMC are mixed with heavy media. Drain and rinse screens separate the products from the media based on particle size. The screens are typically fitted with profile wire screen decks with apertures that are slightly smaller than the bottom size of the feed. For example, if the feed is deslimed at 0.5 mm, then the drain and rinse screens will be fitted with 0.5 nun or smaller screen decks. Two hoppers collect undersize material, media, and water passing through the screen decks. The drain hopper, located under the feed end of the screen, collects heavy media that drain off the product. This flow returns directly to the heavy-media sump without further processing. Process water deluges the product after the drain section, to remove any remaining media that cling to the product. The rinse hopper, extending from the end of the drain hopper to the discharge end of the screen, collects media that are rinsed from the product. Material collected in the rinse portion is diluted with water and requires additional processing to recover and concentrate the media solids.

Media System The media system includes three functional areas: over-dense media, heavy media, and dilute media. The pulp density of the slurry and process function identifies these three areas. The way each area processes material varies from plant to plant, because it depends on operator preference, experience, and the specific application. Each area contains a sufficient volume of slurry and storage capacity to handle fluctuations in process conditions, yet provide responsive and accurate control of flow to the HMC.

Over-dense Media Area. The over-dense media circuit contains media slurry that has a pulp density greater than the heavy media. Over-dense media are gbnerated from two sources, wetdrum magnetic separator concentrates and fresh media solids makeup. In some plants, over-dense media are stored in a separate sump where they can be transferred to the heavy-media area. In others, the over-dense media are sent directly to the heavy-media area. Fresh media can be stored and supplied as a dry powder, or as wet solids to the over-dense media circuit. Wet solids may be supplied at lower cost than the dry powder because drying is not required. A feeder meters controlled quantities of dry or wet fresh media from the supply bin to the over-dense area to maintain the desired pulp density and inventory.

Heavy-media Area. The heavy-media circuit includes a large sump or tank to hold slurry of the desired pulp density. The pulp density is typically 0.05 to 0.10 specific gravity points less than the desired specific gravity of separation. Water and over-dense media are added in controlled quantities to maintain the sump level and pulp density. A portion of the heavy media may be diverted to the dilute-media circuit to reduce the concentration of fine, non-magnetic solids (slimes). Heavy media is mixed with raw deslimed feed and fed to the HMC. Dilute-media Area. The dilute-media circuit receives flows containing media that have been diluted by process water and contaminated with slimes (non-magnetic material that impairs the slurry properties of heavy-media). Classifying cyclones or static thickeners increase the solids concentrations of the dilute flows in order to reduce the volume of slurry that is treated by wetdrum magnetic separators. Process water produced by the classifying cyclones or static thickener is typically used to rinse the HMC products on the rinse section of the drain and rinse screens.

970

Wet-drum magnetic separators, shown in Figure 8, receive the thickened dilute-media flows and recover and concentrate magnetic solids. Concentrated magnetic solids are returned to the overdense or heavy-media circuit depending on the plant’s particular design. Non-magnetic solids often contain valuable fine product, so they are typically directed to the fine-coal recovery circuit for additional processing.

Figure 8. Wet-drum magnetic separator Source: Eriez Magnetics Instrumentation and Control. Elements of the media circuit are monitored and controlled by the plant process control system. The system measures process variables including sump level, flow rate, magnetic concentration, and media density, and controls process elements including makeup water and fresh media supplies. Various strategies can be used to provide precise, stable density control of the media circuit. The control system must maintain a balance of water and fresh media solids to maintain sump levels and pulp densities. Well-designed systems quickly compensate for changes in deslimed feed surface moisture content, changes in recovery, blockages, variable media quality, and inadvertent spills. Operator-machine interfaces display historical and instantaneous information. Alarms and set points can be easily monitored. At least one coal HMC plant has used online analyzers to measure the quality of HMC products. The control system changes the density of the heavy-media to maintain constant quality. This approach provides consistently higher recovery than manual methods as the feed or cyclone characteristics change. CIRCUIT DESIGN CONSIDERATIONS The coal preparation engineer or metallurgist considers a wide range of information before specifying and designing a heavy-media cyclone circuit. The characteristics of the raw feed must be considered with final product quality in mind. The strategy must understand how salable and

971

unsalable materials differ in particle size and specific gravity, in order to specify a process that can effect a separation. For coal applications, a washability analysis of the feed (a characterization of the raw material by particle size and specific gravity) must be carefully examined before proposing a flowsheet or process. The evaluation process becomes more complicated if the feed is intentionally crushed to fine sizes to liberate salable product. Coal applications usually avoid deliberate crushing since the increase in product quality cannot be economically justified. Some projects have been proposed to produce low ash, low sulfur product with special HMC circuits, but as of 2001, these have not been put into commercial operation. Many projects have underestimated the concentration of fines that are present in the feed or have been generated by attrition in the plant. The feed is a geologic material and therefore its washability characteristics can quickly change. The engineer must account for a range of particle size and specific gravity distributions in the initial design to ensure sufficient capacity exists, especiallyi the fines and water clarification circuits. There are many examples of preparation plant modifications undertaken to provide additional capacity or greater efficiency. Economics is the prime consideration for selecting a circuit. In the case of heavy-media cyclones, its selection over other types of processes such as fine coal jigs, hydrocyclones, concentrating tables, spirals, and flotation is usually based on the following: Capacity and corresponding required plant area Product recovery Ability to quickly adjust to changes in feed quality Reliability Operating cost Need for multiple products Requirement to process multiple feed sources The performance of HMC circuits can be modeled by computer to quickly evaluate the recovery of salable product as a function of the specific gravity of separation and changing feed conditions. Partition or distribution curves describe the performance of a HMC for various particle sizes. Governmental organizations and other agencies have published cyclone performance data. Cyclone manufactures have proprietary data about specific cyclone configurations and they should be consulted when the project extends into the detailed design stage. Other sources of information can be found in published literature. The volumetric and feed capacity of heavy media cyclones is listed in Tables 2 and 3.

Table 2 Heavy-media cyclone typical volumetric and feed capacities Source: Krebs Engineers Cyclone Diameter

Nominal Volumetric Howrate

Nominal Capacity

Inches

Meters

US gaVmin

mA3/hr

St/hr

t/hr

20

0.51

1200

213

60

54

26

0.66

1800

409

100

91

30

0.76

2600

590

208

189

33

0.84

3800

863

304

276

40

1.02

5200

1181

416

377

972

Table 3 Heavy-media cyclone typical capacities Source: Multotec Process Equipment (Pty) LTD Cyclone Diameter

Nominal Volumetric Flowrate

Inches

Meters

US gaUmin

mA3/hr

20.1

0.5 1

1000

227

24.0

0.61

1600

363

26.0

0.66

1900

43 1

28.0

0.71

2300

522

31.5

0.80

2900

659

35.4

0.90

3900

886

39.4

1.oo

4900

1113

45.3

1.15

6800

1544

51.2

1.30

9100

2067

57.1

1.45

11800

2680

OPERATIONS AND MAINTENANCE Operators strive to reduce operating costs and improve reliability and performance of the HMC circuit. New materials and designs have provided the operator with more choices and designs than ever before. For example, heavy-media cyclone circuits are incorporating larger-capacity screens and larger-diameter cyclones to reduce unit costs. Three-meter wide screens were introduced about 20 years ago and are now popular because of their high capacity and low unit cost. Improved pump designs are more efficient, able to pump larger particles, and have enabled operators to extend maintenance intervals.

Plant Layout Many new plants and plant additions in the United States are being designed in the “low-profile’’ concept popular for many years in Europe, South Africa, and Australia. The low-profile design makes access to large equipment by overhead cranes possible. Performance Factors Cyclone performance is influenced by wear, media composition, feed quality, and configuration. Plants that process feed materials with a high concentration of abrasive rejects (diamond applications in particular) must pay special attention to wear. Media QuaIity. The quality of the media, either magnetite or ferrosilicon, influences both.cost and performance. Media costs generally increase with decreasing particle size, but may improve cyclone performance. Most suppliers provide a chemical assay, a size distribution, and a magnetic proportion of their products. Magnetite is normally an upgraded mined material, so variability and performance may change from supplier to supplier. Ferrosilicon is a manufactured material with more closely defined specifications.

973

Media become contaminated with fine feed material and change particle size distribution during normal operations. A low concentration of slimes helps to stabilize the heavy-media slurry, but too much slime increases the slurry viscosity, which degrades separation performance. The heavy media slurry may become unstable when its pulp density becomes too low. Obtaining a sample of the cyclone overflow media and measuring its pulp density can provide a quick check of cyclone performance. The measured overflow pulp density is compared with the corresponding feed heavy-media pulp density. Proper operation is indicated when the ratio of the two pulp densities falls within a specified range gained from operating experience. Figure 9 shows the relationship between overflow and feed media pulp density and its effect on viscosity and stability.

1.35

1.40

1.45

1.so

1.55

1.60

1.65

1.70

1.75

1.8(

Cyclone Feed Heavy Media Pulp Density, gmmslcc

Figure 9 The ratio of feed and overflow media pulp density indicates the condition of the heavy media Source: After Multotec Process Equipment (Pty) LTD Adequate wet-drum separator capacity is required to maintain slimes concentrations at acceptable levels. Some researchers have described a bi-modal particle size distribution that reduces media viscosities at high pulp densities.

Feed Conditions. Changes in coal mining conditions can create situations where the reject handling capacity of the cyclone is met or exceeded. For example, a continuous miner section may cut through a fault zone or clean up a roof fall, sending a high concentration of refuse to the plant. Changes in mining methods may drastically alter the particle size distribution of the raw coal. For example, a continuous section may be opened while a long wall section is moved.

974

Cyclone Configuration. The internal configuration of a cyclone (vortex finder size, apex diameter) is chosen to provide the required low- and high-density product capacity while maintaining acceptable separation performance. Selection of the liner material is important to balance purchase price with maintenance costs. Common liner materials include rubber, urethane, and various types of silicon carbide and alumina. Performance Monitoring Comprehensive cyclone performance should be checked periodically by concurrently obtaining samples of low- and high-density products. It is time-consuming and costly to obtain samples and conduct a meaningful washability test. Many operators avoid a full test, claiming that operating conditions change between the time that samples are taken and results are made known. Some operators conduct a modified performance test using a single particle size specific gravity. These results are used as an indication of performance and may signal that a more comprehensive test is warranted. As an alternative to traditional sampling and testing, tracers have been used to quickly measure equipment performance. The tracer technique injects particles that mimic feed particles into the cyclone feed, and the tracer particles are counted as they exit with the low- and highdensity products. Cyclone performance is computed from these counts. Cyclone performance varies with operating conditions and cyclone configuration. Industry experience provides the following general guidelines about performance and process variables. Separation efficiency, expressed by the Probable Error (Ep) value, improves with feed particle size and declines with increasing cyclone diameter and specific gravity of separation. Cyclones generally operate more efficiently at greater feed pressures. Wear and particle size classification become more pronounced at higher pressures, creating higher operating costs and potentially unstable media. The diameter of the apex increases with wear. Probable error significantly increases (decreased performance) with apex diameter. The length of the vortex finder shortens with wear. When compared with optimal length, probable error increases (decreased performance) with decreasing length.

CONCLUSIONS Heavy-media processes provide efficient and cost effective separations for coal, diamonds, scrap, and ores. Recent advances in materials of construction and a better understanding of operating factors have allowed manufacturers to develop large diameter cyclones that can handle feed material once reserved for baths and vessels. The HMC will continue to gain acceptance worldwide as a method of producing high quality products to meet stringent market requirements.

REFERENCES Akers, D. J., Zitron, Z., and Toole-O’Neil, B. 1998. Coal Cleaning for HAP’S Control: Cost and Performance. The Proceedings of the 23rd International Technical Conference on Coal Utilization & Fuel Systems. 77 1 - 1 8 1. Alderman, J. K. 1999. Y2K, SOz, and Coal Washing: Danger and Opportunity. Coal Age (May). 38 - 42. Bastunskii, M. A. , Bogin, V. E., and Melkumov, L. G. 1968. Determining the Process Variables for Automatically Controlling the Viscosity of a Magnetite Suspension. Coke Chem. USSR. 10- 15.

975

Bosman, J. Dense Medium Separation, Does Size Really Count? Multotec Cyclones (Pty) Limited. Buder, M. K., Clifford, K. L., Huettenhain, H., and McGowin, C. R. The Effects of Coal Cleaning on Power Generation Economics. Chaston, I. R. M. and Napier-Munn, T. J. 1974. Design and Operation of Dense-medium Cyclone Plants for the Recovery of Diamonds in Africa. Journal of the South African Institute of Mining and Metallurgy01 (5). 120 - 130. Chironis, N. P. 1987. On-line Coal-tracing System Improves Cleaning Efficiencies. Coal Age (March). 44 - 47. Conzemius, R. J., Chriswell, C. D., and Junk, G. A. 1988. The Partitioning of Elements during Physical Cleaning of Coals. Fuel Processing Technology (19). 95-106. David, D. 1996. HMS Cyclone Development at Argyle Diamonds. The AusIMMAnnual Conference. 265 - 173. Deurbrouchk, A. W. 1974. Washing Fine-Size Coal in a Dense-Medium Cyclone. Bureau of Mines Report of Investigations 7982. Deurbrouchk, A. W. and Hudy, J. 1972. Performance Characteristics of Coal-Washing Equipment: Dense-Medium Cyclones. Bureau of Mines Report of Investigations 7673. Engelbrecht, J. A. The Effect of Changes in Cyclone Design Variables on Dense Medium Separation. Multotec Cyclones (Pty) Limited. Fiedler, K. J., Munro, P. D., and Pease, J. D. 1984. Commissioning and Operation of the 800 tph Heavy Medium Cyclone Plant at Mount Isa Mines Limited. The Aus. I.M.M. Conference. 259 271. Fiscor, S. 2000. Largest U.S. Plant Gets Larger. Coal Age (April). 26 - 28. Fiscor, S. and Lyles, J. 2000. Prep Plant Population Reflects Industry. Coal Age (October). 3 1 38. Flintoff, B. C., Plitt, L. R., and Turak, A. A. 1987. Cyclone Modelling: A Review of Present Technology. CIMBulletin 80 (905). 39 - 50. Gottfried, B. S. 1978. A Generalization of Distribution Data for Characterizing the Performance of Float-Sink Coal Cleaning Devices. International Journal of Mineral Processing 5. 1 - 20. Harrison, S . T., Woodman, J. R., and Peatfield, D. M., LARCODEMS - More Than a Proven De-Stoning Vessel on South African Coals? LTA Process Engineering Limited and Independent Coal Consultant paper. Hollinden, G. A., Wells, W. L., and McGlamery, G. G. 1976. Effects of Coal Quality on the Reliability and Economics of FGD Systems. NCALBCR Coal Converence and EXPO III. Kempnich, R. 2000. Coal Preparation Worldwide. Coal Age (September). 58 - 66.

976

King, R. P. and Juckes, A. H. 1984. Cleaning of Fine Coals by Dense-Medium Hydrocyclone. Powder Technology40. 147 - 160. Knight, J. C., Adams, R. J., and Reeves, R. 1986. Improving Jig Performance Using On-Line Performance Monitoring System. Coal Prep 86 Conference Papers. 90 - 96. Korkmaz, M. and Mondale, K. 1986. Magnetite Recovery Circuitry, Design and Operation. Coal Prep 86 Conference Papers. 35 - 41. Kovatch, K. 1997. Custom Coals Files for Chapter 11, Negotiates with Prospective Buyer. Pittsburgh Business Times (June). Laskowski, J. S . and He, Y. B. Bimodal Dense Medium for Fine Particles Separation in a Dense Medium Cyclone. U.S. Patent 5,819,945, 13 October 1998. Mengelers, J. and Absil, J. H. 1976. Cleaning Coal to Zero In Heavy Medium Cyclones. Coal Mining & Processing. May. 62 - 64. Nicol, S. K. and Buckley, A. N. 1997. Project 5048: Absorption Tracer Technology. Australian Coal Research Limited. August. Osborne, D. G. 1986. Fine Coal Cleaning by Gravity Methods: A Review of Current Practice. Coal Preparation. 207 - 242. Robertson, R., Placha, D., Terry, R., and Watters, L. 1997. Recent Developments in Dense Medium Cyclone Circuit Design. Societyf o r Mining, Metallurgy, and Exploration, Inc. Preprint 97-1 53. Sanda, A. P. 1997. Marfork Plant, Bigger and Better. Coal Age. 3 1 - 34. Supertracer - Radio Tracing System. SCIR. South Africa. Van der Walt, P. J. and Venter, W. P. 1975. The Use of Norwalt and Dynawhirlpool Dense Medium Separators in Coal Preparation. Australian Mining (June). 30 - 38. Vivenzio, T.A. 1980. Impact of Cleaned Coal on Power Plant Performance and Reliability. Electric Power Research Institute. CS- 1400. Research Project 1030-6 (April). Voges, H. C. 199 1. Tests on the Beneficiation of Coal Fines. Journal of the South African Institute ofMining and Metallurgy 91 (2). 41 - 51.

977

Types and Characteristics of Non-Heavy Medium Separators and Flowsheets John K. Alderman’

ABSTRACT Gravity separation technology is employed in commercial mineral processing plants for a range of minerals, industrial minerals, and is the preeminent method of beneficiating coal. Water-based and air-based processes are used to concentrate minerals and coal based upon differences in particle specific gravity. Topics discussed include sample analysis for evaluating gravity separation processes, operating principles for machines to separate coarse, intermediate, and fine particles, and typical flowsheets for commercial processes. INTRODUCTION Concentration of ores using gravity separation is the oldest, most versatile, and most widely used suite of technologies. Unlike heavy medium baths and cyclones, which depend upon suspensions of fine, high-density particles to create an environment for separating materials, non-heavy medium processes (NHM) use water and air, sometimes in combination with centrifugal forces and particle interactions, to produce the physical conditions required for efficient separation by specific gravity. NHM processes have been known since the Middle Ages, as documented by Agricola (Agricola). Figure 1 shows an engraving of a sluice-type separator used to recover gold. Riffles were placed in the bed of the trough to enhance separation of gold from lower-density minerals. Some designs employed the skins of animals to trap particles of gold in the hairs, while rinsing away the lighter sands, and finally scrubbing the gold-impregnated hide in a tub to recover the fine particles of gold. In fact, the legend of the “Golden Fleece” derives from the practice of lining the sluice with sheep hides. Agricola even describes the early application of jigging to mineral concentration. “The jigging sieve has recently come into use by miners. The metalliferous material is thrown into it and sifted in a tub nearly full of water. The sieve is shaken up and down, and by this movement all the material below the size of a pea passes through into the tub, and the rest remains on the bottom of the sieve. The residue is of two kinds, the metallic particles, which occupy the lower place, and the particles of rock and earth, which take the higher place, because the heavy substance always settles, and the light is borne upward by the force of water.”

’

Material Automation Systems and Service, Inc., Castle Rock, CO

978

Figure 1 A launder with riffles (engraving from Agricola) An engraving from Agricola shows hand jigging in the middle ages.

Figure 2 Hand jigging

979

NHM processes are used in the coal industry to process millions of tons of coal every year throughout the world, but are also used extensively in other areas of mineral processing, such as those listed in Table 1.

Table 1 Minerals beneficiated in NHM processes Aluminum Diamonds Mica Asbestos Glass Sand Mineral Sands Barite Gold Monozite Chrome Iron Ore Phosphate Clay Lead Pumice Copper Manganese Sand & Gravel

Tantalum Titanium Tin Vermiculite Zinc Zirconium

Gravity concentration process can be applied to particles as large as 400 mm (ROM jig) or as small 25 microns (centrifugal jig). They employ the principles of stratification, differential settling, hindered settling, consolidation trickling, teeter-bed separation, thin-film separation, and centrifugal force. The significant unit operations in commercial operation in recent years include: 0

0 0 0

0

Wet andDry Jigs Wet and Dry Tables Rising Current Washers Spirals and Cones Centrifugal Jigs

Given the very high separating efficiency available from heavy media processes, one might well ask “Why consider a conventional gravity concentration process?” Some advantages of conventional gravity separation include.

Lower Capital And Operating Costs Heavy media processes are typically more complex and more costly to build and operate than conventional plants. Metallurgical grade magnetite and ferrosilicon, quantities of which are lost in the process, may be difficult or costly to obtain in some regions. Wider Range Of Feed Sizes Can Be Processed Heavy media processes typically handle feeds in the range of 150 mm x 6 mm (baths) or 50 mm x 0.5 mm (cyclones). Cleaning finer sizes requires a substantially larger investment in screens and magnetic separators. Magnetite losses can be high when cleaning particles smaller than 0.5 mm. Conventional gravity separation processes can effectively treat particles both much larger and much smaller than can be treated in heavy media systems. Refractory Dilution The presence of material such as Bentonite clay in the feed to a heavy media process can cause high magnetite consumption, circulating media viscosity problems, and can degrade separating efficiency. Downstream Processes Sensitive To Iron Compounds Some residual magnetite or ferrosilicon will adhere to the concentrate, and these compounds may have an adverse impact on a subsequent process or products. Conventional gravity concentration beneficiates ores without the risk of iron contamination.

980

Water-Averse Situations In some locations process water may be scarce, or unavailable. Removal of water from products and tailings is typically an expensive operation that may be accompanied by a long-tern environmental liability. Dry gravity concentration circumvents the need for process water in appropriate applications. Conventional gravity separation is appropriate for many mineral processing applications and is underutilized as an economic beneficiation tool. The remainder of the discussion will focus on sample testing and the specific characteristics and applications of the various unit operations. SAMPLE TESTING Basic size analysis and fractionization by specific gravity (sink/float analysis) can most often provide sufficient data to predict the performance of conventional gravity separation equipment. Supplemental crushing or milling of intermediate gravity products may be beneficial in determining whether re-treatment of middlings is worthwhile. Screening If little is known about the ore being investigated, test sieves should be selected using a ratio of 2 between sieves, so a sieve stack could consist o f . . .No. 4, No.8, No. 16, No. 30, No. 50, No. 100, and No. 200. Assay of each individual sieve fraction will provide a preliminary indication of ore concentration by size, and which sieve fractions can be beneficially combined for sinklfloat analysis. If wet processing would likely be employed in the commercial plant, then wet sieving should be used in the laboratory tests. For dry processing, laboratory dry sieving is appropriate, but it is also advisable to run a set of wet sieve tests just to determine the percentage of adhering fines not removed by dry sieving. It is not unusual for fines circuits in commercial processing plants to be overloaded due to the fact that the size distribution of the mined ore is frequently finer that the size distribution of core samples, and some ores physically break-down when exposed to water and impact from screening and pumping. Wet-attrition testing can frequently reveal whether an ore is prone to size degradation through wet processing. A wet attrition test developed for coal could be used or adapted to other ores (Australia Standard). SinWFloat Analysis Strategically combined sieve fractions can be fractionated by specific gravity in the laboratory using heavy liquids and sinklfloat analysis. In sinklfloat analysis, sieve fractions are immersed in a series of organic liquids or inorganic salt solutions appropriate to the material being analyzed. For example, in coal analysis, one might select a series of liquids with specific gravities of 1.30, 1.40, 1.50, 1.60, 1.70, 1.80, 1.90, and 2.00. For sand and gravel, the series might range from 1.70 to 2.50. For iron orelsilicate separation, liquids ranging from 2.50 to 3.60 could be used. Some of the heavy liquids used in the past, and present are highly toxic, have associated environmental problems, or are quite expensive. In the 1978 edition of Mineral Processing Plant Design, Clerici's solution (thallous formate and thallous malonate) was recommended for high specific gravity separations, since this solution could reach a specific gravity of 5.2 at 95" C. Clerici's solution is no longer recommended because tests have shown it to be toxic, corrosive, and carcinogenic. ASTM recommends the organic liquids listed in Table 2 for coal, and of course these liquids are also applicable to mineral separation (ASTM).

981

Table 2 ASTM organic liquids Specific Name Gravity Petroleum Spirit 0.70 White Spirit 0.77 Naptha 0.79 Toluene 0.86

Name Perchloroethylene Methylene Bromide Bromofonn Tetrabromoethane

Specific Gravity 1.60 2.49 2.90 2.96

While not on this list of liquids suggested by ASTM, Methylene Iodide has a density of 3.31 g/ml and would therefore be useful in a number of mineral processing applications. Organic heavy liquids have a number of health, safety, and environmental issues that require special handling and disposal techniques. Another family of heavy liquids is based upon tungsten salt solutions and is used in siddfloat analysis. Table 3 provides information on these solutions. Table 3 Meta-tungstate solutions Name Sodium Metatungstate Lithium Metatungstate Lithium Metatungstate LST @ 80' C

Specific Gravity 2.25 - 3.00 < 3.00 < 3.50

While considerably safer to handle than Clerici's Solution or organic heavy liquids, viscosity is a consideration at higher specific gravities, so heating the solution, or using a centrifuge can reduce separating time. These solutions can be very expensive, with LST running over $ 500 per litre, or over $ 180 per kg. When using any heavy liquid, the process engineer should be alert to the possibility of interaction between the liquid and the ore sample being separated. In a recent project involving low rank coal, results from the laboratory appeared to be in error, and the author suspected that the coal absorbed solvent during the sink/float test, altering the specific gravity of the coal. Originally, the sample was immersed in the high-density liquid first, followed by immersion in successively lower density liquids. On retesting, the sample was tested in reverse order, and the assay results were significantly different and in-line with expectations. There are know interactions between the metatungstate solutions and certain minerals types such as chlorides and sulfides, so the process engineer should consult with the laboratory prior to selection of test liquids. Once size and sink/float analysis have been generated for the test samples, many manufacturers of gravity concentration equipment can predict or even guarantee the separation performance and capacity of their equipment based upon computer models of the data. Where there is little to no experience with gravity separation for a particular ore in a particular location, laboratory- or pilot-scale process testing can be of value to confirm process models and evaluate the response of ore constituents to a wet or abrasive environment. Many equipment manufacturers offer free or low-cost pilot testing, especially for relatively new technology. Some manufacturers offer rental evaluation units that can be moved to site and are capable of processing from 1 ton per hour to 20 tons per hour of ore. A number of private commercial firms and universities offer fee-based testing. This alternative provides independence in execution of the test program, and can incorporate multiple processes. While pilot testing adds additional costs to the test program, studies have shown that pilot testing ultimately saves money, especially for new, or unfamiliar processes (Fonseca 1988) by improving confidence in process assumptions and reducing risk associated with technology.

982

WET GRAVITY SEPARATION PROCESSES Commercial wet gravity separation processes in contemporary use include processes based on very old technology (sluices), fairly old technology (jigs), relatively modem technology (tables, rising current washers, spirals), and modem technology (centrifugaljigs). In addition, a number of gravity separation processes have been significantly upgraded over the last few decades by design improvements, improvements in the materials of fabrication, and application of modern instrumentation and control technology. Jigs Jigs are available in a variety of designs, reflecting the type of ore processed and the particle size of the feed. In jigging, ore is introduced onto a perforated plate in the washing compartment and water is pulsed through the perforated bed-plate. The pulsations cause lower density material to stratify on top of the heavier material. Products are separated by a weir device or through the bedplate. The pulsation effect may be created by either forcing water through the perforated bed-plate, or moving the bed-plate in a tank of water. Some early mechanical jigs used the mechanical action of a piston to create pulsations. One versions of this type jig, the Harz jig, is sketched in Figure 3.

Figure 3 Sketch of a plunger jig A major development in jigging technology, the Baum jig, was introduced by Fritz Baum in the late 19' century, and has been in wide use since the early part of the 20thcentury. A sketch of the Baum jig (Baum 1894) is shown in Figure 4.

Figure 4 Sketch of the Baum jig

983

Two features of the Baum jig that are particularly noteworthy are the use of compressed air to produce pulsations, and the geometry of the jig hutch, which is a "j-tube" design. The use of an air valve to introduce compressed air into the vertical leg of the j-tube allows accurate control of not only the expansion stroke, but also the suction stroke. The jig has evolved into a device that exploits the principles of differential acceleration, free settling, hindered settling, and consolidation trickling shown in Figure 5. The geometry of the Baum jig permits an even distribution of force over a wide bed, thereby allowing the construction of single units with capacities over 500 tons per hour. Modem versions of the Baum jig include the Batac Jig, a design with the air expansion chambers directly under the bed, eliminating some of the space occupied by the j-tube design.

I

I

Free Settling

Hindered

Consolidation Trickling

Differential Acceleration

Figure 5 Jigging sequence Plunger jigs, with the exception of the Jeffiey diapham jig, are used in mineral processing, whereas the Baum Jigs were developed for, and are used primarily in, the coal industry. While most plunger jigs resemble the Harz jig, and are small, fairly low capacity units, the IHC Cleveland circular jig is a unique design employing a plunger and diaphram to generate pulsations, and is capable of high capacities. A sketch of the circular jig is shown in Figure 6. Feed slurry, typically from a dredging operation, enters the center annulus, and is distributed radially to the trapezoidal compartments. A significant advantage of this design is the decreasing velocity as slurry moves though the compartment and is discharged to the perimeter launder. The trapezoidal design enhances fines separation.

Figure 6 Top view of circular jig

984

Jigs can accommodate significantly larger particle sizes than other conventional gravity separation equipment, but when properly configured, with sized feed, jigs can also make separations as fine as 0.075 mm for minerals of high specific gravity.

Shaking Tables The basis for the design of the modern shakmg table may be said to have originated with the introduction of the Wilfley table in the late 1800’s. The factors that come into play when separating with a shaking table include particle specific gravity, particle size, particle shape, and particle mass. In addition, the design of the shaking table and adjustments to the table configuration, such as the type of stroke employed, stroke frequency, stroke amplitude, side tilt, end elevation, deck surface material, dressing water distribution, feed solids, and feed rate also impact performance. The shaking table effects a separation by a combination of hindered settling, stratification, and consolidation trickling. A sketch of a typical shaking table is shown in Figure 7. Looking at the sketch, feed solids and water are delivered at the upper right-hand comer of the deck. The deck of the table slopes downward from the upper-right to the lower-right and also pitches downward from the lower-right to the lower-left. As denser particles accumulate between the riffles, the eccentric side-to-side motion of the drive transports the densest particles to the upper-left zone of the deck. Lighter gravity material is carried up and over the riffles by the flow of dressing water and slope of the deck, to the lower-right (Zimmerman, 1950).

Dressing

I

Feed

i

1

Middling

Figure 7 Sketch of a shaking table Tables, once in widespread use for separating medium to fine size particles (19 mm - 0.1 mm), have slowly been replaced by concentrating spirals. Tables require a relatively large amount of floor space per unit capacity, and achieving an even distribution of feed to a large number ofdecks can be a challenge. A large table circuit will also generally require one or more operators to maintain proper operating conditions and make adjustments for feed of variable composition.

Spirals Spiral concentrators, or spirals, have evolved into one of the more widely used gravity separation devices employed in mineral processing, since their initial introduction in the mid-1940’s. Application of improved trough geometry, optimized for various minerals, improved materials of construction, and the introduction of multi-start units are changes that have improved the performance of spirals. A sketch of a triple-start spiral is shown in Figure 8.

985

Feed/

Figure 8 Spiral concentrator A spiral might well be described as a centrihgal sluice. A slurry of fine particles (generally less than 1.O mm) is delivered to the helical sluice at a solids content of approximately 25% solids by volume. Low-density particles are thrown to the outer edge of the helix, while the higher-density material makes its way down the helix along the inner, low velocity region. A splitter device at the discharge of the helix is used to separate the products in to heavy, light, and middling fractions. The middlings fraction frequently contains larger-size, misplaced low-density material. Spirals are simple, have no moving parts, and when applied in multi-start configuration, have a relatively high capacity, versus shaking tables. Spirals also have relatively low maintenance requirements. The efficiency of spirals is relatively low, the performance is adversely impacted by feed variability, and there has been no incorporation of automatic control (to the author's knowledge) to compensate for changes in feed quality automatically, in real time. Efficiency can be improved by close sizing of the feed and re-treatment of the middlings product. As with any circuit consisting of numerous, low capacity devices, feed distribution and control issues can translate to lower efficiency for the circuit taken as a whole, than for any individual unit.

Rising Current Washers A number of gravity separation devices have been introduced over the years based upon the concept of using a dense, aqueous suspension of solids to form a teeter-bed of high-density particles to effect a separation. Contemporary versions of these devices are offered by allmineral and Stokes. A two-stage rising current washer is shown in Figure 9.

986

Feed

1

Figure 9 Two-stage rising current washer In the case of a two-stage rising current washer, the inner cylinder is a high-velocity section to separate coarse, high-density particles from the remainder of the feed. Generally, most of the solids overflow to the outer cylinder where a deep, teeter-bed of high-density particles form a barrier to the lighter particles above. Low-density particles overflow the perimeter of the outer cylinder and are collected in a radial launder. Coarse particles are discharged through a pinch valve at the bottom of the inner cylinder. Heavy concentrate from the secondary separation is discharged through dart valves. The pinch valve and dart valves are controlled by separate pressure sensors. As the teeter-bed builds to the desired control point, the sensors transmit a signal to timer-relays to activate the discharge valves. Dressing water is adjustable for both the primary and secondary separations. Rising current washers have a number of advantages fopseparating particles in the size range between 1 mm and 0.1 mm. First, they are high capacity units capable of treating several hundred tons per hour in a single unit. They can make gravity separations over a relatively wide range. Because of the deep beds of heavy and light particles, they can accommodate process surges relatively well. Built-in automation allows the units to adjust to varying feeds. Separation efficiency is better than that of spirals (Parekh, 2001). Rising current washers are primarily used in the sand and gravel and coal industries, but could also be of use in many mineral dredging applications.

Cone Separators The Reichert cone concentrator is a flowing-film separator based on the principles of the buddle, a device used in the middle ages. The Reichert cone, shown in Figure 10, is capable of capacities of up to 300 tons per hour of solids in a single unit. Feed particle size typically ranges from 0.5 mm to 0.05 mm. Reichert cones are effective as economical pre-concentrators for upgrading low-grade ores, scavengers for treating tailings, and in some cases can be used to make a final product. The Reichert cone has no moving parts and can easily be arranged to provide a multi-stage separation to improve the quality of the final concentrate.

987

Feed

I

Figure 10 Cone separator The separation characteristics of a flowing-film separator, such as the Reichert cones are affected by particle shape and size, as well as particle density; therefore, sizing feed ahead of the unit is beneficial.

Enhanced Gravity Separators Enhanced gravity separators, such as centrifugal jigs (Altair, Campbell, Falcon, Kelsey, Knelson) and tables (Multi-gravity concentrator) began appearing as commercial devices in the 1970's. A sketch of the Falcon concentrator is shown in Figure 11. Enhanced gravity devices use centrifugal force to accelerate the differential settling between fine-size particles of different specific gravity. Commercial devices are capable of generating up to 300 g's of centrihgal force. Feed

I

Lighi 0

Figure

lo0

Enhanced gravity separator

In the case of the centrifugal jig, the feed slurry is introduced at the base of the conical shaped bowl. As the solids move up the wall of the bowl, they are stratified according to density and highdensity particles are discharged through ports located near the upper portion of the cone-shaped bowl. Low-density particles are carried out of the bowl by the flow of slurry through the unit (Abela, 1997).

988

Enhanced gravity separators are increasingly applied to particle sizes once the sole domain of froth flotation. For gold concentration, enhanced gravity separators may be effective down to 5 microns, but machines can also be set-up to handle particles as large as 6 mm. They have been successhlly applied to ores such as chromite, gold, iron, mineral sands, silver, tantalum, tin, titanium, and to coal for reduction of pyrite and other minerals associated with coal. Enhanced gravity separators, while available as batch units, are also available for continuous processing at feed rates of up to 150 tons per hour.

DRY PROCESSES Commercial dry mechanical gravity separation technology was introduced in the early 1900’s and probably reached its zenith in the middle part of the century in the coal industry. Dry mechanical (as opposed to magnetic) separation takes advantage of material properties such as hardness, resilience, and coefficient of friction, as well as specific gravity. Both wet jigs and wet tables have counterparts that employ air instead of water to effect a separation. Although dry separation has been largely discounted in recent years, there have been recent developments resulting in improvements in machine design and control. Dry separation is generally less efficient than wet separation, and is sensitive to the surface moisture content of the feed. So why consider dry separation? The best reason is economics. In desert environments water may be unavailable, or prohibitively expensive for mineral processing. In coal processing, where energy is the final product, water may be as important a contaminant as mineral matter. Air-based gravity separation is a simple, low-cost technology, suitable for mobile applications and free from the environmental and regulatory concerns associated with mill water and wet tailings disposal. In short, where wet processing is impractical, and air-based gravity separation can provide an acceptable financial return, it can and should be considered for ore beneficiation. Presently, air tables and air jigs are employed by the minerals industry. Air Tables Air tables were among the earliest devices to be commercially successful for gravity separation. The Triple-S air table, shown in Figure 12, was introduced in the mid 1920’s in the coal industry. The air table is operationally similar to the wet shaking table, with the exception that air is introduced through the wire mesh deck to provide the fluidizing medium. An eccentric drive provides the shaking motion to the deck. A single machine that occupies approximately 6.5 m2 of floor space can process 5 - 15 tons per hour of material, depending on the size and density characteristics of the feed.

Lighter Middling Heavy Product Product Product Figure 12 Sketch of an air table

989

Stoners, an example of which is shown in Figure 13, use air to fluidize the particle bed and an eccentric drive to provide motion to the deck, but forego riffles. Instead, the shahng motion of the deck causes the heavier particles to be carried upward, against the flow of gravity. The lighter particles, entrained in the upflow of air through the wire mesh deck, flow with gravity and are discharged at the opposite end of the deck fiom the heavy particles. Both the air table and stoner are used in agriculture and recycling, but do have some applications in mineral processing, such as clay, perlite, vermiculite, slag, and gangue. In recent years, stoners and air tables were applied in series to remove pyrite and other mineral matter fiom a dried, synthetic coal product. (USDOE, 1997)

Feed

Light Product

Figure 13 Sketch of a stoner Air Jig Air jigs operate on essentially the same principles as wet jigs, but use air instead of water to fluidize and pulse the mineral bed. There have been two relatively recent developments regarding air jigs, one for the coal industry and one for the gold industry. Allmineral has demonstrated a 1.2 M x 2.4 M air jig capable of processing 50 tons per hour of 50 mm x 0 coal. The jig, shown in the sketch in Figure 14, employs a lower-pressure, high-volume source of air to fluidize the particle bed, and a higher-pressure, lower-volume air source to superimpose a pulsation stroke that stratifies the particles. Like a wet jig, the air jig exploits differential acceleration, hindered settling, and consolidation trickling to effect a stratification of higherdensity and lower-densityparticles. The new air jig differs fiom previous air jigs in the materials of construction, the nature of the jigging action, the principal of deep-bed separation, and automatic control of high-density material discharge. Instead of fine mesh wire cloth for the bed deck, punch plate is used. The advantage is longer wear life, reduced blinding, and improved air distribution. While prior jigs were design to maintain a thin bed depth along the length of the deck, the new air jig is designed to operate with a bed depth of up to 230 mm. A density monitor positioned near the discharge end of the machine provides the means to automatically compensate for variations in feed quality. Another recent development in air jigging technology is the Rotary Air Concentrator, a device demonstrated in a desert environment of Mexico for concentrating gold. The machine can accommodate feeds smaller than 14 mm and can recover heavy minerals down to 80 microns (Piggott, 2000). Feed, screened to minus 14 mm, is delivered to a central hub and distributed to the radial deck by revolving blades. This method of distribution also serves to maintain a level depth of material on the deck. Low-pressure pulsating air is delivered through 2 mm profile wire screens provides the jigging action to stratify the bed of ore. Fine gold-bearing black sand that pass through the 2mm openings of the deck are collected separately fiom the coarser gold particles.

990

Feed Hopper

Vibrator

-Pulsed

Air Inlet Fluidizing Air Inlet

Discharge Heavy Discharge Figure 14 Sketch of an air jig

FLOWSHEETS Flowsheets for wet and dry separation have a number of common features, as shown in the generic flowsheet below.

-

+

L

t

A

Tailings Recovery/Disposal

Figure 15 Generic flowsheet

991

1

For most gravity separation processes it is desirable to size the feed ahead of the concentration device. The concentrate, or product, will generally require subsequent treatment, such as dewatering and sizing, and tailings will also generally require some form of treatment prior to disposal. The most simple flowsheets will be those for dry gravity separation processes. Figure 16 illustrates one such flowsheet.

In a dry process, initial sizing of the feed and subsequent processing of sized feeds improves the separation efficiency. Oversize feed can be crushed and added to the screened coarse-size feed. Fine particles from both the coarse and small circuits are sized in dust cyclones, with the undersize products from the cyclones collected in a baghouse. The cyclone and baghouse products are added to either the product or tailing stream, as appropriate.

4I

Feed Screen

I Dust Cyclone

I

I

Product Belt

To Disposal

Figure 16 Dry separation flowsheet

992

Wet process flowsheets are more complex than flowsheets for dry processing due to the need to handle water in the process. A relatively simple wet process might consist of a Baum Jig for the particles between 50 mm and 1.0 mm, spirals for particles between 1.0 mm and 0.15 mm, and a centrifugaljig for particles finer than 0.15 mm. A flowsheet for this process is shown in Figure 17.

Classifying Cyclone

H e a d o Tank

;

:

Baum Jig

Feed Ore

Fine Dewatering Screen

Coarse Dewatering Screen

1

I

"I

I l

r

kreen

-t

Predict

Cyclone Sump & Pump

Clarified Water Sump & Pump

L-.,-,+

to tailings impoundment

Figure 17 Wet process flowsheet Minus 50 mm ore is wetted and fed to the coarse Baum jig. It is not necessary to size the feed prior to jigging. The heavy fiaction is removed from the jig with a bucket elevator and the lighter gravity fiaction and majority of the minus 1.0 mm particles are washed over the discharge weir of the jig. The light discharge is screened to separate the jigging water fiom the solids and remove the minus 1 .O mm particles. The minus 1.0 mm slurry is sized using classifying cyclones to produce a 1.0 mm x 0.15 nun oversize product. This fraction is processed in spirals. The light fiaction and heavy fraction are mechanically dewatered and dned using inclined vibrating screens and centrihgal driers. The slimes fiom the process are treated in a clarifierhhickener to concentrate the solids for disposal to a tailings impoundment and recycle process water to the plant.

993

REFERENCES Agricola, G., translation by H.C. and L.H. Hoover. 1950. De Re Metallica., New York: Dover Publications, Inc. Australia Standard, AS1661 Method C-2. ASTM Annual Book of Standards. 2001.D 4371-91 Standard Test Method for Determining the Washability Characteristics of Coal, Volume 05.06, p. 398. Fonseca, A.. 1988. Coal Quality and its Effects on Combustion. In Industrial Practice of Fine Coal Processing. Chapter 3 . Littleton, Colorado. Society.of Mining Engineers. Zimmerman, R.E. September 1950. The Cleaning of Fine Sizes of Bituminous Coals by Concentrating Tables, Mining Engineering, pp. 956 - 966. Parekh, B.K. 2001. et al, Innovation in Coal Cleaning, Coal Prep 2001, pp. 191-206,2001. Abela, R.L. June 25 - 26, 1997. Centrifugal Concentrators in Gold Recovery and Coal Processing, Extraction Metallurgy Afiica. Piggott, K. October, 2000. Air Separation - Success in Mexico, Mining Magazine, pp. 32 - 35.

994

THE SELECTION AND SIZING OF CENTRIFUGAL CONCENTRATION EQUIPMENT; PLANT DESIGN AND LAYOUT Andrk R. Laplante’

ABSTRACT A rational approach to the design of gold gravity circuits nested in grinding circuits using semi-

continuous centrifuge units is first presented. The need to understand the economic impact of gravity recovery is stressed, as the size and nature of the gravity circuit should depend greatly on how it affects overall recovery. Gravity recovery depends primarily on the quantity and size distribution of the gravity recoverable gold (GRG), but also on how often gold is recycled in the grinding circuit, which is largely dictated by its product size (Pso). Finally, gravity recovery will also depend on the effectiveness and capacity of the recovery units. These factors are briefly reviewed. Continuous centrifuge units are then presented and typical applications are discussed. Economic and metallurgical criteria must be matched, and potential synergies between the more expensive centrifuge units and cheaper, typically non-centrifuge gravity units must be exploited. INTRODUCTION Full-scale centrifugal gravity concentrators can be divided into two very different types of applications. Firstly, semi-continuous2 centrifuge units are ased for gold recovery, most often from the circulating load of grinding circuits. These units can only achieve very small weight recoveries into the concentrate stream, typically from 0.01 to 0.1%, and their use is limited to high value, low-grade metals such as gold and platinum group elements. Secondly, continuous centrifuge units are also used for the recovery of minerals or metals other than gold, typically cassiterite, zircon, chrome/chromite, tantalum-bearing minerals and gold carriers. Weight recovery then varies from less than one percent to as much as 50% percent. Whereas the semi-continuous units are limited in application, their simplicity, reliability and very low operating costs make them by far the most common centrifuge application, outnumbering continuous centrifuges more than 10 to 1. Accordingly, they will be the main focus of this contribution. Fully continuous centrifuges, however, are inherently more versatile and are likely to gain in popularity over the next ten years. They will be briefly discussed in the second part. This contribution is based on bench and pilot test work discussed in Spiller and Laplante (2002), which ideally should be read before, especially for the section pertaining to semicontinuous units. It includes a significant amount of information relevant for purposes of plant design that will not be repeated here. SEMI-CONTINUOUS UNITS FOR GOLD RECOVERY The recovery of gold by gravity pre-dates flotation or cyanidation, and even amalgamation, by thousands of years. Yet in the context of the modern gold mill, it is often a misunderstood topic, even with those familiar with other applications of gravity recovery or with other gold recovery methods. The misunderstanding stems from the high circulating load of gold, which makes possible the unique approach used for its recovery-i.e. slowly bleeding it from the circulating load of the grinding circuit. Often as little as 2% of the gold in the circulating load is recovered at each pass. Overall gold recoveries of 20 to 50%, or one to two thirds of the gold recoverable by gravity (GRG), are nevertheless achieved. Contrast this with most mineral recovery circuits, in

’ McGill University, Department of Mining, Metals and Materials, Montreal, Quebec, Canada ’ The term “semi-continuous” units is used because the feed and tailing streams are continuous, but the units must be stopped to recover the concentrate.

995

which the feed is exposed to rougher-scavenger recovery only once, and recoveries approaching 100% of the recoverable species of value at acceptable concentrate grade are targeted. Other notable differences include the very high upgrading ratios, very low yields (concentrate weight recoveries), and much smaller size and capacities of the cleaning circuits (gold rooms), when compared to the roughing stage. This presentation will expose the different steps necessary to perform the preliminary design of a gravity circuit: an evaluation of the need for gravity recovery, 0

characterization of the gravity recovery potential of the ore,

0

a prediction of how much will be recovered, selection of an optimum recovery effort (to be defined), and unit selection and circuit design.

This presentation is largely based on research work based at McGill University over the past The most important contribution of this work was the refinement of the concept of gravity recoverable gold (GRG), its measurement in ores and in streams of gravity and grinding circuits, and application to the design and optimization of gold gravity circuits. Also widely available are a large number of practical papers detailing the use of gravity recovery to recover gold from grinding circuits are available and essential for a good understanding of existing practice, as well as equipment manufacturers websites, some of which are included in appendix. These papers not only provide insight into circuit design and operating practice, but also shed some light on the impact of gravity recovery on overall recovery. 15 years (Table 1).

TABLE 1 Some References on the Concepts Summarized in this Work Reference Content First study of a gold gravity circuit using a preliminary concept Laplante et al, of GRG recovery in gravity units 1989 Basic description of how gold responds in ball mills and Banisi et al, 1991 hydrocylones in grinding circuits First study of a gold circuit using the Knelson Concentrator Laplante and Shu, 1992 MD3-based measure of GRG Original presentation on the GRG test Woodcock and Laplante, 1993 First general presentation on practical aspects of gold recovery Laplante et al, 1994 Presentation of the model used to predict gravity recovery from Laplante et al, the results of the GRG test and a description of the behavior of 1995 GRG in grinding, classification and recovery units Study of the performance of recovery units when recovery GRG Laplante et al, from high density gangues 1997 and 1998 Basic guide-lines for the design of gold rooms Laplante et al, 1999 Discussion on the nature of the GRG and other tests used to Laplante, 2000b characterize the recoverability of gold by gravity Second general presentation on practical aspects of gold Laplante, 2000c recovery Laplante and Xiao, A simplified approach to the prediction of gravity recovery using multilinear regressions and the gravity recovery effort 200 1 Upgrade of the Woodcock and Laplante 1993 on the GRG test. Laplante, Huang with a summary of the data bank and Woodcock, 200 1 Application of the GRG-protocol to the study of flash flotation Laplante and in a grinding circuit Dunne, 2002

996

Step one: Understanding the Economic Incentive of Gold Gravity Recovery The economic incentive of recovering gold ahead of cyanidation or flotation circuits is often misunderstood. It is defined here as the net contribution of the gravity circuit to overall recovery, both in terms of actual recovery and decrease in capital and operating costs. It is in fact the answer to the question: what would be the drop in overall recovery or increase in operating costs that would result from the elimination of the gravity circuit? It is seldom more than a 1 to 3 Yo contribution to overall recovery, even when gravity recovery is high (40 to 70%), and can be as low as 0.2%. Ahead of flotation, it can be significantly more, especially for low-grade copper gold ores or for copper-zinc ores, the former because of the very high upgrading ratios and the latter because of the increased selectivity of flotation, which displaces some of the gold to the zinc concentrate. Net smelter return is higher for a gravity concentrate (99.7%+), compared to a copper concentrate (92-98%), and even more so a zinc concentrate (NSR as low as 0%). Other ores that benefit greatly from gravity recovery are severe preg-robbers (Lewis, 1999). Increased overall recovery may not be the only economic factor when considering gravity recovery. Additional factors include decreased costs for the downstream circuit, both capital and operating, reduced gold inventory, and increased security. The latter is a significant factor in the renewed interest in gravity recovery in South Africa. Most Australian mills use coarse grinding and very short retention times in cyanidation circuits, both factors contributing significantly to the economic impact of gravity recovery. Many Australian operations also suffer from very poor water quality (total dissolved solids of 50,000 to 300,000 ppm) and abundant coarse gold (defined here as gold coarser than 300 pm). The first step when considering gold gravity recovery is to estimate the potential economic impact of gravity recovery as a function of how much is recovered. The simplest approach is to consider a linear relationship between gravity recovery and its economic benefit, and to use reported results of similar operations to estimate the relationship. Example of economic benefit. An operation is planned for the production of 4 t of gold per annum. A similar operation reports an impact for gravity recovery of 1% extra gold recovery (based on proposed production) at 40% gravity recovery. What will be the economic impact of gravity recovery? Answer: One percent recovery represents 40,000 giyear, or a contribution of 1,000 g/year per 1% gravity recovery, using the above assumptions. At $USlO/g, a 40% gravity recovery represents $US4OO,OOO/y; each extra 1% in gravity recovery adds $US1 O,OOO/y. Because the economic benefit of gravity recovery can vary significantly from project to project, the extent to which gravity should be used also varies, even when the amount of GRG is relatively constant. Contrast two CIP applications, one in Canada with a 48-h leaching time, a 80% passing 75 pm grind and excellent water quality, to the same application in Australia with a 16-h leaching time, a 80% 175 pm grind and very poor water quality. The benefit of gravity recovery is much more significant for the Australian application. These two cases are not caricatures or oversimplifications, but actually quite characteristic of gold mines in the two countries, the Canadian operation typically a high-grade underground one, and the Australian one a low-grade open cut (or combination underground open cut, with most of the tonnage from the open cut). Step 2: Characterizing the GRG The second step is to estimate the potential of the ore for gravity recovery. There are a number of suggested approaches. It is important at this point to understand the difference between the potential of a gold ore for gravity recovery, and a prediction of how much gold will be recovered. Actual gold recovery is a function of three very different sets of variables, the first ore-specific, the second grinding circuit specific, and the third gravity circuit specific. When characterizing the ore, only the first should be measured, as the final flowsheet is normally not selected yet. It is therefore inappropriate at this point to try mimicking the gravity circuit for the purpose of predicting gravity recovery. The McGill GRG test is detailed in a parallel chapter of this book (Laplante and Spiller, 2002); it is a release analysis that used a laboratory centrifuge, a Knelson Concentrator MD3 (Laplante et al, 2001).

997

Figure 1 shows typical GRG responses, presented as a cumulative retained curve. The curve “Low” shows an ore for which gravity recovery would be inappropriate, because the GRG content is low and fine. This type of ore is often encountered in the American cordillera, either in North or South America, typically in gold-copper porphyric deposits. Other factors that would contribute to a low gravity recovery are a high gangue s.g., gold particles with a high silver content and a coarse grind size. Curve “Average” shows a typical response for a free milling ore, which should yield a gravity recovery of 30% to 40% for many applications, and curve “High” shows an ore very amenable to gravity recovery, which could easily yield gravity recoveries in excess of 50% at full scale.

I 100

”.

~

.. ....

. ,~

..

I”

”‘“-””‘“I

80

..

--t Low

60

-f-- Awrage

40

-A-

High

20 0

10

I

100

1000

Particle Size, pm

FIGURE 1 Typical GRG Responses The GRG test of McGill University is only one of the methods that can be used to determine the gravity recovery potential of a gold ore. Whatever method is selected, it is essential for circuit design that not only the amount, but also the size distribution of the GRG be determined, as it is essential information to choose the aperture of screening equipment ahead of gravity and predict the amount of gold that will be recovered. Step 3: Predicting Gravity Recovery to Achieve an Optimum Recovery Effort In typical recovery circuits, material is fed to the circuit only once. Circulating loads, if they exist at all, are minimized, as the consensus is that high circulating loads are deleterious to overall circuit efficiency. Recovery downstream of the main recovery circuit is minimal or non-existent (i.e. of values reporting to the scavenger tailing stream). As a result circuits are designed with adequate retention time to maximize recovery at optimum concentrate grade (for flotation) or at adequate leaching time (for cyanidation). Gravity recovery from the grinding circuit follows a different logic. Circulating loads of GRG are typically 2,000 to 10,000% in the absence of gravity, an indication that GRG, and in particular coarse GRG, will circulate from twenty to hundred times on average before reporting to the cyclone overflow. In the presence of gravity recovery, circulating loads decrease from 500% to 2000% (the higher the recovery effort, the lower the circulating load). The logic of gravity recovery is therefore to rely on the high number of times GRG goes around the circulating load (i.e. the number of passes) to achieve adequate gravity recoveries. At each “pass”, as little as 1% of the GRG can be recovered, or as much as 15%. The amount of GRG recovered at each pass is in fact a key variable affecting how much gold will be recovered by gravity, and defines what is termed the “gravity recovery effort.” The gravity recovery effort (or simply “recovery effort”) is an important concept because it correlates particularly well with how much GRG is recovered (see Fig. 4), and is the most important factor affecting gravity recovery after the amount and size distribution of GRG.

998

Example of gravity recovery effort. From the cyclone underflow of a grinding circuit 25% (one cyclone underflow out of four) is bled and screened; the screen undersize feeds a centrifuge unit. Screening efficiency is 70%, and the centrifuge unit recovers 30% of the GRG in its feed. The centrifuge concentrate is treated in a gold room with an 80% GRG recovery. What is the gravity recovery effort? Answer: Assuming that all units except the cyclones are in open circuit (which is generally the case), the recovery effort is equal to the product of the recoveries of the various units: 25% for the bleed, 70% for the screen, 30% for the centrifbge and 80% for the gold 25% 70% 30% room, or 5.6% (= x-xx 80% 1. 100%

100% 100%

The third most important variable that affects gravity recovery is “the number of passes” -i.e. the ability of classification to keep GRG in the circulating load until it is recovered. This is priinarily dependent on the partition curve of GRG, which in turn is largely dictated by the target grind size of the ore, or Pso, which can vary between 53 pm and 230 pm. Because the link between the partition curve of GRG and grind size is still not fully understood, it is preferable, for retrofit (“brownfield”) applications, to measure directly the GRG partition curve. This is achieved by determining the amount of GRG for each size class of the cyclone overflow and underflow. Figure 2 shows two typical partition curves for ore and GRG. Such curves are characterized, for the purposes of simulating gravity recovery, by the proportion of GRG in the minus 25 pin fraction reporting to the cyclone underflow, or R-25p,,,. The first partition curve of GRG is extremely high, and for well-operated cyclones and a Ps0 of 75 pm, the amount of GRG reporting to the cyclone overflow above 25 pm is negligible. The second GRG partition curve is obtained when a much coarser grinding is targeted, in this case 63% passing 75 pm. The difference between the two GRG partition curves is very significant, and would result in dramatically different gravity recoveries being achieved, everything else (i.e. the GRG amount and size distribution, the gravity recovery effort) being equal. The partition curve is essentially determined by the required fineness of grind.

100

&

80

+- Ore (coarse) -I-GRG (coarse) --t Ore (fine) -x-- GRG (fine)

3 0 0

0”

-s

60 40

0

20 0 10

100 Particle Size, urn

1000

Figure 2 Typical Partition Curves for Ore and GRG (fine: grinding product of 83% passing 75 pm; coarse: grinding product of 63% passing 75 pm)

The unique behavior of GRG in cyclones explains most of the high gold circulating loads observed (slow grinding kinetics is also a contributing factor) and why it is important to describe such a behavior. Target grind size also affects GRG grinding kinetics and how much fine GRG will be liberated. Both effects are less significant than that of the partition curve. As previously stqted, maximizing gravity recovery should not the main objective of the gravity circuit, as the downstream circuit would recover most of the gold not recovered by gravity. The optimum gravity recovery is dictated by its economic incentive. This will be illustrated later on.

999

Figure 3 illustrates the interaction between the nature of the GRG, the recovery effort and the economic incentive of gravity recovery. Xiao (2001) has attempted to represent the link between the recovery effort, the size distribution of GRG and the fineness of grind by simple multi-linear regressions. Two regressions were produced, one for fine GRGs and one for coarse GRGs. For the fine GRG size distribution data set: RfGRG = -233.09 + 17.10*ln (Re)+3.61*ln (%)*In (~)+60.71*ln(R25pm)-11.92*ln(T) -4.34*ln (GRG.25 pm) -57.77*1n (GRG.75 pm) +55.51 *In (GRG-150,,,I EQ. 1 For the coarse data set: RCGRG = -65.4 + 15.59*ln (Re)+5.49*ln (%)*In (~)+37.81*ln(RZ5 ,,)-17.26*1n (T) -30.04*ln (GRG.75Lun) +12.67*1n (GRG.150b,m) EQ. 2 Where: R‘GRG and K G R G are the GRG recoveries of fine and coarse GRG size distributions, respectively. GRG-25pm is the cumulative GRG content below 25 pm, in % GRG-75pmis the cumulative GRG content below 75 pm, in % GRG-150pm is the cumulative GRG content below 150 pm, in % k is the recovery effort T is the dimensionless residence time in the ball mill as defined in Xiao (200 1) R25pm is the proportion of GRG finer than 25 pm reporting to the cyclone underflow, which is used to represent the partition curve of GRG.

Figure 4 shows that the link between GRG recovery and the recovery effort is far from linear; rather, GRG recovery is generally proportional to the logarithm of the recovery effort’, which implies that there is a rapidly diminishing return to increasing the recovery effort. Use of an incorrect recovery effort either results in a failure to fully tap the economic benefit of gravity recovery, or the bloated capital cost of an oversized circuit. Figure 5 (Xiao, 2001) shows that as long as the GRG content is determined accurately, the recovery effort is properly estimated and a reasonable estimate of the partition curve of GRG is available, the predictive model of Xiao is reliable. The GRG test, when performed on a representative sample, provides the necessary information about the ore, leaving a proper estimation of the recovery effort and the partition curve of GRG as the two most challenging steps. A databank is being presently built of GRG recovery as a function of operating conditions for various gravity recovery units. Earlier work on the unit most used for gold recovery by gravity, the Knelson Concentrator, requires updating as units have evolved (bowl design, rotating speed) and are now fed at higher throughputs than twenty years ago (i.e. 50-70 t/h for a CD30 as opposed to 10-20 tih, as in Laplante et al, 1989). Other units also used for gravity recovery (Falcon SB, Gekkos In-Line Pressure Jig) have been little studied at plant scale on the basis of size-by-size GRG, but have also been proved capable of effective GRG recovery (Gray, 1997; Goulsbra et al, 1998). For design purposes, overall GRG recoveries (recovery effort) and Xiao’s regressions can be used to predict recovery. For optimizing purposes, it is best to measure size-by-size GRG recovery and partition curves. These data can then be used in a more accurate size-by-size GRG recovery model (Laplante et al, 1995). For brownfield applications, either approach could be used. Care should be taken when using an overall recovery, as GRG recovery is extremely size dependent, and overall recovery is strongly dependent on the size distribution of the GRG in its feed. This is illustrated in Figure 6, which shows that recovery is much more sensitive to particle size than operating variables (for Fig. 6, fluidization flow and rotation velocity).

.’The relationship departs from linearity only when GRG recovery exceeds 80%. This is achieved only for coarse GRG distributions at recovery efforts exceeding 10%.

1000

Gravity Circuit cost

Gravity Recovery

Optimum Design

FIGURE 3 Technico-economic Evaluation of Gold Gravity Recovery

1001

I00 ,

S

80

*---__

.

2

'(3

- ...

20

&Coarsest

- 1

G RG ~

_

._

-

A

,

0

-m-

Finest

GRG

10

1

:

.- ..

.. . .

100

Recovery Effort, % _.___

___

_.. -_

FIGURE 4 GRG Recovery as a Function of the Gravity Recovery Effort (Xiao, 2001)

80 60 40 20 0 ' 0

#L . .

~

. .

I

20

40

60

80

100

Measured Gold Recovery, %

Figure 5 Correlation Between Gold Recovery as Predicted by Regression Model and Actual Plant Recovery (Xiao, 2001)

1002

FIGURE 6 Typical Size-by-size Recoveries for a Centrifuge Unit (Knelson XT20) (Legend: first number: ## of g’s; second number: fluidization flow in USGPM) TABLE 2 Measured GRG Recoveries for 20 and 30 inch-diameter Knelson Concentrators Feed Rate, t/h 20-in units 30-in Units 3 93 5 72 35 30 10 90 25 60 25*

50 (*: includes screen efficiency)

Table 2 shows some typical Knelson Concentrator GRG recoveries measured at plant scale. Some of these recoveries include screen efficiency, as the typical Knelson-supplied screen does not allow separate measurement of screening and Knelson efficiency. GRG recovery is also affected by top feed size, gangue specific density, recovery cycle length and rotating velocities. Thus the recoveries of Table 2 should only be used as guidelines. Conditions deviating from “typical” should always be factored in; for example, a high gangue s.g. lowers GRG recovery dramatically (Laplante et al, 1997 and 1998). Nevertheless, since GRG recovery is proportional to the logarithm of the recovery effort, estimates of GRG recovery are much more accurate than estimates of the recovery effort from which they are derived. The last link to the estimation of the optimum recovery effort consists of testing various alternatives to retain the best one based on accepted evaluation techniques (Laplante and Xiao, 2001). Usually three or four scenarios provide a solution that is near optimum. Various scenarios for a free-milling gold ore could be: 0 Nogravity 0 A low recovery effort of 1 to 3% using a “undersized” centrifuge unit processing a roughly screened feed (stationary screen) or a jig, and a bare-bone gold room to treat the gravity concentrate An average recovery effort of 5 to 10% using more centrifuge capacity with adequate feed preparation (i.e. vibrating screens) and a more complex gold room 0 A large recovery effort of 12 to 16% with either very large centrifuge capacity or a coinbination of centrifuges and jigging followed by intensive cyanidation of the gravity concerpte

1003

For gold-copper ores, options would also include flash flotation, or a combination of flash float and gravity (Putz et al, 1993; Laplante and Dunne, 2002). This is particularly appropriate when the GRG content is low (20 to 40%) and fine, or when flotation is used downstream in the main circuit. For “extreme gravity applications”, whre gravity is the only recovery method used, recovery efforts exceeding 20% can yield GRG recoveries exceeding 90% have been reported, but not in detail (the applications take place mostly in Russia and Ukraine, Van Kleek, 2001). Of the options described above, none is intrinsically better than the other: they all reflect different economic incentives, from a contribution as low as 0.2% of the value of the gold in the feed to contributions approaching 100% (i.e. for “extreme” gravity recovery). The approach of Figure 3 will now be demonstrated using two case studies. All cost estimates are approximate, and used only to illustrate the general approach. For each case study only two alternatives will be considered, for the sake of brevity. All costs and benefits will be expressed in US dollars. Case 1: Typical Canadian Underground Operation: Data: 1800 t/d, production 3 tly Au, GRG content of 67% with average distribution (Fig. 2, curve 2), net economic impact of gravity of 0.5% of production at gravity recovery of 40%. Value of gold of $lO/g. Single stage milling, T value of 1.7; a Ps0 of 75 pm for the ore, corresponding to a R-25LL,,, of 78% for gold. Alternatives: modest gravity recovery effort or no gravity. Discounting rate of 12%. Mine life of 8 years. No gravity recovery is the base case with a net present value (NPV) of 0. A modest gravity recovery effort would consist of a static screen feeding a CD30 Knelson Concentrator and a “bare-bone”gold room (80% recovery). A bleed of 50 t/h will be taken from a circulating load of 200 t/h. Screen efficiency is assumed to be 50%, as is the Knelson unit recovery. The capital cost of the modest recovery effort option is evaluated at $300,000. The fine GRG regression requires the YOpassing 25 pm, 26.1%; 75 pm, 56.7%, and 150 pm, 75.9% (provided by the GRG test). The recovery effort is calculated for both options as follows: Fraction of the circulating load treated: Screen efficiency Centrifuge unit recovery Gold room recovery Overall recovery effort 45.2% 30.3%

GRG Recovery (from Eq. I ) Gold Recovery (GRG Rec.*GRG Content)

Net Increased gold recovery ($1~) Net Present Value for 8 years production ($)

$1 14,000 $264,000

The above result is very informative: even a small economic impact can justify a modest recovery effort. Additional benefits that could easily exceed that calculated above would also accrue, such as a decrease in gold inventory, reduced gold losses that are not accounted for in general metallurgical accounting (e.g. gold that reports to spent mill liners and building foundations), and protection against surges in head grade (normally accompanied by a coarser gold size distribution) and carbon fouling. For this first case study, it is very unlikely that a larger recovery effort would be economically justified.

1004

Case 2: Typical Australian Open-Cut Operation: Data: 5000 tid, production 8 tiy Au, GRG content of 85% with coarse distribution (Fig. 2, curve 3), net gravity impact of 3% of production at a gravity recovery of 50% on account of short retention time in the CIL circuit and poor water quality. Use a value of gold of $IO/g. The grinding circuit consists of single stage milling, with a 7 value of 1.7, and a Ps0 of 75 pm for the ore, corresponding to a R-2=,b,,nof 78% for gold. Alternatives: modest gravity recovery effort (same as above) or average effort, with three centrifuge units in parallel fed by vibrating screens and a gold room with two tables and one scavenging unit. The annual operating costs of these two approaches are $70,000 and $150,000, respectively. The capital cost of the modest recovery effort option is evaluated at $300,000. The capital cost of the average recovery effort is evaluated at $700,000. The coarse GRG regression requires the YOpassing 75 pm, 32.2%, and 150 pin, 59.5% (provided by the GRG test). The recovery effort is calculated for both options as follows: Low Effort 15% 67% 40% 80% 3.2%

Fraction of the circulating load treated: Screen efficiency Centrifuge unit recovery Gold room recovery Overall recovery effort

59.3% 50.4%

GRG Recovery (from Eq. 2) Gold Recovery (GRG Rec. * GRG Content)

Average Effort 30% 80% 50% 90% I0.8Yo

8 1.7% 69.4%

Annual Increased gold recovery (‘OOO$) Net Annual Benefit (‘OOO$)

1,512 1,442

2,082 1,933

Net Present Value for 8 years prod. (‘OOO$)

6,863

8,897

The intermediate recovery effort yields a higher NPV than the modest one, even though the modest effort yields more than two-thirds the gravity recovery of the average one. If the recovery effort is further increased from the average case, hrther gains in NPV value are very modest. For example, installing intensive cyanidation for a gold room recovery of 99% yields a slightly lower NPV than that of the intermediate case when additional capital ($200,000) and operating costs ($3O,OOO/y) are factored in. What would make intensive cyanidation attractive in this case would be the increased security of the “hands-off’ operation, as well as the guarantee that a very high gold room recovery can be achieved (typically 97% or more), which removes a significant source of uncertainty when predicting circuit performance. Intensive cyanidation would have a major impact if gold room recovery were much below 90%, which is often the case for fine GRG. Equipment Selection, Scale-up and Layout Equipment selection: Typical applications would consist of installing a centrifuge unit in the primary circulating load. In the presence of significant amounts of GRG above 105 pin, the use of a .jig in conjunction with a centrifuge can be advantageous, especially when high recovery efforts are targeted (e.g. Kundana Mine, West Australia). The choice of screen ahead of the primary recovery unit is important, especially if relatively fine screening (which for gold applications would range typically from 800 to 1200 pm) if needed. The author advocates the use of horizontal or near-horizontal vibrating screens, which are less likely to bypass the very fine fraction that often has the highest gold content. The choice of primary recovery unit would probably hinge on a

1005

centrifuge (two suppliers) or jigs. The InLine Pressuge jig offers definite advantages over the traditional duplex jig, in particular the ability to use the bedplate motion to control bed dilation, rather the fluidization flow (Gray, 1997). Gold room recovery can be effected with the traditional shaking table approach, which can be improved with the appropriate flowsheet including magnetic separator (preferably drum rather than belt), screen, centrifuge unit to scavenge fine GRG from the table tail (Laplante et al, 1999). A small ball mill can be used to re-process table middlings if GRG is poorly liberated (mill discharge back to table). Alternately, intensive cyanidation can be used (Gray and Katsikaros, 1999; Lethlean, 2000) and offers many advantages, including that of eliminating scale-up problems and performance uncertainty associated with gold room performance (since recovery typically exceeds 97%). When choosing equipment, the nature of the equipment is important (e.g. semi-continuous centrifuge vs. jig, etc.. .), but the choice of supplier can also be critical. The quality of servicing is as important as the equipment itself. This sometimes tips the balance in favor of the largest supplier. For semi-continuous centrifuge units, the experience of Newcrest in Australia is interesting, as the have successfully relied on Falcon’s SB unit to process flash flotation concentrates and Knelson Concentrators to process cyclone underflows (Gregory et al, 1996). There is a perception in the industry that the latter is a more rugged unit than the former, but very few users are willing to come up with actual comparisons. For very large throughputs, the 48-inch Knelson offers capacities that the largest Falcon SB cannot match. Scale-up: As shown above, scaling-up gold gravity performance should not be based on bench scale performance, particularly for semi-continuous centrifuge units. Rather, established plant unit performance should be combined with GRG data and the recovery effort based on existing full-scale circuits. Circuit Design: Some design criteria are outlined in Laplante (2000~).They include: Built-in reliability is as important as performance, especially is there is a single primary recovery unit. Reliability is very often linked to screening and pumping/piping rather than the primary recovery unit itself. Reliability is often poor at plant start-up, when cash-flow is particularly important and coarse gold most susceptible to settle in poorly mixed areas. Do not neglect the efficiency of the gold room, which can be very low. The use of intensive cyanidation should be examined, and is probably the best option in the absence of significant concentrations of cyanacides in the gravity concentrate. Do not scale-up on the basis of bench scale performance. Screening can be critical, especially for difficult applications (fine GRG, high gangue s.g.). A more expensive screen may be required to feed enough tons of the correct size distribution to the primary recovery unit. Create effective circulating loads. A primary recovery unit capable of high recoveries with a modest upgrading ratio in closed circuit with a centrifuge unit is a good example of these useful circulating loads. Layout: For primary recovery, the primary circulating load (often the sole circulating load) should be targeted first, either at the cyclone underflow or ball mill discharge. The former is preferred for lower recovery efforts, typically with 20 to 30% of the circulating load diverted to the screen ahead of the primary recovery unit. The latter is preferable for larger recovery efforts, which may tap 50% of the circulating load or more. This may require additional pumping, but has the merit of insuring that all the cyclone underflow is directed to the ball mill. Some Australian operations tap the first portion of the trommel undersize, the rationale being that this material is upgraded (i.e. the trommel acts as a rough pre-concentrator). Gold room layout should include adequate headroom to insure gravity flow between the various units. The reader is referred to Laplante et al(l999) for more details. It is important to ensure that the primary concentrate can be pumped to the gold room, rather than being recovered in tote boxes that overflow and cause very significant fine gold losses. The primary concentrate reservoir should overflow to a secondary reservoir that is flushed between concentrate dumps, once the fine gold has settled, again to minimize fine gold losses.

1006

CONTINUOUS CENTRIFUGE UNITS Equipment Selection The four centrifuge separators to be discussed all have non-centrifuge equivalents. The first, the Kelsey centrifugal jig, is now distributed by Roche via Mineral Technologies. The second, the Falcon C (Continuous) Concentrator, is in fact a centrifuge Reichert cone. The third, the Mozley MGS/MeGaSep, is a centrifuge table (with an optional shaking action). The fourth, the Knelson CVD, is a relatively new comer whose recovery principle is much like a hydrosizer. The four units are discussed in Laplante (2000). Broadly speaking, the centrifuge units have the strengths and weaknesses of their “1 g” counterparts. For example, just as tables offer the best performance of the four but suffer from low capacities, so does the MGS, which can only treat 2 to 5 t/h. The MeGaSep is a larger version of the MGS that can treat roughly ten times the throughput, but the ability of its rake mechanism to move large quantities of concentrate is still unproven (whereas the capabilities of the MGS 902 are well established). The Falcon C generally yields very low upgrading ratios, much like its I -g counterpart, the Reichert cone. Because they excel over different size ranges, these four units are seldom in competition with each other. The MGSIMeGaSep is best in the 10 to 50 pm range, whereas the Knelson CVD, with a continuous addition of fluidisation water, works better above 50 pm. The Kelsey and Falcon C fit between these two units, with capacities similar to the Knelson and much above the MGS and higher than the MeGaSep. These two units were evaluated for tin recovery at CVRD (Costa e Silva et al, year unknown), the Kelsey recovering slightly more than the Falcon, but at a much higher enrichment ratio. Similar results were also obtained with iron ore tailing streams (Laplante et al, 1993). The MGS/MeGaSep is used to treat conventional flotation concentrates, and seems to be the only continuous centrifuge capable of doing so4. It can also be used to scavenge fine very heavy minerals (i.e. s.g. of 7 to 9) from gravity tailing streams. The Falcon C normally yields very low upgrading ratios, which makes it an ideal scavenger. for example at Tanco (Deveau, 2000). It is also used at Echo Bay’s Republic Mine to preconcentrate gold and gold carriers (mostly pyrite) for an intensive cyanidation. It has been tested for sulfur rejection from coal, with some degree of success (Luttrel et al, 1995). The Kelsey jig has also a number of applications in tin (mostly) and tantalum recovery, both in Australia and South America (Wyslouzil, 1990; Beniuk et al, 1994). It is also used to treat beach sdnds, either for fine heavy minerals recovery or the rather difficult zircon-kyanite separation (s.g. of 4.65 and 3.63, respectively). A recently commissioned circuit in Western Australia will scavenger gold and gold carriers from a deslimed cyanidation tailing, using three Kelseys 51800 fed at approximately 80 t/h each. The higher capacity than originally rated was achieved at the expense of concentrate grade, but the concentrate is a relatively fine product that can easily be upgraded in the existing spiral circuit. This interaction between the more costly centrifuge units and the cheaper I-g units is also used at Tanco, as the concentrate of a two-stage Falcon C circuit is recycled back to the main gravity circuit. Note that when a more costly centrifuge unit is used in series with a cheaper I-g unit, the latter should not necessarily be used as rougher. That centrifuge units should often see roughing duties may appear counterintuitive, given their higher cost. Mathematically, however, it can easily be shown that when two units are used in closed-circuit series, a much higher overall recovery is achieved if the unit with the highest recovery as rougher (Laplante, 2000~). Thus if the centrihge unit is used for its ability to recover fine heavies, it should be used as rougher (e.g. Kelsey jig feeding spirals for iron ore or gold carriers recovery; Falcon Cs feeding tables at Tanco). If the centrifuge unit is used for its upgrading capabilities, it should perform the cleaning stage (e.g. Reichert cones feeding Knelson CD30 at Omai Mines). The main competitors of centrifuge units are often non-centrifuge units that can in some cases offer very similar performance at lower cost. For the vast majority of applications, a spiral or I -g jig circuit can yield acceptable recoveries above 106 pm, and should be the preferred route.

‘There are a number of applications of semi-continuous centrifuges to treat either flash or rougher flotation concentrates (in regrind circuits).

1007

Exceptions are generally linked to very small density differentials and (possibly) shape-based separation. Scale-up and Layout As discussed in Laplante and Spiller (2002), scale-up is generally not an issue because these units are generally tested a full-scale or near full-scale because of their relative newness. Typical nearscale test work involves unit that generally have about one-fifth the capacity of the full-scale unit, and scale-up should be based on a constant specific feed rate 4 . e . feed rate per unit surface. For most applications, it would be unwise to go directly from bench scale to full-scale without the demonstration stage. Layout problems again are not likely to arise from these units because of their relatively small size, hence sinall footprint and headroom requirements. Further, a single stage is generally used, because of unit cost. The typical layout should include screening ahead of most if not all of these units, and in the case of the Kelsey jig screening of the tailing to recover the ragging. Because gravity products are often difficult to pump, gravity should be used as often as possible to direct products to the next unit, which makes headroom more important than footprint. Headroom is also important for easy servicing. CONCLUSIONS Gravity recovery of gold represents a special application of gravity recovery in that it is often a “safety net” upstream from the main recovery circuit. Another unique feature of gold gravity circuits is the low unit efficiency but the very high circulating loads of GRG in all grinding circuits whose main classification is hydrocyclone-based. Thus how much gold will be recovered is as much a function of the grinding circuit than it is of the fundamental recoverability of gold in an ore. A third component affecting gold gravity recovery is the nature of the gravity circuit itself. The optimum or near-optimum design of gold gravity circuits from an economic perspective must rely on a good understanding of all three components. Additional work required to improve the approach pertains mostly to improved databases. First, the unit recovery database must be built up, with particular emphasis on operating conditions for Knelson Concentrators, the most commonly used unit, and at least a minimum base for the Falcon SB and Gekko IPJ units. Second, the link between the partition curve of GRG and that of the ore must be better understood. There is limited but compelling evidence that shows that when classification efficiency is poor, as diagnosed with the ore partition curve, the ability of hydrocyclones to keep GRG above 37 pm in the circulating load can be decreased. Even a drop in the recovery to the cyclone underflow from 99.9% to 98-99%, as small as it is, has severe consequences for gravity recovery. Continuous centrifuge units have not been used as widely as semi-continuous units, but are clearly gaining acceptance, can treat a wider selection of ores, and have potentially more economic impact. The mineral processor must become familiar with these units and be on the lookout for potential applications. This includes developing a good understanding of capital and operating costs as well as possible synergies with cheaper unit that can lower overall costs. REFERENCES B a n i , S., A.R. Laplante and J. Marois. 1991. A Study of the Behaviour of Gold in Industrial and Laboratory Grinding. CIM Bull., Nov. 1991, pp. 72-78. Beniuk, V.G., C.A. Vadeikis and J.N. Enraght-Moony. 1994. Centrifugal Jigging of Gravity Concentrates and Tailings at Renison Limited. Minerals Engineering, Vol. 7(5/6), pp. 577-589. Deveau, C. 2002. The Evolution of Falcon Continuous Concentrators at Tantalum Mining Corporation of Canada. Proc. of 32’ld Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 2000, pp. 1-18. Costa de Silva, E., N.A. dos Santos, V. de Macedo Torres. No date. Concentrifugal Concentrators - A New Era in Gravity Concentration - The Experience of CVRD Research Center. On Website of Geologics, 10 pp.

1008

Gray, A.H. and N. Katsikaros. 1999. The InLine Leach Reactor - The New Art in Intensive Cyanidation of High Grade Centrifugal Gold Concentrates. Randol Gold and Silver Forum, May 1999,5 pp. Gregory, S., R. Dunne, P. Gelfi, V. Martins and A. Goulsbra. 1996. Gravity Concentration at the Telfer and New Celebration Gold Mines. Randol Gold Forum ‘96, Olympic Valley, April 1996, pp. 79-85. Laplante, A.R., L. Liu and A. Cauchon. 1989. Mineralogy and Flowsheet Changes at the Camchib Mines Inc. Mill, Chibougamau, Quebec. Process Mineralogy IX: Applications to Mineral Beneficiation, Metallurgy, Gold, Diamonds, Ceramics, Environment and Health, Eds. W. Petruk, R.D. Hagni, S. Pignolet-Brandom and D.M. Hausen, TMS Pub., pp. 247-258. Laplante, A.R. and Y . Shu. 1992. The use of a laboratory centrihgal separator to study gravity recovery in industrial circuits. 24lh Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 1992, Paper 12, 18 p. Laplante, A.R., M. Buonvino, A. Veltmeyer, J. Robitaille and G. Naud. 1993. Case Studies with the Falcon Concentrator. 25‘” Ann. Meet. of Canadian Mineral Processors. Ottawa, Jan. 1993, paper 25,22 pp. Laplante, A.R., A. Putz, L. Huang and F. Vincent. 1994. Practical Considerations in the Operation of Gold Gravity Circuits, Proc. of 26‘” Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 1994, Paper 23,2 1 pp. Laplante, A.R., F. Woodcock and M. Noaparast. 1995. Predicting gravity separation gold recovery, Minerals and Metallurgical Processing J., May 1995, pp. 74-79. Laplante, A.R., B. Zhang, L. Huang, J. Ling, P. Cousin and L. Racine. 1997. Difficult Gold Gravity Separations. Proc. of 29‘” Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 1997, Paper 3 I , pp. 421-434. Laplante, A.R., J. Ling, and J. Xiao. 1998. Difficult Gold Gravity Separations -- An Update. Proc. qf 30’”Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 1998, Paper 35, pp. 619-638. Laplante, A.R., L. Huang and B.G. Harris. 1999. The Upgrading of Primary Gold Gravity Concentrates. Proc. of 31”‘ Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 1999, pp. 2 I 1-226. Laplante, A.R. 2000. Centrifuge Units for Gravity Separation - An Update. Proc. qf3.2’”‘ Ann. Meet. of Canadian Mineral Processors, Ottawa, Jan. 2000, pp. 475-488. Laplante, A.R. 2000. Testing requirements and insight for gravity gold circuit design. Rand01 Gold Forum, Vancouver, April 2000, pp. 73-83. Laplante, A.R. 2000. Ten do’s and don’ts of gold gravity recovery. Randol Gold Forum, Vancouver, April 2000, pp. 107-1 17. Laplante, A.R., F. Woodcock and L. Huang. 2001. Laboratory procedure to characterise gravityrecoverable gold. SME Trans., Vol. 308, pp. 53-59. Laplante, A.R. and R. Dunne. 2002. The GRG Test and Flash Flotation. Proc. of 34‘” Ann. Meet. qf Canadian Mineral Processors, Ottawa, Jan. 2002, pp. 105-124. Laplante, A.R. and D.E. Spiller. 2002. Bench Scale & Pilot Plant Testwork for Gravity Concentration Circuit Design. Mineral Processing Plant Design -Update 2002. D. Barratt Ed., Vol. 3, B-5. SME Leathlean, W. and L. Smith. 2000. Leaching Gravity Concentrates Using the Acacia Reactor. Randol Gold Forum, Vancouver, April 2000, pp. 93-100. Lewis, G. 1999. Increased Recovery from Preg-Robbing Gold Ore at Penjom Gold Mine. Randol GoldandSilver Forum, May 1999, pp. 105-108. Luttrel, G.H., R.Q. Honaker and D.I. Philips. 1995. Enhanced Gravity Separators: New Alternatives for Fine Coal Cleaning. Proceedings of I2lh International Coal Preparation Conjerence, Lexington, Kentucky, pp. 28 1-292 Putz, A,, A.R. Laplante and G. Ladouceur. 1993. Evaluation of a Gravity Circuit in a Canadian Gold Operation, Randol Gold Forum, Beaver Creek, Sept. 1993. pp. 145-149 Van Kleek, D.M. 2001. Knelson Concentrators Extreme Gravity. Knelson Concentrators Web Site (www.knelson.com), 2001, 7 pp.

1009

Woodcock, F. and A.R. Laplante. 1993. A Laboratory Method for Determining the Amount of Gravity Recoverable Gold. Randol Gold Forum, Beaver Creek, Sept. 1993. pp. 15 1- 155. Wyslouzil, H. 1990. Evaluation of the Kelsey Centrifugal Jig at Rio Kemptville Tin. Proc of22"" Annual Meeting of Canadian Mineral Processor, Ottawa, Jan. 1990, pp. 46 1-472. Xiao, Z. 200 1. Developing Simple Regressionsfor Predicting Gravity Recovery in Grinding Circuits, M.Eng. Thesis, McGill University, Oct. 2001, 138 pp. Websites: Falcon Concentrators: InLine Leach Reactor (Gekkos) Kelsey Jigs (Roche Mining) Knelson Concentrators MGS and MeGaSep

http://www.concentrators.net/ htb:i/www.gekkos.com.au/ htb://www.geologics.com.au/ http://www.www.knelson.com http://www.mozley.co.uk/mg.htm

1010

Sizing and Selection of Heavy Media Equipment: Design and Layout Donovan F. Symonds and Shawn Malbon

ABSTRACT Heavy media circuits are used extensively to separate coarse particles (100 mm to 0.5 mm) according to their differences in specific gravity. The most common applications are in the coal industry where heavy media processes have become the dominant method for beneficiating coal due to their high separating efficiencies. The paper describes the five steps involved in heavy media separation namely: sizing and pre-treatment, separation (heavy media vessels or cyclones), draining and rinsing, media recovery, and dewatering and crushing. Detailed calculations are presented for selecting and sizing the various components of a heavy media cyclone circuit from sizing the raw coal feed to dewatering the clean coal and reject products. Basic flowsheets, together with a layout of a typical heavy media cyclone circuit, are presented. INTRODUCTION The vast majority of heavy media plants are designed to process coal. However, the techniques and calculation procedures can be applied in general to non-coal minerals that perform a separation based upon the difference in specific gravity between the mineral to be concentrated and the gangue material. Most of the formulae presented in this chapter are based upon empirical data collected and proven in many successful heavy media circuits for over fifty years. They usually relate to factors such as mean particle diameter, surface area and particle density and can therefore be applied to other non-coal applications. The following discussions relate to codrock separation, which generally take place at effective separating gravities between 1.35 and 1.8 sp. gr. It should be noted that a few coal preparation plants are washing raw coals below 0.5 mm (28 mesh). However, this is not common and therefore is not discussed in this chapter. Special design features and considerations are required to achieve satisfactory plant operations and cleaning results for these plants. The deviations of these flow sheets compared to the plant designed herein, in terms of plant equipment, do not significantly affect the sizing and selection criteria. The typical calculations presented at the end of the chapter refer to a heavy media cyclone circuit. Similar calculations were developed earlier by one of the authors for a heavy media vessel circuit.’ Throughout this chapter all quoted volumetric and solid capacities are in metric (S.I.) units unless otherwise noted. DESCRIPTION OF TYPICAL HEAVY MEDIA CIRCUITS There are two types of heavy media circuits, one for treating coarse coal, i.e., vessels, and one for treating small coal, i.e., cyclones. Both these circuits have the following common elements, (see Figures 1 and 2 for a cyclone circuit and a vessel circuit respectively): 0

0 0 0 0

sizing and pre-treatment separation, (heavy media vessels or cyclones) draining and rinsing media recovery crushing and dewatering.

‘

Symonds, D.F.1986, “Selection and Sizing of Heavy Media Equipment” Design and Installation of Concentration and Dewatering Circuits Published by Society of Mining Engineers, Littleton CO, 250 -260

1011

i

1

1 FEED

I

1

1 I

I I

,

,CLEAN

THICKENING CYCLONE

CYCLONE I

I

I

i

I

I I

-05mm TO OTHER CIRCUITS

I

I

r ” :

L - - - SECONDARY MAGNETIC SEPARATOR

I I I I

- 1

I

‘I

i CLEAN WATER

21

I

I

TO FINE COAL CIRCUIT

r

I I

CIRCULATING

L - - - -

H M CYCLONE PUMPJBE-

L

_

Figure 1 Typical heavy media cyclone circuit

_

_

_

_

_

_

- _ . _ -

_

_

1013

I

. I

E

2

SIZING AND PRE-TREATMENT Screening of the feed material to either a vessel or cyclone is required in order for the separator to achieve maximum efficiency at the desired separating gravity. It is of particular importance that the amount of undersize material passing to the heavy media separator be kept to a minimum. This is especially true in the case of a heavy media cyclone circuit. The misplacement of these fines has a two-fold effect in the separator: 0

0

the fines report to the floats fraction thereby minimizing efficiency and diluting the product quality the fines increase the viscosity of the medium, causing efficiency to be reduced, which adversely affects the product quality.

Coarse Coal Circuit Sizing Screens In the case of a coarse coal heavy media vessel circuit, the feed is preferably wet screened. This ensures two things: it maximizes screening efficiency it pre-wets the feed material to the vessel thus allowing a predictable flow of water into the separator and prevents the coal from “rafting” on the surface of the media. The relationship between mean particle size and surface moisture following wet screening is given in Figure 3. When the plant flow sheet incorporates dry screen extraction of undersize material, it is always advantageous to wet screen the resulting product prior to its introduction to the vessel. Both wet and dry screening operations are carried out generally on vibrating screens, either inclined linear motion type, or “banana” type screens. In some cases the screens are double decked to minimize the floor area required to screen the required capacity. A separate chapter in this book discusses the methods used in sizing and selecting these screens. The amount of water required for efficient wet screening is typically between 0.4 and 0.75 m3/h/ t of feed depending on the size consist and the nature of the feed material. Consultation with the screen manufacture is strongly recommended even if the above factors are available. Banana type screens incorporate five differing contiguous sloping screen sections each decreasing from the feed end to the discharge end. Because of this, the velocity of the material traveling along its length decreases. The majority of the finer undersize material passes through the screen deck in the earlier stage of its travel over the screen; this leaves the more difficult job of screening out the near sized underflow material to be conducted on the remainder of the screen. Due to the progressive lessening angle of the screen deck and therefore the slower rate of travel, the material has a longer residence time on the screen and screening efficiency is enhanced. Banana screens, dependent on its duty, are stated to have 30% to 50% more feed capacity than conventional linear motion type screens.

Pre-Wet Screens In the case where a dedicated pre-wet screen is used, the required width of the screen can be calculated from: Width (m) = Feed Rate (th) 60 x sp. gr. Where sp.gr. is the specific gravity of the feed material to be pre-wetted. Screens 3.65 or 4.88 m (12’ or 16’) in length are usually employed for this application and are fitted with 0.5 mm aperture wedge wire profile screen decks.

1014

17.5

-

15.0 -

12.5 -

zg

10.0 -

i F:

3

!i

5.0

-

2.5

-

0-

~

I

I

I

1

I

I

I

I

I

I

I

~~~

Figure 3 Surface moisture of dewatered coal (based on bituminous coal -800 kg/m3

1015

It should be noted that the pre-wet screen is not sized to effect screening, only to precondition the feed material. It is recommended that the maximum average bed depth be no greater than three times the mean particle diameter. For pre-wetting applications the amount of water used depends upon the size consist of the feed. Usually this ranges from 0.50 m3/t for +20 mm feed material to 1.40 m3/t for -20 mm sized feed.

Small Coal Circuit. In the case of a heavy media cyclone circuit that is treating a pre-classified feed, i.e., where a coarse coal separator is being used in the same flow sheet, the sizing of the cyclone feed is normally carried out at 0.5 mm, using sieve bends. The sieve bends are then immediately followed up with horizontal linear motion dewatering screens for moisture control. Again “banana” type screens can be used to both size the feed at 0.5 mm and subsequently dewater the product. Desliming Sieve Bends. For efficient screening the solids concentration to the sieve bends should be around 30%. Based on this, the volume of solids and water can be determined and the size of sieve bend(sj can be calculated. The amount of +0.5 mm fraction in the feed has no influence on the area of the sieve bend required. The sieve bend is sized based upon the volume of material passing through the device. The capacity of sieve bends is a function of open area, which depends upon the wedge wire profile width and the aperture. The sieve aperture required to deslime at 0.5 mm is 0.75mm. For sizing raw coals with a top size greater than 15 mm, the profile of the wedge wire is usually 2.2 mm wide x 4.5 mm deep. For smaller feed sizes, the wedge wire profile is reduced to 1.5 mm wide x 4.0 mm deep. Ty ically for desliming at 0.5 mm the capacity of a sieve bend is between 87.5 to 112.5 m3mm.

F

Since sieve bends come in various widths and radii, as well as arc angles of 45 and 60 degrees, the required area of sieve bend can be achieved to suit the layout of the plant. Manufacturers, or their representatives, should be consulted to determine the actual capacities of sieve bends for any specified application.

Desliming Screens. As stated above, the sieve bend@)discharge onto a dewatering screen so that the moisture content of the feed to the heavy media circuit can be predicted. Since the main function of this type of screen is not classifying but rather to dewater, the screen open area is not a factor. Water, in the form of sprays, can be used on the screen to maximize undesirable fines passing to the cyclone circuit. The width of the screen, which is usually 4.88 m (16’) long and fitted with 0.5 mm aperture wedge wire profile screen decks, is calculated using the following formula: Screen Capacity =19((dm)’x( sp.gr.)’)”O. 33 t/h/m Where dm is the mean particle size and sp.gr. is the specific gravity of the feed.

SEPARATING EQUIPMENT As mentioned above, there are two types of heavy media separators used in coal beneficiation, namely vessels for coarse coal, and cyclones for small coal. Vessels In general there are two types of heavy media vessels, open bath and drum. Both types achieve similar separating efficiencies for any given feed size and separating specific gravity. The preference of one type over the other is a dependent on items such as maintenance, plant layout

1016

considerations and an assessment of the costs of operating and maintaining the equipment. The following narration is applicable to either type. The top size of the feed particles to a heavy media vessel is limited by two main factors, namely liberation size and the practical ability of mechanical equipment to handle large sized particles. In terms of coal cleaning, the larger the size of particle the greater the chance it has to contain bands of middlings or non-coal material resulting in coal being lost to the reject. Also, large particles require the use of large, robust and generally costly materials handling equipment, These limitations therefore restrict the top size of feed particles to heavy media vessels to approximately 100 to 150 mm (4 to 6"). The bottom size is essentially determined by viscosity and settling effects, i.e., the ability of the smallest reject particle to fall through the medium suspension and report to the reject collection system within the residence time of the particular vessel. In most heavy media applications this bottom particle size is between 10 and 6 mm (318 and W'). It is important that quiescent flow of media through the vessel is achieved, otherwise misplacement of products will result. When considering the use of a particular vessel, two major questions should be asked of the supplier. What are the projected Ep (ecart probable moyen) and the guaranteed Ep for the desired duty? To answer these questions the supplier will require the following data: 0 0 0

the nature of feed material (i.e., raw coal, middlings, etc.) the size consist and maximum feed rate (a) the washability data and projected effective separating gravity.

Acceptable Ep values for heavy media vessels range from 0.015 to 0.05 depending on the above data. A lower value indicates a more efficient separation. The feed capacities of open bath coal cleaning vessels vary according to the size consist of the feed. As the geometric mean size of the feed increases, so does the capacity. The following formula for calculating the floats capacity per m width or diameter is applicable to both open bath and drum type vessels: Floats Capacity (t/h/m) =(Dxd)"O.5 Where D and d are the top and bottom size of feed respectively. Media flows entering the vessels of between 150 and 200 m3/h per m width are correspondingly used.

Cyclones The top size of the feed to a heavy media cyclone is generally determined by the diameter of the cyclone inlet opening. Typical ranges of maximum feed particle sizes are a function of the cyclone feed opening, i.e., approximately 113 of the cyclone inlet diameter. The maximum feed particle size and cyclone diameters are as follows in Table 1.

1017

Table 1 Heavy media cyclone diameter and maximum particle size Source: Krebs-Engineers Cyclone Diameter Max Particle mm in mm in 500 20 35 1.25 600 24 40 1.5 750 30 50 2.0 75 3.0 840 33 1020 40 100 4.0 The bottom size is dictated mainly by the lowest size at which efficient screening and draining and rinsing can take place using conventional sieve bends and screens. Practical experience has shown that this size is approximately 0.50 mm for cyclones up to 0.66 m diameter. Some cyclones are treating sized coal down below this particle size with efficient separations. However, specialized circuits have to be designed for these particular instances. This chapter only deals with the washing of +0.50 mm particles. The configuration of the cyclone, i.e., inlet, overflow and underflow openings can greatly affect the efficiency of the cyclone separation. Standard diameter ratios of the inlet, overflow and underflow orifices to the cyclone diameter are approximately 0.2, 0.4 and 0.3 respectively. The cyclone has in the past has always been fed tangentially and k a t angle of repose of 10 degrees to the horizontal. However, more recently, Krebs Engineers developed a larger inlet orifice resulting in an involuted configuration. This increases their capacity of up to 40% over the old configured cyclones and still achieves similar Ep’s. To increase cyclone operation with minimal maintenance, cyclones utilize ceramic lining and replaceable cone liners and apex inserts in lieu of ni-hard, in the lower high wear regions of the cyclone. Separating efficiencies of heavy media cyclones are less than heavy media vessels. Ep values of between 0.03 and 0.06 is not uncommon for raw coal separations treating between 50 mm and 0.5 mm sized feed at specific gravity of separation between 1.35 and 1.80. The cyclone pulp capacity, (media and coal) of a cyclone is primarily a function of the pressure that at which it operates. The more normal operating range is between 150 and 250 kPa (7 to 21 psig). For efficient separations, the media to coal ratio of the feed to a cyclone should not be less than 3.5: 1 but more ideally 4: 1. The following table shows typical dry tonnes throughput capacities: Table 2 Heavy media cyclone diameter and capacity -~ Source: Krebs-Engineers Cyclone Diameter mm in 500 20 600 24 750 30 840 33 1020 40

Capacity Dry t/h 45-68 95- 127 175-200 250-300 350-400

Capacity pulp m3/h 235-300 390-420 590 865 1180

Apart from the feed capacity, the cyclone has a limit to how much reject material it can handle. This is due to the restriction caused by the apex. The cyclone supplier, or his representative, upon receiving data as to the nature, the size consist, the washability of the feed, and the desirable separating specific gravity, will size the apex accordingly and supply a range of expected and guaranteed Ep’s for the application. The

1018

manufacture will also furnish the feed pulp capacity and the required operating pressure for the application. Heavy media cyclones can be either pumped or gravity fed depending on site-specific limitations. However in either case the correct operating pressure must be adhered to. An overflow from the cyclone feed tank must be included so that when coal is introduced to the tank it can displace its own volume of medium back to the medium sump to maintain a constant head on the feed pump and constant pressure at the cyclones. The specific gravity of the medium obtainable with magnetite is dependent on the grain size of the magnetite. For the usual operating specific gravity range of between 1.4 and 1.8 the grain size of the magnetite ideally should contain 95% of -50 micron particles. Should this grind not be achieved, the media will have a tendency to classify within the cycIone and therefore impair efficiency.

Media Circuit For heavy media vessels the use of a head box to provide constant volumetric flow to the vessel is advisable. This eliminates any decreasing flow to the vessel as the circulating media pump wears. It is necessary to “bleed” some circulating medium out of the system to the medium recovery circuit. This helps to keep the circulating medium free from the misplaced fines, which pass to the system from the inefficient sizing operation of the feed. It also aids in reducing the viscosity effect the misplaced fine particles have on the circulating medium Due to the volume of coal and its associated water being fed to the vessel or cyclone, the media is initially diluted and the total system volume is increased. Therefore room must be made available to add in more magnetite, usually as overdense media, to maintain the medium’s specific gravity. If the volume of newly added magnetite is less than the volume of water being introduced into the system with the coal, more water, in the form of circulating media level control, must be used. Circulating Media Tank The circulating media tank should be capable of storing at least the equivalent of two minutes of the circulating media pump capacity, i.e.: Capacity (m3)= 4/30 Where Q is the circulating media pump capacity in m3/h. It should be lined with some wear resistant coating such as epoxy paint. There should be an allowance of 0.3 m from the maximum media level in the tank when full to the overflow level. Where the storage capacities of the tanks are prohibitive for steel tanks to be used economically, rectangular, 100 mm thick pre-fabricated concrete sumps have been used in lieu of steel conical tanks.

Cyclone Pump Tube Should the heavy media cyclones be fed via a pump tube, the diameter of the pump tube is sized according to the feed capacity of the cyclones, i.e.: Diameter (mm) = 37.5 x 1.15 x (Q)1/2 Where Q is the cyclone feed pump capacity in m3/h. To calculate approximately the pressure head required for pumps, a simple plant layout depicting elevations, similar to Figure 4, is quickly sketched from which projected heads can be estimated.

1019

Figure 4 Preliminary plant elevation layout

Piping and Valves All circulating media and cyclone feed piping is usually thick walled SDR 11 high-density polyethylene for gravity flow lines, wired armored flexible hose or schedule 80 mild steel for pressure lines.

DRAINING AND RINSING The purpose of draining and rinsing the product from either a vessel or a cyclone is to recover as much magnetite media as possible for re-use. For both vessel and cyclone circuits the draining and rinsing of media is done using sieves and vibrating screens. In the case of a vessel a fixed sieve is used whereas in a cyclone circuit a sieve bend is used, in both cases they are employed to recover between 80-90% of the free flowing media passing with the products from the separator. The drain and rinse screen is positioned directly after the sieve, and on the first 1.2 m of its length, any surplus media from the sieve is drained. The sieve and screen drain products are collectively known as the correct media.

Static Sieves and Sieve Bends The static sieve typically is fitted in the inclined discharge chute between the vessel and the screen and is comprised of wedge wire profile bars spaced at 0.5 mm intervals. In some instances the bars run in parallel to the stream to ensure maximum media recovery. The wedge wire bars should be of sufficient width to stand the duty of coarse coal or refuse passing over them. At least 3.4 mm wide x 6.5 mm deep profile wedge wire must be used when processing greater than 50 mm particles. The following table illustrates the throughput capacity of 3.4 mm wide wedge wire profile screens:

Table 3 Throughput capacity of 3.4 mm wedge wire screen Capacity Aperture (mm) 0.5 1.0 m3/h/m2 Static Sieve @ 15 degree 51 81 Vibrating Screen 88 170 The sieve bends employed for media drainage are as those for desliming, see above, with the exception that the aperture between the wedge wire profile bars is 1 mm. Typically for media recovery, the capacity of a sieve bend is between 105 to 135 m3/h/m2of sieve bend surface.

Drain and Rinse Screens The remaining 3.6 m of screen, since drain and rinse screens are usually 4.8 m long, is used to remove the adhering media from the products. The rinsing is aided by the use of water, being sprayed onto the passing product on the screen. Two rows of sprays are used to remove the adhering media from the product. The sprays are arranged to cover the full width of the screen. The collected underflow from this section of the screen is aptly called dilute media. The amount of spray water used depends on the size of the product being rinsed. Typically for coarse coal, 27 m3/h per m width of screen is required, of which 213 is used as primary spray water and 1/3 is used as secondary water. For heavy media cyclone circuits a total of 36 m3h per m width of screen is used in total, with 75% used as primary and 25% used as secondary water. The amount of total spray water used should be increased when raw coals containing high amounts of clay material are processed. The capacity of coarse coal drain and rinse screens is obtained using the following formula: Capacity (t/h/m) = 60 x sp.gr.

1021

Where sp.gr. is the specific gravity of the product.

For cyclone drain and rinse screens the following formula is used: Capacity (Wm) = 12 ((dm)2x (sp.gr)’)”0.33 Where dm is the mean particle size in mm and esp. is the specific gravity of the product. Should double deck screens be used for coarse coal, the top deck should be computed as if a sizing screen and should be checked so that the maximum average bed depth does not exceed three times the mean particle size. As in the case of sizing screens, banana type screens are currently being used to replace both sieve bend and linear motion type screens. In the case of the former, particular attention must be paid to the delineation of the correct and dilute medium split beneath the screen. The water and adhering media from the drain and rinse screens from the product screens is collected together and typically passes to the dilute media tank, where the magnetite is recovered for re-use in the circulating media circuit.

MEDIA RECOVERY CIRCUIT There are basically two variations of media recovery circuitry. The decision on which system to employ is based upon the magnetite concentration in the dilute media system. For maximum magnetite recovery efficiency, the magnetite solids concentration fed to magnetic separators should not be less than 100 gA. For magnetite concentrations less than 100 fl,thickening cyclones are used prior to the magnetic separators to concentrate the feed. For magnetite concentrations of 100 gA and above, the dilute media is pumped directly to the primary magnetic separators. In the latter case an intermediate magnetite thickening system, i.e., flow sheet employing both primary and secondary magnetic separators, may be used in conjunction with secondary magnetic separators. In both flow sheets the thickening cyclone overflow is invariably used as the primary spray water on the product media recovery screens. The following circuit calculations utilize to the latter media recovery circuit. Magnetic Separators Magnetic separators come in drum diameters as small as 380 mm and as large as 1220 mm, and lengths range from 305 mm through to 3050 mm. Generally, only 750 mm 900 mm and 1220 mm diameter separators ranging from 900 mm to 3050 mm long are used in coal preparation. Primary magnetic separators have recently being configured as counter-rotation units, i.e., the rotation of the drum is counter to the direction of the flow of slurry. This configuration was developed to handle high solids loadings. Primary magnetic separators rotate at approximately 11 rpm. Permanent ceramic or rare earth magnets are used. The control of the strength and depth of the magnetic field is due to design of the poles and pole pieces, the size and number of magnets, and the arc covered by the magnet assembly. All these factors have been improved upon in recent times resulting in increased magnetite recovery and better separation of magnetic and nonmagnetic particles. The capacity of magnetic separators, in terms of slurry volume, is dependent on the drum diameter. The following capacities are typical:

1022

Table 4 Magnetic separator capacities Drum Diameter 750 mm 900 mm 1220 mm

Capacity m3/hlmMagnetite 55-65 60-80 80-95

Capacity thlm 16 19 23

In a media recovery circuit in which the dilute media is pumped directly to the magnetic separators, it is advantageous to utilize a head box to ensure constant flow to primary magnetic separator. The recovered magnetite from primary and secondary magnetic separators is re-used in the media circuit either directly or indirectly via a magnetite tank and pump.

Thickening Cyclones Effluent from the primary magnetic separator, which contains approximately 0.3 g/l of magnetite, requires thickening prior to passing to secondary magnetic separators. Cyclones configured to thicken the magnetite solids are employed to increase the solids concentration of the magnetite slurry prior to feeding the secondary magnetic separators The innovation of long bodied cyclones has improved the performance of the classifying effect of the cyclone and should be used for this duty. It is recommended that no cyclones bigger than 375 mm in diameter should be used for this application due to the fine size consist of the magnetite. Typical cyclone feed capacities are shown in Table 5 . Table 5 Thickening cyclone capacities Cyclone Diameter 250 mm 315 mm

Capacity m’/h 55-90 90- 180

The overflow from the cyclone can be used as the primary spray water on the product drain and rinse screens. A portion of this overflow water is used to maintain the volumetric feed requirements of the selected secondary magnetic separator.

Secondary Magnetic Separators The secondary magnetic separator is sized according to the above given capacities based, primarily, on the total volume of the secondary spray water used on the product rinse and drain screens. Usually the secondary magnetic separator rotation, at 5 rpm, is concurrent with the feed slurry flow. This design is most effective for producing an extremely clean magnetic concentrate from relatively coarse materials. The effluent from the secondary magnetic separator is typically discharged to the fines (-0.5 mm) circuit if applicable or to the plant effluent stream. Dilute Media and Magnetic Separator EMuent Tanks These tanks should be capable of storing at least the equivalent of a minimum of 1% but preferably two minutes of their respective pump capacity, i.e.: Capacity (m3) = 4/30 Where Q is the pump capacity in m3h. It should be lined with some wear resistant coating such as epoxy paint. There should be an allowance of 0.3 m from the maximum media level in the tank when full to the overflow level.

1023

The capacity of 60 degree conical tanks for various tank diameters is given in Figure 4 and should be constructed in mild steel plate with a minimum thickness of 8 mm.

Piping And Valves All media recovery piping is usually rubber lined, flexible hose or schedule 80 mild steel. CRUSHING AND DEWATERING Coarse Coal Plant Coal Particles greater in size than that required by the coal purchaser, is crushed more commonly in either a swing hammer type or roll type crushers. The selection of either is dependent upon the friability and hardness of the coal to be crushed together with the size reduction required. It is not possible to quote typical capacities for crushers since they are sized according to the following factors: 0 0

0

the ratio of the size reduction required the top size of feed material the size distribution of the material to be crushed.

Small Coal Plant The clean coal product from the cyclone drain and rinse screen(s) is finally dewatered in centrifuges before shipment to the coal purchaser. There are two types of centrifuges currently on the market, one group in which the basket revolves about its horizontal axis and another in which the basket revolves about its vertical axis. The selection of either is primarily a function of cost, maintenance, and plant layout considerations. Typical capacities for centrifuges shown in Table 6. Table 6 Centrifuge capacities Type Horizontal Horizontal Vertical Vertical

Feed Size 25 mm x 0.5 mm 12.5 mm x 0.5 mm 25 mmx0.5mm 12.5 mm x 0.5mm

Capacity t/h 125-155 115-145 127-215 70-145

Product Surface Moisture 3.5-4.5% 5-6% 4.5-5.5% 3.5-4.5%

The product surface moisture given above is for a typical clean coal, and is dependent on the size distribution of the product. These surface moisture contents appear to be anomalous, since, in some cases, the finer coal has lower surface moisture than the coarse coal. This is due to the type of centrifuge used and the g-force associated with each machine. Fine coal centrifuges operate at higher g-forces, have lower throughput capacities and are generally more costly to purchase and operate. One of the main criteria governing the selection of the required number of plant equipment for the plant design is the layout of the plant. In general, one large sized piece of equipment is more economical that using two smaller sized units to perform the same duty. In the case of upgrading an existing plant, physical constraints may dictate the use of a certain number of specified equipment, i.e., a 2.44 m (8’) wide linear motion sizing screen requires less head room than on 2.44 m wide “banana” type sizing screen. An appreciation of the site constraint should be known before any equipment can be selected. It would be prudent for the designer to utilize the same size and manufacturer of new equipment that is already being used in the existing plant. This allows the mine to minimize spare parts inventory.

1024

In a “green field” application, the plant designer has much more freedom in selecting the size and number of pieces of equipment to achieve the required duty. The number and size of equipment of any part of the circuit(s) is somewhat dictated by the number and size of preceding and following equipment. For example, it does not make sense to feed one 0.76 m (30”) diameter heavy medium cyclone from three 1.5 m ( 5 ’ ) wide desliming screens and pass the cyclone clean coal to two 0.91 m (3’) wide clean coal drain and rinse screens. It should be more practical and economical to use two 2.44 m (8’) wide desliming screens and one 2.44 m wide clean coal drain and rinse screen. Standardization of equipment size can be beneficial, i.e., the use of one 2.44 m (8’) wide desliming screen in conjunction with one wide clean coal drain and rinse screen even though the calculations suggest that one 1.8 m (6’) wide screen is required. Therefore, the calculation to determine plant equipment size and number should not be finalized until the circuit equipment calculations have been completed. SUMMARY The information for sizing and selecting heavy media equipment presented in this chapter is based on empirical data collected from a number of sources over the last five decades. A well designed heavy media plant will result in a highly efficient plant with very little loss of saleable product or magnetite. It should be emphasized that prior to the finalization of the plant design and the ordering of equipment, manufacturers of the equipment should be consulted so that they can confirm the actual sizing for the given duty. REFERENCES The technical information pertaining to circuit design was obtained from many sources but two publications stand out as primary sources of the empirical formulae used in the design of heavy media circuits. They are: Dutch State Mines. 1965. The Heavy Media Cyclone Washeryfor Minerals and Coal, Netherlands National Coal Board. 1978. Code of Practice, The Design of Coal Preparation Plant. UK The equipment capacities were obtained from catalogues, web sites and other publications issued by equipment manufactures. These include: Allis-Chalmers, Appleby, WI Baker Hughes, Baker Process, Salt Lake City, UT Centrifugal and Mechanical Industries, St. Louis, MO Conn-Weld Industries Inc., Princeton, WV Krebs Engineers, Tuscon, AZ Sterns Magnetics Inc., Cudahy, WI Ludowici Mineral Processing Equipment, USA, Pittsburgh, PA Multotec Process Equipment, Kempton Park, South Africa Birtley Engineering Ltd, Chesterfield, UK The Daniels Company, Bluefield, WV CALCULATIONS EXAMPLE OF TYPICAL SMALL COAL HEAVY MEDIA CYCLONE CIRCUIT CALCULATIONS Coarse Coal Plant Feed Rate Size of Raw Coal Medium Circuit

400 t/h 15 mm x 0.5 mm 1.60 Separating gravity

1025

-

~~

Th Average 425 300 222 77 125

FEED Cyclone Feed Clean Coal Discard -0.5 mm in raw coal

Th Maximum 425 300 250 150 125

sp. gr.

dm

1SO7 1.482 1.357 2.014 1.573

5.8 1 5.81 5.8 1

DESLIMING WATER REQUIREMENTS Required Feed solids concentration for efficient sizing - 30 5% - 425/1SO7 Volume of solids - 282.02 m3h - 940.06 m3h Total volume of slurry to sieve (s) 658.04 m3h Therefore water required SIEVE BEND Sieve to pass -0.5 mm solids and water = Volume of -0.5 mm solids

-

Total volume to pass through the sieve(s)

-

Sieve capacity (1.5 mm w/w profile) Sieve area required

=

-

DESLIMING SCREEN Maximum screen capacity Hence required width is

79.47+658.04 737.51 m3h 112.5 m3h /m2 737.5M12.5 m2 6.56 m2 (.75 mm Apt. 60" Arc)

Select from the following:

a) c)

12Y1.573 79.47 m3h

Radius (m) 0.76 1.22

= = = = =

Width (m) 8.20 5.10

Width (ft) 26.90 16.73

19((5.81)2x (1.482)2)*.33 79.82t/h/m 300fl9.82m 3.76 m 12.33'

Screen are manufactured in certain widths, therefore the selection of two 2.4 m (8') x 4.8 m (16') long screens each with two, 2.1 m (7') wide x 1.52 m ( 5 ' ) radius sieve bends preceding them.

HEAVY MEDIA CYCLONES Feed to primary cyclones is Maximum refuse to handle

300 t/h 150 t/h

1026

CYCLONE Dia. mm 500 mm 660 mm 750 mm 840 mm 1020 mm

KREBS Design 20B-S 26B-S 30B 33B 40B

Allowable Tonnage Capacity t/h 45 127 189 276 378

Actual Number Required 7 3 2 2 1

Allowable Actual Pulp Pulp Capacity Capacity m3/h* m3/h 235 1645 397 1191 590 1180 865 1730 1180 1180

Medium Coal Ratio 7.13 4.88 4.83 7.55 4.83

At a medium to coal ratio of 4:1, the maximum amount of feed slurry to the cyclones is 1010 m3h. Since two, 750 mm diameter cyclones can be fed from two 2.4 m (8') wide desliming screens, and that the total pulp volume is the same for one, 1020 mm diameter cyclone, and that the maximum particle size of the feed is 15 mm, then two, 750 mm diameter cyclones is a suitable selection.

CLEAN COAL DRAIN AND RINSE SCREEN Capacity = 12((1.357)2x (5.81)2)"1/3 t/h/m = 47.54t/h/m Required width = 250/47.54 = 3.16m = 10.35' CLEAN COAL SIEVE BEND Sieve to pass 80% of circulating medium =

lOlOx.8m3h 808.0rn3A-1 135 m3h /m2 808/135 m2 5.99 m2

Sieve capacity (1.5 mm w/w profile) Sieve area required

= = = =

Select from the following

(1 mm Apt. 60" Arc)

a) c) d)

Radius (m) 0.76 1.22 1.52

Width (m) 7.52 4.69 3.76

Width (ft) 24.68 15.37 12.34

From the above calculations two, 2.4 m (8') wide x 4.8 m (16') long drain and rinse screens can be selected each having one, 1.8 m (6') wide x 1.52 m ( 5 ' ) radius sieve bend preceding them,

CLEAN COAL CENTRIFUGE Number of centrifuges required for 250 t/h of 15 mm x 0 mm clean coal is: 250/160 = 2 REFUSE DRAIN AND RINSE SCREEN Capacity = =

12((2.014)2 x (5.81)2)"1/3 t/h/m 61.84t/h/m

1027

=

Required width

=

150/61.84 2.43 m 7.96’

= =

1010x.2m~/h 202.0m3/h

Sieve capacity (1.5 m m w/w profile)

=

135 m3/h/m2

Sieve area required

= 202/135 m2 = 1.50m2 (1 mm Apt. 60l Arc)

=

REFUSE SIEVE BEND Sieve to pass 20% of circulating medium

Select from the following

a) c) d)

Width (m) 0.76 1.22 1.52

Radius (m) 0.76 1.22 1.52

Width (fi) 6.23 3.88 3.12

From the above calculations, one, 2.4 m (8’) wide x 4.8m (16’) long drain and rinse screens can be selected having one, 1.8 m (6’) wide x 1.52 m ( 5 ’ ) radius sieve bend preceding it.

MEDIA CIRCUIT Surface Moisture of Feed (with 5.81 mm mean grain size) Water Diluting Medium

12 %

=

(300/.88) - 300 40.91 m3/h

However in practice it has been found that 6.5% of the feed’s surface moisture does not dilute the medium = (300/.935) - 300 20.86 m 3 k Therefore the volume of water diluting the medium = 40.91-20.86 m3/h = 20.05m3/h Adhering Medium The amount of medium adhering medium sprayed off the products to the medium recovery circuit when using a combination of sieves and screens is calculated as follows: Volume

=

0.95/(dm x sp. gr.) m3/t

Using the worst case scenario of maximum refuse loading: For clean coal

= =

For refuse

= =

(0.95 x 222)/(5.81*1.357) 26.75m3/h (0.95 x 150)/(5.81*2.014) 6.25 m3/h

The specific gravity of the separation is different to that of the specific gravity of the medium,

1028

The separating gravity is higher. Likewise the specific gravity of the overflow and underflow from the cyclone is not the same as the medium specific gravity. Separating specific gravity Medium specific gravity Overflow specific gravity Underflow specific gravity

= = = =

1.60 1.526 1.376 2.126

Assume overdense specific gravity is 2.10. Flow In Water Diluting The Medium Overdense Or Recovered Medium Flow Out Clean Coal Adhering Medium Refuse Adhering Medium Circulating Medium To Dilute Medium

m3/h 20.05 Y

sp. gr. 1.oo 2.10

26.75 6.75 X

1.376 2.126 1.526*

-

Y + 20.05 = 26.75 + 6.25 + X 2.1Y + (20.05 x 1.O) = (26.75 x 1.376) + (6.25 x 2.126) + 1.526X

Volume balance Mass balance Multiplying volume balance by 2.1 then

-

Therefore Y

2.1Y+ (20.05 x 2.1) = (26.75 x 2.1) + (6.25 x 2.1) + 2.1X

Es 1 Eq2

Es 3

(20.05 x 1.1) = (26.75 x 0.724) -(6.25 x.03) + 0.574X

Subtracting Eq 2 from Eq 3 therefore and X

-

-

22.055 - 19.2045 = S74X 4.97 m3/h 26.75 + 6.25 + 5.01 - 20.05 17.92 m3lh

CIRCULATING MEDIA PUMP The bleed of circulating medium to dilute medium must be accounted for in sizing the circulating media pump capacity, if the bleed is taken off of the feed supply. Therefore the capacity of the circulating media pump: 1010 + 5.01 = 1015 m3/h If the cyclones are being fed by a pump tube, then an allowance of 25 m3/h of head box overflow. Therefore the capacit of the circulating media pump is: 1015 + 25 = 1040 mY/h

CIRCULATING MEDIA TANK The minimum volume stored in the circulating media tank is: 2 x (1040/60) = 34.67 m3 If the cyclones are being fed by a pump tube then an allowance of 25 m3/h of head box overflow should be catered for, resulting in the pump being sized to handle 1015 + 25 = 1040 m3/h

1029

PUMP TUBE The diameter of the ONE pump tube is: 37.5 x 1.15 x (1010)1/2 = 1341 mm The diameter of the TWO pump tubes is: 37.5 x 1.15 x (505)1/2= 969 mm Magnetite Balance The amount of magnetite passing to the medium recovery system is calculated as follows:

Clean Coal Refuse Bleed Total

m3/h 26.75 6.25 4.97 38.01

sp. gr.

1.76 2.26 1.526

* 414 1239 579

t/h 11.06 7.74 2.88 21.68

gA* For every 0.1 specific gravity increment above 1.0, the medium contains approximately 110 gms/liter of magnetite, i.e., at 1.50 the medium contains 550 grams/liter of magnetite or 0.55 tons/m3. The amount of magnetite passing to the circulating medium system from the overdense system is m3/h 17.92

sp. gr. 2.10

gn 1210

t/h 21.68

Secondary 38 19 58

Total 154 77 230

Therefore the magnetite transfer balances SDray Water Reauirements 32 m3/h /m width of screen Using a total of 24 m3/h /m width of screen Primary 75% 8 m3/h /m width of screen Secondary 25% Clean Coal Refuse

m 4.80 2.40

Primary 115 58 173

Total dilute medium is: Total spray water + adhering medium + bleed - water lost on products 230 + 33 + 4.97 - 40.91 thus = 227.46m3/h Magnetite Conc. 21.6W227.46 t/m3 0.10 t/m3 This magnetite solids concentration of the dilute medium is suitable to be pumped directly to primary magnetic separators

DILUTE MEDIUM PUMP If a head box is used to feed the primary magnetic separator then an extra 25 m3h should be added to the dilute medium pump capacity.

1030

Therefore the capacity of the dilute medium pump is: 252.72 m3/h

DILUTE MEDIA TANK The minimum volume stored in the circulating medium tank is: 1.5 x (253/60)=6.32m3 PRIMARY MAGNETIC SEPARATORS Allowable Slurry Ca acity Required Width Diameter m /h/m m 750 mm 60 3.80 900 mm 70 3.25 1220 mm 85 2.68

P

Allowable Mag. Cap. t/Nm 16.0 19.0 23.0

Required Width

m 1.36 1.14 0.94

From the above calculations the selection of a suitably sized magnetic separator can be selected based on the plant layout and the cost differences of the suitable units. For this example, say two, 2.13 m (7’) wide x 750 mm (30”)diameter.

MAGNETIC SEPARATOR EFFLUENT PUMP For design purposes the effluent from the primary magnetic separators is taken to be the same as that feeding them. Therefore the thickening cyclones employed to thicken up the magnetite solids concentration lost by the primary separators is in this particular example have to handle 256.32 m3h of slurry. MAGNETIC SEPARATOR EFFLUENT TANK This is the same volumetric capacity as the dilute media tank. However the cyclone feed pump requires more head to feed the thickening cyclone(s). THICKENING CYCLONES Cyclone Diameter 250 mm 375 IILm

Allowable Slurry Capacity m3/h/m 55

90

Required Number m3/h/m 4.1 2.5

Allowable Slurry Capacity 90 180

Required Number 2.5 1.3

From the above calculations the selection of suitably size and number require of cyclones can be selected. Consult with manufacturer to determine the suitability of the selection since the pressure head on the cyclone affects greatly the performance of the cyclone.

SECONDARY MAGNETIC SEPARATORS The cyclone overflow is directed to the primary spray water on the product drain and rinse screens. This volume is 173 m3/h.Therefore the secondary magnetic separator feed is 227.72 - 173 m3/h = 55 m3/h. Diameter 750 mm 900 mm 1220 mm

Allowable Slurry Capacity m3/wrn 60 70 85

1031

Required Width m 0.92 0.78 0.65

From the above calculations the selection of a suitably sized magnetic separator can be selected based on the plant layout and the cost differences of the suitable units.

EQUIPMENT LIST No. Desliming Sieve Bends Desliming Screen Primary HM Cyclones Clean coal Sieve Bends Clean Coal Screen Clean Coal Centrifuge Refuse Sieve Bends Refuse Screen Primary Mag.Separators Thickening Cyclones Secondary Mag.Separat0r.s Dilute Medium Pump Mag.Sep.Effluent Pump Circulating Medium Pump HM Cyclone Feed Pump Pump Tube Circulating Medium Tank Dilute Medium Tank Mag.Sep.U/flow Tank

2 2 2 2 1 2 1 1 2

3 1 1 1 1 2 2 1 1 1

Size 7' x 5' radius 8 ' x 16' 30 inch dia. 6' x 5' radius 8' x 16' 125 t/h each 6 x 5' radius 8' x 16' 7' wide x 36" dia. 10 inch dia. 3' wide x 36" dia. 113 gpm 1,113 gpm 4,580 gpm 2,224 gpm 38 inch dia. 9,160 gallons cap. 2,226 gallons cap. 2,226 gallons cap

1032

(SI) 2.18mx 1.52mradius 2.44 m x 4.88 m 760 mm dia. 1.83 m x 1.52 m radius 2.44 m x 4.88 m 125 t/h each 1.83 m x 1.52 m radius 2.44 m x 4.88 m 2.18 m wide x 914 mm dia 250 mm dia 914 mm wide x 914 mm dia 253 m3/h 253 m3/h 1,040 m3/h 505 m3/h 970 mm dia 35 m3capacity 6 m3capacity 6 m3capacity

Photometric Ore Sorting Bo Arvidson

ABSTRACT Of the electronic ore sorting machines, the photometric, a.k.a. optical, sorting machines are the most common. Due to the rapid development of computing technology, major advances in recognizing and quantifying rock surface characteristicsare now made possible. Another benefit is increased programming flexibility and substantially reduced capital cost. One of the developments resembles artificial intelligence in the way the sorter machines are “taught” which rock particles to sort. Other enhancements are in the area of optical resolution and improved electronic component reliability. The basic method to move the rocks by air jets has not changed; the enhancements of this part of a sorting machine are mainly in the manufacturing design, blast driver electronic components and the somewhat improved air blast pattern precision. Recent technical advances are reviewed in this chapter. The most common sorting criteria are also discussed. Since some suppliers will not reveal their sorter technical features due to confidentiality concerns, the interested parties need to contact these for more detailed information.

INTRODUCTION Sorting coarse rock manually is an ancient method still practiced in many countries. The usual purpose is to reject barren waste or divide ore components into high-grade or low-grade portions. Sorting machines based on photometric principles were the first developed to replace manual sorting. Later developments included radiometric sorters (for uranium and golduranium ores), and conductivity and magnetic sensing methods (sometimes in combination with photometric technology). One sorter was based on elemental analysis, and processed a copper ore and later a manganese ore for a short time. Other technologies, such as microwave radiation and laserinduced fluorescence, were also attempted. However, with respect to commercial success, no other method has matched that of the sorters based on photometric principles alone. The first optical sorters were developed for agricultural applications in the 1950s, e.g. beans, peanuts, cereal flakes, potato chips, etc. In the 1960’s, these sorters were adapted to sorting black and white rocks. Driven by the need to process gold ores in South Africa and magnesite ore in Greece (Schapper 1976; Barton 1977), the 1970s brought forth the development of the fvst highcapacity sorter for ores with enhanced efficiency. The mechanized sorting proved more economic than hand sorting and many additional applications were developed in the calcite ore, phosphate, pegmatite, and even lignite coal areas (Arvidson June 1986; Arvidson October 1986; Arvidson 1987; Arvidson April 1988; Arvidson October 1988; Arvidson and Reynolds 1995) in addition to the afore mentioned. An optical sorter has four basic components: A feed presentation system, an optical system, a computing system, and a rock separation system.

1033

Feed Presentation System The feed presentation system may consist of several steps, such as sizing, washing, feed rate control, particle alignment, wetting, acceleration, and stabilization before passing the particles through the sorting zone. In nearly all applications, after sizing the rocks into the desired fractions, or in the screening process, the rock particle surfaces are spray-washed. Exceptions are lignite coal and rock salt. In very cold climates, a steadwater spray may replace water washing to avoid any ice formation on sub-freezing surfaces. Most feeding systems use vibratory feeders. Sometimes a sorter utilizes a slide chute to direct feed material into a particular path. For the fast-moving feed belt in laser sorters (4.1 d s ) , accelerating the rocks to high speed, then stabilizing them to ensure defined positions in the separation zone requires a high degree of moving parts synchronization and an elaborate rock stabilization system, see illustration in Figure 1. Most sorters use a slower speed feeding system, most often directly fed from a vibratory feeder into a free-fall area, or onto a slow-moving conveyor belt (0.5-1.5 d s ) . The optical system may be mounted above, or the rocks may drop into a free-fall zone where cameras view the rocks as they fall, see Figure 2. Soft Accelerator Roller

Slide Plate

soft Pullev

Main Belt Idlers Tail Pulley

Figure 1 M16 feed presentation system Materials to besorted FUNCTIONAL DIAGRAM

Feeding Devlee

Figure 2 Two-sided viewing using CCD line scan cameras and white light

1034

The vibratory feeders have a feed hopper associated with them, designed for the intended rock size fraction(s) and in accordance to the need for surge capacity.

Optical System There are two primary options for optical sorting systems: Either a scanning laser beam or a line scanning video camera. Various types of lasers are used, such as the common red laser or the (more rare) blue laser. The scanning video cameras range from simple b/w to color with very high resolution. The latter permits recognition of a wide range of surface properties in addition to the common brightness level of determination. A scanning laser beam system uses a multi-faceted, high-speed mirror wheel to provide the scanning feature, see illustration in Figure 3. Rocks are scanned as they leave the fast-moving conveyor belt, and photocells register reflected monochromatic light. The reflectance level, i.e., the brightness, and the rock size are then calculated as average values. Furthermore, the randomly positioned rock coordinates are determined so that a blast pattern can be accurately generated.

Mirror Drum With 20 Facets

Laser Ream

Photo Multiplier

Figure 3 M16 laser scanning principle Because of the monochromatic nature of a laser, the recognition of color variation is impossible. Using a red laser would result in equivocating light red rocks to the level of white rocks. If this is a problem, a blue laser could prove advantageous over red, though the blue laser is more costly and has a shorter life span. Scanning video cameras were used in several of the early photometric sorters and are commonly used today in foodstuff sorters and the most advanced ore sorters. High resolution permits recognition of subtle variations in surface properties along with brightness levels (e.g., quartz strings in dark rock), color differences (e.g., red and green colored limestone), and potentially small variations in colored stone (e.g., talc). Hence, line-scan video cameras can be used for all types of optical sorting, not only monochromatic applications. The most advanced cameras can sometimes “see” differences that the human eye may have a problem detecting - and at much greater speed.

Computing System For high-speed ore sorting in the early 1990’s, dedicated processors, such as the laser sorters and monochromatic line scan cameras, were still faster and more suitable than microcomputers. With the rapid advance of computer technology, the situation changed. Huge volumes of data could be processed at a faster rate and much lower cost. The volume of data to be processed in real color

1035

sorting is several orders of magnitude greater than for monochromatic laser sorting. Several new suppliers are taking advantage of advanced computing technologies that also enable features to make the machines more “user friendly.”

Rock Separation System To move rocks out of their natural trajectories, three systems have been applied: Mechanical, air jets, and water jets. The mechanical system may be a variety of pistons and plungers, while the fluid media use fast-moving valves. Obviously, the mechanical systems are limited concerning speed. In one “home-made” sorter plant, about six two-stage machines using modified tomato sorter plungers and over-band cameras were separating pebble-size rocks with less efficiency than a single M16 sorter. There have been attempts to move rocks from their natural trajectories using water jets to achieve desired separation; however, air is the only medium in use today. For small rocks (say below 15 mm), only a one-stage valve opening/closing is required, making the valve construction simple and inexpensive. For large rocks (over 50 mm), a two-stage valve may work best, while in separating intermediate size rocks (15-50mm) either system will work well. When the air blast system is used for large rocks, the high-pressure air in combination with the short blast duration (to achieve sufficient momentum) causes a high sound level. For small and intermediate size rock sorting, the sorter plant is quieter. For large rocks sorting, special sound dampening is needed. If the brightness (or other selected surface characteristics) meets predetermined criteria, air blast valves are activated based on rock position and calculated size to create a matching blast pattern. In most cases, the rocks that constitute the smallest portion of the feed will be blasted, unless there are technical limitations in properly recognizing the rocks. The greatest sorter operation cost is associated with the consumption of compressed air, often in the range $0.5 - l/ton of feed. Hence, it is important to minimize air usage. TECHNICAL CRITERIA FOR ORE SORTING FEASIBILITY Particle Size The maximum sortable size is determined by the maximum achievable air impact. For a standard compressed air system and an optimum air blast pattern, the size should be about 200 mm (8”) for a rock with a density of around 3g/cc. However, in practice the maximum size is between 150-160 mm due to economic design considerations, such as distance between falling rocks and the blast manifold, or wear on chutes and stabilizing components (if applied). It is assumed that when size is specified, we are considering a square mesh size with one side indicating the size of the rock. Obviously, some rocks will have a larger dimension than the size of the square mesh (“slabby” rock). To obtain the best grade-recovery relationship, the maximum size is determined by achievable liberation of the valuable component. This does not mean the rocks need to be totally liberated from a mineralogical viewpoint. In many instances, the economic liberation may be determined by the maximum tolerable amount of valuable material found in the waste portion. The smallest sortable particle size can be down to around 1 mm, but the economic limit may be much higher. Modified food sorters with channelized feed systems can sort rock salt at a few millimeters (bean and peanut sizes), and modified grape sorters are used for talc. These types of sorters operate at relatively low capacities (typically 1-5 t/h) and are relatively inexpensive being based on mass-produced food sorter equipment. For lower-value commodities, such as limestone, the economic limit may be closer to 15 mm. In one such application (essentially a black and white sorting case), high-intensity magnetic separation, a much lower cost process, performed at the level of an optical sorter-up to around 25 mm. To obtain an efficient physical separation of rocks with a minimum of “misplaced” particles, it is necessary to process narrowly sized material. Typically, the coarsest size fractions should have a top size to bottom size ratio of 2: 1 or less, and 3: 1 for fine size fractions. However, there is

1036

an example of a 5:l ratio for limestone, where the sorting precision requirement could be relaxed. In reality, mass distribution may also play a role in adjusting the size limits up or down. A typical application may have one sorter for 80-120 mm rocks, two sorters for 40-80 mm and three sorters (or more) for 15-40 mm. Surface Condition The particle surfaces must be sufficiently clean for an optical system to “see” the surface adequately. Often, sorting inefficiencies are caused by masking of surfaces in the viewing area by dust, dirt, or mist. In addition, in most applications, a moist rock surface exhibits the most distinct optical properties, while a dry surface distorts the rocks true characteristics. Still, an extremely wet surface may also mask the surface properties by having a reflectivity that is so high it “blinds” the optics. One known exception is lignite coal (7). Adding water to this material would turn it to slurry. Here, the surfaces are kept in their natural state while avoiding dust covering as much as possible. As mentioned, in a freezing environment, ice formation can be avoided if a steam spray is used. This may not provide the same level of surface cleanliness as washing/moisturizing during nonfreezing conditions, but is still a reasonable compromise. Color Criteria For black (or very dark) and white particle sorting, blasting the white rock from the feed stream is best. The optical system may not be able to distinguish the black (or very dark) rock from the reference surface. Though compared to the first generations of sorters the optical systems are more refined and are available with better reference media that enable more accurate rock color recognition, this problem still exists, especially when the reference medium can possibly become dirty. Fortunately, for the common application of magnesite ore sorting, the magnesite (i.e., white) particles constitute less than 50% of the feed fraction. Only when a small particle size fraction concentrate needs to be cleaned (i.e., removal of a small fraction of black rock) will there be a problem. However, since 1992 new economic options are available for the removal of such rock and may be complementary to optical sorting (8). For sorting of differently colored rock, we have to consider monochromatic and full light spectra sorting separately. Monochromatic Sorting Since color filtered light is no longer a needed option, only the laser light source sorting will be discussed here. There is currently only one supplier of the laser-type sorter since the demise of OSNA (Ore Sorters North America Inc.). For many years, the laser-type sorter (model M16)was a dominating high-capacity optical sorter. Some of this sorter technology, first developed in the 1970’s, has been modernized, but the light source limitation remains: It is monochromatic. However, it is a very high intensity light source, enabling sharp distinctions of even relatively small differences in average brightness levels. Another limitation is that only one side of the rocks can be viewed. This is usually of little or no consequence if the rocks are reasonably homogenous. However, as will be discussed in the section for actual applications examples, problems may arise if the rocks have distinctly different appearance depending on which side is exposed. The same limitation also applies to all one-sided viewing camera sorters.

Fuli (White Light) Spectra Sorting Several different light sources may be used in a sorter application. Some may provide more uniformity with regards to the spectra, while others will be biased towards one end of the spectrum. This subject is rather complex and is considered beyond the scope for this chapter. The main point is that for some applications, the selection of the light source may be essential to

1037

achieve maximum sensitivity. This applies primarily to applications involving rocks that have a substantial blue-spectrum component, for example green talc. When using advanced color-line scan cameras, a large amount of data can be obtained from the optical appearance of the rocks to be sorted. In addition to color differentiation, intensity and hue can be discerned over a large range. The possibilities to distinguish one rock from another seem endless. Therefore, it is important that suitable algorithms be applied in data processing to enable all relevant information for rock sorting. This is the main difference between suppliers of full- spectra sorting equipment. Full-spectra sorting possibilities are illustrated in Figures 4-9. Figure 4 shows rocks of different colors. The three on the left have the same brightness. The colors represent different characteristics of the ore. A full-spectra sorter proved capable of identifying reliably each of the differently colored rocks, while both a black and white, and monochromatic optical system failed. These systems coufd only distinguish between the three rocks on the left (light) and the two rocks on the right (dark).

Figure 4 Rocks with brownish-tan tint, pink color, gray (middle) and black colors. Figure 5 shows limestone rocks that are high-brightness white. In the next picture, Figure 6, the same rocks are turned over revealing the color difference of both sides. A sorter with one-sided viewing only has a 50% probability of recognizing the colored rock simply based on its orientation. With two-sided viewing, such rocks can be rejected. In some instances, the red color is more detrimental than other colors, so a full-spectra sorter can selectively reject the “worst” rocks if desired.

Figure 5 Limestone rocks will best side up

1038

Figure 6 Limestone rocks turned over ( U O O ) Additionally, shown in Figure 7, some rocks have the same average brightness level and similar colors, but are distinctly different in features such as contrast between darker and lighter areas. Although monochromatic sorters can be used to recognize “spikes” in brightness, the fullspectra sorter can more easily quantify the areas of contrast and hence generate more accurate sorting of such rocks. In this case, the rock with the greatest contrasts contains gold.

-

-

Figure 7 Gold ore rocks with gold- rock on right containing a piece of gold Figures 8 and 9 show the products from a Spectrasort machine. Obviously, the rocks in Figure 8 are of the best quality. In Figure 9, most of the rocks have a visible coloration. The rocks that only appear to be of good quality would generally have a discoloration on the other side.

Figure 8 High brightness limestone product

1039

Figure 9 Limestone reject product TESTING AND PLANT DESIGN The only testing facilities known at present are in supplier’s facilities. Typically, initial technical feasibility tests are done with 25-50 kg samples that have been sorted by the client into the appropriate desired products. The rock size can be of various fractions, for example 30 to 60 mm and 60 to 100 mm. The selection of a test sample of ore to be treated requires good professional geological knowledge including the mineral composition and optical recognition of the minerals, and the approximate liberation size. Once the proper sorting criteria have been established a largescale test is normally conducted. This can range from a few 100 kg for small particles to several tons for coarse rock sizes. In some cases, on-site trials may be required. With the advance of fuzzy logics and advanced programming, the sorters can be “trained” to first recognize particular categories of rock characteristics then distinguish rocks based on similarities within a category. Hence, the extensive test procedures of the recent past are no longer required. (The reader is advised to contact the relatively few suppliers of ore sorter equipment for further consultations.) If an application requires processing small particles (less than about 15 mm) at moderate capacities (less than 10 t/h), several suppliers manufacturing foodstuff sorters also offer “rock” sorter. However, it is this author’s advice to consult with the suppliers who have extensive experience in industrial ore sorter installations to avoid problems unknown to the foodstuff industry. It is often said that 90% of the ore sorting process success is dependent on proper feed presentation rather than the sorter machine. Besides properly designing the feed presentation system, here are a few pointers that may assist in the engineering design of a sorter plant, see the next section. Sorter Plant Design In the following, a few points are raised to consider when designing a sorter plant. Ore sorter suppliers usually provide engineering support with required detailed information. 0

Dust and moist air extraction. The blasting of rocks with high-pressure air jets produce an environment in the machine that is filled with small particles and mist. In order to keep the optical system operating as desired, this dirty air must be extracted and cleaned for adequate environmental control.

0

Adequate compressed air system. Design for a short piping length, sufficient buffer volume, and low-pressure drop system. The normal air cleaning (i.e., traps) to remove water and avoid freezing must be included. Depending on the compressor type and the valve system, oil mist may have to be added, not removed.

1040

0

0

Monitoring of feed and product streams. Many plants have sound insulated, temperature-controlled control room positioned so that the feed can be observed, but unfortunately not the product streams. CCTV or other on-line monitoring systems may be incorporated to view product streams. Water and particle collection system. Water spraying of the feed material and fastmoving rocks may cause substantial spillage in some operations. Collection and disposal should be engineered for easy housekeeping and maintenance.

These points may seem obvious for a seasoned plant operator, but there are many examples where they have not received proper attention. ECONOMIC CRITERIA FOR ORE SORTING FEASIBILITY Main Economic Benefits of Sorting Technology The major incentives for applying coarse rock sorting are: Rejection of barren (or low-grade) waste before transport to concentrator plant, saving transport cost for worthless material. Reducing comminution cost, a major cost element in all mineral processing operations. Reducing environmental impact by rejecting material that can either be used for construction purposes or stockpiled in a dry form, i.e., less fine particle waste is generated. Enhances the downstream processing efficiency due to more uniform and improvedgrade feed. Enhancing ore reserves since the mine cut-off grade can be reduced due to inexpensive ore grade upgrading. In some cases, rejection of waste material containing bad components (such as arsenic) that may be more expensive to remove after fine grinding. In some cases, final grade products or intermediate products may be achieved instead of using more expensive beneficiation methods (i.e., flotation). In other words, cost saving is the number one benefit of applying preconcentration by sorting (followed by the benefit of competitive cost for the production of final grade materials). Capital cost discussion Since a high-capacity sorting plant requires a substantial quantity of additional equipment and a high-capital cost sorter machine, the overall capital cost is a considerable factor in the total economic picture. In the 1980’s,when the M16 sorter sold for $520,000 to 650,000 (excl. compressor), the cost for a complete single sorter plant could be estimated at a factor of 2 times the sorter cost in the USA.Today, the sorter cost in real terms is much lower, typically 25% for the same capacity. Construction cost and the cost for mature equipment such as compressors and screens are relatively the same when inflation is considered. Hence, the multiplier based on a single sorter is currently much higher, in the range of 5-8depending on the location and the sorter machine selected. Based on constant dollars, the capital cost per ton of sorter feed has been reduced by nearly 50% in 20 years.

1041

The capital cost per unit of product is determined mainly by the plant capacity and the yield of the desired product or products. The capacity of the sorter plant is determined by the particle size range and the portion of the feed that needs to be blasted, see an illustration in Figure 10 for a typical high-capacity sorter. As shown, the capacity is an almost linear function of the average particle size for a narrow size fraction. SORTER CAPACITY TPH as fuiictioii of rock size and percent reject

AVERAGE PARTICLE SIZE IN MM

Figure 10 Assuming a capital cost of 10%per year, the capital cost per ton of feed as a function of the average particle size in narrowly sized fractions can be approximated as shown in Figure 11. If the yield is known, the capital cost per ton of product can be calculated by dividing the cost per ton of feed by the product yield fraction. For an intermediate particle size magnesite application, the yield may be about 20%. Hence, if the capital cost per ton of feed were $0.30, the capital cost per ton of product would be $1SO.

1042

Approximate Capital Cost in US$ Per T/H Feed (7,200 hrsla)

Average Rock Size In mm

Figure 11 Operating cost discussion The dominating cost for sorting lies in operation. The possible exception would be small particle (- 10 mm, - 3/8”) sorting, where the capital cost may dominate. Compressed air consumption comprises the dominating operating cost element. Blasting small rocks requires somewhat more air per ton of blasted material than large rocks, but the difference is insignificant. Quoted numbers are about 40 m%on for large rock and 50 m%on for small rock. In the US with relatively low energy cost ($0.03-0.04/kWh), the compressed air cost may be approximately $0.35 to 0.70/ton feed rock. (The cost needs to be prorated according to the electric energy cost.) Maintenance cost can vary from roughly $0.35 per ton feed for some large rock (100-140 mm) M16 sorter plants to a few pennies per ton for less maintenance-intensive, free-fall sorters processing intermediate size rocks (less than 50 mm). APPLICATION EXAMPLES The following optical sorting applications are in operation or have been in operation:

1. A. 1.Magnesite: Greece, Australia, Turkey 2. A.2.Limestone: Finland, Italy, and USA. 3. A.3 Phosphate: USA

4. A.4 Pegmatite: USA, Italy 5 . A S Gold Ore: South Africa, Japan

6. A.6 Lignite Coal: Hungary 7. A. 7Talc: France, USA 8. A.8 Lead-Zinc Ore: Greenland 9. A.9 Pyrite Ore: India

1043

Similar ores are shown to be sortable in other countries, but have not been implemented, or there is no public information documenting such plants. In one case of limestone sorting, the only high-capacity, large particle sorter commercially available at the time of plant implementation was the laser sorter. This ore was of the “slabby” nature, meaning the limestone could be comprised of layers of good white limestone and strongly colored layers, usually red or green, with some having very dark to black components (amphiboles). These colored components reduce the final product brightness and may create issues with the yellow color spectrum specification. Since the laser sorter could only “see” rocks from one side, there was a 50-50 chance that discoloration on one side of a flat rock would go unnoticed. To compensate for such a sorting “error,” the average brightness level was set at such a high level that only about 30% of the rock was blasted into the high-grade product sending the rest to the low-grade fraction, which fortunately had some value. It was shown that if a two-sided viewing were applied, about 75% of the rocks would be acceptable as a high-grade product, while rejecting 25% to the lower-grade products (still having value).

PHOTOMETRIC ORE SORTERS ON THE MARKET The best known suppliers of optical sorters are: Ultrasort (Australia). Web site: www.ultrasort.com.au

Figure 12 Sorter from Ultrasort

1044

Mogensen GmbH (Germany). Web site: www.mogensen.de

Figure 13 Sorter illustration from Mogensen GmbH

AIS Sommer (Germany). Web site: www.ais-sommer.de Spectrasort (Italy/Switzerland). Web site (under construction): www.spectrasort.com

Figure 14 Spectrasort installation (processing limestone)

1045

PLANT INSTALLATION EXAMPLE The largest plant in the world using optical sorters is Grecian Magnesite with about a dozen largecapacity sorters operating in the plant in Yerakini, Greece. All of the sorters work on the brightness level principle and process ore from 25 to 120 mm. The first two sorters were commissioned in the late 1970’s, which pioneered the use of laser sorters in a large-scale operation. For more details, see reference 6. ECONOMICS OF SORTING The author has reported several examples on the economy of sorting. Here is one of the examples: Table 1 Magnesite example BASIC MINE DATA 35,0 ROM Grade Magnesite % Kiln feed tons per annum 60,000 1300 300 Operating days/yr. Mining cost per per ton ton for for first first 100 Is1 $11,00 ining cost 1,00 kta CALCULATION OF MINING RATE Sortable portion % Sorter feed grade % Magnesite recovery % Mining rate TPD CALCULATION OF ECONOMICS Marginal min. cost $/ton Sorting op. cost $/ton feed Capital cost $x000 Interest rate % Depreciation, years Capital cost/year Feed prep. $/ton feed Waste transportation cost/ton Value of kiln feed $/ton Total cost/year Revenue/year Value of extended mine life $ 8,00 per ton Value in favor of m/c sorter/year

I HAND SORTING 45,0 29,0 90,0 1702,9

MACHINE SORTING 55,0 33,0 95,0 981,5

7,00 3,40 200 10,00 10 $ 32,549 $1,80 $0,50 $70,00 $ 6.890,416 ($ 2.690,416)

7,00 1,10 1100 10,00 10 $ 179,020 $1,80 $0,50 $70,00 $3,611,221 $ 588,779 $ 1.208,480

17,00 10,90 |1400 |10,00 110 ||$ 227,844 |$ 1,80 I $0,50 |$ 70,00 |$ 3.001,909 |$ 1,198,091 |$ 1.651,151

$ 4.577,675

|$ 5.539,659

1046

M/C SORT & MAG. SEP. 80,0 33,5 95,0 1785,5

1

SUMMARY Improved optical components and computing technology advances over the last ten years have enabled substantially reduced costs for sorting equipment. New suppliers offer equipment mainly for small or intermediate size particles (10-30 mm), and one new supplier offers sorters over the entire size range that can be handled by high-capacity sorting equipment. Sorting can now be based on more criteria than just brightness levels. These include some aspects of surface texture and color spectra. Two-sided viewing enhances the efficiency in some applications. An equipment comparison overview is shown in the table below. FEATURE Rock size

SCANNING MONOCHROMATIC Large size (over 100 mm)

No Double-sided viewing Contrast in single rock Limited High capacity Adaptable to varying conditions Installation base in nonfoodstuff industry Sort by brightness Sort by color cost

SCANNING FULL SPECTRA One for large size, most for moderate to small Yes - one supplier Yes Yes - for most Nearly unlimited

Yes Limited

-

Large if old models are Large if glass sorting is included expanding in ore applications included Yes Yes Nearly unlimited Very limited Lower than old technology Lowest

CONCLUSIONS Except for diamond ores, photometric sorting is a relatively unknown technology in the mining industry. Very few applications and a lack of information are probably the main reasons; however, when technical conditions permit effective sorting, the economic advantages can be substantial. As for any coarse rock beneficiation process, the liberation level of the valuable and waste component(s) is of outmost importance to establish the technical feasibility. Advances in optical systems and computing technology have drastically reduced the cost for machine sorting in the last decade and greatly enhanced the application possibilities.

1047

REFERENCES Arvidson, B.R.: New Applications of Ore Sorting for precious Metals Recovery. Randol Gold Workshop, Beaver Creak, Colorado, USA, June 103, 1986. Arvidson, B.R.: Sorter Applications for Precious Metals. Paper Presented at the American Mining Congress International Mining Show, Las Vegas, Nevada, USA, October 5-9, 1986. Arvidson, B.R.: Economics and Technical Features of Preconcentration Using Sorting. Society of Mining Engineers Annual Meeting, Denver, Colorado, USA, February, 1987, Preprint No. 87125.

Arvidson, B.R.: Industrial Minerals Beneficiation by Ore Sorting. Paper presented at Industrial Minerals International Congress, Breakaway Processing Session, Boston, April, 24-27, 1988. Arvidson, B.R.: Sorters for Coal and Lignite. Coal Sorting Seminar, Oroszlany, Hungary, October 11-12, 1988.

Arvidson, B.R., Reynolds, M.S.: New Photometric Ore Sorter for Conventional and Difficult Applications. Presented at Processing for Profit (Industrial Minerals), Session 4, Paper 1, Okura Hotel, Amsterdam, Netherlands, April 26-27, 1995. Barton, P.J., Schmid, H.: The Application of LaserPhotometric Techniques to Ore Sorting Processes. Paper presented at XI1 IMPCS, Meeting 3, Paper 1, Sao Paulo, Brazil, 1977. Schapper, M.A. : Beneficiation at Large Particle Size, using Photometric Sorting Techniques. Presented at 2" IFAC Symposium on automation in Mineral and Metal Processing, Johannesburg, September, 1976, and Australian Mining, April, 1977

1048

Electrical Methods of Separation Andrew L. Mular', P. Eng., Distinguished Member SME-AIME, Fellow CIM ABSTRACT The principles of electrical methods of separation are reviewed, where electrical properties of solids become important. Relevant characteristics of electric fields are summarized along with charging processes such as by contact charging, by induction in the field, by conductive induction on a field electrode and by ion spray in a corona field. Electrical forces, combined with gravitational and centrifugal forces, act on charged particles and are responsible for mineral separation. Usually, commercial separators employed in mineral processing are of the drum-type, while free fall separators are typically used in potash processing. INTRODUCTION Electrical separation involves the separation of usually dry solids from each other or from a suspending medium such as air based on differences in electrical properties under the influence of electrical, gravitational and/or centrifugal fields. Typical separations include:

Coal from shale Feldspar from quartz Zircon from ilmenite Halite from sylvite

Solids from gaseous effluents Fe,O, from silica Diamonds from silica SnO, from scheelite

For separating solids from effluent gas, there are no lower size limits; upper sizes are determined by particle diameter, specific gravity and retention time in an electric field. For mineral separation, lower limits are around 200 mesh. Feed can be about 100% passing 14 mesh. Flowsheets that involve electrical separation may include roughers, scavengers and cleaners in the normal manner. Electrical separation is used for industrial minerals and there is widespread use in the food industry. Separation of dusts from gases by means of ionic field separators is widely accepted (Cottrell precipitators). Relevant Electrical Properties of Solids Solids have differences in electrical resistivity (resistance to the flow of electric charges) and dielectric constant (measure of the ability to polarize in an electric field). Differences serve as a basis for separation and provide a means to classify solids, namely, as conductors, non-conductors and semiconductors. Shown on the next page (Table 1) is a chart that lists minerals as either conductors or non-conductors. Conductors can be separated from non-conductors. Conductors. Conductors have resistivities of about ohm cm. They permit the instantaneous flow of electric charges under the influence of very small electric fields (potential differences). If, by some means, charges are transferred to a conductor: i. There will be no internal electrical fields (if there were, the conductor would

immediately allow the flow of charge so as to reduce the field to zero). ii. Excess charge resides uniformly on the surface which is at the same potential at any point (if there were a difference in potential, charges would flow immediately to reduce the difference to zero). 'Mineral Processing Professor Emeritus, Dept. of Mining, (Jniversity of B. C., Vancouver, B. C., Canada

1049

Table 1: List of typical conductors and non-conductors

Anhydrite Apatite Barite Bastnasite Beryl Calcite Chrvsotile Corundum Diamond Epidote Feldspars Fluorite Garnet Gypsum Hornblende Kyanite

Non-Conductors Mica (Biotite) Mica (Muscovite) Monazite Olivine Perovskite Quartz Scheelite Siderite Sillimanite Sphene Stgaurolite SulDhur Topaz Tourmaline Xenotime Zircon

Acmite Augite Brookite Cassiterite Chalcopyrite Chormite Columbite Copper Davidite Euxenite Ferberite Franklinite Galena Gold Graphite Hematite

Conductors Ilmenite Ilmenite (hi iron) Ilvaite Koppite Magnetite Marmatite Molybdenite Pyrite Rutile Samarskite Sphalerite Tantalite Wolframite Wulfenite

Conductors have dielectric constants that are extremely large (infinite for perfect conductors) and resistivities are very small. Non-Conductors have resistivities of about lOI4 ohm cm. They, theoretically, do not permit the flow of electric charges in their interiors. If, by some means, charges are transferred to a non-conductor: i. Internal electric fields arise (charge moves relatively slowly to reduce fields to zero). ii. Charges reside on the surface (not uniformly to start) but potential differences can

exist between any two points. It takes time to reduce them to zero. The dielectric constant of a non-conductor is close to 1 (water is about 81).

Semiconductors. Semiconductors have resistivities of from about 1 to about lo4 ohm cm. These solids have electrical properties intermediate between conductors and non-conductors. Since many of our minerals can be classed as semiconductors, it is useful to consider them in some detail. The band theory of solids is the basis for explaining differences between insulators and semiconductors. Permitted energy levels (levels in between are forbidden to electrons) of atoms combine into bands of permitted levels in a solid. Thus, we have a K-band, an L-band and so on just as an atom has a K-shell, an L-shell and so on. Forbidden levels in atoms combine into bands of forbidden levels in the solid. At absolute zero, valence electrons occupy a band of levels in the solid (called the valence band), which is separated by a forbidden band of levels (referred to as an energy gap) from the band of empty but permitted levels just above it (called the conduction band). If the valence band is filly occupied by electrons and if the energy gap is small enough, thermal energy at room temperatures can easily excite electrons into the conduction band which has plenty of empty levels. This leaves the valence band with immobile positive atoms (holes or sites). Electric fields easily excite electrons residing in the conduction band and positive sites residing in the valence band and conduction occurs. The solid is called an intrinsic semiconductor, because the number of conducting electrons in the conduction band is equal to the number of conducting sites in the valence band. If donor impurity atoms (such as A1 atoms in ZnO) occupy lattice positions or interstitial sites, they introduce electrons whose energies lie in the energy gap just below the conduction band. At room temperature, these electrons are easily excited into the conduction band. The semiconductor is called n-type, because negative electrons in the conduction band are charge carriers. If acceptor impurity atoms (such as Cu atoms in ZnO) occupy lattice positions or interstitial sites, they introduce empty energy levels in the energy gap just above the valence band. Electrons in the valence band are easily excited, at room temperature, and occupy these empty levels where

they are tightly bound. As a result, many valence band atoms, now deficient in electrons, have net positive charge sites that readily "move" in an electric field (the motion of a positive site in one direction, means that valence band electrons are jumping in the opposite direction om one atomic valence site to the next). The semiconductor is p-type, because positive sites in the valence band act as charge carriers. Clearly, depending upon the concentration and type of impurities present, electrical properties of semiconductors can be similar to either highly conducting or highly insulating solids or to something in between. Surface Characteristics. The surface resistivity may determine the apparent electrical properties of a solid. Suppose a variety of ZnS is an insulator. If it is treated in dilute copper sulfate solution and then dried, it exhibits conductor-like behavior. Selective surface treatment can, in principle, convert conductors into insulators and vice-versa. Surface resistance has an important role in charging and discharging of particles (see subsequent sections).

RELEVANT CHARACTERISTICS OF ELECTRIC FIELDS Electrostatic separators employ a generator to produce electrostatic fields. The field electrodes of the separator are shaped to create special fields which enhance the separation of minerals. Generation of An Electric Field For many years, the workhorse in electric field generation was the mercury arc and high vacuum tube rectifier, which have now been replaced by solid-state selenium or silicon (preferred) rectifiers. Vacuum tubes with about a 20,000 hr life were immersed in an oil bath, where the oil (transformer oil) was a very high grade dielectric capable of withstanding field voltages of at least 40 kv. Solid state units do not require oil baths. DC generators (we no longer employ Van de Graf generators) take normal AC voltages and convert them to DC voltages via a combination of high voltage step-up transformers/rectifier/filters. High voltage output can be half-wave rectified or full-wave rectified with filtering to obtain a more constant voltage. One post of the generator is grounded, while the other post (usually negative in polarity) is attached to the insulated field electrode of a separator. Figure 1 below shows a typical connection. 120V AC

P P

I'

W

if

Insulated Field Electrode

Grounded Field Electrode I

I

Note that either side of the generator can be grounded to obtain the desired polarity of the insulated field electrode.

1051

Types of Electric Fields Recall that the shape of an electric field is determined by the geometry of the field electrode surfaces facing each other. One of the field electrodes is grounded; the other is connected to a generator pole of choice (either + or polarity), although the pole of negative polarity is preferred as we shall see. The other generator pole must be grounded to complete the circuit. The field electrode and its generator pole of choice must be isolated fiom ground and connected via shielded cable. In general, the electric fields thus created can be either uniform or non-uniform. Uniform fields. Uniform fields, which are rarely employed for electrostatic separators, are generated when flat field electrodes are parallel to each other and connected to a DC generator. Characteristics are shown in Figure 2.

-

?%=

Ey = Lldatrength along y atcoMtantx

,I"

UNIFORM FIELO

Since the capacity, C, of the field electrodes is the ratio of the charge, q, on an electrode to the potential difference, AV ,across the electrodes, then:

where A is the surface area of the electrode plate exposed to the field and epsilon is the dielectric constant of the medium between the plates (for air it is 1.00001). This shows that the charge on an electrode is proportional to the voltage and varies inversely with d. Non-Uniform Fields. These fields are shaped by electrodes which have curvature that is either homogeneous or non-homogeneous. Examples of homogeneous fields are shown in Figure 3.

1052

HOYOOENEOUS NOWNIFORM FIELDS

Figure 3: Homogeneous non-uniform electric fields To determine the capacity of the above field electrodes, the expression for a uniform field must be made a function of the x and y coordinates (fields are uniform along z except for edge effects). If the curvature of the plate follows a known functional form, then field gradients in the x and y plane can be determined as a function of x and y. In other words, lines of equipotential are effectively parallel to the surface curvature of the field electrodes. Lines of force are always at right angles to lines of equipotential and are perpendicular to the surfaces of field electrodes. For the cylindrical electrodes, which are mirror images, the capacitance is given by:

where d is the distance between the cylindrical centers, r is the radius of a cylinder, and C is the capacitance per unit length of cylinder. Note that Cosh refers to the hyperbolic cosine. Typical non-homogeneous non-uniform fields are shown in Figure 4.

NON-HOMOGENEOUS, NON-UNIFORM FIELDS WITH AND WITHOUT CORONA.

Figure 4: Typical non-homogeneous, non-uniform electric fields

1053

The essential feature is that field electrodes have curvature but are not mirror images of each other. For the configuration on the left side (Diagram I , where two cylindrical electrodes are of different radius) the capacitance per unit length of cylinder is given by:

Diagram 2 (upper right In Figure 4) shows what is known as a corona field, where the radius, of one cylindrical electrode has been shrunk to about 0.5 mm or less. In this situation, there is a finite flow of ionic current generated by breakdown of air molecules that ionize around the wire and flow in a broad swath towards the opposite electrode. In the corona field, the average current flow in coulombs per second is approximated by:

[

i =2xsk E,

r'l

+-

(4)

where k is the ionic mobility, Er is the field gradient at a distance r from the wire and J is a constant for a given set of conditions. Note that r I h , where h is the distance from the surface of the drum of radius rl to the wire. Usually, the corona wire diameter, r2, is very small compared with rl the large drum radius. In this case, the capacitance is given by:

To direct the ionic current to the large drum in a narrow beam, a small drum,of radius R is connected electrically to the wire (lower portion of Figure 4). To estimate the extent (area) of coverage on the large drum, draw tangents to the small drum that cross at the corona wire and extrapolate to the surface of the large drum . Corona fields have some interesting properties. At high enough potentials, the field gradient near the wire is so large, that air molecules (mainly 02)are zapped with electrons to form ozone and other ions. These travel towards the field electrode of opposite polarity to form a current. In doing so, they carry air along, which is equivalent to an "electric wind". At potentials of about 8kv and above, flows of current are of the order of 0.1 to 1 microamp. As the potential increases, flows of around 100 microamp are observed prior to breakdown (the breakdown strength of air is around 27 x l o 6 coulomb/mete? ). Near the threshold voltage (prior to any sparking) a positive corona (wire is + relative to opposite electrode) is noisy (hisses) and intermittent. It goes over to a steady blue glow at somewhat higher potential and becomes very irregular and electrically noisy. Suddenly streamers appear and sparking occurs. A negative corona (preferred) is not as erratic and is stable to just above the threshold. It is then interrupted at a regular pulse rate (a pulse lasts for around lo-' seconds). The rate of pulsation increases to about lo6 when breakdown occurs. Near the breakdown point, the "electric wind" is often felt. Non-Uniform fields are desirable in electrostatic separators, because a dielectric force acts on particles to lift them preferentially towards a field electrode, as discussed in a later section. ELECTRICALLY CHARGING PARTICLES AND BEHAVIOR IN FIELDS The electrical force that acts on a charged particle in an electric field is the product of the charge, q, on the particle and the field intensity, Er, or F = qEr . If q is in coulombs, and Er is in dynes per coulomb, F is in dynes. Often, field strength is expressed in Newton per coulomb, where 1 Newton = lo5 dyne. Thus, means to deliberately charge particles is important to electrostatic separation, if we wish to take advantage of the magnitude of F.

1054

Particle Charging Mechanisms Particles are charged by contact, by ion spray in a corona field, by induction and by conductive induction on a grounded field electrode. Charging By Contact. Contact charging occurs when two dissimilar solids make surface contact. Charge arises due to a difference in work functions of the two solids. Remember, the work. function of a solid is a measure of its electron affinity. A high work function means a high electron affinity. When two dissimilar solids are in contact, a contact potential difference arises. Electrons transfer from the solid of lowest work function (lowest electron affinity such as a metallike solid) to the solid of highest work fimction (highest electron affinity such as an insulator). The contact potential difference is proportional to the difference in work functions. Because the work function depends upon the dielectric constant, it has been shown that:

E ~E ,, ~ ,k are, respectively, the surface charge density on the conducting particle, the where oC, dielectric constant of the conductor, the dielectric constant of the non-conductor, and k is a constant for the mineral combination. Minerals of high work function have low dielectric < E~ , electrons constants; those of low work function have high dielectric constants. Since transfer from c (conductor) to nc (non-conductor). c is left with a net + charge. Charging By Induction In An Electric Field. If a conducting particle enters an electric field, the field electrodes can charge the particle by induction. The particle will polarize. Negative charges will tend to orient towards the positive electrode; positive charges will orient towards the negative electrode. If the particle was uncharged to begin with, there are an equal number of positive and negative charges. The particle is equivalent to a dipole, which is attracted to the nearest electrode in a homogenous field . If the field is non-homogeneous, the dipole is drawn to regions of highest field gradients. Various minerals may have permanent electric dipoles, so that these reinforce those which are induced in an electric field. Charging By Conductive Induction On Grounded Field Electrode. The essential idea behind conductive induction is illustrated in Figure 5.

-

d u c t i n g partide polarizes by induction

+Bcbodb conducts -charge to particle in field

particle conducts + charge to negative electrode

-

CONDUCTIVE INDUCTION

Figure 5: Charging by conductive induction

1055

Y

Conducting particles enter a non-homogeneous non-uniform field. In the field, particles polarize by induction but electrical forces are insufficient to cause attraction to the positive electrode by overcoming the influence of gravity. Particles make contact with the negative grounded electrode and begin to lose their positive charges via conduction. If the particle remains in contact for a sufficient period, it begins to assume the potential of the negative electrode as conduction continues in the reverse direction (electrons to particle). Figure 5 shows an equivalent circuit for the particle in contact with the grounded electrode. The net charge, qci ,on the particle at any time, t, is:

The term - is called the leakage rate constant, while the product CpAV is the maximum RPCP charge acquired by the conducting particle. The particle resistance, R,,, is comprised of the particle bulk resistance, Rb and particle surface resistance, R, such that

suggesting that the two resistances are in parallel. Conductors with properties more analogous to semiconductors will have small leakage rate constants and will leak charge more slowly to the grounded electrode. Charging By Ion Spray in Corona Field. Figure 6 illustrates the influence of ion spray charging by a corona field. The essential idea is that the small cylindrical electrode, attached to the corona wire and isolated from ground, acts to beam the ion flow onto the larger cylinder which is grounded. A particle residing on the grounded cylinder is charged positively by the ionic flow of current (1 to 100 microamps). Both conductors and insulators are charged indiscriminately. However, insulators hold their charge for a much longer time. Depending upon times of contact, a conducting particle will be influenced by conductive induction to a greater extent. The magnitude of the ion current, qi , is estimated from:

where E is the dielectric constant, r is particle radius, Er is field strength at the particle position, n is concentration of ions at the particle surface, e is charge on the electron, t is time and f(RpCp) is a leakage function.

1056

Figure 6: Influence of ion spray charging by a corona field Electrical Forces Acting on Particles in Electric Fields The electric force, F,,, acting on a particle at a point in a field is given by:

where s means surface, A is the surface area of the particle (cm2), Er is the field strength at the particle position (newtodcoulomb), is the surface charge density on the particle (coulomb/cm2), F, is the electric force acting on charged particles, and Fd is the dielectric force. The latter force is discussed in subsequent paragraphs. If the total charge, qp, on the particle is known, the electric force is:

Fe,

= Fc+Fd=

Erqp +Fd

The total charge arises from the following sources: Induction associated with induced and permanent dipoles

= qi

Conductive induction Contact charging Ion spray bombardment

= %i

-

- qc -

1057

qis

The total charge on the particle at any time is the algebraic sum of the above:

Contact Charging Force, Fc, Acting on Particles. A charged particle not in contact with a field electrode is acted upon by a force Fc that steers it towards the field electrode of polarity opposite to that of the particle. A positively charged particle will migrate towards the negative electrode. However, if the particle is in contact with a field electrode, the force Fc is either lifting or pinning in nature. The essential idea is illustrated in Figure 7, which shows two particles in contact with a grounded positive field electrode. One particle has a net negative charge; the other is charged positively.

imlated fmm ground

T

Fc is l i n g

Fc is pinning

rmgaiive Mki electrode grounded

Figure 7: Illustrating electric force due to contact charging The particle with a net positive charge experiences a pinning force Fc of attraction to the grounded electrode (or of repulsion by the positive insulated electrode). A particle with an equal but opposite net charge (- charge) experiences a liftingforce Fc of attraction towards the electrode above it (or of repulsion by the grounded negative electrode). Pinning implies that the particle is attracted by the electrode on which it rests in the field; lifting implies that the particle is repulsed by the electrode on which it rests in the field. Force, Fd, Acting on Particles Due To Dielectric Constants. Figure 8 illustrates the influence of dielectric constants on particle motion in a field.

1058

L20-V AC

S M n g lines of twcs in the nonunifom

INFLUENCE OF DIELECTRIC FORCE ON PARTICLES IN NON-UNIFORM FIELD

Figure 8: Influence of dielectric force on particles in non-uniform field Particles of dielectric constant higher than the suspending medium are attracted into regions of highest field gradients regardless of their charge, because they have the highest specific inductive capacity and will accommodate the field of force more easily than the suspending medium. They are "more permeable" to the field than the medium. In contrast, particles of dielectric constant lower than the suspending medium will migrate towards regions of lowest field intensity because they are "less permeable" to the field than the medium. Most minerals have a dielectric constant greater than air and will experience a dielectric force that tugs them towards regions of highest concentrations of lines of force. However, those minerals that have a dielectric constant greater than others will experience a larger force. The difference in dielectric constants of minerals is often much greater than the differences in density, magnetic susceptibility or electrical conductivity, so that dielectric separations are entirely feasible. Image Force, F,,, Acting on Charged Particles Outside the Field. The idea of an image force, Fi,, is illustrated in Figure 9.

d-p.mabdm RdidabrdeapomdloWd

d**on

T m a * w

In(o(hsms(.lOfaU~Rd*..

Figure 9: Image force acting on particle outside of field

1059

Referring to Figure 9, the assumption is made that a charge of -q is concentrated at the centroid of a particle that is not resident in the electric field. Nevertheless, the charge -9,when placed near a conducting sheet, experiences an attraction of +q units of charge located one particle radius into the metal (based on the mirror image principle of electrostatics). From Coulombs law: 2 2 2

-9'- - --n o d F.im = Ed

E

where 0 is the surface charge density of a particle of diameter d. This shows that Fim is proportional to the square of the surface charge density and the square of the particle diameter. It is the image force that binds non-conductors tightly to the drum in drum-type electrostatic separators. A brush and a special wiper electrode are employed to dislodge the particles. While in the field, particles are pinned or lifted depending upon the particle and its charge. Outside of the field, particles that have retained a charge will now experience the image force. Thus, Fc in the field has become F, outside of the field. Gravitational and Centrifugal Forces Acting on Particles in Electrical Fields Depending upon the design of the electrostatic separator, either gravitational forces or centrifugal forces or both will act on particles in electric fields. Consequently, at any one time, a force balance must be considered to estimate the consequences of the various forces involved. Figure 10 below illustrates the idea, where particles are in contact with a large rotating cylindrical field electrode facing both a corona and a static field.

Fel =total electric forw ading on particle = Eq + Fd Fcf = centrifugal force = mrw(w) Fn = normalcomponent of Fg = fg Sin A Fg = force due to gravity

Figure 10: Gravitational and centrifugal forces acting on particles in electric fields In the field, forces acting on a conducting particle are FeI which consists of Fc and Fd with both acting to lift the particle away from the rotating electrode, FCfthe centrifugal force which is equal to mn? (r is particle radius, m is particle mass and w is the rpm of the drum), and F, the normal component of the force, F, due to gravity. Note: F, = mg and F, = F, Sin A. If the particle is an insulator, Fel will likely act opposite to that shown in the diagram and cause pinning, depending on the rpm. If the insulator is outside the field, then F, becomes Fim and the image force acts to retain it to the drum. The force balance for the particle shown is: mg sin A = &

+

wp+ Fd 1060

where Erq, is either lifting or pinning in the field. Outside the field, the charged particle generates an image force consistent with whatever value qp is at that point. For the case considered, namely, that the particle is a conductor, F,, will likely cause the particle to lift rapidly with reinforcement from the centrifhgal force. If the particle drops below the horizontal and is still held to the drum electrode, the image force is responsible. It can hold insulator-like particles tightly, so that special brushedwiper electrodes may be required to dislodge them. ELECTROSTATIC SEPARATORS Static field separators are those which use normal electric fields; ionic field separators employ fields in which there is a flow of ionized gas molecules/atoms. Some texts refer to static field separators as electrostatic separators and ionic field separators as electrodynamic separators or high tension separators. Rather than confuse the issue, the words "electrostatic separator" will refer to any device that uses electrical fields (static and/or ionic) to separate minerals in air. Other terms will be used to better define the separator (dielectrophoretic separator, for example, or electrostatic precipitation) in special cases only. Electrostatic separators often have the following characteristics: 1. 11.

iii. iv. V.

vi. vii. viii.

A way to present dry ore to the separator (ore temperature may be raised). A way to generate electric fields across metallic field electrodes. A combination of metallic electrodes that introduce ionic and/or static fields. Field electrodes assist in the transfer of particles through fields. A means to ensure that particles exit electric fields unhindered. A means to split the exiting stream into a concentrate, middling and tailing. A way to keep field electrodes clean. Feed hopperddistributors and discharge hopperdconveyors.

Electrostatic separation plants will have drying, dust collection and storage bins. Dusting can be a problem that must be resolved during early phases of plant start up. A circuit diagram for iron ore cleaning at Wabush Mines is shown below. ONVEYOR TRIPPER

HIGH TENSION

FINAL CONCENTRATE

Figure 11: Iron ore cleaning flowsheet at Wabush Mines

1061

Drum Type Separators Drum-Type electrostatic separators employ either a conductance field (Figure 4, Diagram 1 ) or an ionic field (Figure 4, Diagram 2 or 3) or a combination of both (See Figure 10). A common design in Mineral Processing is the Carpco separator, where the two fields are combined in whatever way enhances gradelrecovery. Figure 12 is a schematic of the laboratory model with the rotor brush and AC wiper electrodes not shown.

ntcn VOLTAGE METER

.

METER SWITCH INPUT MlGH VOLTAGE b OUTPUT

nran VOLTAGC CONTROL MOMENTARY CONTACT SWITCH ON DELAY R C U I FOR MlQH VOLTAGE MIGM VOLTAOC POWER SWITCH

SLIDING DOOR

HlGM VOLTAGC RLCTlClER

b MAIN W W E R SUPPLY SWITCH AND CUSL .OX

IN THIS SECTION

Figure 12: Laboratory model of drum-type electrostatic separator Because lab models are identical in cross section to a full scale unit, scale up is obtained by multiplying the capacity of a lab test unit by the field electrode length of an industrial unit, selecting lengths of up to 8 ft. Different diameters of rotating field electrodes (called rotors) can be evaluated in the lab with a research model (recall that the centrifugal force acting at the surface of a rotor at a given rpm, is proportional to the drum radius. Hence retention time in the field and FCfcan be altered to suit). Capacities are reported as lbs per hour per linear length of rotating field electrode ( rotor). A rule of thumb for capacity is that a unit should handle about 100 lb per hour per inch of rotor. Remember: In the Figure 12 research model, about 1/4 of an inch on each side of the rotor does not contribute to the capacity; in a commercial unit, about 2" on each side does not contribute. Accurate scale up should reflect this. The research drum separator employs a metallic feed hopper for uniform distribution of -10 mesh ore onto the rotor. Normally, the ore is first deslimed (-325 mesh taken out in some way), unless this has been done by nature (beach sands for example) and then dried. An infrared lamp is employed to heat the drum. A low voltage AC generator applies an AC field between the rotor and small wiper electrodes located at the bottom left of the rotor. This serves to dislodge particles held by the image force. A sweeper brush at the left side of the rotor keeps the rotor dust free.

1062

Cutters can be adjusted to divert a tailings (non-conductors), a middling (semiconductors and locked particles) and a concentrate (conductors) into corresponding collection hoppers. The cutterhopper collectors can be replaced with a special series of about 20 bins that permit of an assessment of "best" cutter locations. Normally, the rotor is the grounded field electrode, while the positive pole of the DC generator is likewise grounded. The isolated negative pole connects to the ionic field beam electrode and the conductance electrode. These are "hot" and should not be touched by hand. The best way to understand the behavior of the separator shown in Figure 12 is to trace the charge on a conducting and non-conducting particle as they progress from the feed hopper to a discharge point outside of the fields. Tracing Charge on Conductor Passing Through Drum Separator. Figure 13 illustrates important features of a conducting particle being fed to the drum separator (see Figure 12).

-

-

0 CONDUCTING PARTICLESPASSING THROUGHCARPCOSEPARATOR

PositivelyGrounded Rotor at w rpm

I I A

I

I

B

% +

HYPOTHETICALCONDUCTINGPARTICLE

Figure 13: Conducting particle passing through drum separator The conductor particle, with a low electron affinity, gives up negative charge to the hopper via contact and gains a net positive charge. When the particle touches the positive rotor (point A), more + charge is gained, but the negative corona is beginning to charge the conductor negatively to cause a pinning effect (point B to C). As the particle travels farther to the right it becomes more

1063

negative. At point D, the corona field has weakened and conductive induction occurs. The particle picks up a strong positive charge fiom the rotor. At point E, the particle is lifted by the negative field electrode in the conductance field. The particle, in combination with centrifugal force, continues upwardly and out. At point F, the particle is losing + charge to the atmosphere and falling downwardly under the influence of gravity. Tracing Charge on Non-Conductor Passing Through Drum Separator. The corresponding effect on a non-conductingparticle is shown in Figure 14.

I

::q-;,. Jl Ion Beam

DC Generator-

/ I

-

-

0

Pinned by corona

Conductance Elsc(rode

-

Positively Grout-d8d Rotor at w rpm NONCONDUCTINGPARTICLES PASSING THROUGHCARPCOSEPARATOR

A

B

C

D

E

F

HYPOTHETICAL NONCONDUCTINGPARTICLE

Figure 14: Non-Conducting particle passing through drum separator A non-conducting particle has a high electron affinity and becomes negatively charged (point A) by taking electrons from the metal feed hopper. At point B, the particle is charged more negatively by the negative corona field. At point C, the particle begins to enter the conductance field, so that at D and E the force due to electric charge is keeping the particle pinned to the rotor. By the time the particle reaches point F, it is out of the field and is being held by the image force.

1064

Wabash Mines High Tension Plant Drum Configuration. Figure 15 shows typical field electrode settings for electrostatic separators at Wabush Mines (hematite from quartz). Note the use of dual ion beam fields at the expense of conductance fields. The plant at Wabush uses Humphrey spirals to produce concentrates that are dried and then fed to the electrostatic separation plant (called a high tension plant). Final concentrates can be produced, which contain less than 1.7% SOz. Typical mass balances around the Wabush plant, during an investigation of electrode positions, are shown in Table 2. DC generators produce 28 kv across the electrodes of all units. Some general comments on the electrostatic separator shown in Figure 12 are: 1) Conductors: leak charge rapidly to the rotor; lifted by the field; image force not a factor. 2) Non-Conductors: leak charge slowly to rotor; usually pinned by field and by image force. 3) Conductors are easy to separate from non-conductors if they are the same size. 4) Particle size and shape: determines the amount of charge on a particle and magnitude of centrifugal force. 5 ) Specific gravity differences: affect magnitude of mechanical forces that act. 6) An increase in rotor speed at constant voltage: faster rotor cuts charging time, increases the centrifugal force and decreases time spent in contact with rotor. Improves separation up to a point. 7) Increase corona voltage at constant drum speed: improves separation to point. At a high enough charge on conductors, they will not lift. 8) Increase rotor diameter: increases retention time in fields. Useful for separating non-conductors where leakage rate constants are important. 9) Relative humidity affects resistance of non-conductors and causes them to conduct. Can improve separation by heating the feed to drive off moisture (T OF = 1.84 (% R.H.). 10) Use roughers, scavengers, cleaners, etc. in the usual manner. 11) Typical feed rates: 90 to 120 Ibhnch of rotor lengthhour. Equivalent rotor length = 3". 12) Typical rotor speeds: rougher step 175 to 200 rpm for 14" rotor; cleaner for conductors about 100 rpm or less; cleaner for non-conductors about 200 to 500 rpm. 13) The standard DC voltage should be 28 kv. Stay in the range of 25 to 30 kv. 14) First, find settings that are best for making grade; then find settings for making recovery. Increase voltage, decrease rotor speed, decrease feed rate to cut nonconductors in the concentrate; 15) The feed hopper should be at about 12 o'clock for the 14" rotor. Free Fall Separators and Other Designs The book by Ralston, 0. C., entitled Electrostatic Separation of Mixed Granular Solids, Elsevier, shows a variety of electrostatic separators, including free fall designs. The latter are used in the Potash industry. Particles fall between field electrodes by gravity (separation of sylvite from halite). Ralston (1961) shows a large number of progressive designs---even moving belts are employed as field electrodes. Few of these have caught on in mineral processing, the mainstays being the drum and free fall designs. Figure 16 shows an interesting design wherein two parallel electrodes are made of tubes placed side by side. In this way, the field is homogeneous but not uniform. Surfaces of electrodes can be rotated and cleaned continuously to avoid particle build up. These have been used in raw potash processing. Triboelectric charging (charging by friction) is employed at temperatures of 70 to 160°C (Knoll & Taylor, 1984).

1065

SPLITTER SEMI-CONDUCTORS (Middling

Figure 15: Wabush Mines high tension separator configuration Table 2: Wabash Mines high tension plant mass balance examples

1066

CR(wHLD POTASH ORE -1.5 mm

m?oR M

TRIBOELECTRK CHAR@#@

I E

YUTt

CO”TRATC

woll

C W L

C-TWWO

cowoITIwlN0

t l u c Y U t ACID

TRleWRQlQtlC

WUTL

wm

ntn

rlUL CONCENTRATE

UIO

Figure 16: Parallel “tubes” free fall separator ;charging by triboelectrification Another interesting approach (Jordan & Weaver, 1977) is s h o h in Figure 17. A liquid of intermediate dielectric constant (kerosene or a mixture of alcohol and carbon tetrachloride for example) is used as a suspending fluid. Particles of dielectric constant lower than the medium and higher than the medium can be separated fiom each other due to differences in dielectric forces. Particles of dielectric constant less than the medium are attracted to regions where the field gradient is weaker (to the screen electrode of the diagram), while particles of dielectric constant greater than the medium are attracted to regions of strong field gradients (to the surface of the drum electrode of the diagram). A splitter serves to divide the particles at the bottom. Either DC SAC fields can be employed at low voltages (1000 V).

Figure 17: Rotating drum dielectrophoreticseparator

1067

SUMMARY After a brief exposure to relevant electrical properties of solids, the characteristicsof electric fields have been reviewed. Methodology involved in charging conductors and non-conductors were discussed and the behavior of charged particles in the presence of fields generated by field electrodes has been summarized. The effect of centrifugal and gravitational forces in drum-type separators was considered. Electrostatic separators commonly encountered in mineral processing are the drum-type and free fall units. Several interesting designs were briefly described. In general, electrical methods of separation will continue to be important to the industry. REFERENCES Ralston, 0. C., (1961), Electrostatic Separation of Mixed Granular Solids, Elsevier Publishing Company, New York, NY Dyrenforth, William P., (1978), Electrostatic Separation, Ch. 23 in Mineral Processing Plant Design, Eds. A. Muiar and R. Bhappu, SME, Littleton, CO Knoll, F. S., J. B. Taylor, (1986), Selection and Sizing of Electrostatic Concentrating Equipment, Ch. 14 in Design and Installation of Concentration and Dewatering Circuits, Eds. A. Mular and M. Anderson, SME, Littleton, CO Knoll, F. S., J. B. Taylor, (1984), Advances in Electrostatic Separation, SME Preprint 84-71, SME-AIME Annual Meeting, Los Angeles, CAYFeb. 26-Mar. 1, 1984 Romero, Diego L. and Joy P. Romero, (1989), Electrostatic Separation Circuits, in Operation and Maintenance in Mineral Processing Plants, Ed. P. G. Claridge, CIM Special Volume 40, Canadian Institute of Mining and Metaliurgy, Montreal, Quebec, Canada Ghosh, A. K., and V. F. Daughney, (1988), High Tension Separation At Wabush Mines, Canadian Mineral Processors of CIM, Ottawa, ON Carpenter, J. Hall, (1970), Electrical Methods For the Separation of Minerals, Minerals Science and Engineering, January, 1970 Jordan, C. E. and C. P. Weaver, (1977), Development of a Continuous Dielectrophoretic Mineral Separator, SME Preprint 77-H-323, SME Fall Meeting and Exhibit, St. Louis, MO, October 19-21, 1977 Lawver, J. E., (1985), General Principles and Types of Electrostatic Separators, Section 6, Ch. 3 of SME Mineral Processing Handbook, SME, Littleton, CO Lawver, J. E., (1973), Electrostatic Separation, Ch. 10 in Electrostatics and Its Applications, Wiley, New York, NY

1068

Selection and Sizing of Magnetic Concentrating Equipment; Plant DesignLayout Daniel A Norrgran' and Michael J. Mankosa'

ABSTRACT Over the last decade, magnetic separation has become prevalent in the minerals industry due to advancements in the design and application of magnetic separators. Recent technological advancements in the field of magnetic separation were developed through two different avenues of approach. First, precise magnetic circuit modeling and optimization is now carried out on the computer using sophisticated multidimensional finite element analysis. This technique is applied to the design of both permanent and electromagnetic circuits. As a consequence, essentially any type of magnetic separator can be developed (or redesigned) with a very high level of confidence and predictability. Second, the recent development of rare earth permanent magnets has revolutionized the field of magnetic separation. The advent of rare earth permanent magnets in the 1980's provided a magnetic energy product an order of magnitude greater than that of conventional ferrite magnets. This has allowed the design of high-intensity magnetic circuits operating energy fie and surpassing the stragth and effectiveness of electromagnets. New applications and design concepts in magnetic separation have evolved. New types of magnetic separators have been developed making many older style separators obsolete. This paper is an attempt to provide an overview of the available technology for production plant design. Specific applications will be discussed as well as the design fundamentals for the basic types of magnetic separators. Guidelines are also provided for sizing and selection of magnetic separation equipment. INTRODUCTION The science of magnetic separation has experienced exceptional technological advancements in the last decade. As a consequence, new applications and design concepts in magnetic separation have evolved. This has resulted in a wide variety of magnetic separator designs, highly effective and highly efficient, ranging a wide spectrum of industrial applications. Magnetic separation may be used as an integral part of the primary process, or as a secondaryor scavenger operation to produce a mineral concentrate. Magnetic separation may also be used in applications involving the removal of tramp metal prior to crushing and grinding systems. The applications of magnetic separation are very diverse and employed in many different industries. It may be used to remove tramp metal such as shovel teeth or chain m a crusher line, or micron sized iron of abrasion &om a high purity industrial minerals process stream. It may also be used to selectively concentrate magnetic components such as magnetite or hematite minerals in an iron ore concentrator. It is also used in removing deleterious magnetic elements to purify non-magnetic elements such as in the manuhdme of ceramics or refi.actories. Also, recycling and secondary recovery applicationsare being developed with increasing interest using magnetic separation as an explicit method for recovering residual values fiom various process streams and hazardous constituents fiom waste streams. MAGNETIC SEPARATION SYSTEMS Magnetic Field Generation There are two distinct methods to generate a magnetic field The first method, which has a long history, is the use of an electromagnet. Electromagnets employed in magnetic separatotsare almost Eriez, Erie, Pennsylvania

1069

exclusively solenoid coils. The second method is the use of a permanent magnet. Both methods of magnetic field generation have application in the advanced magnetic separation techniques of today. Each type of magnet, with its associated circuit design, has many parameters of distinction. In most instances, a specific type of magnetic circuit is selected to accommodate the application.

Mode of Separation There are two basic modes of magnetic separation systems. The first mode is the use of magnetic separation for the removal of tramp iron fiom a process stream. The second mode is the use of magnetic separationas an integral unit operation in the recovery process. Tramp Metal Removal The removal of tramp metal &om a process stream is the most common application of magnetic separation. This application is straightfixward and well documented An overview of this methodology will be presented. There is a basic premise that tramp metal or ferrous contamination is present to some extent in all bulk materials. This contamination must be removed to 1) protect down stream equipment fiom potential damage,2) assure product quality, and 3) protect against fire or explosions in the process lie. Tramp metal may commonly originate in the f m oE 0

0

0

0

Metal parts fiom earthmoving equipment. Wire, cable, chain, reinforcing bar or support structures fiom mining or quanying operations. Nails, clips, strapping, or other hardware fiom shipping containers. Welding rod, nuts, bolts, and washers, or other hardware inadvertently introduced fim maintenance and repair procedures. Rust, chips, and fine iron of abrasion continually eroding &om pipe lines, chutes, bins, and process equipment.

When there is only an occasional piece of tramp metal or a very low level of ferrous contamination, a manual clean magnetic separator will suffice. These separators are stationary systems utilizing either permanent or electromagnets to generate the magnetic field The magnet effectively captures the tramp metal and holds it until it is manually removed. A relatively large amount of ferrous contamination present in the process stream will necessitate more fiequent removal of the collected ferrous tramp. At this point a selfcleaning separator such as a magnetic drum or a magnetic pulley is recommended.

Integral Separation Magnetic separation as an integral unit process is used in many industries. Most notably are minerals, abrasives, ceramics, and secondary recovery processing applications. In these types of applications, the magnetic h c t i o n and the non-magnetic hction may occur as a homogeneous blend The magnetic separator is therefore subjected to continuous operation performing a continuous separation. In one case the magnetic fiaction may represent the product and upgradiig may take the f m of the magnetic collection and recovery of the magnetic minerals. In another case, the non-magnetic Won may represent the product and upgradiig may take the f m of the magnetic removal of deleteriousiron beariig constituents fiom non-magnetic minerals. In some applications the magnetic h c t i o n represents the value. This is most prevalent in the and hematite concentration and recovery of iron bearing magnetic minerals such as magnetite @go4) (Fe&). Iron ore concentrators typically treat thousands of tons per hour of milled ore recovering the magnetic iron bearingminerals fiwn the non-magnetic host rock Magnetite and fenosilicon (FeSi) are commonly recovered fiom heavy media Operations using magnetic separators. Other common magnetic minerals of value that are treated on large economies of scale are ilmenite (FeTiQ) and chromite (FeCr204). In other applications the non-magnetic fiaction represents the value. Most notably is the magnetic cleaning of industrial minerals such as silica, quartzite, feldspar, nepheline syenite, and fluorspar used as glass batch feedstock material. In these types of applications the product has to be essentially fiee of iron bearing particles. Increasingly, there is also a demand for higher purity

1070

feedstock materials used in the manufacture of items such as specialty ceramic components, insulators, substrates, refractories, electrical components, specialty glasses, optical fibers, and grinding and polishing compounds. The integral unit process necessitates a self-cleaning separator. Magnetic drums, pulleys, and roll type separators are widely used as well as high-intensity and high-gmdient matrix type separators. For all practical purposes magnetic separator selection can be reduced to the decision tree illustrated in Figure 1. The first node indicates that the separation will be either tramp metal removal or an integral separation. The integral separation has subsequent nodes to indicate if the mineral will be treated as either a slurry or as dry and fiee-flowing and the magnetic field strength necessary for effective separation. Figure 1. Magnetic Separation Decision Tree I

2

Permanent Stationary

b

Tramp Collection

Suspended (Conveyor Belt)

L

-

I

Plate Grate Trap Permanent Electromagnetic

FF=F? Low-Intensity

Separation Application

RareEarth Drum

High-Intensity

1 I

Integral Separation

Rare Earth Roll

I Wet Drum

ri--l Rare Earth Wet

WHIMS HGMS

MAGNETIC SEPARATORS FOR TRAMP METAL COLLECTION Stationary Permanent Magnetic Separators for Tramp Metal Collection The most enduring design incorporates simplicity. This is certainly the case when addressing stationary permanent magnetic separators. These separators, specifically plates, grates, and traps, represent the first industrial application of magnetic separation and perhaps the most prevalent type of separator in use today. These separators perform very well in removing trap metal fiom a process prOteCting down stream equipment as well as a s d g product quality. These Stationary permanent magnetic separators are illustrated in Figures 2,3, and 4.

1071

Figure 2. Plate magnet consisting of a permanent magnet in a stainless steel fixture. The senamtor i s imtal1p.d in a rhiitp

Figure 3. Grate magnet consisting of permanent magnets in stainless steel tubes. Material flows around the tubs when placed in a dry process stream

Figure 4. Trap magnet consisting of a permanent magnet in stainless steel tubes. This separator is placed “in-line’’ with a sluny stream. The slurry flows around the

Figure 5. Suspended electromagnet. This separator has a deep magnetic field to effectively remove tramp ferrous iron.

tUbeS.

Stationarypermanent magnetic separators are configured with the common names of plate, grate, or trap. These are simply enclosed barium ferrite or rare earth permanent magnets. The process stream flows over, around,or through the permanent magnets and ferroustramp metal is collected and held. The prevalence of stationary permanent magnets can be attributed to several characteristics. They provide inexpensive insurance in the removal of tramp metal protecting downstream equipment such as mills, crushers, shredders, granulators, mixers, and fine screens fiom potential damage. They are relatively low cost compared to most other types of magnetic separators. Permanent magnets are used to generate the magnetic field requiring no utility for operation - there is no w i n g cost. Thae are no moving parts virtually eliminating maintenance costs. Suspended Magnetic Separators When addressing the magnetic collection of tramp metal, the Suspended Electromagnet (SE) is undoubtedly the industry workhorse. A photograph of a “standard“ manual cleaning SE magnetic is shown in Figure 5 . The SE magnet, providing tramp metal collection of conveyed materials, is a widely used magnetic separator. The electromagnet is mounted or suspended over a conveyor belt to remove relatively large pieces of tramp metal that represent a potential hazard to downstmim

1072

crushers, mills, and grinders. The SE consists of an aluminum conductor coil with a cylindrical steel core. The coil magnetically induces the steel core, which in turn configures the magnetic field for the collection of tramp metal fiom a moving conveyor belt. The coil is housed in a steel box which has the dual purpose of protecting the coil and completing the steel circuit. The coil is submerged in an oil bath within the housing to provide convection cooling. A rectifier is used to supply direct current to the magnet. Specificationsof selected suspended electromagnetsare shown in Table 1.

Belt Width (Meteraches) 0.9136

Weight (KilogramsPounds) 82311815

1.5160

373 118225

Power (Watts) 3800

I

9940

Applications of Suspended Magnetic Separators The SE magnet removes ferrous tramp metal f b m moving conveyors to protect downstream equipment such as crushers, mills, shredders, and presses. The largest market for SE magnets is in coal mining, hardrock mining, and aggregate products removing shovel teeth, cable, tools, and bolting prior to cxushing and grinding. The foremost factor in SE magnet selection is the burden depth of the material on the conveyor belt. (Notethat the burden depth accounts for the belt speed, belt width,capacity, and bulk density.) The burden depth determines the suspension height of the magnet and consequently the effective magnetic field strength at the belt surface. SE magnets are mounted in one oftwo positions over a umveyor belt as shown in Figure 6.

In pasition 1, the magnet is mounted over the conveyor belt prior to the head pulley. This position requires higher magndic field strengths to attract the ferrous component, shift the direction of momentum,and pull it through the bed of material. In position 2, the magnet is mounted just over the stream of material leaving the head pulley. This position utilizes the full potential of the magnet as it reacts with the material in suspended trajectory. Tramp metal is easily pulled through the suspended burden. Further,the flow of material is directed

1073

toward the magnet hce. Collection of the tramp metal fiom the material flow does not necessitate a change in direction. At conveyor belt speeds of less than 350 fpm, the suspended trajectory of the material is minimal and becomes near vertical. In this case the magnet must be shifted to a position approaching directly over the head pulley. SE magnets are available as a manual cleaning style or a self-cleaning style. Manual cleaning magnets are best suited where only small amounts or occasional pieces of tramp metal are encountered. The manual cleaning magnet must be periodically turned off in order to remove tramp iron accumulation fiom the magnet hce. Self-cleaning magnets employ a cross-belt running around the magnet h e to provide continuous removal of collected tramp iron. When tramp metal is attracted to the magnet hce, the cross-belt intercepts it and discharges it to the side away fiom the conveyor belt. The selfcleaning magnet is best suited where a high level of tramp iron or large pieces of tramp iron are anticipated. The cross-belt is continuously driven around the magnel on a system of four pulleys driven with a small gearmotor. In most applications the magnet is suspended 5 to 8 an (2 to 3 inches) over the top of the burden. Experience dictates that the operating magnetic field strength or the magnetic field strength at the belt surface should be 400 to 600 gauss for adequate fenous collection. As the magnet width increases to accommodate the belt width, the magnetic field strength increases at any given distance. This is simply the case that a wider magnet has a larger core and coil. There are exceptions to the sizing of the SE magnet. The response of the magnet is influenced by the following ictors: Belt Speed - As the belt speed increases it becomes increasingly difficult to attract and collect ferrous components. This is especially difficult when the magnet is situated in position 1 where the momentum of the ferrous component has to change direction. Burden Depth - As the burden depth increases, an increase in the magnetic field strength is required. A ferrous component situated on the surfhce of the belt, buried under a heavy burden, requires an increased magnetic attractive force for collection. Size of Ferrous Component - The suspension height of the magnet must be increased if relatively large ferrous components are anticipated. The suspension height of the magnet, with the fmous component attracted to the he,must clear the burden on the conveyor belt and not impede the flow of material. The higher suspension height requires a stronger magnet to maintain the magnetic field strength at the working distance. Shape of Ferrous Component - The shape of the ferrous component is also a hctor. Steel plate for example has a high wrhce area relative to its weight. This configuration reacts with a high magnetic attractive force when exposed to a magnetic field In mtmt, a sphere has the lowest surfacearea relative to its weight. This mliguration reacts with a minimal magnetic attractive force. Permanent suspended magnets are also available where operational conditions suffice. The advantage of a permkent suspended magnet is that it operates enerb fiee without peripheral equipment. The drawback of a permanent magnet is that it cannot genmte the depth of magnetic field required for treating high burden depths on the conveyor belt. The range of the burden depth is fiom 13 to 23 an (5 to 9 inches) corresponding to a maximum suspension height of 20 to 30 an (8 to 12 inches). At these operating distances the permanent magne4 rivals the electromagnet. INTEGRAL SEPARATION - MAGNETIC SEPARATORS FOR THE CONCENTRATION AND RECOVERY OF MINERALS

Magnetic Drum Separators The magnetic drum separator consists of a stationary, shaft-mounted magndic circuit completely enclosed by a rotating drum. The magnetic circuit is typically comprised of several magnetic poles that span an arc of 120 degrees or more. When material is introduced to the revolving drum shell (concurrent at h e 12 o'clock position), the non-magnetic material discharges in a natural trajectory. The magnetic material is attracted to the drum shell by the magnetic circuit and is rotated out of the non-magnetic particle stream. The magnetic material discharges fiom the drum shell when it is rotated out of the magnetic field A magnetic drum separator is shown in Figure 7.

1074

Figure 7. Drum type magnetic separator treating garnet concentrate.

Separation Variables The magnetic attractive force generated by a drum type separator is opposed by centriiigal force. The primary variables affecting separation efficiency are the magnetic field strength, feed rate, linear speed of the separator surflice, and particle size. An effective separation requires an equilibrium among the variables described as follows: A balance must be struck between an economic feedrate, product specifications, and recovery. As the feedrate increases, the layered particle bed on the separator surface increases in height and the collection of magnetics decreases. The linear speed of the drum is also a primary variable related to the feedrate. As the linear speed is increased, the layered particle bed decreases in height responding with an improved collection of the magnetic particles. The centriiigal force exerted by the drum or roll d c e is the critical bctor in providing sepration. Beyond the critical speed, the centrifugal force overcomes the magnetic attmtive force and the Separationefficiency deteriorates. Particle size will also effect separation efficiency independent of all other variables. Coarse particles provide a relatively high burden depth on the separator surface and respond with a relatively high magnetic attractive force. Coarse particles typically provide high unit capacities with high separation efficiencies. Fine particles with a relatively low mass respond detrimentally to electroStatcforces. As a consequence, precise magnetic separations balancing magnetic forces againstcentrifugal forces deteriorates. Dry - Low-Intensity Drum Type Magnetic Separator Low-intensity drum magnetic separators are effective in collecting ferromagnetic materials and are used in various applications in the minerals processing industries. These separators combine the attributes of the permanent magnetic field with the self-cleaning feature. Magnetic drum type separatorsare a staple in many mineral processing industries and are commonly used in two different modes of operation The separator is equally effective in a treating a process stream for the purpose of 1) Concentrating and recovering a magnetic mineral. 2) Rejecting magnetic minerals to produce a "clean" non-magnetic mineral product. An array of magnetic elements are available. Agitating magnetic elements are used in mast integral applications where a "clean" magnetic product is desirable. The magnetic poles have alternating polarity to provide agitation The agitating type element minimizes the loss of non-magnetics to the magnetic product (termed misplacement) due to physical entrapment. There are two basic types of magnetic circuits used in drum type magnetic separators. Presented in Figure 8 is a schematicof a High Gradient magnetic element.

1075

Figure 8. High Gradient magnetic element consisting of alternating polarity magnetic poles. This element produces a high magnetic gradient near the surface of the drum.

Figure 9. Interpole magnetic element consisting of main poles with interpoles. The interpoles oppose the polarity of the main poles to produce a “deep” magnetic field.

The High Gradient element, as the name implies, is designed to produce very high magnetic field gradient and subsequently high attractive force. Several identical magnetic poles comprise the element. The poles are place together minimizing the air gap between them to produce the high &ce gradient. Due to the high gradient, the attractive force is strongest closer to the drum making it most effective when utilized with a relatively low material burden depth on the drum surface and lower unit capacity, The Interpole style element utilizes a true ceramic “bucking”magnetic pole or “interpole“b e e n each main pole. The magnetic field of the bucking elements are charged to oppose both of the adjacent main poles resulting in maeased magnetic field at depth. Therefore, the Interpole elements allow relatively high material burden depths on the drum surfkce and thus higher unit capacity or improved separation efficiency in difficult applications. Presented in Figure 9 is a schematic of an Interpole magnetic element. There are several variations of these two basic magnetic circuits dependent on the application. Presented in Table 2 are general specifications of magnetic drum separators. Magnetic drum separators incorporating either a radial or agitating magnetic element are commonly used to collect tramp metal fiom a mineral stream. The element typically contains large strontium-ferritepoles to provide depth of field as well as a high surface gradient. This is necessary to attract and hold ferromagnetic components when operating at a high burden depth. These separators can operate at high capacity when utilized for tramp metal such as nuts and bolts. A 0.6 meter (24 inch) diameter drum can operate at a unit capacity up to 40 TPWmeter of drum width. The unit capacity is reduced by 50 percent for the collection of iron of abrasion. Low-intensity drum separators for mineral separation applications combines a large diameter drum 0.9 to 1.2 meters (36 to 48 inches), with an agitating element and a high drum speed. This combination provides a high capacity separation with a high level of precision. The separator typically incorporates an axial agitating (alternating polarity High-Gradient style) magnetic element that spans 180 to 210 degrees. Strontium-ferritemagnets are used for the collection and recovery of ferromagnetic materials. The high drum speed in combination with many agitating magnetic poles provides a high level of agitation as the fmomagnetics are transferred along the drum surface. This results in a near complete rejection of non-magnetic material subsequently producing a very clean ferromagnetic concentrate. A schematic of a 0.9 meter (36 inch) diameter low-intensity drum magnetic separator is shown if Figure 10. The magnetic drum separators operate completely enclosed in housing as shown in Figure 10. The housing incorporates a feed bin with an adjustable gate to deliver a controlled feedrate across the width of the drum. The housing also has magnetic and non-magnetic product discharge chutes. The only utility necessary is the electrical input for the drive system. The parallel Shafl gearmotor that drives the drum is mounted on the housing. The drive system uses a V-belt drive which offers protection to both the motor and the drum shell m the event of a jam from oversize material. In many casesthe drive is controlled with a variable fiequency controller to allow variationsin the drum speed

1076

Table 2. General Specifications of Magnetic Drum Separators Drum Type Radial or agitating magnetic circuit for tramp metal collection

Attributes Radial or high gradient agitating magnetic element using strontium ferrite magnets. Magnetic element designed to produce a "deep" magnetic field. 1200 gauss magnetic field strength on surface. 0.8 m/s drum speed. 0.4 to 0.8 meters diameter.

Agitating magnetic circuit for coarse material up to 25 mm (linen) topsize

High-Gradient agitating magnetic element using strontium ferrite magnets. 1200 gauss magnetic field on drum surface. 10 to 15 magnetic poles with wide (8 to 10 cm wide) spacing. Magnetic element designed to produce a "deep" magnetic field. Up to 2.5 m/s drum speed. 0.9 meter diameter.

Agitating magnetic circuit for coarse material 6 mm (1/4 inch) topsize.

Applications Widely used drum throughout industries. Functional for tramp metal and iron of abrasion removal. Effective in removing large tramp metal such as nuts and bolts at high capacity. Effective in removing fine iron of abrasion at relatively low capacities. Provides preconcentration of coarse ores up to 25 mm (1 inch) top size. Applications include magnetite recovery from iron ore and metallics recovery from slag.

Unit Capacity Tramp metal applications can operate at high capacity. 0.4 meter diameter drum operates at 20 to 26 TPH/meter width. 0.6 meter diameter drum operates at 30 to 40 TPH/meter width. Reduce capacity by 50 percent for fine iron of abrasion. Coarse minerals 6 mm to 25 mm (1/4 inch to 1 inch) can generally be treated at high capacities. Unit capacities ranging up to 80 TPH/meter width.

High-Gradient agitating magnetic element using strontium ferrite magnets. 1100 gauss magnetic field on drum surface. 20 magnetic poles with wide (5 to 8 cm wide) spacing. Magnetic element designed to produce a "deep" magnetic field. Up to 5.0 m/s drum speed. 0.9 meter diameter

Provides cleaning or concentration of minerals in the 100 mesh to 6 mm top size. Applications include cleaning industrial minerals such as alumina and silicate minerals. Also magnetite recovery from iron ore and metallics recovery from slag.

Coarse minerals (6 mm can be treated at unit capacities up to 65 TPH/meter width. Minerals in the 100 mesh , range can be treated at unit capacities up to 33 TPH/meter width

Agitating magnetic circuit for fine material (-100 mesh)

High-Gradient agitating magnetic element using strontium ferrite magnets. 1000 gauss magnetic field on drum surface. 30 to 40 small magnetic poles with narrow (2 to 4 cm wide) spacing. Up to 7.6 m/s drum speed. 0.9 meter diameter

Provides cleaning or concentration of minerals in the -100 mesh size range. Applications include upgrading magnetite from milled preconcentrates, magnetite recovery from flyash, and upgrading iron carbine.

Salient Pole Rare Earth

Salient pole agitating magnetic element using rare earth magnets. 7000 gauss magnetic field on drum surface. 0.4,0.6,0.9, and 1.2 meter diameter. Magnetic element employs 5 to 12 magnetic poles spanning a 110 degree arc.

Effective in magnetically collecting an entire range of paramagnetic minerals such as hematite, specularite, and ilmenite. Effective in cleaning industrial minerals such as silica, quartzite, feldspar, nepheline syenite, alumina and zircon.

1077

High drum speed maintains a low burden depth allowing unit capacities up to 33 TPH/meter width.

Minerals in the 100 mesh range can be treated at unit capacities of 10 to 16 TPH/meter width on a 0.4 m diameter drum, 16 to 23 TPH/meter width on a 0.6 m diameter drum, and 26 to 33 TPH/meter width on a 0.9 m diameter drum

FOR D W C WAD ADD 1% TO STATIC IDAD

The 0.9 and 1.2 meter dim- drums typically have drum widths up to 3.0 meters (120 inches). The drive motor ranges fim 3.7 to 18.6 kilowatts (5 to 25 horsepower) dependent on the magnetic circuit mfigumticn drum speed, and width of drum. Came particles in the -25 mm to +6 mm (-1 inch + 1/4 inch) size range responds well to a magnetic circuit employing large wide magnetic poles. This magnetic circuit generates a “deep” magnetic field effectively capturing coarse fenmagnetic or composite minerals. Coarse iron ore (magnetite) can be treated at a relatively high unit capacity. As an initial magnetic Separation, the separator rejects a non-magnetic product relatively &ee of iron while recovefing the magnetite as a “rougher“ concentrate- This concentrate representsan intmediate grade Containing locked particles and requires milling to fivtherliberate the iron minerals.

1078

Crushed copper slag also responds well to this separator. The major component in this slag is iron bearing faylite which constituents approximately 90 percent of the feed. The faylite is recovered as the magnetic hction while the metallic copper k rejected as a non-magnetic concentrate. Coarse slags and iron ores are treated at a unit capacity of 80 to 115 TPWmeter of drum width on a 0.9 metex diameter drum. Fine particles in the - 100 mesh size range responds well to a magnetic circuit employing small narrow magnetic poles. The magnetic circuit may consist of up to 40 separate magnetic poles providing a very high degree of agitation as the material is transferred through the magnetic field Applications in these size ranges include iron ores as well as intermediate grade concentrates, slags, flyash, ferrosilicq and iron carbide. Iron ores and concentrates are treated at a unit capacity of up to 33 TPWmeter of drum width on a 0.9 meter diameter drum. Flyash which is very fine and has a low bulk density is treated at a unit capacity of 16 to 25 TPWmeter of drum width on a 0.9 meter dmeter drum.

-

-

Dry High-Intensity Drum Type Magnetic Separator Rare Earth Drum The Rare Earth Drum magnetic circuit is comprised of segments of alternating rare earth magnets and steel pole pieces. The steel poles are induced and project a high-intensity, high-gradient magnetic field. The peak magnetic field intensity on the drum is approximately 7000 gauss and is effective in removing many paramagnetic constituents. The Rare Earth Drum was developed to meet several operational objectives. Initially, the project focused on the dry treatment of ilmenite and hematite. It was anticipated that the separator would provide high capacity with relatively low capital and operating cost. Test work carried out on the prototype separator demonstrated that this design will effectively recover ilmenite fiom a heavy mineral concentrate and hematite fiom siliceous gangue. The unit capacities fhr exceeded the typical capacities of cross-belts and induced magnetic roll separators. The applications utilizing the Rare Earth Drum magnetic separator to date have been very diverse. While many of the units are used primarily for the recovery and concentration of paramagnetic minerals, most are used for removing deleterious iron bearing minerals fiom process streams such as alumina, silica, and feldspar. These various applications have confirmed that the Rare Earth Drum is effective in treating a wide range of particle sizes and operates at relatively high capacities. Presented in Table 3. are the unit capacities of specific applications using the Rare Earth Drum magnetic separator. Table 3. Applications and Unit Capacities of the Rare Earth Drum

Application

Preconcentrating Hematite Recovering Specularite fi-om a Gravity Concentrate Recovering Final Ilmenite Product - 35 Mesh fiom Heavy Mineral Concentrate Recovering “High-Iron” Ilmenite fi-om - 35 Mesh ‘cL~~-Irm’7 Ilmenite Recovering ILmenite 6-om Roasted -4 Mesh Concentrate Tabular Alumina -6 Mesh +48 Mesh Cleaning Tabular Alumina -48 Mesh +325 Mesh Cleaning Tabular Alumina -48 Mesh Cleaning Silica and Feldspar -28 Mesh Cleaning Garnet Collecting Nickel Bearing Pyrrhotite fi-om Crushed Nickel Ore Recovering Metallic Nickel €rom

0.6 0.6

10-13 13-16 20

0.9

33

0.4 0.4 0.4 0.4

0.4

-48 Mesh -100 mm (- 4 Inch)

0.6 0.4 0.9

23 16 10 10-13 13-16 13-16 66

- 10 mm (-318 Inch)

0.9

40

1079

-

Dry - High-Intensity Roll Type Magnetic Separator Rare Earth Roll The Rare Earth Roll, generating peak magnetic field strengths in excess of 2.1 Tesla (21,000) gauss is very effective for concentrating or removing weakly magnetic minerals fiom a dry process stream. The Rare Earth Roll magnetic separator is designed to provide peak separation efficiency and is typically used when a high-purity product is required. The roll is constructed of discs of neodymium-boron-iron permanent magnets sandwiched with steel pole pieces. The steel poles are magnetically induced to the saturation point of approximately2.1 Tesla (21,000 gauss). Magnetic roll diameters are typically 75, 100, and 150 mm (3,4, and 6 inches), although separators as large as 300 mm (12 inch) diameter are available. The separator is configured as a head pulley in the separator. A thin belt, usually fiom 0.125 to 0.50 mm (5 to 20 mils) thick is used to convey the feed material through the magnetic field. When feed material enters the magnetic field, the non-magnetic particles are discharged fiom the roll in their natural trajectory. The paramagnetic, or weakly magnetic, particles are attracted to the roll and are deflected out of the non-magnetic particle stream. A splitter arrangement is used to segregate the two particle streams. The magnetic rolls are constructed up to a width of 1.5 meters (60 inches). Typical feed rates to the separator range fiom 2 to 8 TPWmeter (100 to 400 pound/hour/iich)of roll width. Coarse 6 mm (1/4 inch) material, such as chromite or hematite, can be treated at a very high rate, while fine material, such as silica, feldspar, nepheline syenite, or alumina, must be treated at lower rates. Sized silica sand can typically be treated at a rate of 4 to 6 TPWmeter (200 to 300 pounds/hour/iich) of roll width. The separator can be configured with s e v d magnetic rolls in series to provide a multiple stage separation. Two to three magnetic rolls are typically placed in Series (modular design) to provide optimum separation efficiency. When cleaning a non-magnetic material, the rolls are usually arranged in a non-magnetic repass configuration. The non-magnetic material fiom the first stage of separation is delivered to the second roll to fiuther remove residual paramagnetics. The rolls can also be arranged in a magnetic repass configuration to provide a cleaning stage of the magnetic hction. A schematic of the Rare Earth Roll magnetic separator is illustrated in Figure 11. This separator is configured with a two stage separation. The Rare Earth Roll is enclosed in a housinglhework. All product contact areas are fabricated fiom 304 stainless steel. The h e w o r k incorporates an electromagnetic feeder to provide a umsistent feed rate aaoss the entire width of the roll. A hopper is mounted over the feeder to umtain and deliver incoming feed. There are also magnetic and non-magnetic product discharge chutes. A wide range splitter is used to segregate the magnetic h d i o n from the non-magnetic produd A dust hood is mounted over each roll. The only utility necessary is the eledrid input for the electromagnetic vibratory feeder and the drive system. The vibratory feeder extends the entire width of the magnetic roll to provide a u n i f m distribution of feed. The vibratory feeder on a separator utilizing a 1 meter wide magnetic roll will draw approximately 0.8 kw. Each magnetic roll is driven independently. A 1 meter wide magnetic roll will utilize a 0.75 kw (1 HP) gearmotor. The roll speed is variable with a VFC control. Belt speeds in a production mode typically range fiom 0.6 to 0.9 d s . Treating silica sand on a 1 meter wide roll at a feed rate of 5 TPH and a belt speed of 0.8 d s results in a burden depth of approximately 1 mm (0.05 inches). Wet - Low-Intensity Drum Type Magnetic Separator The wet drum magnetic separator is used extensively in iron ore and heavy media applications. It is also used for collecting magnetite or “black sands” fiom gold beariig ores and tailings as well as pynrhotite in sulfide mineral applications. The wet drum magnetic separator is well established with literally thousands of separatorsin current operation. The wet drum magnetic separator consists of a rotating magnetic drum situated in a tank that receives the feed slurry. The magnetic chum consists of a stationary, shaft-mounted pennanent magnetic circuit completely enclosed by a rotating drum. The magnetic circuit is unnprised of segments of alternating permanent magnets that spans an arc of 120 degrees. The peak magnetic field intensity on the drum is approximately 2000 gauss and is effixtive in removing fmomagneticminerals.

1080

Next Page

The magnetic drum is mounted in a tank. Slurry is fed to the tank and subsequently flows through the magnetic field generated by the drum. The magnetic minerals are attracted to the drum shell by the magnetic circuit and are rotated out of the slurry stream. The magnetic mineralsdischarge &om the drum shell when rotated out of the magnetic field. There are thee basic styles of wet drum tanks. Tank selection is based on the specifics of the application. In a wet drum separator, the magnetic force acting on a ferromagnetic particle is predominately opposed by hydrodynamic drag f m . This fame, when properly applied, provides the vehicle of separation washing away the nonmagnetic particles while the faromagnetic particles are collected in the magnetic field The hydrodynamic drag force is also responsible for any losses of fmmagnetics. The variables affecting the collection of ferromagnetics in a wet drum magnetic separator are: Magnetic field strength. The magnetic field strength must be sufficient to effectively collect ferromagneticminerals.

1081

Overview of Flotation Technology and Plant Practice for Complex Sulphide Ores N. W. Johnson’ and P.D. Munro’

ABSTRACT Complex ores are differentiated from polymetallic ores by the fine grain size of the minerals and the intimacy of their textural association. For example zinc-lead-silver ores from Broken Hill in New South Wales, Australia and Mount Isa, Queensland, Australia are both polymetallic but the Mount Isa material is classified as complex because the constituent minerals’ small grain size requires fine grinding for adequate liberation. Recent developments in regrinding technology mean that it now is feasible to economically regrind to 7 to 10 microns (80% passing size) if required. Flotation separation of the fine particles necessary for mineral liberation may result in significant inefficiencies compared with those in “normal” plant practice. The large surface area per unit weight of fine particles combined with prolonged exposure to grinding processes, long flotation residence times and the use of recirculated water means that the chemical environment can also be important. This paper extensively draws on the experiences of treating the complex zinc-lead-silver ores of the Mount Isa Inlier in Australia, which have arguably received the most intensive research and implementation effort since 1980.

INTRODUCTION The authors define complex ores as those characterised by the fine grain size of the minerals and by an intimate textural association. In this context complex ores are differentiated from those which are polymetallic but still achieve a good metallurgical performance. For example zinc-leadsilver ores from Broken Hill in New South Wales, Australia and Mount Isa, Queensland Australia are both polymetallic, geological opinion assigns them a similar genesis (Guilbert and Park 1986) but they have substantially different flotation performances. Processing of the Broken Hill sulphide ore with a flotation feed sizing of 55% -74 microns and virtually no regrinding typically produces a lead concentrate over 70% Pb grade at +90% recovery and a zinc concentrate grade of 52% Zn at +90% recovery. The zinc sulphide mineral at Broken Hill is marmatite containing 10-11% Fe (Lawrence 1968) putting a theoretical upper limit of around 54% Zn for the concentrate grade.

1. Professor of Minerals Engineering Department of Mining, Minerals and Materials Engineering The University of Queensland St Lucia, Queensland 4072, Australia

2. Principal Consulting Engineer Mineralurgy Pty. Ltd. PO Box 818 Toowong, Queensland 4066, Australia

1097

By contrast the Leadzinc Concentrator of Mount Isa Mines Limited makes a lead concentrate approaching 55% Pb grade at +80% recovery and a zinc concentrate at 52% Zn at 7 5 4 0 % recovery (Young and Gao 2000) from a flotation feed of 80% -37 microns and after grinding to 80% - 12 microns for zinc regrinding and to 80% -7.5 microns for zinc retreatment regrinding. The Mount Isa zinc-lead-silver ore does have a larger amount of pyrite present that is not a factor at Broken Hill but the absence of iron sulphide minerals such as pyrite, pymhotite and marcasite does not necessarily make an ore "free milling" or "simple". Iron sulphide minerals are not significant in the zinc-lead-silver ores for the No. 2 Orebody at McArthur River, Northem Territory, Australia and at Century, Queensland, Australia but the concentrators treating these ores have to regrind rougher concentrates to -6 to 8 microns to make saleable products; a zinc-leadsilver bulk concentrate at McArthur River and a zinc concentrate at Century. Different forms of iron sulphide minerals such as pyrite have different implications for processing. For example coarse grained euhedral pyrite is relatively easy to depress in flotation whereas much more severe conditions are required to deal with the fine "framboidal" pyrite (-8 microns) found in the zinclead-silver ores at Mount Isa, Hilton, George Fisher and other deposits. This pyrite has a graphitelike carbon content rendering it naturally floating and it forms "atoll" structures with galena (Grondijs and Schouten 1937; Croxford, Draper and Hamaway 1961). Croxford and Jephcott (1972) stressed that when comparing Broken Hill, Mount Isa and McArthur River in the sedimentary context they become much easier to explain. In the case of McArthur River, a very difficult metallurgical problem exists because negligible metamorphism has occurred. Consequently, the original fine-grained nature of the McArthur sulphides has been preserved. For example 80% of the galena in No. 2 Orebody exists as -3 microns grains and allowing only 25% recovery into a 50-55% Pb grade concentrate. The lack of metamorphism has preserved illite and bituminous material and the "smearing" of these components during the very fine grinding needed for mineral liberation probably adversely affects flotation. Similarly the different degrees of metamorphism explains the differences in galena beneficiation performance between Mount Isa and Broken Hill and even within the Mount Isa orebodies themselves such as the Racecourse versus Black Star orebodies (Davey and Slaughter 1970). There are quantitative and semi-quantitative mineralogical methods available to determine the degree of complexity of an ore as well as usefully comparing potential metallurgical problems of a new deposit against comparable existing operating mines. Jackson, Gottlieb and Sutherland (1988) used QEM*SEM, which is based on an automated scanning electron microscope, to compare grain sizes for a range of Australian zinc-lead ores by the characteristic of Phase Specific Surface Area. Carson (1995) has developed a method based on optical microscopy that is used by Noranda to assess the "in situ" metallurgical complexity of a deposit. The "middlings rating" from this method is semi-quantitative and ranges from 1 (coarse) to 6 (very fine) the grain sizes of specific mineral pair intergrowths that can form middling particles during grinding of the ore. Examples of some well-known complex ores are listed below. 0

0

Stratiform sediment hosted zinc-lead-silver ores such as Sullivan in British Columbia, Canada, Red Dog in Alaska, U.S.A. and those of the Carpentaria or Mount Isa Inlier in Northern Australia. These include Mount Isa, Hilton, George Fisher, Lady Loretta and Century in Queensland and McArthur River in the Northern Territory. Pyritic copper-lead-zinc-silver deposits such as Brunswick Mining and Smelting, Heath Steele and Caribou in New Brunswick, Canada. Geological opinion tends to assign these deposits as to having been formed by similar processes to the stratifom sediment hosted deposits mentioned above (Guilbert and Park 1986).

1098

0

0

Iberian Pyrite Belt deposits such as Aznalcollar-Los Frailes in Spain and Aljustrel in Portugal. Volcangenic massive sulphide deposits such as Hellyer in Tasmania, Australia and Woodlawn in New South Wales, Australia. Epigenetic replacement deposits such as Elura, New South Wales, Australia.

All of the above require very fine grinding and in several cases metallurgical results are only modest without ultra fine regrinding below 10 microns. Interestingly zinc is the dominant commodity produced from the treatment of complex ores. Omitted from the above list are the "Kuroko-type" volcanogenic massive sulphide deposits of Japan, as these do not require ultra fine regrinding for acceptable metallurgical results. Excluded from consideration are ores where the difficulties in flotation are a result of chemical reactions in the orebody rather than intricate mineral textures. That is, where geological processes such as weathering and penetration by water, have formed complex oxides, carbonates, sulphates etc. Tsumeb in Namibia has such an orebody (Boyce, Venter and Adam 1970) where beneficiation by flotation was made difficult by varying mineralogy. Chemical issues from soluble species in the ore, such as the activation of sphalerite by copper ions, can be a concern when treating the supergene portions of copper-zinc deposits such as Thalanga in Queensland, Australia (Wong, Breen, Doherty and Phelan 1991; Nice and Brown 1995) but are not discussed further. LIBERATION The essential requirement of liberation of the valuable mineral(s) in a complex sulphide ore, as for all types of ore, was always well known. However, in the experience of the authors, traditionally less attention was paid to quantification of this essential step than for the separation step. Reasons for such diminished attention included the cost of obtaining the necessary liberation data, uncertainties in the meaning of the liberation data, the necessity to interface with related disciplines and the need for attention to detail in many steps in sequence to provide reliable samples for liberation measurement. It must be noted that the means of obtaining the liberation data existed via the traditional method of point counting of mounted particles in a polished briquette by an operator using an optical microscope (Amstutz 1961). For lower grade samples such as the valuable mineral in the tailings, the point counting method was supplemented by area counting. In the experience of the authors, it was relatively common for metallurgists to infer the extent of liberation from the performance in the separation stage. Such an approach became increasingly risky for regrinding stages with very fine target sizings eg 80% passing sizes of 5 to 15 microns. Obtaining a sound separation performance, and ultimately the best possible separation performance, often required many tests covering several weeks at the laboratory scale. Sometimes, even if there was a large increase in liberation from regrinding, inadequate persistence in the testwork gave a poor separation after regrinding as the best result achieved. The inference from the resulting poor separation could be that there was no benefit from regrinding. Obtaining liberation data for the feed and product of regrinding, in such circumstances, will show an increase in liberation and provide conviction that the optimisation phase for the separation process must continue until the benefits of the improved liberation are observed in the separation results. For complex sulphide ores, greater attention to quantification of the liberation step (in initial grinding and regrinding) occurred in the last 20 years of the twentieth century. This happened because metallurgists realised that technical understanding of the liberation issue needed improvement for the complex sulphide ores that were difficult to liberate, and the change was assisted by the emergence of automated and semi-automated instruments to acquire the liberation data. These instruments were largely based on the electron microscope although, in some cases, optical image analysers played a role. Examples of instruments using electron microscopes were

1099

the QEM*SEM developed by the CSIRO in Australia (now known as LEO QEMSCAN), the MPSEM-IPS image analysis system at CANMET in Canada, and more recently the Mineral Liberation Analyzer (MLA) from the Julius Kruttschnitt Mineral Research Centre (JKMRC). The IBAS (Kontron) system also mentioned in the literature, uses an electron beam based image analysis system, which allows the discrimination of minerals based on their gray level signal, by direct chemical analysis (EDX analysis) or a combination of both methods. These instruments did not decrease the cost of obtaining liberation data for a given size fraction. However, they did increase the rate of production of liberation data and, in principle, decreased the subjectiveness of the liberation data in comparison with a human operator. Such instruments could produce misleading data from briquettes with inadequate preparation quality (such as incomplete polishing) or by unexpected properties of the mounted particles. However, in general, the ability to obtain liberation data improved. However, it must be stressed that the major change was an increased willingness for metallurgists to use the liberation data. Liberation data were used to determine the increase in liberation across each grinding and regrinding step. For flotation circuits containing regrinding, the acquisition of liberation data on all exiting streams (concentrate(s) and tailing(s)) provided independent information on the liberation achieved as a result of all size reduction within and preceding the separation circuit. The liberation values for minerals in the stream representing the summation of the exiting streams (this stream was sometimes known as the “recalculated feed”) can be compared with the liberation values for the actual plant flotation circuit feed to obtain this information (Johnson 1987).

To obtain the liberation value achieved for a ground ore, liberation data for a series of size fractions must be obtained. The industry standard method for particle size analysis with collection of the individual size fractions in the sub-sieve size fraction is the Warman Cyclosizer, a water elutriator device which is now favoured over the Haultain infrasizer which used air (Anon 1981; Finch and Leroux 1982). For many complex sulphide ores, the liberation value for the valuable mineral has effectively ceased to increase by the Cyclosizer C5 fraction on examining the values from the coarse fractions to the Cyclosizer C5 fraction. This is fortunate because the Cyclosizer C5 fraction is usually the finest fraction for which liberation data can be acquired easily by optical microscopy or by instrumental methods. For such ores, the liberation values in the C5 fraction may be assumed to apply to the finer fractions also. For some particularly fine grained complex sulphide ores, such as the McArthur River and Century ores, the liberation values have not reached a plateau by the Cyclosizer C5 fraction and additional liberation data must be sought at least from the next fraction, nominally from five to ten microns, where ten microns is taken as the fine end of the Cyclosizer C5 fraction. Examination of this fraction poses greater difficulty for optical microscopy and for the instrumental methods in general. Physical barriers associated with the limitations of optical microscopy and the diameter of the region generating x-rays in instrumental methods prevent the examination of size fractions smaller than approximately 3 microns, with the limit depending on the minerals present and their distribution. In an initial grinding stage, and assuming the case where the liberation occurs in the initial grinding stage (and not in regrinding), the following ranges can be recognized. Liberation Value*(%) For Valuable Mineral (* uncorrected 2D value)

Expected or Theoretical Separation Possible

< 70 70 to 80 > 80

Poor separation results Sound separation results Good separation results

1100

Empirical evidence supporting the above relationship between the liberation values and subsequent metallurgical performance can be seen in papers such as Petruk and Schnarr (198 1) and Petruk et a1 (1991). If a low level of liberation is achieved in initial grinding (for example, less than 50%) and a major contribution to overall liberation occurs in regrinding, then as a guide the total liberation achieved in all size reduction stages needs to reach the values indicated. Achievement of a suitable liberation value provides a basis for optimisation of the relevant conditions for maximising the efficiency of separation. In addition to the head grade of the valuable mineral in an orebody, the amount of energy required for size reduction of a complex ore (a combination of the work index of the ore, the fineness of target product sizings and the tonnages of solid to be ground) largely determine if processing of the orebody is economic. For a complex sulphide ore, the use of liberation data is important in the design of circuits for new ores, for monitoring the on-going operation at a concentrator on, for example, a monthly basis from a liberation perspective, and for devising means of improving the performance of a concentrator through a development program. The monitoring of liberation levels in the actual and “recalculated” plant feed is used in such cases. It must be noted that this activity directed at liberation levels must be distinguished from the supplementation of recovery-size data with liberation data for more detailed analysis of separations that will be discussed in a later section. In such analysis, the distribution of liberation-based species between the products from a separation is examined in general terms.

CHEMISTRY OF PREPARATION AND SEPARATION In the last two decades, the role of pulp potential on the conditions during preparation ie grinding or regrinding and during the actual flotation separation has become much better understood and used to a greater extent by industrial practitioners, particularly in process development and in solving process problems. The pulp potential in grinding and regrinding can be altered by use of balls, which are more inert than mild steel, ie balls with elevated chromium levels, pebbles or ceramic materials. Mild steel balls in a conventional grinding or regrinding stage produce a reducing environment. By implication, oxidation of the sulphide minerals present will be lessened. In addition, the mild steel grinding balls, which provide ferrous irons have one clearly detrimental impact for many systems. At the alkaline pH of most grinding or regrinding systems, the iron species will be precipitated indiscriminatingly on the valuable and gangue minerals, because the solubility product for the iron hydroxides is exceeded. For many complex sulphide ores, this is a very detrimental property because the implication of very fine target sizings in grinding or regrinding is that, in conventional technology, there will be large quantities of adsorbed iron hydroxides due to the long grinding residence time required. To create hydrophobic surfaces on the valuable minerals, collector has to sufficiently displace these iron hydroxides on the valuable minerals. In contrast, steel grinding media with elevated chromium levels, pebbles or ceramic media would not provide reducing conditions resulting in much lower or no iron hydroxide precipitates.. Hence, the added collector has much less iron hydroxides to displace and it is much easier to establish hydrophobicity of the valuable minerals, possibly at lower collector additions. In the design of some of the newer types of regrinding mills, air cannot enter the grinding chamber, except being dissolved in the water in the feed. This feature can also impact on the conditions inside such mills. A common theme for all sulphide minerals is that their state of oxidation may change to some extent in the grinding step, dependent on whether oxidising or reducing conditions exist. This is a continuation of the oxidation of sulphide minerals (Abramov and Avdohin 1997), which commences with mining, and may predate even this step. Most separation stages utilise air as the flotation gas and oxidation of sulphide minerals can continue. The following picture of parallel reactions emerges (Lauder et a1 1994). In addition to the quantities of metal ions such as magnesium and calcium and anions dissolved in “new” water to

1101

the plant, hrther calcium and magnesium ions and other ions enter the aqueous phase in the sluny from slight dissolution of ore minerals (the solubility product of each mineral must be satisfied). The reagents can further increase the supply of some species eg calcium ions from lime additions. In addition, sulphide minerals are subjected to oxidation ie the introduction of a range of metal cations and sulphoxy anions. The use of sulphoxy reagents eg sulphur dioxide, sodium sulphite or metabisulphite will further increase the levels of sulphoxy species in the aqueous phase. Hence, the condition of mineral surfaces and the aqueous phase is not static throughout a circuit treating sulphide minerals. The pH in sulphide circuits is usually alkaline. The implication is that the high hydroxide ion concentration in conjunction with a relatively low concentration of metal ions will cause the solubility product for each metal hydroxide to be exceeded and metal hydroxides to be precipitated, except for magnesium hydroxide which requires a pH of approximately 10.5 and calcium hydroxide (>12). These high pH values are sought in many complex sulphide circuits. Much of the metal hydroxides exist as hydrophilic debris on the surface of particles. All minerals tend to acquire the hydrophilic deposits. Hence, it can be seen that metal ions from one mineral are likely to travel to the surface of other minerals. In some cases, this transfer causes unintended activation of minerals (Guy and Trahar 1985; Basilio et a1 1996). For example, in a galena flotation circuit, sphalerite may be inadvertently activated by copper ions from copper minerals or the recycled water, by lead ions from concurrent oxidation of the galena or silver ions from oxidation of silver minerals. Pyrrhotite can also be activated by such copper ions. It follows that analysis of the aqueous phase alone (after such precipitation occurs) will give a misleading picture as the majority of many species exist on the surfaces in some form. This can be addressed by analysis of surfaces by instrumental methods, provided that the surface is presented to the instrument retaining its original condition. In contrast, a method has been developed which uses classical chemical analysis methods to assess the concentration of metal ions in the aqueous phase, by removing a portion of the water from the slurry. A complexing agent for metal ions (EDTA) is then used to retum metal ions in surface deposits to the aqueous phase (Rumball and Richmond 1996). The increase in concentration of each metal ion is related to the quantity existing on the surfaces of minerals. Hence, general information is provided, as the method cannot indicate the minerals on which the deposits existed. Surface analysis of particular minerals by instrumental methods with a strong ability to focus on individual particles is required to provide such information. However, the EDTA based method requires simple equipment and does give the relative increase in the quantity of precipitated species from the start of the circuit to the exit points. The chemical conditions in the preparation step are known to have an effect on the separation, which can be achieved in the following flotation circuit (Trahar 1984). Further, in the flotation process that follows, for a given set of preparation conditions, the separation that is achieved is dependent on the pulp potential which is obtained. (Trahar 1984; Richardson and Walker 1985). Air is added usually as the flotation gas. It happens that the “air-set” steady-state pulp potential in a flotation bank is often suitable for the flotation separation to proceed. Sometimes, the “air-set’’ steady-state pulp potential will not be reached at the start of a rougher bank, if reducing conditions existed in the grinding stage and air is added in an unplanned way during transportation from the grinding mill to the rougher flotation bank. It is possible for the pulp potential to be less that the value needed for adequate collector adsorption, causing diminished flotation rates for the valuable mineral. For some separations, modification of the “air-set” pulp potential to another value may give a more efficient process. The pulp potential can be altered by the use of oxidising and reducing agents, by the use of nitrogen or nitrogedair mixtures to halt the increase in pulp potential after leaving a reducing environment in a grinding mill containing mild steel balls, and by direct

1102

connection to a power supply. The concurrent use of an oxidising or reducing agent and a suitable gas can lower the addition rate of the agent needed to achieve a target pulp potential. The value of the pulp potential in the flotation separation is important because the electrochemical mechanism is very well established as the principal means for uptake of collector on sulphide minerals and native elements. The implication of this mechanism is that the major uptake of collector on a particular mineral only occurs when the potential of the system reaches an appropriate value where the reaction involving the uptake of the collector can occur. Equally, if the potential continues to be raised, a potential will be reached where the collector will not be stable on the surface and another species will dominate on the surface. Many authors such as Woods (1984), Ralston (1991) and Witika and Dobias (1995) have reviewed the dependence of the flotation process on pulp potential. Examples of relationships between the recovery of various valuable sulphide minerals and pulp potential exist in the references. Monte et a1 (1997) recently briefly summarised the electrochemical mechanism: “It is well established that flotation with thiol collectors is achieved through a double mechanism of cathodic reduction of oxygen which consumes the electrons generated by anodic oxidation of collectors. The anodic oxidation of xanthates may involve different sub-processes such as metal xanthate formation at the surface, chemisorption of the xanthate ion at the sulphide surface, and catalytic oxidation of xanthate ions giving rise to the formation of dixanthogen, which adsorbs on the surface, rendering it hydrophobic. In this last case, the surface of the mineral provides the transfer of electrons from anodic sites to cathodic sites but does not participate in the reaction itself.” Water Issues The quality of water available at the site of a processing plant can vary greatly depending on the climate zone in which it is located and the origin and history of the water available in adjacent rivers or underground. It can range from very high quality, soft water derived from melting snow, to poor quality, hard water (high levels of dissolved calcium and magnesium) from underground aquifers, and to even lower quality sea water (very high levels of dissolved species including sodium chloride). It can be noted that some underground water is more saline than seawater as well as containing high levels of dissolved calcium and magnesium. On storage, the water will revert to the ambient temperature at the site, which can be less than 10°C or which could exceed 4OOC. In addition, there is usually a considerable difference in the ambient temperature between summer and winter at a given site. Occasionally, a different source of water is utilised in the wet and the dry seasons.

In semi-arid climates and many other sites, sustainability issues dictate that water in the plant products is retrieved and returned to the plant to minimise the requirement for “new” water from the local source. This recycling of water always provides the opportunity for chemical species derived from oxidation of the ore, from oxidation of grinding media or from the added reagents (and their degradation products) to increase in concentration in the water added to the plant. Such species can have deleterious effects on the performance of the flotation process (Rao and Finch 1989; Levay et a1 2001). It should be noted that recycling of water from the products could considerably increase the already deleterious quantities of species in the “new” water at the site. For example, the calcium ions in hard, “new” water are often augmented by calcium ions derived from lime additions to raise the pH of the pulp in the plant and possibly derived from the minerals in the ore itself. There can be four approaches to the accumulation of deleterious species in the recycled water. One is to accept the reduction in performance as the most cost effective approach. Adding the “new” water to more sensitive parts of the plant eg the cleaners can mitigate some of the effects.

1103

Another is to process a portion of the recycled water to remove such deleterious species. A fourth approach is to use zero or minimal recycling. In this case, some processing of the water for environmental reasons may be needed before it is discharged from the site, The temperature of the pulp in a plant is higher than the temperature of the water added to the plant. Grinding and regrinding steps in particular raise the temperature of the slurry. Hence, in some parts of a plant where the summer ambient temperature is 35 to 40°C,the temperature in parts of a grinding circuit may exceed 50°C. Such extremes in temperature may cause instability of reagents or have unexpected effects on the solubility of compounds. The phosphate family of compounds is important in manipulation of water quality in mineral processing because of the wide range in chemical behaviour of its members (Rashchi and Finch 2000). For plants using hard water (eg high calcium ion levels) and desiring to use sodium carbonate (soda ash) as a modifier, a concomitant problem has traditionally been the continual precipitation of calcium carbonate (calcite). This causes a continual reduction in the internal pipe diameter and eventually its blockage. Small additions of linear polyphosphates inhibit the precipitation of thick layers and become an enabling technology at sites seeking to use sodium carbonate in the presence of hard water. In general, the phosphate radical precipitates metal ions but polyphosphates complex metal ions and lower the effect of calcium and magnesium in hard water ie act as a water softener. State of Dispersion

The extent of dispersion of a slurry cannot be measured directly at present. Inferences on the extent of dispersion can be made by seeking to disperse further the particles by various means and seeking to observe a change in a property of the system, indirectly indicating that a change had occurred in the state of dispersion. For complex sulphide ores, two situations can be recognised. The first is the state of dispersion of the pulp in the initial roughing, scavenging and cleaning stages where the 80% passing size of the solid may be typically in the range 40 to 100 microns. The second is the state of dispersion of the pulp after each regrinding step. After the last regrinding step, the 80% passing size of the solid may be in the region 7 to 10 microns. If fine gangue particles adhere to intermediate or coarse particles containing the valuable mineral, the flotation rate constant of the valuable particles can decrease. If fine valuable particles adhere to intermediate or coarse gangue particles, there will be little chance of recovery of the adhering valuable particles. In contrast, if fine valuable particles adhered to intermediate or coarse particles containing the valuable mineral, the probability of recovery of the fine valuable mineral may be raised. This condition may happen in an unplanned way in some systems. Some researchers have created this situation at a small scale through selective flocculation, although this has not become an industrial flotation process (Dippenaar 1985; Fuerstenau 1988). In systems with very fine regrinding, the state of dispersion of the slurry has not received close attention, because adequate separations have generally been achievable. The tendency for particles of different minerals to adhere is driven by the magnitude and sign of the zeta potential of the minerals. However, in real systems, the observed zeta potential may be quite different from those observed for the pure mineral because precipitates or deposits on the surfaces eg metal hydroxides confer their zeta potential on the mineral (Rashchi et al, 1998; Sui et a1 1998). Inferences on the state of dispersion of a slurry are possible by measuring the apparent size distribution of the solid and seeking to observe a finer size distribution by addition of dispersing agents or sonic energy. A change in sizing may be deduced by settling tests or by a laser sizing. A rheometer can also be used to deduce information with dispersant additions to the slurry.

1104

High intensity conditioning can be used in the laboratory also to indirectly investigate the state of dispersion by examining if a significant change in flotation performance is possible after its use. Such changes may be related to the shearing apart of adhering particles (or possibly adhering precipitates or deposits). The slurry must not return to its original condition quickly after completion of the high intensity conditioning step. The high intensity conditioning process was adopted industrially at Hellyer (Holder 1994). In a PhD thesis on high intensity conditioning, Engel (1 999) indicated that, between 1945 and 1982, only nineteen articles with some connection to conditioning devices before flotation and after grinding were published for sulphide ores. Engel provided reasons for the inactivity on this topic after the 1930’s and 1940’s: “Firstly, the purpose of conditioning devices was subverted by the evolution of low operating-cost pulp mixers, designed by engineering companies mainly to suspend pulps rather than vigorously condition them. Secondly, different types of flotation cells can provide significant power inputs over the time of flotation, and are often used without recognising the influence of this power, above a minimum required for mineral suspension.” The work indicated that the mechanism by which high intensity conditioning affected a flotation system was the shearing of slime particle (or other deposits) from the surface of larger valuable particles, improving the flotation response of the cleaned valuable particles. An important example of industrial application of high intensity conditioning was for the Hellyer complex sulphide ore (Lane and Richmond 1993). Staff at Lakefield Research wrote several papers on the benefits of various applications of high intensity conditioning with flotation reagents in the 1980’s and 1990’s. In general, those papers did not directly investigate the mechanism of the effect. However, it was postulated that the creation of fine particle aggregates occurred and that these were readily floatable (Bulatovic and Salter 1989). Reagent Practice - Collectors Two decades ago, the xanthate family of collectors was the most commonly used family of collectors and the dithiophosphate family was the second most commonly used family. Dithiophosphates were considered to be almost as versatile as xanthates and more selective in the presence of gangue sulphides such as pyrite. Both xanthates and dithiophosphates were observed to have an affinity for both sulphide minerals and precious metals over a wide range of pH including the alkaline region, which is of practical importance. It was relatively common for metallurgists to seek beneficial effects from mixtures of these two families ie from synergism. Some additional types of collector have been developed in the last two decades and are available commercially. As in the past, these collectors require testing on each complex sulphide ore to compare their performance with the traditional collectors to establish if there are improvements in performance and if these improvements are cost effective. The cost of these new collectors has varied from region to region of the world and, in the regions where the price was high, industry has made less use of the new reagents. The new collectors exist in the following families: 1. 2. 3. 4.

Dithiophosphinates Monothiophosphates Modified thiourea eg alkoxycarbonyl ally1 thiourea Modified thionocarbamates eg alkoxycarbonyl thionocarbamates

A collector which has found significant applications for complex sulphide ores is Aerophine 34 18A (sodium diisobutyl dithiophosphinate) which belongs to the dithiophosphinate family, a

1105

relative of the well-known dithiophosphate family. This collector has found use as an auxiliary collector at low addition rates for the flotation of minerals containing precious metals (Mingione 1990). It has also found considerable application in Australia as a principal collector for complex sulphide ores under conditions of high ambient pulp temperatures and poor water quality where there have been issues with the use of collectors such as xanthate. In the acid range, monothiophosphates exist in the thiol form and collect sulphide minerals, as for example other thiol collectors (xanthates and dithiophosphates) in the alkaline region. The thione form of monothiophosphates exists in the alkaline region (from tautomerism of the molecule) and this form retains the ability to interact with precious metals such as gold and largely loses its ability to interact with sulphide minerals. Nagaraj et a1 (1992) reported that the adsorption of monothiophosphates on gold increased with the percentage of silver in the gold. Nagaraj (1997) discussed the modified thiourea and modified thionocarbamates. “Cytec Industries introduced ethoxycarbonyl thionocarbamate in 1985 and ethoxycarbonyl thiourea in 1989 (the AERO 5000 SERIES). They are now used widely in copper and precious metals (Au, Ag, and platinum-group) flotation in Australia, Canada, Chile, Europe and the U.S. The hexyl ethoxycarbonyl thionocarbamate, introduced in 1991, has already gained a rapid commercial acceptance in the industry. These modified thionocarbamates and thioureas are stable compounds quite resistant to oxidation. They are more selective against iron sulfides than the simple dialkyl thionocarbamates even at pH < 10. The alkoxycarbonyl thionocarbamates and thioureas were both developed as selective collectors for operation at reduced pH values, and indeed they have proven this in large-scale usage thus affording substantial lime savings. They have excellent shelf life, hydrolytic stability in a wide pH range and they are readily dispersed in water.” Nagaraj (1 997) commented, “the modified thioureas show a greater ease to float chalcopyrite than the corresponding thionocarbamates. The alkoxycarbonyl thionocarbamates, on the other hand, appear to float the copper-rich minerals such as bornite, covellite and chalcocite more effectively that chalcopyrite. These differences appear to be kinetic in nature and equilibrium recovery of the minerals may sometimes be the same for both classes of collectors”. This gives an indication of general properties, which may be recognized from testwork with the new molecules collectors. The alkoxycarbonyl thionocarbamates are modifications of the dialkyl thionocarbamate molecule, well known in the past as 2200.

Reagent Practice - Depressants Organic depressants used to depress talcose and carbonaceous minerals include derivatives of hydroxycellulose, guar gums, dextrin and mixtures of dextrin with azo dye intermediates (Bartrum, Dobrowolski and Schache 1977). A development in general depressant technology has been the introduction of new, synthetic depressants in the polyacrylamide family (Nagaraj 2000). These are water-soluble polymeric depressants of medium molecular weight in contrast to the higher molecular weight polymeric depressants from natural sources such as polysaccharides. Various functional groups can be incorporated in these synthetic polymers and the performance of each form can be evaluated (Boulton et al 2001). Increased application of this group of depressants to complex sulphide ores may be observed in the future. Evaluation of most of the cheap and practical inorganic compounds as flotation depressants in sulphide systems has occurred in the past, making identification of any new or better inorganic modifiers unlikely (Nagaraj 1994). However, investigations have continued on improving understanding of the mechanisms of common inorganic modifiers and apparently anomalous observations for some ores. One important example for complex sulphide ore is the role of cyanide in galena flotation. Rey (1958) indicated that cyanide could be an accelerator for galena. Hall et a1 (1990) reviewed several industrial cases. Prestidge at al(l993) used solution and surface analysis to confirm a basis for the anomalous observations and provided a mechanism where the extent of acceleration of the galena

1106

flotation rate from interaction between the cyanide and the collector, ethyl xanthate, was dependent on the pulp potential of the system. There have also been improvements in understanding solution reactions of sulphoxy depressants and the mechanism of depression of the sulphoxy depressants (Grano 1999). It is well known that, when sulphur dioxide, sodium metabisulphite or sodium sulphite are added to a system, the dominant form of the reagent (SO2, HSO; or SO3=)is determined by the pH of the system irrespective of the reagent added. The sulphoxy reagents are viewed as complex in their solution properties and in competitive adsorption on sulphides. Yamamoto (1980) found that sulphate ions react with xanthate ions in the aqueous phase. The rate of this reaction was temperature dependent (Sheldon and Johnson 1988) and, for elevated pulp temperatures of 35”C, could affect the flotation rate of galena on which the xanthate was adsorbing in laboratory tests. Various ways of countering this effect were devised (Grano et al, 1997). However, it was also shown that, while this effect existed in the plant, a more serious issue was the deposition of deposits identified as calcium sulphate on the minerals (Grano et a1 1995). The important point was that high pulp temperatures and poor water quality created conditions for precipitation of interfering overlayers on the valuable mineral, with which the collectors competed with difficulty to adsorb and create a hydrophobic surface.

Reagent Practice - Modifiers While lime has been the traditional pH modifier, there has been an increasing awareness of its deficiencies, which arise in plants with high levels of dissolved calcium, and magnesium in the water supply and where the water is recycled. The calcium ions in the lime further elevate the concentration of calcium ions in the recycled water and may provide a basis for precipitation of calcium salts or calcium hydroxide where the solubility product of the salt or hydroxide is exceeded. Magnesium salts or hydroxides may have their solubility product exceeded also, although the use of lime as a reagent does not contribute. One likely location is in cleaner sections where, with the use of lime, both the calcium ion concentration and the hydroxide ion concentration are raised. Such precipitates are often capable of random uptake on the surface of all minerals. This can diminish the performance of the critical cleaning section. If rougher sections are operated at such elevated pH values, similar effects can occur. The benefits of adding the limited amounts of “new” water to cleaning sections can be recognised, as the calcium and magnesium concentrations will not be as high. The benefit of using sodium hydroxide as a pH modifier can also be recognized as it does not contribute additional calcium ions to the circuit water. Sodium carbonate also does not contribute calcium ions to the circuit water. However, its effects are more complex as it increases the carbonate ion concentration, which allows the solubility product for metal carbonates to be exceeded, noting that sodium carbonate has been used with lime for water treatment in other industries. Brunswick Mining and Smelting has used both lime and sodium carbonate since the 1960’s (McTavish 1980), but the solution chemistry behind the use of the complementary dual modifiers was not well understood (Nesset 1998). The performance of the zinc circuit deteriorated unusually in the summer of 1995. A comprehensive investigation followed and the most likely explanation emerged: “As the summer progressed, the quality of recycled water was affected with increasing sulphate, acidity and calcium levels. The increase in calcium levels in the recycle water was likely related to the increased use of lime to offset the acidity caused by oxidation of the ore plus oxidation of sulphides in the tailings area during the hot summer. Regardless of the reason for the need for more lime, the effect of increased Ca in the mill water due to the recycle stream would be to decrease the effectiveness of sphalerite activation by CuSO4.”

This paper revealed the significance of seasonal events at one site and the increase in difficulties arising when the quality of recycled water is diminished. In this case, in an extreme

1107

summer, the water quality deteriorated briefly to a level that was probably still superior to the water at many other operations in semi-arid areas throughout the year. The paper reported that the calcium and carbonate ion concentrations were the most important in the system, rather than the sulphate concentration. The carbonate concentration had to be sufficiently high to remove adsorbed calcium from sphalerite (and pyrite) surfaces as precipitated calcium carbonate in the aqueous phase. This removal increased the effectiveness of sphalerite activation. The paper also reported “plant testing in a zinc rougher circuit revealed that a sequence of C03 addition, followed by Cu activation and then Ca addition (via soda ash, CuSO4 and lime) prior to flotation was the best combination for effective sphalerite activation and pyrite depression. Carbonate addition on its own improved the Cu activation of both sphalerite and pyrite and therefore did not result in improved selectivity for zinc.” Lauder et a1 (1994) described the benefits from the introduction of sodium carbonate as a modifier to the Hilton Concentrator, which was seeking, improved activation and performance of liberated sphalerite, in the presence of high concentrations of calcium ions (in the region of 1000 ppm). Later, these experiences lead to the progressive introduction of sodium carbonate to various sections of the Mount Isa Leadzinc Concentrator (Young, Pease and Fisher 2000). However, it must be stressed that the introduction of sodium carbonate to a system and understanding its behaviour at a particular site, is a large and difficult task. Ammonia was used as a pH modifier at the Geco Concentrator (Brooks and Barnett 1978). Reagent Practice - Frothers The main developments in frother practice were increased understanding of the benefits of mixtures of frothers and a limited increase in the chemicals that were used as frothers. The traditional frothers (polyglycols and methyl isobutyl carbinol ie MIBC) have been supplemented with longer chain alcohols than MIBC, esters and an alkoxy-type frother (triethoxybutane). Some processing plants have worked collaboratively with one of a small number of companies supplying frothers seeking a product (often a mixture of frothers) suited to that plant. The increasing use of very large tank cells has probably contributed to this trend. For complex sulphide ores with very fine regrinding, the fine hydrophobic particles stabilise the froth phase and may provide excessively stable foam in pump boxes, creating pumping difficulties. Elsewhere in the plant, a stable froth containing relatively coarse composite particles in a traditional scavenger step may be required. The stability of the froth phase depends on the fiother family, the size and hydrophobicity of the particles, the quantity of dissolved solid in the circuit water, the presence of other reagents in the system and other factors. Dippenaar (1 982) explained that froth instability occurs as the contact angle of the valuable minerals increases to very high values approaching 90 degrees. Intermediate levels of the contact angle are preferable for froth stability. Klimpel (1999) reviewed the findings for plants from Riggs (1986) and Klimpel and Isherwood (1991) and the following is quoted from Klimpel(l999):

0

0

different frother components are most effective on certain particle sizes; to a first approximation, blending of appropriate large- and small-particle frother components, gives an additive effect of broadening out the total size range recoverable; many more types of chemical structures can be used as blending components than were previously thought based on using a given component just as a stand alone frother; and it is possible to effectively blend frother components not only for particle size but for rate of flotation effects as well.

1108

The alcohol based frothers are generally regarded as being more effective for achieving higher concentrate grades at the expense of recovery, as a less stable froth with effective drainage results. This may make them suitable for complex sulphide ores requiring very fine regrinding. With recycling of water at a plant, the frother concentration may increase over many days or weeks to an equilibrium level greater than the requirement of the plant, removing the ability to adjust its addition rate. This position is unacceptable. A stable frother with low volatility and low tendency to adsorb on solid in the system may give this outcome. It is believed that frothers containing large molecules tend to provide this outcome. It is the experience of the writers that MIBC, which is a relatively small molecule, is suited to a site with recycling of water and high ambient temperatures, because it does not accumulate excessively in the closed circuit water system. Two important volumes, Frothing in Flotation and Frothing in Flotation II, exist on the topic. A poorly understood aspect of frothers in industrial practice is the extent of the role of coadsorption of frother and collector molecules at a surface and the extent to which a positive interaction assisting the attachment step may arise. Some fundamentals on this topic are provided in the volumes. Entrainment and Its Minimisation The detrimental entrainment mechanism for recovery of liberated gangue minerals is caused by the physical transportation of liberated gangue minerals from the pulp zone of a flotation cell to the concentrate launders. From examination of the severity of the mechanism or a size-by-size basis (Johnson, McKee and Lynch 1974), it was found that minus ten micron fractions of gangue are transported and recovered by the mechanism very effectively. The severity of the mechanism declined with increasing particle size, becoming negligible by 50 microns. Various authors have subsequently reported a large amount of corroborating findings. It was shown in a text book (Lynch et a1 1981) that, for perfectly mixed gangue minerals in a pulp zone, the following relationship existed connecting the behaviour of the water to the related behaviour of the liberated gangue mineral for a given size fraction: k, where

k, kw CFi

*

----

-

CFi

-

first-order rate constant for recovery of liberated gangue in size fraction i first-order rate constant for recovery of water Transportation efficiency factor, approaching 1 for very fine size fractions and declining to 0 for fractions above approximately 50 microns. (also known as classification vector and classification function)

=

k,

1

Because the rate constant values are directly related to the recovery of the mineral or species in question via the rate equation it can be seen that, when the transportation efficiency factor has a quite high value of 0.8 for a fine size fraction, the recovery of the gangue component will be 0.8 of the water recovery. Hence, if the water recovery from a cell was 5%, the recovery of the liberated gangue mineral will be 4% for the size fraction in question. Complex sulphide ores by definition, require the achievement of grinding or regrinding product sizings that are very small, perhaps with 80% passing sizes below 10 microns. The inevitable consequence is that a high proportion of the feed exists in size fractions where the value for the transportation efficiency factor is quite high, and therefore the contribution of the entrainment mechanism for recovery of gangue minerals must be substantial.

1109

Equation 1 can be rewritten as shown in equation 2 by using the definition of a first-order rate constant. Recovery Rate of Gangue

=

CFi

Recovery Rate of Water

I

----

2

Mass of Water

Mass of Gangue

Where recovery rates refer to discharge rates from cells and masses refer to masses in the pulp zone. Rearranging this equation provides the following equation, which provides the process variables that influence the recovery of a liberated gangue mineral in a size fraction. Recovery Rate of Gangue

=

CFi

*

*

Recovery Rate of Water

----

3

Mass of Gangue Mass of Water

One approach from equation 3 is to dilute the pulp to a greater extent. For complex ores, sufficient roughing or cleaning capacity must be installed to provide the necessary residence time required by the valuable mineral. For cleaning systems, multiple stages of closed circuit cleaning at low pulp densities are often employed to allow repeated application of the benefits of such “dilution cleaning”. Another approach is to reduce the transfer of water from the pulp zone to the concentrate launder to zero. This is achieved by froth washing where sufficient water is added evenly to a froth zone to provide a small positive bias ie enough water is added to supply all the water needed in the concentrate and the small excess is transferred from the froth zone to the pulp zone. Hence, the “particle laden” water from the pulp zone is replaced by an appropriately added, definite quantity of water containing no particles.

FLOTATION PERFORMANCE Circuit Types - Recycle The effectiveness of a separation in a bank with properly designed capacity is determined by the differential between the rate constants for the valuable and gangue minerals. Manipulation or improvement of the rate constant differential between liberated valuable minerals and liberated gangue minerals by alteration of chemical and physical variables is continually undertaken in plants. However, for a circuit containing a number of banks, it must be recognised in circuit design or circuit improvement that, for a given set of rate constants for valuable and gangue minerals in each bank (and therefore differentials between rate constants), a range of circuit efficiencies will result for various arrangements of the banks ie the position of the overall grade recovery curve for the total circuit is dependent on the type of circuit (open circuit or various types of closed circuit layouts). Lauder (1992) discussed the benefit of circuits with recycle in comparison to an open circuit layout. He concluded that “recirculating systems, correctly designed, effectively increase selectivity and the attainment of grade and recovery targets, ie the system that incorporates a recirculating stream correctly will achieve more cost effective performance”. He also considered the comparison of various circuits, each of which contained recycling. He concluded that “many different recycling systems may achieve the same performance, but some will utilise less stages and capacity than others”.

lit0

For complex sulphide ores, a practising metallurgist must be reminded that if the rate constant differentials are difficult to improve in the various banks, improvements may be possible in overall performance with the existing rate constant differentials if the circuit layout was not initially well selected. An example is provided (Lauder et a1 1994) where the destination of a retreatment tailing stream was changed, in conjunction with improved selectivities in individual banks, to provide improved overall performance. Lauder’s approach was in accordance with the principles defined by Glembotskii et a1 (1961). It is important to mention the typical effect of an addition of a depressant to a system. This class of reagent typically improves the rate constant differential between the valuable and some gangue minerals up to some addition rate. However, it often also decreases the rate constant for the valuable mineral. Hence, if a depressant addition is increased or if an addition is introduced for the first time in a bank of fixed and fully utilised capacity, a decrease in recovery of the valuable mineral is to be expected at a higher concentrate grade than previously. In open circuit systems and particularly closed circuit systems, the use of depressants must be compatible with the installed capacity. Excessive use of depressants in closed circuit systems will result in increased circulating leads of minerals, which will decrease the residence time even further. Hence, in selection of bank or stage capacities, the role and effect of depressants must be carefully considered. In general, it is particularly important for closed circuit systems that pump and bank capacities match the necessary values dictated by that recirculating system. Circuit Types - Direction of Composites to Regrinding Complex sulphide ore requires grinding at the commencement of the circuit to very fine product sizings to achieve adequate liberation. Alternatively, a cheaper option in terms of operating and capital costs can be taken where a somewhat coarser rougher feed is prepared and only the composite particles warranting regrinding are directed to the regrinding mills. By this approach, a greatly lessened tonnage of solid has to be reground to the necessary very fine size required. In general, there are three broad methods by which the composite particles containing the valuable mineral(s) can be directed to a regrinding mill.

1 2.

3.

Regrinding of cleaner section feed. Regrinding of cleaner tailing as a result of a clear mechanism for directing relevant coarse composite particles containing the valuable mineral to the cleaner tailing. Regrinding of cleaner tailing in a circuit designed with large flotation and pumping capacities to allow a large circulating load, thereby ensuring a significant portion of all the coarse particles (both barren and containing the valuable mineral) reach the regrinding circuit.

If a cleaner section feed stream is reground all coarse particle (liberated valuable minerals composite particles bearing valuable minerals and barren particles) must be reground. This approach provides no difficulty in directing the desired particles to the regrinding circuit but it does mean that liberated valuable minerals are reground and that considerable amounts of barren particles undergo regrinding. Coarse particles can be accessed via the cleaner tailing. There are two general mechanisms by which the coarse particles of interest (composite particles bearing valuable minerals) can be directed to a cleaner tailing stream, by use of only the flotation properties of the cleaner block (Johnson et a1 1982). In the most desirable mechanism, the recovery of valuable mineral in the coarse fractions across the cleaner block is significantly lower than for the intermediate size fractions and this is caused often by the reagent practice in the cleaner block. Examples are the use

1111

of elevated pH or a depressant that affects the behaviour of the more easily influenced coarse fractions. Such reagents can also be expected to decrease the cleaner block recovery of the valuable minerals in the minus 10 micron fractions if the reagents are used in excessive quantities. The other mechanism for directing coarse particles to the cleaner tailing does not rely on the recovery in the coarse fractions being considerably less than in the intermediate fractions. In this “blunt” approach, the plant must be designed to operate with very high circulating loads. This means that a high percentage of valuable bearing particles and barren particles in 4 size fractions of the cleaner feed will report to the cleaner tailing. The coarse fractions therefore become accessible to regrinding by a “brute force” method. Circuit Types - Preflotation Preflotation provides a means of removal of gangue minerals (sulphide or non sulphide), which are recoverable in the absence of collector at the commencement of a flotation circuit. Its relevance for complex sulphide ores is that these ores can contain gangue minerals such as talc (Jackman, Scamardella and Tilyard 1994), carbonaceous pyrite (Croxford, Draper and Harraway 1961; Johnson and Jowett 1982) and others which are moderately hydrophobic and which can therefore be recovered with frother as the only reagent. Cleaning of the preflotation rougher concentrate may be included to assist in removal of valuable mineral that inadvertently reached the rougher concentrate by entrainment and possibly some additional mechanisms. Developments in preflotation technology in the last two decades have centred on minimisation of the recovery of valuable mineral with the preflotation concentrate, whether rougher only or for the rougher and cleaner stages. The possibility of froth washing technology (see section on entrainment) now exists to reduce the loss of valuable mineral by entrainment in the minimum number of flotation stages. Increased awareness of the role of pulp potential in the flotation of valuable minerals by collector has resulted in consideration of setting the pulp potential in a preflotation stage at a value which may lower the recovery of the valuable mineral due to minor collector concentrations in the circuit water or due to an ability by the valuable mineral to acquire some hydrophobicity in the absence of collector. Some applications of the use of nitrogen to manipulate pulp potential to assist some closely related processes are given in Johnson (1988). Nitrogen gas may be used to halt the increase in pulp potential in the preflotation section after leaving a grinding circuit, thereby providing conditions less conducive for accidental recovery of the valuable mineral. Reagents may also be used to set the pulp potential, and a combination of nitrogen and reagents is also possible. Circuit Types - Reverse Flotation Reverse flotation circuits exist after cleaning systems and therefore exist at the extremities of flotation circuits. The valuable mineral is rendered hydrophilic by changing the chemical conditions while at least one gangue mineral remains hydrophobic. Liberated particles containing the hydrophobic gangue mineral, or composite particles containing the hydrophobic gangue mineral, may be removed by flotation. Cleaning and/or regrinding of the concentrate containing the removed gangue mineral concentrate can be conducted. The gangue mineral in question could be one that is recoverable in the presence of frother only. Therefore, the flotation stage can perform a similar function to the preflotation step. In addition, the reverse flotation step may be used to remove a gangue sulphide mineral such as pyrite which was recovered due to collector adsorption in the preceding conventional rougher and cleaner stages, in the conventional competitive collector adsorption and flotation system. Brunswick Mining and Smelting employed reverse flotation in the 1970’s and 1980’s in production of both zinc and lead concentrates and was a leader in this technology (McTavish 1980). The use of these reverse flotation circuits was discontinued because of a desire to simplify

1112

the complex flotation circuit and lower operating costs (Hendriks and Ounpuu 1985) by reviewing and improving the liberation and separation steps in the conventional part of the circuit. A lead reverse flotation system was employed at the Woodlawn Concentrator to remove pyrite from the lead concentrate (Bums, Duke and Williams 1982; Jackman, Scamadella and Tilyard 1994). A combination of high temperature and sulphur dioxide (or a closely related reagent) has been used to depress the galena in the applications at Brunswick Mining and Smelting and Woodlawn. At Mount Isa Mines Limited, it was found to be possible depress the galena by raising the pH. At this plant, depression of the galena allowed removal of carbonaceous pyrite, which possessed moderate hydrophobicity and did not require the addition of collector for its flotation from the galena. The much less powerful and less costly conditions for depression of pyrite at this plant are believed to reflect the less adequate adsorption of the collector on the galena in the conventional part of the flotation circuit. Tilyard (1999) reviewed reverse flotation practices for zinc concentrates in seven industrial plants. High temperatures and sulphur dioxide (or a closely related reagent) were used to depress the sphalente. This concentrate is the most common for application of reverse flotation because of the relatively high grade zinc concentrate sought by smelters. The quantity of troublesome sulphide gangue minerals in certain zones of an orebody may also result in intermittent use of this technology. Usually, pyrite is the sulphide gangue being removed to improve concentrate quality. The use of Eh-pH data from an operating zinc reverse flotation process is valuable in diagnosis of problems when establishing such processes at a new site and investigating approaches for obtaining the necessary conditions more cheaply (Johnson and Munro 1988). Xu, Finch and Rao (1 992) corroborated these findings. Analysis of Flotation Data The position and shape of recovery-size curves (valuable and gangue mineral recovery on the yaxis and the logarithm of size on the x-axis) have become a common starting point for analysis of the performance of a flotation bank or section (Cameron et a1 1971; Trahar 1981; Frew 1982). It is desirable that the recovery of minerals is determined so that the entire solid in the system is included in the analysis. The inclusion of water recovery values assists in analysis of the contribution from entrainment, particularly for complex sulphide ores requiring very fine regrinding. The analysis of flotation data on a sized basis may lead to consideration of split conditioning (Heyes and Phelan 1988) (and combined flotation) or other technologies. The Mt Keith concentrator of WMC Resources was designed to deslime nickel ore at 5 to 7 microns for split flotation to improve metallurgical results (Clark and George 1994). It should be noted that there is a tendency to assume that, if a class of particles reports to a given size fraction in recovery-log size curves, the class of particles existed in that size fraction in the flotation process ie that the pulp was fully dispersed. The addition of liberation data to recovery-log size data eliminates the need for deductions and inferences about the dominant state of liberation of each mineral in each size fraction of each product. Examples of such an analysis for the Mount Isa zinc-lead-silver ore are provided by Johnson (1987). The data can be collected for a bank or for the inputs and outputs of a plant. The liberation state of gangue minerals in the concentrate and valuable minerals in the tailing can be determined. It has become increasingly common for liberation data to be used by metallurgists in flowsheet design and to support operating plants for the analysis of swaration data. As discussed earlier, liberation data are also important in determination of the increase in liberation of minerals across size reduction steps. The use of recovery-size data in conjunction with liberation data will sometimes reveal the recovery of a liberated gangue mineral in quantities that indicate it has been recovered by the

1113

normal flotation mechanism. It cannot be explained by entrainment as the recovery value is much greater than the yardstick provided by the water recovery and there is no evidence of significant recovery of the mineral in tests without collector. Further, for the valuable mineral, there may be cases of unexpectedly low recovery of the liberated valuable mineral. In principle, further clues to the causes of these two types of process weakness can be sought by examination of surfaces. The availability of suitable instruments and centres providing surface analysis has increased slowly from approximately 1970. Industrial metallurgists working with complex sulphide ores have made increasing use of such facilities, particularly since 1985 (Stowe, Chryssoulis and Kim 1995; Greet, Netting and Skinner 1997). EQUIPMENT Grinding The primary grinding sections of concentrators treating complex ores range from installations with conventional rod mill-ball mill arrangements such as Mount Isa to those with semi-autogenous grinding (SAG) mills such as Brunswick Mining and Smelting (retrofitted over an original rod mill-ball mill arrangement) and Century Zinc which is a large plant treating over 5 million t/y of ore. The original grinding section at McArthur River Mining was unusual in that a fine flotation feed was prepared in a single stage SAG mill though this has now been supplemented by a tower mill or vertimill. Both the Red Dog (Kral 1992) and Hellyer (Richmond 1993) plants, which were designed by Cominco Limited, installed tower mills as tertiary grinding units following SAG mill-ball mill combinations. An attraction of the tower mill for this duty is its reported better energy efficiency compared with a ball mill in grinding below say 50 to 60 microns coupled with a smaller floor "footprint". One issue with the tower mill is that it is relatively closed to air compared with a conventional tumbling mill and the lower amount of oxygen introduced to the pulp can give a more electrochemically reducing environment. This can be ameliorated by the use of high chromium or ceramic grinding media but could still affect flotation (see CHEMISTRY OF PREPARATION AND SEPARATION above) The major innovation in grinding equipment which complex sulphide ore treatment has pioneered for the base and precious metals mineral industries is ultrafine grinding. Product sizing from such equipment is currently 80% - 6 to 8 microns. This technology originated in the industrial minerals industry grinding materials such as pigments and china clay. It uses a stirred mill with fine rock or coarse "sand" media. Currently there are two commercial units on the market, the IsaMill and Metso Stirred Media Detritor. Enderle et a1 (1997) and Johnson et a1 (1998) discussed development of the IsaMill while its application is covered by Young and Gao (2000) and Gao, Young and Allum (2002). Background on the Metso Detritor is covered by Smith (1999) while Burgess, McGuire and Willoughby (2001) and Davey (2002) discuss performance in sulphide mineral operations. The IsaMill is a horizontal high speed stirred mill operating at high power intensities up to 350kW/m'. After prototype development at Mount Isa Mines the first production IsaMill units were installed at McArthur River Mining, which commenced treating zinc-lead-silver ore with four 1MW units on regrinding duty in 1995. The largest installation to date is the eight units for lead and zinc regrinding in the Lead/Zinc Concentrator at Mount Isa Mines Limited. Another application is regrinding gold concentrate at Kalgoorlie Consolidated Gold Mines. The Metso Detritor is a vertical stirred mill with the largest unit currently available rated at 355kW. The Detritor's first production use in a sulphide ore concentrator was at the Elura zinclead-silver mine in New South Wales, Australia while the Century Zinc operation uses 15 of these machines for regrinding zinc first cleaner concentrate and three for regrinding zinc scavenger concentrate. The Thalanga copper mine in Queensland Australia successfully employs a Metso Detritor on copper regrinding.

1114

One benefit from these ultrafine grinding mills is that mineral surfaces are cleaned as commented on by Davey (2000). In fact these mills are probably the most successful examples of high intensity conditioning (see State of Dispersion above). Flotation Most of the recent technical literature on flotation machines has discussed the design and operation of large cells focusing on reducing capital costs and power costs. Papers on so-called “flash flotation” have emphasized the desirability of suspending and recovering very coarse particles, which is the opposite of that required in complex ore treatment where fine particles are the target. Most of the technical literature supports the view that small bubbles improve flotation kinetics. Ahmed and Jameson (1 989) observed up to a 40 times increase in flotation rates of 4 to 42 micron quartz and zircon particles as the bubble size was decreased from 655 microns to 75 microns. They suggested that the ultimate flotation rate depends on the balance between bubble-particle attachment and detachment; both of which are inversely proportional to the bubble diameter and increase with particle density, size and agitation. They felt that, in general an optimum bubble and particle size could be found to maximise recovery but concluded that separation performance could deteriorate if the feed had a wide particle size distribution. Power, Franzidis and Manlapig (2000) reported hydrodynamic data such as superficial gas velocity, bubble size, bubble surface area flux, air flow velocity etc. for 25 different flotation machine installations in South Africa and Australia. From this some usehl implications can be drawn on the possible application of the machines to fine particle flotation. Unfortunately to date there is a paucity of published data relating the hydrodynamic data to the relative performance of different flotation machines when treating very fine particles. Very large flotation machines have found little application in most concentrators treating complex ores. Exceptions are Red Dog and Hellyer, which originally used Maxwell cells for roughing while Century Zinc, was commissioned in 1999 with all conventional machines having 71 Outokumpu OK-100 loom3 cells and five Outokumpu OK-50 50m’ cells (Barnham and Kirby 2001). Reasons for the persistence of small to medium sized flotation machines are the age of the plants and possible concerns about a poorly shaped residence time distribution and short circuiting if there are only a few machines in series on some duties on plants with modest throughputs. A real problem with large flotation cells treating very fine particles can be the relatively low ratio of lip length to cell volume when froth discharge becomes rate limiting. Also very long residence times from using large machines could result in detrimental effects on mineral surfaces. Gorain and Stradling (2002) analysed the performance of various cells at the Red Dog concentrator using the K = PSbRf flotation model developed at the JKMRC. They found one cell of the three cell types examined capable of operating over a far larger range of operating conditions than the other two resulting in a greater range of recovery and selectivity. A machine with such flexibility is to be preferred over the alternatives. The superiority of flotation columns to reduce entrainment and thus improve concentrate grade when treating fine particles has been amply demonstrated (Espinosa-Gomez, Finch and Johnson 1988) and these devices have been used on some concentrate cleaning duties at Mount Isa, Hilton, Red Dog and Hellyer. One potential drawback with using flotation columns to treat very fine particles is that they tend to produce larger bubbles than mechanical cells. Also columns do not have the high-energy turbulence zone provided by the impeller in a conventional mechanical cell, which could reduce the attachment efficiency for fine particles. Another limitation on the use of flotation columns is that kanying capacity“, which is defined as mass of solids to overflow per unit of time per unit of column cross sectional area, decreases linearly with decreasing 80% passing particle size (Espinosa-Gomez, Yianatos, Finch and Johnson 1988; Espinosa-Gomez, Finch, Yianatos and Dobby 1988). Fairweather (1989) concluded that the efficiency of froth washing to reduce entrainment decreased with decreasing particle size and worsened as the concentrate recovery rate approaches the maximum value estimated by the Espinosa froth carrying capacity relationship. This could make columns less economic than conventional cells when treating very fine particles. The issue of having to cope with the large volume of column tailings

1115

due to the addition of wash water when running at a positive bias may be no worse than having to operate conventional cells at low pulp densities to minimise entrainment. Ancillary Operations One of the disadvantages of very fine particle flotation is the resulting stability of the froth and the pumping problems associated with it. This was a major issue for the first years of operation at McArthur River Mining (Bowen 1999) where spillage materially affected metallurgical performance. Young, Pease and Fisher (2000) reported that a new Warman centrifugal slurry pump design developed at McArthur River and Mount Isa Mines improved the flotation circuits' abilities to cope with the tenacious froths without "air locking" or causing spillage. These tenacious froths are difficult to thicken and filter with the entrained air exacerbating capacity problems caused by the very fine particle sizing. Inadequate concentrate dewatering capacity can result in returning "dirty" concentrate thickener overflow and "cloudy" filtrate back to the process causing further deterioration in separation performance. Attrition grinding produces some sub-micron sized particles and these ultrafines probably account for the reported rheological problems when pumping thickened slumes. Slurry chemistry can be very important under these circumstances and careful studies of agglomeration and flocculation may be warranted to determine the best conditions to minimise these rheological effects.

PLANT PRACTICE It is not intended to discuss the details of past and current plant practices, which are well documented for some of the concentrators treating complex ores. The reader should study the references given to gain an appreciation of the "drivers" of the methodologies and technologies mentioned above. Good review articles on Brunswick Mining and Smelting are by Neumann and Schnarr (1970), Hendriks and Ounpuu (1985) and Shannon et a1 (1993). Zinc-lead-silver ore treatment practice at Mount Isa has been discussed by Davey and Slaughter (1970), Bartrum, Dobrowolski and Schache (1977), Munro (1993), Young et a1 (1997) and Young, Pease and Fisher (1997). Rohner (1993) described treatment of the closely related ore from the Hilton deposit. Simkus and Egan (1994) covered improvements to operation of the Sullivan plant while data on equipment and operations at Red Dog is available in Mian and Ehrsam (1996), Fairweather, Lee and Mian (1997) and Lacouture and Hope ( 2002). The interesting evolution of plant practice at Woodlawn is discussed in Roberts, Bums and Cameron (1 980), Bums Duke and Williams (1982), Williams and Phelan (1985), Tilyard, Bowen and Holborow (1993) and Jackman, Scamardella and Tilyard (1994). Another volcanogenic massive sulphide ore of interest is Hellyer, which has been reviewed by Richmond (1993) and Platts, Quilliam and Wolski (1994). A perusal of these articles on concentrators treating complex ores provides some interesting common themes, which are also relevant to less well-documented operations such as McArthur River and Century Zinc. 0

Plant start-up performance has been almost uniformly poorer than that predicted from pre-production test work. Nearly all data sighted confirm the conclusion of Brown (1995) that it takes four years to achieve "plateau metallurgy" which is less than that originally planned. Preflotation sections to remove talcose and carbonaceous minerals were necessary in many of the plants.

1116

0

0

Making bulk concentrate proved to be only a short to medium term alternative to the production of single mineral concentrates by fine grinding and extremely fine regrinding. Concentrators such as Brunswick Mining and Smelting, Hellyer and Mount Isa found that it was more economical to install extra grinding capacity. The exception has been McArthur River, which to date has only produced a bulk concentrate. Finer grinding and regrinding allowed circuits to be simplified reducing circulating loads and making flotation sections easier to operate. Much of the complexity and associated circulating loads came from the addition of retreatment and upgrading sections attempting to compensate for inadequate liberation. Stowe (1992) commented that the treatment of a complex ore did not necessarily imply having a complex flowsheet. He also observed that in Noranda's experience, no case of circuit simplification proved detrimental to metallurgical efficiency.

CONCLUSIONS Commencing in the 1970's there has been over two decades of intensive work on the treatment of complex sulphide ores quantifying mineral liberation and both the physical and chemical aspects of mineral separation by the flotation process. In the 1990's this culminated in significant improvements in metallurgical efficiency when treating these hitherto refractory ores. These improvements resulted from a willingness (and in some cases the use of the new stirred mill technology) to grind tine enough for adequate liberation. The authors commend minerals processing engineers designing or operating flotation plants for complex sulphide ores to carefully study the wealth of data in the technical literature. Much of the data is available in greater detail than for other sulphide flotation operations eg mineral behaviour by size and liberation class down to sub-sieve size fractions. This represents an enormous effort in collection and analysis that can be usefully drawn upon to improve the metallurgical outcomes when treating such ores.

ACKNOWLEDGMENTS The authors would like to thank their many colleagues in Australia, Canada and other countries for their fruitful collaboration in over thirty years of floating complex sulphide ores. Special gratitude is due to those men and women who have worked with them over that time on the zinclead-silver deposits of the Mount Isa Inlier.

REFERENCES Abramov, A. A. and Avdohin, V. M. 1997. Oxidation of Sulphide Minerals in Beneficiation Processes. Gordon and Breach Science Publishers. Ahmed, N. and Jameson, G.J. 1989. Flotation Kinetics. Mineral Processing and Extractive Metallurgy Review. Vol. 5.77-99. Amstutz, G.C. 1961. Microscopy Applied to Mineral Dressing. Fijtieth Anniversary of Froth Flotation in the U.S.A.,2ndof 2 Vols, Quarterly of The Colorado School of Mines, Vol. 56 NO. 3.443-481. Anon. 1981. Cyclosizer Instruction Manual Particle Size Analysis In The Sub-sieve Range, Warman International Inc., Bulletin WCSR, 2ndrevision. Barnham, M. and Kirby, E. 2001. The Design and Commissioning of the Pasminco Century Process Plant, at SME Annual Meeting 2001,26-28 Feb 2001, Denver, CO, USA.. Bartmm, J., Dobrowolski H.J. and Schache I.S. 1977. Developments in Milling of Silver-LeadZinc Ores in the Mount Isa Area Since 1970. Lead-Zinc Update, eds D.O. Rausch, F.M. Stephens and B.C. Mariacher. SME of the AIME, New York. 157-182. Basilio, C I, Kartio, I J, Yoon, R H. 1996. Lead activation of sphalerite during galena flotation. Minerals Engineering, Vol. 9 No. 8. 869 - 879. Boulton, A, Fornasiero, D and Ralston, J. 2001. Selective depression of pyrite with polyacrylomide polymecs. Int. J. Mineral Process. 6 1. 13 - 22.

1117

Bowen, P. 1999. The Operational and Metallurgical Development of the McArthur River Concentrator, in Flotation Colloquium - 28 April 1999, Sir James Foots Institute of Mineral Resources - University of Queensland, Australia. Boyce, J.H., Venter, W.J. and Adam, J. 1970. Beneficiation Practice At The Tsumeb Concentrator. AIME World Symposium on Mining & Metallurgy of Lead & Zinc Volume I Mining and Concentrating of Lead and Zinc. eds D.O. Rausch and B.C. Mariacher. SME of the AIME, New York. 542-570. Brooks, L S and Barnett, C, 1978. Noranda Mines Limited - Geco Division. In Milling Practice in Canada, ed. D.E. Picket The Canadian Institute Of Mining and Metallurgy. 182 - 184. Brown, P.J., 1995. Startup Problems In Lead-Zinc Plants - A Personal Opinion. G & T Metallurgical Services, Kamloops, British Columbia, Canada. Burgess, F., McGuire I. and Willoughby, R. 2001. Operation of Sand Mill Detritors at Pasminco Operations. Fine Particles Processing and Tailings Summit, 2-4 July 2001, Perth, Australia. Bulatovic, S. and Salter, R.S. 1989. High Intensity Conditioning - a New Approach to Improving Flotation of Mineral Slimes. Proc. Int. Symposium of Complex Ores, Halifax, eds G.S. Dobby and S.R. Rao. Canadian Institute of Mining and Metallurgy. Pergamon. 169-181. Bums C. J., Duke, P. J. and Williams, S. R. 1982. Process development and control at Woodlawn Mines. Proceedings XIV International Mineral Processing Congress. Vol. 4, The Canadian Institute of Mining and Metallurgy. 18.1 - 18.14. Cameron, A.W., Kelsall, D.F., Restarick, C.J. and Stewart, P.S.B. 1971. A Detailed Assessment of Concentrator Performance at Broken Hill South Limited. Proc. AusIMM. Vol. 240. 53-67. Carson, D.J.T. 1995. Geological- Mineralogical Controls on the Metallurgical Response of Ore. In Short Course on Due Diligence. Prospectors and Developers Association of Canada, Toronto. Clark, I..J. and George, C.W. 1994. Development and Design of an Alternative Flowsheet for the Treatment of Mt Keith Ore. Proceedings Fifth Mill Operators ' Conference, Roxby Downs 10-14 October 1994. The Australasian Institute of Mining and Metallurgy: Melbourne. 27 1275. Croxford, N.J.W., Draper, N. and Harraway, D.H. 1961. Some Aspects of the Carbonaceous Fraction of Mount Isa Lead Concentrate. Proceedings of the AusIMM. No. 197. 149-161. Croxford, N.J.W. and Jephcott, S. 1972. The McArthur River lead-zinc-silver deposit, NT. Proceedings of the AusIMM. No. 243. 1-26. Davey, G. 2002. Ultrafine and fine grinding using the METSO Stirred Media Detritor (SMD), Proceedings 2002 - 34th Annual Meeting of the Canadian Mineral Processors, Canadian Institute of Mining Metallurgy and Petroleum, Montreal. 153-169. Davey, J.M. and Slaughter P.J. 1970. Changes In Lead-Zinc Flotation Practice at Mount Isa Mines Limited, 1955-1970. AIME World Symposium on Mining di Metallurgy of Lead & Zinc Volume I Mining and Concentrating of Lead and Zinc, eds D.O. Rausch and B.C. Mariacher. SME of the AIME, New York. 401-427. Dippenaar, A. 1982. The destabilization of froth by solids 1. The mechanism of film rupture. Int. J. Mineral Processing. 9. 1-14. Dippenaar, A. 1985. Shear flocculation of Fines for Improved Flotation. Mintek Report M230, December. Enderle, U., Woodall, P., Duffy, M. and Johnson, N.W. 1997. Stirred mill technology for regrinding McArthur River and Mount Isa zincllead ores. X X International Mineral Processing Congress, Aachen, Germany, September 21-26. Vol. 2. 71-78. Engel, M.D. 1999. High intensity conditioning for flotation. PhD Thesis. University of Newcastle, New South Wales, Australia. Engel, M.D., Middlebrook, P.D. and Jameson, G. J. 1997. Advances in the study of high intensity conditioning as a means of improving mineral flotation performance. X X International Mineral Processing Congress, Aachen, Germany, 21 - 26 September, (3). 391 -407. Espinosa-Gomez, R., Finch J.A. and Johnson N.W. 1988. Column Flotation of Very Fine Particles, Minerals Engineering, Vol. 1 No. 1. 3-18. Espinosa-Gomez, R., Yianatos, J., Finch, J. and Johnson, N.W. 1988. Carrying Capacity Limitations in Flotation Columns. Column Flotation '88, ed. K.V.S. Sastry. SME of AIME. 143-148.

1118

Espinosa-Gomez, R., Finch, J.A., Yianatos, J.B. and Dobby, G.S. 1988. Technical Note: Flotation Column Carrying Capacity: Particle Size and Density Effects. Minerals Engineering. Vol. 1 NO. 1.77-79. Fairweather, M.J. 1989. Process mineralogy of column flotation. Processing of Complex Ores Halifax, eds G.S. Dobby and S.R. Rao, Can. Inst. Min. Met. And Pet., Pergamon, 369-378. Fairweather, M.J., Lee, K. and Mian, N. 1997. Lead Circuit Flotation at the Red Dog Mine. Proceedings - 29th Annual Meeting of the Canadian Mineral Processors, Canadian Institute of Mining Metallurgy and Petroleum, Montreal. 307-323. Finch, J.A. and Leroux, M. 1982. Fine Sizing by Cyclosizer and micro-sieve, CIM Bulletin. Vol. 75 NO. 839. 235-240. Frew, J.H. 1982. Variation of Flotation Rate Coefficients in Zinc Cleaning Circuits. Int. J. Min. Proc. Vol. 9. 173-189. Frothing in Flotation (A Volume in Honour of Jan Leja), ed. J. S . Laskowski, Gordon and Breach Science Publishers, 1989. Frothing in Flotation II, eds J. S . Laskowki and E.T. Woodbum, Gordon and Breach Science Publishers, 1998. Fuerstenau, D.W. 1988. Enhancement of Fine Hematite Flotation by Shear Flocculation and Carrier Flotation. Flocculation in Biotechnology and Separation Systems, ed. Y.A. Attia. Elsevier. 695-706. Gao, M., Young, M and Allum, P. 2002. IsaMill fine grinding technology and its industrial applications at Mount Isa Mines. Proceedings 2002 - 34th Annual Meeting of the Canadian Mineral Processors, Canadian Institute of Mining Metallurgy and Petroleum, Montreal. 171188. Glenmbotskii, V A, Klassen, V. and Plaksin, I.N. 1961. Flotatsiya (translated as Flotation). Primary Sources: New York, 1963. Gorain, B. and Stradling. A. 2002. An integrated approach to modelling, design and optimization of the flotation plant at the Red Dog concentrator. Proceedings 2002 - 34th Annual Meeting of the Canadian Mineral Processors. Canadian Institute of Mining Metallurgy and Petroleum, Montreal. 279-296. Grano, S R, Wong, P, Skinner, W, Johnson, N W and Ralston, J. 1995. Detection and control of calcium sulphate precipitation in the lead circuit of the Hilton concentrator Mount Isa Mines Limited, Australia. Proc. XIX Int. Min. Process. Cong. Am. Inst. Min. Metall. Eng. Publ, San Francisco. Chapter 29. 171 - 179. Grano, S R, Johnson, N W and Ralston, J. 1997. Control of the solution interaction of metabisulphite and ethy xanthate in the flotation of Hilton ore of Mount Isa Mines Limited Australia. Minerals Engineering. 10 (1). 17 - 39. Grano, S.R. 1999. Mechanisms for the action of sulphide and carbonate ions in theflotation of the Hilton ore of Mount Isa Mines Limited, PhD Thesis, University of South Australia, Australia.. Greet, C.J., Netting, A.K.O. and Skinner, W. 1997. What is on the surface? An examination of the surface composition of galena from the lead primary rougher circuit of Mount Isa Mines' LeadIZinc Concentrator. Proceedings Sixth Mill Operators ' Conference. The Australasian Institute of Mining and Metallurgy, Melbourne. 133-139. Grondijs, H.F. and Schouten, C. 1937. A Study of the Mount Isa Ores. Economic Geology. 32, June-July. 403-450. Guilbert, J.M. and Park C.F. 1985. Deposits Related to Submarine Volcanism and Sedimentation. In The Geology of Ore Deposits. Chapter 14. New York. W.H. Freeman. Guy, P.J. and Trahar, W.J. 1985. The effects of oxidation and mineral interaction on sulphide flotation, Flotation of Sulphide Minerals, ed. K.S. Forssberg, Elsevier. 91-1 10. Hall, C I, Johnson, N W, and Quest, K B. 1990. The interaction of ethyl xanthate and cyanide in the lead roughing of a Mount Isa Ore, AusIMM Proc., Vol. 295, No. 2,39 - 43. Hendriks, D W and Ounpuu, M. 1985. Modification to zinc flotation circuit of Brunswick Mining and smelting in Symposium on Complex sulphides - Processing Ores. Concentrates and Byproducts: 1985, eds A.D. Zunkel, R.S. Boorman, A.E. Morris and R J. Wesely, The Metallurgical Society: New York, 271 - 292.

1119

Heyes, G.W. and Phelan, J.M. 1988. The Application of Separate Conditioning to Improve Zinc Metallurgy at Woodlawn Mines, Third Mill Operators’ Conference, Cobar 9-12 May 1988, The Australasian Institute of Mining and Metallurgy: Melbourne, 85-89. Holder, R.K. 1994. Improvements in Copper and Silver Flotation at Hellyer Using High Energy Conditioning, Proceedings Fifth Mill Operators ’ Conference, Roxby Downs 10-14 October 1994, The Australasian Institute of Mining and Metallurgy: Melbourne, 153-159. Jackman, R, Scamardella, M and Tilyard, P.A. 1994. Metallurgical practice and process developments at Woodlawn Mines Concentrators, Proceedings Fifth Mill Operators ’ Conference, The Australasian Institute of Mining and Metallurgy: Melbourne, 129 - 136. Jackson, B.R., Gottlieb, P. and Sutherland, D.N. 1988. A Method for Measuring and Comparing the Mineral Sizes of Ores from Different Origins, Third Mill Operators ’ Conference, Cobar 9-12 May 1988, The Australasian Institute of Mining and Metallurgy: Melbourne, 61-65. Johnson, N.W., McKee, D.J. and Lynch, A. J. 1972. Flotation rates of non-sulphide minerals in chalcopyrite flotation processes, AIME, Vol256, 1974. Johnson N.W. and Jowett, A. 1982. Comparison of lead primary rougher behaviour of several Mount Isa lead zinc ores, Proc. Of Mill Operators’ Conference N W Q’ld Branch AusIMM, Mount Isa, 271- 285. Johnson N.W., Wickham, P. and Watsford, R.M.S. 1982. Analysis of Leadzinc Concentrator flotation performance a Mount Isa Mines Limited, Proc. OfMill Operators’ Conference, N W Q ’Id Branch AusIMM, Mount Isa, 3 13 - 326. Johnson, N.W. 1987. Application of Kinetics and Liberation Data to Analysis of an Industrial Flotation Process, Vol. 80, No. 899, CIMBulletin, March, 113-117. Johnson, N.W. 1988. Application of electrochemical concepts to four sulphide flotation separations, Electrochemical Society, Proc. From Conf: On Electrochemistry in Mineral and Metal Processing II, Atlanta, 131 - 149. Johnson, N.W. and Munro, P.D. 1988. Eh-pH measurements for problem solving in a zinc reverse flotation process, Bull. Proc. Australas. Inst. Min. Metall, 293 (3), 53 - 58. Johnson, N W, Gao,M, Young, M F and Cronin, B, 1998, Application of the IsaMill (a horizontal stirred mill) to the Lead Zinc Concentrator (Mount Isa Mines Ltd) and the Mining Cycle, in Proceedings AusIMM ‘98 Annual Conference, The Australasian Institute of Mining and Metallurgy, 291-297. Klimpel, R.R. and Ishenvood, S. 1991. Some industrial implications of changing frother chemical structure, Int. J. ofMineral Processing, Vol. 33,369 - 381. Klimpel, R.R. 1999. Industrial experiences in the evaluation of various flotation reagent schemes for the recovery of gold, Minerals and Metallurgical Processing, Vol. 16, No. 1 February, 1 11. Kral, D. 1992. Red Dog: Cominco‘s arctic experience pays off again, Mining Engineering, Vol. 44 NO. 1,43-49. Lacouture, B. and Hope, G. 2002. The Red Dog VIP mill optimization project, Proceedings 2002 - 34th Annual Meeting of the Canadian Mineral Processors, Canadian Institute of Mining Metallurgy and Petroleum, Montreal, 61 1-621. Lane, G., Richmond, G. and Quilliam, R. 1991. Flotation Cells at Hellyer - Horses for Courses, Fourth Mill Operators‘ Conference, 10-14 March Burnie Tasmania 1991, The Australasian Institute of Mining and Metallurgy, Melbourne, 37-44. Lane, G.S. and Richmond, G.D. 1993. Improving fine particle flotation selectivity at Hellyer, XVIII Int. Min. Proc. Cong., 23-28 May, Sydney, Australia, AusIMM, 4, 897 - 904. Lauder, D.W. 1992. The recycle mechanism in recirculating separation systems, Mineals Enineering, 5 (6), 63 1 - 647. Lauder, D.W., Rohner, P., Burford, B.D. and Johnson, N.W. 1994. Transfer of zinc from LGM concentrate to zinc concentrate at the Hilton Concentrator of Mount Isa Mines Limited, AusIMM Proceedings, Vol299, No.2, 15-20. Lawrence, L.J. 1968. The minerals of the Broken Hill district, Broken Hill Mines - 1968, eds M. Radmanovich and J.T. Woodcock, The Australasian Institute of Mining and Metallurgy, Melbourne, 103-136. Lynch, A J, Johnson N W, Manlapig E V and Thorne C G, 1981. Mineral and coal flotation circuits; their simulation and control. Elsevier.

1120

McTavish, S , 1980. Flotation practices at Brunswick Mining, Canad. Min. and Metall. Bull., Vol. 73, No. 814, Feb., 114 - 120. Mian, M.N. and Ehrsam, C.O. 1996. "Northern exposure" The Red Dog experience, Canad. Min. and Metall. Bull., Vol. 89, No. 1005,97-90. Mingione, P.A., 1990, Use of Aerophine 3418 A Promoter for Sulphide Mineral Flotation. In Proceedings - 22nd Annual Meeting of the Canadian Mineral Processors, Canadian Institute of Mining Metallurgy and Petroleum, Montreal, 485-508. Monte, M.B.M., Lins, F.F. and Oliveira, J.F. 1997. Selective flotation of gold from pyrite under oxidizing conditions, Int J. Miner. Process, 51 255 - 267. Munro, P.D. 1993. Lead-zinc-silver ore concentration practice at the Lead-Zinc Concentrator of Mount Isa Mines Limited, Mount Isa, Qld in Australasian Mining and Metallurgy Vol 1 The Sir Maurice Mawby Memorial Volume Second Edition, (eds. JT Woodcock and JK Hamilton), The Australasian Institute of Mining and Metallurgy, Melbourne, 498-503. Nagaraj, D.R. 1994. A critical assessment of flotation reagents. In Reagents for Better Metallurgy, eds P.S. Mulkutla and W.A. Hancock, SME, Littleton CO, 81 - 90. Nagaraj, D.R. 1997. Development of new flotation chemicals, Trans. Indian Znst. Met., Vol. 50, No. 5, October, 35 - 363. Nagaraj, D.R. 2000. New synthetic polymeric depressants for sulfide and non-sulfide minerals, Int. Min. Proc. Congr., Rome. Nesset, J.E., Kim, J.Y., Li, M,, Sui, C., Cooper, M. and Chryssoulis, S.L. 1998. The effect of soda ash and lime as pH modifiers in sphalerite flotation, Proceedings - 30th Annual Meeting of the Canadian Mineral Processors, Canadian Institute of Mining Metallurgy and Petroleum, Montreal, 459 - 481. Neumann, G.W. and Schnarr, J.R. 1970. Concentrator Operations at Brunswick Mining and Smelting Corporation Limited - No. 12 Mine, AIME World Symposium on Mining & Metallurgy of Lead & Zinc Volume 1 Mining and Concentrating of Lead and Zinc, eds D.O. Rausch and B.C. Mariacher, American Institute of Mining, Metallurgical and Petroleum Engineers Inc., New York, 500-525. Nice, R.W. and Brown, P.J. 1995. The Design of a Base Metals Separation Process, Proc. XIX Int. Min. Process. Cong. (Am. Inst. Min. Metall. Eng. Publ.), San Francisco, Chapter 23, 137143. Petruk, W. and Schnarr, J.R. 1981. An Evaluation of the Recovery of Free and Unliberated Mineral Grains, Metals and Trace Elements in the Concentrator of Brunswick Mining and Smelting Corp. Ltd.", Canad. Min. andMetall. Bull., No. 74, Vol. 833, 132-159. Petruk, W., Wilson, J.M.D., Lastra. R. and Healy, R.E. 1991. An Image Analysis and Materials Balancing Procedure for Evaluating Ores and Mill Products to Obtain Optimum Recoveries, Paper 19 in Proc. 23rd. Canadian Annual Meeting of Canadian Mineral Processors, Canadian Institute of Mining Metallurgy and Petroleum, Montreal. Platts, A.M., Quilliam, R.T. and Wolski, M. 1994. Operational Improvements at Hellyer Concentrator 1992 to 1994, Proceedings F@h Mill Operators ' Conference, The Australasian Institute of Mining and Metallurgy: Melbourne, 141 - 151. Power, A., Franzidis, J-P and Manlapig, E.V. 2000. The Characterisation of Hydrodynamic Conditions in Industrial Flotation Cells, Proceedings Seventh Mill Operators' Conference, 12-14 October 2000 Kalgoorlie Western Australia, The Australasian Institute of Mining and Metallurgy, Carlton, 243-255. Prestidge, C A, Ralston, J and Smart, R StC. 1993. The role of cyanide in the interaction of ethyl xanthate with galena, Colloids and Surfaces A: Physicochemical and Engineering Aspects, 82,103 - 119. Ralston, J. 1991. Eh and its consequences, in sulphide mineral flotation, Minerals Engineering, Vol. 4, NOS.7 - 11, 859 - 878. Rashchi, F., Xu, Z. and Finch, J.A. 1998. Adsorption on silica in Pb - and Ca - SO4 - C03 systems, Colloids and Surfaces, A : Physicochemical and Engineering Aspects, 132, 159 171. Rashchi, F. and Finch, J.A. 2000. Polyphosphates: A review - their chemistry and application with particular reference to mineral processing, Minerals Engieeering, Vol. 13, No. 10 - 11 1019 - 1035.

1121

Rey, M. 1958. Differential flotation of lead - zinc ores, Trans. Int. Min. Dress. Cong., Stockholm, 525 - 537.

Richardson, P.E. and Walker, G.W. 1985. The flotation of chalcocite, bormite, chalcopyrite and pyrite in an electrochemical cell, Proc. XV Int. Min. Process. Congr. , Cannes France, Vol. 2. 198 - 210.

Richmond, G.D. 1993. Zinc-lead ore concentration by Aberfoyle Limited, Hellyer, Tas in Australasian Mining and Metallurgy Vol 1 The Sir Maurice Mawby Memorial Volume Second Edition, eds J.T. Woodcock and J.K. Hamilton, The Australasian Institute of Mining and Metallurgy, Melbourne, 530-535. Riggs, W.F. 1986. Frothers - an operator’s guide, Chemical Reagents in the Mineral Processing Industry, eds D. Malhotra and W.F. Riggs, Society of Mining Engineers of AIME, Littleton, CO, 113-116. Roberts, A.N., Bums, C.J. and Cameron, A.W. 1980. Metallurgical developments at Woodlawn Mines, Australia, Complex Sulphides, ed. M. J. Jones, Inst. Min. Met. (London), 128-134. Rohner, P. 1993. Lead-zinc-silver ore concentration practice at the Hilton Concentrator of Mount Isa Mines Limited, Mount Isa, Qld in Australasian Mining and Metallurgy Vol 1 The Sir Maurice Mawby Memorial Volume Second Edition, eds J.T. Woodcock and J.K. Hamilton , The Australasian Institute of Mining and Metallurgy, Melbourne, 504-507. Rumball, J.A. and Richardson, O.D. 1996. Measurement of oxidation in a base metal flotation circuit by selective leaching with EDTA, Int. J. Mineral. Process., 48, 1 - 20 Shannon, E.R., Grant, R.J., Cooper, M.A. and Scott, D.W. 1993. Back to Basics - The Road to Recovery Milling Practice at Brunswick Mining. Paper No. 2 in Proceedings - 25th Annual Meeting of the Canadian Mineral Processors, Canadian Institute of Mining, Metallurgy and Petroleum, Montreal. Simkus, R.S. and Egan, J.R. 1994. New Circuits for the Treatment of Fine Grained Zinc/Lead Ore at the Sullivan Concentrator. Paper No. 14 in Proceedings - 26th Annual Meeting ofthe Canadian Mineral Processors, Canadian Institute of Mining, Metallurgy and Petroleum, Montreal. Smith, C. 1999. Svedala Stirred Media Detitor. in IIR Conference Crushing and Grinding ‘99, March 29-30, 1999. Stowe, K.G. 1992. Noranda’s Approach to Complex Ores - Present and Future, in Australian success and challenges in fine particle technology, 1992 AMIRA Technical Meeting, Adelaide. Stowe, K.G., Chryssoulis, S.L. and Kim, J.Y. 1995. Mapping of Composition of Mineral Surfaces by TOF-SIMS. Minerals Engineering, Vol. 8 No. 4E.421-430. Sui, C., Rashchi, F., Xu, Z., Kim, J., Nesset, J.E. and Finch J.A. 1998. Interactions in the sphalerite -Ca - SO4 - C03 systems, Colloids and S u ~ a c e s ,A: Physicochemical and Engineering Aspects. 137,69 - 77. Tilyard, P., Bowen, P. and Holborow, R. 1993. Copper-lead-zinc ore concentrator practice at Woodlawn Mines of Denehurst Ltd, Tarago, NSW. In Australasian Mining and Metallurgy Vol 1 The Sir Maurice Mawby Memorial Volume Second Edition, eds J.T. Woodcock and J.K. Hamilton. The Australasian Institute of Mining and Metallurgy, Melbourne. 527-530. Tilyard, P. 1999. The use of reverse cleaning in testwork on Currawong copper-zinc ore, AusIMM Proceedings, No. 1.29 - 35. Trahar, W.J. 1981. A Rational Interpretation of the Role of Particle Size in Flotation, Int. J. Min. Proc., Vol. 8.289-327. Trahar, W.J. 1984. The Influence of Pulp Potential in Sulphide Flotation. Principles of Mineral Flotation (The Wark Symposium), eds M.M. Jones and J.T. Woodcock. The Australasian Institute of Mining and Metallurgy, Melbourne. 117-135. Williams, S.R. and Phelan, J.M. 1985. Process Development at Woodlawn Mines. Complex Sulphides, eds. A.D. Zunkel, R.S. Boorman, A.E. Moms and R.J. Wesely, TMS, New York. 293-304.

Witika, L.K. and Dobias, B. 1995. Electrochemistry of Sulphide Minerals, Flotation Science and Engineering. Marcel Dekker. 157-177.

1122

Wong, K.Y., Breen, A.M., Doherty, R.J. and Phelan, J.M. 1991. Flotation Treatment of Supergene Complex Sulphide Ores at Thalanga. Fourth Mill Operators' Conference, 10- 14 March Burnie Tasmania 1991. The Australasian Institute of Mining and Metallurgy, Melbourne. 205-21 1. Woods, R, 1984. Electrochemistry of sulphide flotation. Principles of Mineral Flotation (The Wark Symposium), eds M.M. Jones and J.T. Woodcock. The Australasian Institute of Mining and Metallurgy, Melbourne. 91 -1 15. Xu, Z., Finch, J.A. and Rao, S.R. 1992. Sphalerite Reverse Flotation Using Nitrogen. Proceedings of Third International Symposium on Electrochemistry in Mineral and Metal Processing, eds R. Woods and P.E. Richardson. The Electrochemical Society. 170-189. Yamamoto, T. 1980. The mechanism of depression of pyrite and sphalerite by sulphite. Complex Sulphide Ores, ed. M. J. Jones. Inst. Min. Met. (London). 71 - 78. Young, M.F., Pease, J.D., Johnson, N.W. and Munro, P.D. 1997. Developments in milling practice at the lead-zinc concentrator of Mount Isa Mines Limited from 1990. Proceedings Sixth Mill Operators ' Conference. The Australasian Institute of Mining and Metallurgy, Melbourne. 3 - 12. Young, M.F. and Gao, M. 2000. Performance of the IsaMills in the George Fisher flowsheet. Proceedings Seventh Mill Operators' Conference, 12-14 October 2000 Kalgoorlie Western Australia. The Australasian Institute of Mining and Metallurgy, Melbourne. 75-8 1. Young, M.F., Pease, J.D. and Fisher, K.S. 2000. The George Fisher Project to Increase Recovery in the Mount Isa Leadzinc Concentrator. Proceedings Seventh Mill Operators' Conference, 12-14 October 2000 Kalgoorlie Western Australia. The Australasian Institute of Mining and Metallurgy, Melbourne. 157-163.

1123

Overview of Recent Developments in Flotation Technology and Plant Practice for Copper Gold Ores Art Winckers’

ABSTRACT Recent technological developments and current plant practices in the processing of copper gold ores are discussed in the context of the mineralogy of the two most important types of copper gold deposits which are currently mined. These are the porphyry copper gold deposits and the hydrothermal iron oxide copper gold deposits. Five major producing mines in each category are listed, the typical mineralogy of these deposits is described and a metallurgical response profile of copper and gold for the two types of deposits is presented. The flotation circuit design, operating practices and metallurgy of two recently constructed mills processing ores from these two types of deposits are reviewed in detail. These include the Alumbrera and Cadia concentrators which process porphyry ores and the Candelaria and Ernest Henry concentrators which process iron oxide ores. The paper concludes summarizing the recent developments and trends in gold preconcentration, rougher flotation cell technology, regrind mill selection, cleaner flotation circuit design, process control technology, collector usage and overall operating philosophy. INTRODUCTION The most important economic copper gold mineralization occurs in the following types of deposits: Alkaline and calcic-alkaline copper gold porphyry deposits Iron oxide copper gold deposits Copper gold skarn deposits Differences in mineralogy between these types of deposits influence the selection of process design parameters and affect their metallurgical response. A description of the characteristics of the most important copper gold ore types is, therefore, believed to be a necessary introduction to the review of the technological developments in the design and plant practices of copper gold concentrators. CHARACTERIZATION OF THE MOST IMPORTANT CU AU ORES Production Rates And Head Grades Of Major Cu-Au Mines Table 1 lists some of the major mines currently processing ores from these three types of deposit:

‘A.H. Winckers & Associates, Mineral Processing Concultants, North Vancouver, B.C.

1124

Table 1 Major mines processing copper gold ores Milling rate Location Tonnes/day Type of Deposit

Head Grade c u Yo

Head Grade Au g/t

Porhyry Cu-Au Ores Alumbrera Batu Hijau Cadia North Parkes Kemess PT Indonesia Average

Argentina Indonesia Australia Australia Canada Indonesia

80,000 100,000 47,000 14,000 45,000 238,000 88,000

0.74 0.65 0.18 1.76 0.25 1.oo 0.76

0.92 2.60 0.72 0.81 0.56 1.08 1.11

Iron Oxide Cu-Au Ores Aitik Candelaria Ernest Henry Olympic Dam Osborne Punte del Cobre Average

Sweden Chile Australia Australia Australia Chile

47,000 62,000 29,000 24,000 4,000 3,000 32,000

0.35 I .09 1.10 3.02 2.58 1.70 1.65

0.21 0.26 0.55 0.52 0.89 0.48 0.53

Papua New Guinea

85,000

0.80

0.60

Au-Cu Skarn Ores OK Tedi

The average milling rate of porphyry deposits is significantly higher than that of the iron oxide deposits, indicating economies of scale required to develop the lower grade porphyry deposits. The copper to gold ratio varies widely between individual deposits. The relative importance of gold is higher in the porphyry and skarn deposits than in the iron oxide copper gold deposits which have a higher copper to gold ratio. On a world scale, the concentrate production from copper-gold skarn deposits is of minor importance. The copper grade is often very low in these deposits, with the exception of deposits such as OK-Tedi. Ore from these deposits is mostly processed by cyanidation. The processing of these ores will therefore be excluded from the review. The mill design and operating practice of the porphyry and iron oxide copper gold deposits will be discussed in further detail. The design of mills processing porphyry type ores frequently includes coarse free gold recovery circuits to improve the overall gold recovery. As such, it is important to include a review of this technology. Supergene alteration is common in both types of deposits but in most cases only represents a small percentage of the ore reserves.

Mineralogy Copper gold porphyry deposits can be described as large (>20 million tomes) low grade (>0.1% Cu) deposits which are typically hosted in or around polyphase alkalic or calcalkalic intrusive rocks and or extrusive equivalents. The sulfide mineralization is disseminated to partially fracture controlled. The hypogene sulfide minerals include pyrite, pyrrhotite, chalcopyrite, bornite, gold and silver sulphides. Gold frequently occurs as inclusions in or adjacent to pyrite, bornite and chalcopyrite. The amount of “free” gold varies between deposits (McMillan 1991). The sulfide mineralization is often coarse-grained. Economic mineralization is usually accompanied by phyllic and potassium alteration The iron oxide copper gold deposits which are currently mined range from small high-grade deposits to larger lower grade deposits. These deposits are more diverse than the Cu-Au porphyries and can occur in a wide variety of lithologic types and tectonic environments. This

1125

class of deposit is characterized by an abundance of iron oxide minerals (mainly magnetite and hematite) and a relative lack of iron sulphides and may contain elevated REE, P or F levels. Sulphide minerals include chalcopyrite, pyrite, pyrrhotite and sphalerite. Minor amounts of cobalt and arsenic sulphides occur in some deposits. Mineralization may range from disseminated to fracture or vein hosted and or locally semi-massive to massive in nature. Gold occurs as micron sized grains commonly associated with chalcopyrite and as inclusions in pyrite (Hitzman 2000). Metallurgical Response A survey was done to determine the copper and gold recoveries obtained at currently operating

mines. The results of this survey are shown in Figure 1.

Porphyry Cu-Au Ores ,Au ffi vs Rec

Porphyry Cu-Au Ores,Cu ffi vs Rec

Figure 1 Metal recoveries from porphyry and iron oxide copper gold ores

The copper recoveries in the iron oxide copper gold deposits vary widely, probably due to the diverse nature of morphologic and lithologic types represented; the porphyry ores show a fairly consistent head grade to recovery relationship. The gold recoveries of the iron oxide copper gold deposits also show a large fluctuation particularly at the lower head grades. The gold recoveries for porphyry deposits ranging between 70 and 80% do not exhibit a strong head grade to recovery relationship. A number of concentrators processing porphyry ores employ gold gravity or flash flotation circuits in their grinding circuits indicating the presence of a higher level of coarser grained gold in these ores. The only iron oxide copper gold operations that have gravity circuits are the high grade Osborne and Gecko mines in Australia. In summary, this survey indicates that the iron oxide copper gold deposits show a higher level of variability in metallurgical response than the porphyry copper gold deposits. PORPHYRY COPPER GOLD ORE PROCESSING The copper mineralogy in the primary zones of the porphyry copper gold deposits is relatively simple. Chalcopyrite, the predominant copper mineral is usually relatively coarse grained resulting in high rougher recoveries at relatively coarse primary grinds.

1126

Supergene alteration leading to the formation of oxide and secondary copper minerals is common in the upper parts of these deposits. The metallurgy is more variable and lower recoveries are experienced. The presence of sericite and kaolinite minerals which are formed as a result of the alteration can have a significant effect on the froth volume and stability. This is a factor which is often overlooked in the sizing of concentrate pipes, launders and pumpboxes. A design froth factor of three used in a number of plants was found to be inadequate. The gold content of these deposits is of significant economic importance. Gold occurs adjacent to and as inclusions in pyrite and chalcopyrite. The amount of “free” gold varies between deposits and is typically higher in the supergene ores. The presence of “free” gold in these ores has led to the installation of gravity and/or flash flotation equipment in the grinding circuits of these concentrators. Information on the benefit of this equipment to the overall gold recovery is, in most cases, not available. Recent data published on the Tintaya mine provide some insight. Tintaya ore reflects a mixture of porphyry and skarn types of mineralization and may therefore not be entirely representative. The flotation gold recovery from this ore grading 1.6% Cu and 0.35 g/t Au is about 60%. The installation of a gravity circuit consisting of Knelson centrifugal concentrators increased the overall gold recovery by 5% (Choquenaira et a1 2002). About 50% of the gravity gold recovered at Tintaya is plus 100 micron, a size which is not readily recovered by flotation. Centrifugal gravity concentrators have been used for many years to recover gold from grinding circuits. However, the use of flash flotation cells to recover gold from copper gold mill grinding circuits is a fairly new application of the technology. The relative merits and potential applications of gravity and flash flotation circuits are described in detail in a paper by Laplante and Dunne (Laplante and Dunne 2001). The paper also makes reference to flash flotation test work done at the Aitik Cu-Au mine in Sweden. Trials at this mill indicated a potential increase in gold recovery of about 10%. Laplante recommends that flash flotation should be considered when: a) The gravity recoverable gold (GRG) content is significant but does not justify a relatively expensive gravity circuit. b) Flotation is already the main recovery process. c) The GRG content is high but fine. These three conditions were met at Cadia. The gold recovery circuit at Cadia, which combines flash flotation and gravity concentration, will be described in further detail. Alumbrera The Alumbrera project is located in North Western Argentina in the Province of Catamarca. The 80,000 tonne per day operation was commissioned in 1997. Alumbrera is porphyry copper gold deposit that averages 0.57% copper and 0.65 g/t gold. Currently, the mill is being expanded to 100,000 tonneslday. Mineralogy. The ore contains native gold and the following sulphide minerals: chalcopyrite 2%, chalcocite 0.07%, covellite 0.03%, and pyrite 6.6%. The mineralization occurs in porphyritic and andesitic rocks. The iron sulphide to copper sulphide ratio varies between two and five. Oxidation of copper sulfide minerals is prevalent in some of the upper parts of the deposit. The copper mineralization is relatively coarse grained. Coarse Gold Recovery. About 10% of the cyclone underflow of each grinding line is screened to prepare -2 mm feed for the two Knelson concentrators installed in each line. The gravity roughing concentrates from the primary grinding and concentrate regrinding circuits are upgraded in two stages of tabling prior to melting and refining. The combined gold recovery from these circuits is about 5%.

1127

Flotation Roughing Circuit. The target primary grind size was determined by liberation analysis using QEM*SEM and confirmed with locked cycle tests. The results of these tests indicated a 65% copper sulphide liberation at a primary grind p80 of 150 micron. The grinding circuit currently produces a roughing flotation p80 feed size of about 190 microns. While this is coarser than the p80 design feed size of 150 micron, no adverse effect on the copper roughing recovery has been observed (Holborow 2002). There are two roughing flotation circuits, one for each grinding line. Each circuit consists of two parallel banks of eight OK100 tank cells arranged in sets of two on one level. The design retention time is 2 1 minutes at a pulp density of 35% solids. The minimum retention time required to maintain recovery is 17 minutes. The amount of metal in the feed, however, has a significant effect on the roughing kinetics. With head grades projected to decrease in the future from 0.7 to 0.5% copper, the installation of additional roughing capacity was deemed unnecessary for the planned mill expansion. Regrinding Circuits. The roughing concentrate of each line is fed to its own regrind circuit. Regrind ball mills with 3,360 kW drives are installed to obtain a target regrind p80 of 40 micron. The regrind level target is determined by the design requirements of the 315 km concentrate pipeline. A coarser cleaner feed would have been beneficial in reducing slimes losses (Holborow 2002). Each regrind circuit includes a secondary gravity gold recovery circuit consisting of two Knelson concentrators. The concentrates from these units are upgraded in the primary gravity cleaner circuit. Cleaner Circuit. The unusual aspect of this circuit is the use of the Jameson cells. The decision to select Jameson cells was based on the results of pilot plant test work, which indicated that the dewatering characteristics of Jameson cell concentrates compared favourably with those of column cells. Details on the design and operation of Jameson cells are provided elsewhere in this section. The circuit consists of two lines of first and second cleaners. The first cleaners are a two-stage unit, comprising two sets of cells in series. Each set is fed by dedicated pumps, providing the 150 kPa feed pressure required for the operation of these cells. The first cleaner tailings are pumped to a set of scavenger cells, producing a tailing which is recycled to the primary grinding circuit and a concentrate which is fed to the regrind mills. The second cleaners consist of a single Jameson cell in each line. The operation of the cleaner circuit has not been without problems, in particularly low initial recoveries, which were remedied by the installation of internal launders to provide more lip length and improvements in operating practices. Optimisation of the Jameson cell copper cleaner circuit has been difficult due to variable feed grades and feed rates combined with-the lack of “visual” froth control. Ongoing operational improvements however, have increased cleaner circuit recovery to over 95% (Holborow 2002). The flowsheet is shown in Figure 2. Reagent Conditions, Consumption. Lime is used in the roughers and cleaners for pyrite depression. The respective pH ranges are 9.5-1 1.0 and 10.8-1 1.8, depending on ore conditions and pyrite levels. The primary collector is Cytec 7249A (a blend of sodium butyl bithiophosphate, monothiophosphate and modified thiocarbamate) and is added at a rate of 20-30 gltonne. In order to promote gold recovery, SF506 and PAX are added at about 3-5 glt to the primary grinding and regrinding circuit cyclone underflow streams. SF506, mercaptobenzo-tiazol is a specific gold collector. A 1: 1 frother mixture of Dow 1012 and MIBC is added at a rate of 10-16 gltonne. Process Control. A Foxboro DCS system provides the process control hardware. On stream analyzers linked to the DCS provide the input data for on line metallurgy calculations. The flotation circuit regulatory control loops include: Air flow and pulp level of first four roughing cells Jameson cell air flow Reagent flows Expert Systems for supervisory control of the roughing and cleaners are being implemented.

1128

PARALLEL GRINDING AND FLOTATION LINE

I

n

L

PRIMARY GRlNDlND AND GRAVITY

‘

I

CLEANER-2 JAMESON (1)

4

t CLNR SCAV. JAMESON (1x2)

-

r-II

TAILINGS THICKNEFV DAM

Figure 2 Alumbrera mill process flow diagram

I

Metallurgy. The copper metallurgy is affected by the degree of oxidation of the ore and the chalcopyrite to pyrite ratio. A concentrate grade of 28% is targeted to optimise the recovery and to maximize the metal transport capacity of the 3 15 km concentrate transport pipeline. Since start-up in 1997, the copper recoveries have improved steadily from 87 to 93% through de-bottlenecking of the regrind and cleaning circuits and improved process control. The gravity gold recovery is directly related to the particle size of the free gold in the mill feed. The gold flotation recovery has been improved with the combined addition of SF506 (a gold collector), and PAX to the hydrocyclone underflow of the primary grinding and regrinding circuits and the recycling of the cleaner scavenger tailings to the roughers. The typical metallurgy for primary ore is shown in Table 2. Table 2 Alumbrera, typical primary ore metallurgy

Head grade %, g/t Concentrate grade %, g/t Concentrate recovery % Dore recovery %

cu

Au

Ae

0.62 27.7 92.6

0.90 29.5 70.0 4.0

1.60 65.0 80.0

--

--

Cadia The Cadia Hill project is located in New South Wales, Australia. The 47,000 tonne per day operation was commissioned in 1998. Cadia Hill is a relatively low grade copper gold porphyry deposit containing 0.17% copper and 0.74 gtt Au. Mineralogy. The minerals of economic interest indude chalcopyrite with lesser bomite, native gold, and gold associated with pyrite. There are five ore types classified on the basis of sulphide mineral phase abundance: bomite, pyrite/chalcopyrite, chalcopyritetpyrite, chalcopyritehomite and pyritehomite. The pyritetchalcopyrite ore which represents 60% of the mineralisation has the highest gold recovery, while the chalcopyritetpyrite ore representing 20% of the mineralization has the highest copper recovery. Coarse Gold Recovery. The Cadia ore has a gravity recoverable gold (GRG) content of up to 60% but the gold is relatively fine and difficult to recover in full-scale centrihgal gravity concentrators. Bench and pilot scale test work led to the selection of flash flotation cells for primary gold recovery, followed by gravity cleaning of the flash flotation concentrate (Laplante and Dunne 2002). The grinding circuit consists of a SAG mill with two parallel ball mills in closed circuit with cyclones. Approximately 50% of the cyclone underflow from each ball mill reports to an Outokumpu SKI200 flash flotation cell prior to returning to the ball mill. The combined concentrate from the two flash flotation cells is treated in a single Falcon SB38 centrifugal gravity separator. The Falcon concentrate is further cleaned on two Gemini tables in series and processed into dore bars. The tailings from the Falcon flow to a bank of two Outokumpu OK5 “coarse” flotation cleaning cells, the concentrate from these cells is combined with the final flotation concentrate The coarse cleaner tailings are combined with the rougher scavenger and cleaner scavenger concentrates ahead of regrinding. (Dunne et al. 1999). Flotation. The flotation feed size p80 is 175 micron with a typical range from 150 to 200 microns depending on the ore type and throughput rate. The grind recovery relationship is relatively insensitive up to 200 micron. Two parallel banks of rougher scavenger cells, each bank having four rougher and three scavenger Outokumpu OK150 Tank cells, treat cyclone overflow product. The design residence time per bank at 34% solids is 20 minutes. The scavenger concentrate and a portion of the rougher concentrate is or can be reground in a Svedala Verti Mill (VTM 400) to a design regrind p80 of 38 micron.

1130

The rougher concentrate plus reground scavenger concentrates and coarse cleaner tailings are cleaned in six Outokumpu OK30 Tank cells followed by four cleaner scavenger cells of similar dimensions. The cleaner scavenger tailings can either be recycled to the rougher circuit or open circuited to the tailings thickener. The cleaner scavenger concentrate is recycled back to the front of the cleaner with the option of regrinding it with scavenger concentrate, whilst the cleaner concentrate is pumped to a recleaner circuit consisting of a bank of four Outokumpu OK8 flotation cells. The flexibility in the design of the cleanerhegrinding circuit allows the operators to tailor the process conditions to the incoming feed mineralogy. Thus, the metallurgy for each ore type can be optimized, for instance, by avoiding unnecessary regrinding of already liberated minerals. Tailings Thickening, Water Consumption. The rougher and cleaner scavenger tailings gravitate to a 53 meter diameter high rate thickener. The fresh water consumption is about 0.55 cubic meters per tonne. The Cadia flotation circuit flowsheet is shown in Figure 3. Reagents. A simple reagent scheme has been developed. Lime is added to maintain a rougher pH of 9.5-9.7 and a cleaner pH of 9.6-9.9. The reagents are added at the following rates: MIBC Frother 10 g/t; collectors 6 g/t; and lime 0.5 kg/t. Collectors include S701, a ethylthiooctane base gold collector and S8761, a monothiophosphate. S8761, a partial replacement for S701, was found to improve the gold recovery and reduce the reagent cost. The collector additions are low compared to other operations. Process Control. A high level of process control is employed at the Cadia Hill concentrator. The major process streams are analyzed with an Outokumpu Courier 30 XP on-line analyzer. A unique feature of the instrument is its ability to measure low levels of gold with acceptable accuracy, even flotation tailings containing 0.15 g/t Au (Hart et al. 2000). Regulatory control loops include pH, air and reagents. Collector addition to the SAG mill is manipulated by feed forward control based on mill feed tonnage and copper head grade. “Floatstar” (Mintek software) is used to control the pulp levels in all flotation cells. The control is feed forward and takes into account process disturbances and interactions between cells One of the novel flotation process control strategies is the successful application of froth imaging systems to optimize the rougher metallurgy. Seven Frothmaster TM units have been installed, one on each of the seven cells of a rougher/scavenger bank. The instruments make use of machine vision technology to measure the froth characteristics. The instruments produce an extremely reliable and accurate measurement of the froth speed, as well as other parameters such as bubble size and froth stability. The system has the following objectives: 1. Control of the mass recovery of two Tank Cells in series on the same level where differences in the hydrostatic head between the two cells result in the cells pulling inconsistently with respect to each other. 2. Control of the concentrate grade (percent copper) from the first rougher. 3. Reduction in the frother consumption in the rougher circuit by monitoring the froth speed. A simple DCS controller was designed to react to disturbances in the same way that human operators react. The controller has two functions: Stabilization The stabilizing controller uses three manipulated variables (level, frother addition rate and aeration rate) to control the froth speed to a setpoint. Optimisation The optimising controller adjusts the froth velocity setpoint based on deviations of the rougher concentrate grade (measured in real-time by the OSA) from a setpoint. The plant metallurgist adjusts the grade setpoint based on information about the ore. The results from a two month trial in the rougher scavenger flotation circuit were as follows: The ratio of concentrate mass recovery between two cells in series on the same level has been successfully balanced. A 2.5 to 5.6% increase in rougher scavenger recovery was obtained. A 50% reduction in the standard deviation between actual concentrate grade and the grade setpoint was achieved.

1131

(Van Olst et al, 2000). Metallurgy. Typical Cadia Mill metallurgy is summarized in Table 3.

-

Table 3 Cadia Metallurgy First Year of Operation

Head grade %, g/t Concentrate grade %, g/t Overall recovery % DorC. recovery %

cu

Au

0.19 26.3 77.9

0.77 81.0 71.2 11.3

--

IRON OXIDE COPPER GOLD ORE PROCESSING The iron oxide copper gold ores are characterized by an abundance of iron oxide minerals and a relative lack of iron sulphides. Both primary and supergene copper mineralisation can occur. The iron oxide copper gold ore characteristics are similar to those described for porphyry class of ores. The sulphide mineral texture, however, tends to be more complex and a greater variety of minor base metal sulphides can be present. In general, gold is more closely associated with the sulphide minerals, hence, separate gold recovery circuits, either gravity or flash flotation, are less common in iron oxide copper gold concentrators. Candelaria The Candelaria Mine is located in Region I11 of Chile. The original concentrator, designed to treat 28,000 tonnes per day was commissioned in 1994 and reached a production level of 30,000 tonnes per day a year later. An expansion project to double the concentrator capacity was completed in October 1997. Candelaria is a fairly high grade, large tonnage hydrothermal iron oxide copper gold deposit with an average grade of 1.l% Cu and 0.26 glt of gold. Mineralogy. The minerals of economic interest include chalcopyrite and minor gold and silver minerals. Chalcopyrite, the only copper mineral of importance in the Candelaria deposit, typically has a relatively coarse grain size averaging around 500 micron. It can also occur as irregular to rounded inclusions in pyrite, pyrrhotite, magnetite, sphalerite and silicates. Gold is present mainly as micron-sized grains associated with chalcopyrite, especially where the chalcopyrite has replaced pyrite. Sphalerite is generally a minor phase in the deposit, but is widely distributed and may make up more than 10% of the ore minerals in some areas. Galena is a minor phase in the deposit and is usually associated with sphalerite. Magnetite is common and makes up about 10 to 15% of the ore. The pyrite content in the deposit is relatively low. It is normally associated with chalcopyrite, magnetite, or pyrrhotite, but it also occurs alone in thick veins associated with chlorite (Ryan et al, 1995). Flotation. The flotation feed is ground to a p80 size of 130 micron. During the 1997 mill expansion, a complete second line of processing equipment was installed. The flotation cell and regrind mill sizing of the original design and the expansion are discussed below. Rougher Circuit. Ten Eimco 120 m3 cells were installed for rougher flotation, compared with fourteen Eimco 85 m3 cells in the original concentrator. This provided exactly the same rougher flotation capacity and residence time of 20 minutes as in the original concentrator. Similarly Eimco 120 m3 cells were used as cleaner scavengers compared with the Eimco 85 m3 cells which were installed in the original concentrator. The Eimco 120 m3 cell uses an identical mechanism and drive as the Eimco 85 m3 cell which means that spare parts and maintenance requirements for the two are the same. The decision to change from the Eimco 85 m3to the Eimco 120 m3 cell was justified on the basis of equivalent metallurgical performance, operating cost savings (based on 27% power savings) and slightly lower installed capital cost for the larger cells (Marsden and Ogonoski 1998).

1133

Regrinding Circuit. A Svedala VTM-800 Vertimill with a 600 kW drive was installed to regrind rougher and scavenger concentrates to 75% minus 44 micron. The original concentrator included a Svedala 4.3 m diameter by 6.7 m ball mill with a 1,865 kW drive motor for the regrind application. (Note: The average power draw for this mill was only 600 kW, except during initial production when high grade ore was processed). The decision to use the Vertimill was made based on an estimated 25 to 30% lower specific power consumption compared with conventional grinding in this application. In addition, there were substantial capital cost savings (Marsden and Ogonoski 1998). Cleaning Circuit. The cleaning circuit consists of two nearly identical process lines. The cyclone overflow from each regrind mill is pumped to a set of four 3.66 m diameter by 14.0 m high Pyramid Resources column flotation cells, configured in parallel. Each column cell is equipped with 16 air spargers and a wash water distribution system on top of the cell, designed to drop wash water droplets onto the froth at a low velocity for optimal cleaning action. The column cell tailing of line-1 is pumped to a bank of eight 85m3 Wemco scavenger cells, configured in series in a 2+3+3 cell arrangement. The column cell tailings of line-2 are pumped to a row of six Eimco 120 tank cells of identical design as the rougher cells. The scavenger concentrate is pumped to the regrind cyclone feed sump (Marsden et a1 1998). The throughput and metallurgical performance of both circuits are nearly identical indicating that the 45% increase in rougher cell volume had no negative effect on the recovery. The flowsheet is shown in Figure 4. Tailings Disposal and Water Consumption. Flotation tailings are dewatered in two 120 meter diameter thickeners prior to disposal to a rock fill facility operated to provide sub-areal deposition conditions. Tailings decant water is returned to the mill. The fresh water consumption is 0.37 m3 per tonne treated. Reagents. The reagent scheme for flotation of chalcopyrite (and associated gold and silver minerals) requires the addition of two collectors and a frother. The first collector, Shell SF-323 (isopropyl ethyl thionocarbamate), is added to the ball mills. Frother (methyl isobutyl carbinol, MIBC) and a second collector, Hostaflot LIB (di-isobutyl sodium dithiophosphate), are added to the rougher flotation feed box. The rougher feed slurry is maintained at a pH of 10.5 - 11.0. Secondary additions of frother and the second collector can be made into the feed to the third bank of rougher cells. The average consumption of the three reagents are 7.5 g/tonne of SF-232, 4.0 g/tonne of Hostaflot LIB and 7.0 g/tonne of MIBC. Average lime consumption is 1.2 kg CaO/tonne. Process Control. The flotation circuit is controlled by a Pyramid Resources expert control system utilizing information provided by two Outokumpu Courier on-stream ray analyzers. One analyzer is used for the lower grade samples and the second analyzer for the higher grade streams. Metallurgy. The average metallurgy is shown below: Table 4 Candelaria metallurgy

Head grade %, g/t Concentrate grade %, g/t Recovery %

cu

Au

Ag

1.o 30.0 94.0

0.24 5.5 70.0

3.4 70.0 65.0

Ernest Henry The Ernest Henry Mine is located in the Mount IsdCloncuny district of Queensland in Australia. The 29,000 tonne per day concentrator was commissioned in August 1997. Ernest Henry is a relatively high grade iron oxide copper gold deposit with 1.1% Cu and 0.55 g/t Au. Mineralogy. Mineralisation can be divided into two main zones: the supergene and the primary. Supergene ore makes up 15% of the ore body and primary ore the remaining 85%. The primary ore mineralogy is quite simple. The ore assemblage is dominated by chalcopyrite

1134

LINE CYC.

AlLlNGS

Y I CLEANER SCAVENGERS (8- WEMCO 85 cubm) CONC.

LINE2 CYC. O/F

ROUGHERS (10 - WEMCO 120 cubrn)

U

CLEANER SCAVENGERS (6 - WEMCO 120 cubm) CLEANER COLUMNS (4)

-%-

FINAL CONC.

Figure 4 Candelaria mill process flow diagram

within a magnetite-carbonate gangue. There are no other oxides or sulphides of economic importance. The mean magnetite content of the primary ore is 20-25% weight. Gold shows a strong positive correlation with the chalcopyrite, although hematite and pyrite may be important sites for gold mineralisation. Gold is primarily contained within chalcopyrite. Pyrite is abundant, but decreases with increasing chalcopyrite in the higher-grade zones. The supergene ore zone is much more mineralogicaly complex. The predominant copper species in the supergene ore are chalcocite, secondary chalcopyrite, bornite and native copper. Native copper occurs in two distinct forms: a very fine grained disseminated distribution, and a coarse grained variation. There is no apparent relationship between gold and copper in the supergene zone. Gold is usually extremely fine grained and has been noted in interstitial gangue sites and in the sulphides (Strohmayr 1998). Native Copper Recovery Circuit. Native copper was estimated to make up approximately 10% of the supergene ore zone. After start-up, a small pilot plant comprising of screens and a small ball mill was built to test native copper recovery. The aim of this circuit was that coarse particles would be flattened in the ball mill and could then be recovered over the screens. A full scale plant was never built as the economics were not attractive (Tew 2002). Flotation. The flotation circuit feed is typically ground to a p80 size of 160 micron. Efforts to maximize throughput have increased the feed size p80 to 180 micron. The rougher circuit consists of nine 127 cubic meter Wemco “Smart Cells”, providing a residence time of 27 minutes at a pulp density of 38-42% solids. The Wemco cells are equipped with 150 kW motors drawing 135 kW. The first cell however, has a 185 kW motor installed which draws 160 kW. The rougher concentrate, reground in a 1-MW Vertimill to a p80 size of 45-50 micron is fed to a three stage cleaner circuit. The first cleaner consists of eight 50 cubic meter OK 50 tank cells, the second cleaner of eight 15 cubic meter OK16 cells arranged in two banks of four cells. The third cleaner consists of five 16 cubic meter OK 16 cells which are configured in a 2+3 cell arrangement. The first cleaners produce a final tailings product which joins the rougher tailings for dewatering in a 55 meter diameter high density thickener. The flowsheet is shown in Figure 5. Reagent Conditions and Consumption. The rougher pH target ranges between 10.5 - 11.0 depending on ore type. The cleaner pH is maintained at 11.5. Reagent use and consumption are as follows: Lime 830 glt 5 glt Frother, OTX- 140 Collector, SIBX 20 glt Process Control. A Yokogawa Centum CS Distributed Control System provides the process control infrastructure, linked to this is a 12 stream Amdel MSA. The rougher circuit control objective is to maximize rougher recovery by maximizing collector addition until the froth begins to collapse. The Wemco rougher cells are self-aspirated and hence have little air control. However, cell levels are maintained via the use of level controllers to the minimum froth depth possible without pulping the cells. Cleaner circuit strategy is based on achieving a target concentrate grade of 28-29% Cu at a maximum recovery. Typically, pulp levels are fixed in the cleaning circuit with air rates adjusted to maintain grade targets. The facility is available to control air addition rates based on concentrate grade, however, preference is given to operating personnel making these moves. All reagents are measured through flow controllers (lime, frother and collector). Typically, the lime is automatically adjusted to a target pH (Cascade loop), whilst frother and collector are controlled to a set point in AUTO mode as input by operating personnel. There is the facility for feed forward control of frother and collector addition based on the copper content of the mill feed, however, this has been unsuitable whilst treating varying ore types (Tew 2002). Metallurgy. The “Mine to Mill” operating practice has been successfully adopted to enhance the overall mine performance. At Ernest Henry Mining, the ore mineralogy was the underlying basis for plant design, and now plays an important role in the day to day operations of the

1136

CRUSHED ORE STOCKPILE

concentrator. Geological staff determine on a daily basis the anticipated feed type mineralogy which is then used as a guide, or predictive tool for the concentrator operations. Mineralogy is used to predict throughput, flotation grades and recoveries, and tailings thickener capacity as well as being a diagnostic tool to analyse past plant performance. For instance, the ratio between magnetite and hematite in the supergene ore was found to be a reliable indicator of the copper recovery, while the chalcopyrite and secondary copper mineral content could be used to predict the concentrate grade. The recovery of native copper which occurs in the supergene ore is estimated at around 80% for material finer than 50 micron and dropping off above this size (Tew 2002). The ratio of supergene to primary ore in the mill feed has a significant influence on the metallurgy as shown by the metallurgy of the two ore types in Table 5.

Table 5 Ernest Henry, metallurgy by ore type Supergene Ore cu Au Feed grade %, git Concentrate grade %, g/t Recoverv %

0.56 10.5 72.0

1.2 29.0 82.0

Primary Ore Au

cu

--

--

28.5 90.0

10.0 70.0

SUMMARY The mineralogy and the processing of ores from the two most important types of copper gold deposits, the porphyry and the iron oxide deposits, were reviewed.

Gold Preconcentration The most significant difference between the processing of the two ore types is the inclusion of separate gold recovery circuits in concentrators processing porphyry type ores. These circuits are estimated to increase the overall gold recovery by 5-10%. The use of flash flotation in grinding circuits for first pass gold recovery from ores containing fine GRG is a new, relatively low cost option to improve gold recovery and should be considered at the process design stage. Rougher Flotation The porphyry and iron oxide copper gold mines are among the larger base metal mines in the world. They have been leaders in the drive to reduce capital and operating unit costs. This effort has resulted in increased mill design capacities, as well as the development of larger flotation cells for rougher flotation duty. This has led to an evolution in cell design from conventional cells installed in a bank, to ever larger unit cells. The unit or tank cells are typically more cost effective in the plus 50 cubic meter size range. The largest tank cells currently in operation are 150 cubic meter cells. At Cadia, for instance, two lines of eight cells of this size were installed. The alternative would have been the installation of three lines with seven 100 cubic meter cells. At the Candelaria expansion, the rougher cell volume was increased by 40% from 85 to 120 cubic meters without adverse affects on the metallurgy. Savings resulting from reduced footprint area and services, in addition to lower power requirements per unit volume of the larger cells, are significant. The cell design must provide adequate mixing and air dispersion characteristics in order to achieve high recoveries from relatively coarse flotation feeds. As indicated in Table 6 , air can be supplied either by an external blower or the air is self-aspirated. The scale-up and design aspects of large flotation cells are discussed in the literature by Lawrence, (1993) and Bourke (1999), but will also be the subject of further review in papers DE-5 and DE-6 of this section. The rougher cell selection of the four operations discussed in this paper is shown in Table 6.

1138

Table 6 Rougher cell selection details

Porphyry Cu-Au Ores Alumbrera Cadia Construction year Mill feed, tonnedday Flotation feed, p80 micron Rougher flotation lines t/d/rougher line Cells, #/size m3 Volume, m3/k tonne/d Air supply Motor, inst. kW Motor, drawn kW

1997 84,000 190 2 42,000 8x 100 19 Blower 112 107

1998 48,000 175 2 24,000 7x150 44 Blower 175 135

Iron Oxide Cu-Au Ores Candelaria Ernest Henry 1997 60,000 130 2 30,000 8x120 32 Aspirated 150

1997 29,000 180 1 29,000 9x 127 39 Aspirated 150 135

Regrinding and Cleaner Flotation Tower mills are used increasingly for rougher concentrate regrinding because of their higher power efficiency. Typically, regrind levels are in the order of 40 micron. Flotation cells used in cleaning circuits include conventional cells, column cells and Jameson cells. In the case of column cells, conventional mechanical cells are used as cleaner scavengers. This may also be desirable when Jameson cells are used. The use of Jameson cells in copper cleaner circuits is a relatively new development. Comparative performance data with mechanical and column cells is not readily available. Jameson cells appear to be more selective than conventional and column cells with regard to gangue mineral rejection, but are more difficult to operate and control. Meredith et al. (1999) describe the flotation circuit design aspects utilizing Jameson cells. Flexibility in the design of cleanerhegrind circuit allows operators to tailor the process conditions to the incoming feed mineralogy and take further advantage of the “Mine to Mill” operating strategy. In some mills the concentrate from the first rougher and cleaner cells can for instance be made to bypass the regrind circuit, to avoid over-grinding of already liberated copper minerals. In other plants the cleaner or cleaner scavenger tailings can either be recycled to the rougher circuit or open circuited to the final tailing stream. The latter appears to be the case in mills that process porphyry type ores for the purpose of improving the gold recovery. The flotation kinetics of gold is often slower than that of copper minerals. Process Control Progress control systems and strategies vary between concentrators. The DCS or the hybrid DCSPLC, linked to an on-stream analyzer is the preferred hardware configuration in most concentrators. Regulatory control loops for air, pulp level and reagents are commonly used. In one of the concentrators a “Floatstar” (Mintek softare) system is used for feed forward control of flotation cell pulp levels taking into account process disturbances and interactions between cells. Expert Systems are used for supervisory control of equipment such as column cells. One of the most important developments is the successful application of on line digital viewing technology to measure froth characteristics. The analog signal from the instrument is fed to a DCS controller to stabilise and optimise the rougher flotation metallurgy. Reagent Conditions, Selection The rougher pH is controlled to 10.5-11.0 at three of the mills. Thiophosphate type collectors are used as primary or secondary collectors in combination with ethyl thionocarbamate (SF-372), mercapto-benzotiazole (SF-506) and ethylthiooctane (S-701). The latter two belong to a series of recently developed collectors, the F-series chelation collectors and the S-series elecrochemical

1139

collectors designed respectively as a specific gold collector and a chalcopyrite collector with high selectivity over pyrite .The development, evaluation and application of these collectors is described by Klimpel(l994) Operating Philosophy The “Mine to MiIl” concept is used at one of the mines to predict mill throughput and metallurgy from the anticipated feed type mineralogy supplied daily by the geological staff. Mine mill integration is an important step forward in maximizing the returns from a mineral resource. ACKNOWLEDGEMENTS The author would like to thank all those who generously provided infoxmation on the design and operation of the concentrators which were reviewed in this paper. These include: Alumbrera, Mr. R. Holborow of MIM and Mr. J. Miranda of Minera Alumbrera Ltd.; Cadia, Mr J. Dioses, Mr S . Hart and Mr. R. Dunne of Newcrest Mining Ltd.; Candelaria, Mr. J. Marsden of Phelps Dodge Mining Company; Ernest Henry, Mr. R. Holborow of MIM and Mr. A. Tew of Ernest Henry Mining. The author acknowledges Mount Isa Mining, Newcrest Mining Ltd and Phelps Dodge Mining Company for their permission to publish the information. I also would like to thank Dr Moira Smith for her assistance with the classification of the copper gold ore types, the description of the mineralogy and lithology of these ores and for her critical review of the manuscript.

1140

REFERENCES Bourke, P. 1999. Recent Developments in Outokumpu Flotation Technology; Mineral Processing and Hydrometallurgy Plant Design; The Australian Mineral Foundation, October 1999; pp 89-202. Brown, N., Dioses, J., Van Olst, M. 2000. Advances in Flotation Process Control at Cadia Hill Gold Mine Using Froth Imaging Technology. Proceedings SME Annual Conference 2000, Salt Lake City, Utah. Choquenaria, Bombilla, V., Alvarez Munos, 0. Gold Gravity Recovery in Copper Circuits at BHP Tintaya. Proceedings of the 34lh Annual Meeting of the Canadian Mineral Processors. January 22-24, Ottawa. Paper No 6. Dunne, R., Chittenden, R., Lane, G., and Morrel, S. 1999. The Cadia Gold Copper Project Exploration to Start Up, SME Conference, March 1, 1999, Denver. Hart, S., Dioses, J., Reed, M., Geff, P., Clement, B., Valery, W., and Dunne, R. 2000. Cadia Mines Reflections After 1 Years Operation . SME 2000, Salt Lake City, Utah, U.S.A. Hitzman, M.W. 2000. Iron Oxide Cu-Au Deposits: What, Where, When, and Why in Hydrothermal Iron Oxide Copper Gold and Related Ore Deposits, A Global Perspective. ed. T.M. Porter, Australian Mineral Foundation Inc. Holborow, R. 2002. Private communication. Keran, V.P., Zumwah, F., Palmes, J. 1998. Designing the Minera Alumbrera Concentrator Circuit. Mining Engineering. September 1998. Klimpel, R .R 1994. Some new Flotation Products for Improved Recovery of Gold andPlatinum. Proceedings of the 261h Annual Meeting of the Canadian Mineral Processors. January 1994,Ottawa.PaperNo 32. Laplante, A., Dunne, R.C. 2002. The Gravity Recoverable Gold Test and Flash Flotation Proceedings of the 34'h Annual Meeting of the Canadian Mineral Processors. January 22-24, Ottawa. Paper No 7. Lawrence, G.A. 1993. Confronting Future Mineral Processing Challenges with Super Large Flotation Cells. Flotation Plants: Are They Optimized? D. Malhotra ed., Chapter 10; SMME Inc. Littleton, Colorado. MacMillan, W.F. 1991. Porphyry Deposits in the Canadian Cordillera in Ore Deposits Techonics and Metallogeny in the Canadian Cordillera, ed. W. MacMillan et al; Canadian Geological Survey Paper, 1991-4. Marsden, J.O., Ogonowski, D.L. 1998. The Candelaria Concentrator Expansion Project, SME Annual Meeting Denver Colorado, March 1999. Reprint No 99-139. Marsden, J.O., Pennington, R.I., Rocher, W., Miranda, D. 1995. The Candelaria Copper-Gold Concentrator. Proceedings of Cobre'95. Volume 1. CIM. Meredith, G.P., Harbort, G.J., Murphy, A.S. 1999. Flotation Circuit Design Utilizing the Jameson Cell. Mineral Processing and Hydrometallurgy Plant Design; The Australian Mineral Foundation., October 1999. pp 271-225. Ryan, P.J., Lawrence, A.L., Jenkins, R.A., Matthews, J.P., Zamoro, J.C., Marino, E., Urqueta Diaz, I. 1995. The Candelaria Copper-Gold Deposit, Chile in Pierce, F.W. and Bolm, J.G., eds., Arizona Geological Society Digest, pp 625-645. Strohmayr, S., Barns, K., Brindley, S., Munro, P. 1998. Mineralogy Controlling Metallurgy at Ernest Henry. Proceedings Mine to Mill 1998 Conference, (The Australian Institute of Mining and Metallurgy) Brisbane, pp 13-18. Tew, A. 2002. Mill Superintendent Ernest Henry, Private Communication. Van Olst, M., Brown, N., Bourke, P., and Ronkainen, S. 2000. Improving Flotation Plant Performance at Cadia by Controlling and Optimising the Rate of Froth Recovery Using Outokumpu FrothMaster TM. Seventh Mill Operators Conference. Kalgoorly, Western Australia. October 2000.

1141

An Overview of Recent Developments in Flotation Technology and Plant Practice for Nickel Ores Andy Kern(’)

ABSTRACT Nickel sulfide ores invariably contain significant levels of deleterious minerals such as pyrrhotite or magnesium silicates and sometimes both. These gangue minerals have a detrimental effect on the concentrate grade and, in the case of the magnesium silicate minerals, on the ability to smelt the concentrate. This paper highlights some of the recent developments in nickel flotation practice that have significantlyimproved the flotation selectivity of the nickel minerals over these gangue minerals. Aspects of circuit design and equipment selection relevant to the selectivity gains are also discussed. Reference is made to nickel concentratorsin Canada, Russia and Westem Australia.

INTRODUCTION Nickel plays an important role in modern society. Growth in nickel demand (which averages about 4% per year), has been largely driven by the growth in stainless steel whose production accounts for about two thirds of primary nickel demand Other important consuming sectors include the aerospaceindustry, coins and a growing demand for nickel in battery applications. Economic deposits of nickel occur as one of two types; sulfidic and latexitic. While 70% of the land based nickel resources are contahed in latente deposits, the majority of the current world’s production of nickel still comes fiom sulfidic sources (Bacon 2000). Of the 1.027 Mt of nickel produced in 1999, 56% came fiom sulfidic ores. However, there has been significant attention focused in recent years on the recovery of nickel fiom lateritic deposits, most notably the pressure acid leach processes. These will, in time, produce a larger share of the world’s nickel As the demand for nickel continues to grow, there has been a concurrent trend towards lower metal prices. This, along with the threat of significant new sources of nickel coming on line fiom the lateritic deposits, has challenged the sulfide nickel mining operations to achieve breakthrough efficienciesin their operations.

It goes without saying that the mineral processing of nickel sulfide deposits plays a key role in the economic viability of a deposit. Improving recoveries, producing a high grade concentrate for smelting while at the same time controlling or reducing costs continues to be the focus of the nickel mineral processor. This paper discusses current flotation practices at some of the worlds largest nickel sulfide milling operations. It will focus on how individual operations have met and overcome the mineralogical challenges of their ores to stay competitive in the business. (1) INCO Ltd., Clarabelle Mill, Copper Cliff,Ontario, Canada, POM 1NO

[email protected]

1142

NICKEL SULFIDE DEPOSITS Nickel sulfide deposits can be classified according to the types of ultramafic and mafic orebodies With which they are associated. The simplified classification recognizes two principal classes: (Alcock 1988)

Class I Intrusion-related: deposits associated with layered mafic or lenses of ultramafic rocks in igneous intrusive complexes. Class II Volcanic related: deposits associated With ultramafic volcanic flows Examples of some of the better h o r n deposits in each category are listed in the table below.

~

Intmion Related Ultramafic Lenses Layered Mafic Bodies Thompson Sudbury, Norilsk, Mt Keith Octyabr’sky, Lynn Lake Talnald.1, Ungava Peninsula (Raglan) Bushveld, Jin Chum, China Stillwater, Duluth Voisey’sBay

Volcanic Related

S. Windarra LangmUir

The host rock for deposits associated with layered igneous bodies is generally gabbro (composed essentially of feldspar and pyroxene) whereas those associated with the intrusive ultramafic lenses and the volcanic deposits are hosted by dunite or peridotites (composed of olivine and pyroxene) or more commonly their hydrated alteration derivatives serpentine and chlorite. NICKEL SULFIDE ORE MINERALOGY The dominant nickel mineral in sulfidic deposits is pentlandite - a nickel iron sulfide (NiFe)9S8.It accounts for 90% of the nickel mined fiom sulfide ores. In addition, ores may also contain minor amounts of millerite (NiS), Violarite (Ni2FeS4) and arsenides such as niccolite (NiAs) and gersdofite (NiAsS).

Pentlandite almost invariably occurs in association with varyhg quantities of the iron sulfide mineral - pyrrhotite. The iron to s u l k ratio in pyrrhotite is usually less than one-to-one and the formula is often written F%-,&. Typically the composition of pyrrhotite is given by Fe&. Small amounts of nickel may substitute for iron in the pyrrhotite lattice, the nickel content of such pynhotite can be up to 1.5%. In many deposits, this accounts for a significant amount of the total nickel. For example, in the Sudbury area deposits, about 10% of the nickel is in pyrrhotite. Rejection of nickeliferous pyrrhotite to tails can result in significant loses in overall nickel recovery. Fyrrhotite occurs in two crystallographic forms - monoclinic which is ferromagnetic and hexagonal which is para- or weakly magnetic. The nickel content of hexagonal pyrrhotite is generally higher than that of monoclinic pyrrhotite. The form of pyrrhotite in a deposit is an important determinant of the preferred flowsheet.

1143

The pyrrhotite is usually much more abundant in nickel sulfide deposits than the pentlandite. Ratios of pyrrhotite to pentlandite can vary from 1:l to 101. In such ores, the rejection of high levels of pynhotite is essential both to reduce sulfur emissions to the environment and to produce a smelter feed With a high enough nickel content to make it viable. The challengeof rejecting high levels of pyrrhotite is made all the more difficult by the presence of fine, flame-like intergrowths of pentlandite in the pyrrhotite grains (Figure 1).

Hgme 1. Photo of fine pentlandite flames in pyrrhotite - Sndbnry Area Ore These ex-solution flames are typically 1-10 microns wide and are generally regarded as unrecoverable from the pyrrhotite. Inevitably, the presence of these pentlandite flames and the nickel in solid solution means that rejection of pyrrhotite carries with it a penalty in nickel recovery. In essence, one of the difficult challenges for the nickel sulfide mineral processor is the separation of a nickel-iron sulfide (pentlandite) from an iron-nickel sulfide (pyrrhotite). Alteration of the primary sulfide assemblageresults in complex intergrowths that can give rise to problems in flotation recovery. One of the most common alterations encountered is the breakdown of pentlandite to violarite and magnetite. This is usually accompanied by the transformation of the associated pyrrhotite to nkkeliferous pyrite plus magnetite. The violarite grains often retain the associated magnetite as a rim - inhibiting recovery (Alcock 1988). In sulfide deposits, the nickel often occurs along with economic concentrations of copper typically as chalcopyrite (CuFeS3 and to a much lesser extent, cubanite (CuFe2S3). The concentration of copper in an ore plays a significant role in ease or complexity of the separation of pentlandite fiom pyrrhotite.

Nickel also frequently occurs along with economically recoverable concentrations of cobalt, gold, silver and the platinum group elements (PGE’s) - plathum, palladium, rhodium, ruthenium, iridium and osmium. In some ores, such as at the Noril’sk operation in north-west Siberia, and in

1144

the Bushveld Complex of South M c a , the nickel is recovered as a by-product of the mining of the FGE ores. ROCK MINERALOGY Sulfides associated with layered mafic intrusions are hosted by gabbro, which consists of feldspar and pyroxene and only minor amouuts of mica. High levels of rock rejection (>95%) are readily achieved fiom these ores in milling circuits Without the need for rock depressants. Nickel assays in such rocks are in the parts per m a o n range and thus there is no loss of nickel other than in unliberated sulfides in the rock reject. In contrast, sulfides in ultramafic lenses and in volcanic related deposits are hosted by dunite and peridotites or, more usually, their serpenthized equivalents. These include fibrous asbestiform silicates such as chqstotile serpentine or platy silicates such as talc or chlorite. These silicates present the mineral processor With significant challenges. Talc and chlorite have natural hydrophobicity and can dilute the concentrate With high concentrations of magnesia. If the magnesia content is too high, the bath temperature in the smelting fiunace has to be elevated to maintain a slag at a reasonable viscosity. Furnace brick life is shortened and metal losses in slag are increased. Furthennore, some forms of serpeutines tend to slime the surface of the pentlandite resulting in higher losses to tails. For all these ores, dispersants andor rock depressants are essential to make the deposit viable. In addition to the problems associated with magnesium containing minerals, many of the ultramafic rocks also have a significant nickel content. Typical background assays are 0.06% nickel while values up to 0.25 to 0.35% are observed.

From the mineral processors perspective, the design of the flowsheet and selection of reagents Will depend on the requirements for:

u u u

Pyrrhotite rejection Rockrejection Copper-nickel separation

In this section, the technologies used to achieve the above objectives at a number of the world largest nickel sulfide milling operations will be described. Focus will be on the flotation technology and readers are directed to the refeaences at the end for a more comprehensive description of the respective milling operations.

Information relating to the followingmiUing operations will be presented. Thompson Strathcona Clarabelle Noril’sk -gla

Mt Keith

(INCO) (Falconbridge)

(INCO) (Noril’sk Combine) (Falconbridge) (Westem Mining Corporation)

Information presented has been extracted from the literature. No “inside ‘‘ information is presented unless the parties involved have &ven permission. PYRREIOTITE REJECTION In the fillowing section, the technology used to reject pyrrhotite is described for sevaal mills spanningthe range iiom simple to the complex.

1145

The flotation properties of nickeliferous pyrrhotite are significantly different fiom that of nonnickeliferous variety in that pulp oxidation causes activation (Wells et al 1997). Activation by oxidation is now known to be caused by the presence of nickel and copper ions adsorbed on the pyrrhotite surface (Kelebek, Wells and Heinrich W6). These ions have been shown to catalyze the oxidation of xanthate to dkinthogen. As a result, the complexity of the required technology to reject pyrrhotite depends to some extent on the presence and subsequent removal of these activating ions. While pyrrhotite becomes activated by oxidation and hence more floatable, the reverse is true for pentlandite. This presents a challenge to the mineral processor where the very act of flotation causes a progressive deterioration in pentlandite kinetics and an increase in pyrrhotite kinetics. One of the key strategies therefore employed is to float the pentlandite as quickly as possible so as to limit oxidation. In some flowsheets, specific c h d c a l s or strategies are used to inhibit the oxidation of pyrrhotite. Thompson Mill The Thompson Mill is located in northean Manitoba, Canada. It was built in 1958 by INCO and currently treats about 9,000 mtpd of ore fiom two underground mines - Thompson and Birchtree. The ores are batch processed separately through the same flowsheet from separate bins. Recoveries fiom concurrentprocessing of the two ores are lower due to the poisoning effect of the ullramafic component of the Birchtree ore on the Thompson ore. This aspect is discussed later. The Thomson orebody is hosted by metasedimentary and metavolcanic rock with variable ultramafic rock associations. Ultramafic rocks are mostly serpenthites although there is little serpentinite in the Thompson orebody. Sulfide mineralization consists of pyrrhotite and pentlandite a 2.3:l ratio, and minor amounts of chalcopyrite. The pyrrhotite is monoclinic and has an average solid solution nickel content of about 0.6%. The head grade is typically 2.6% Ni, 0.16% Cu and 1O%S. The ore is crushed in a conventional open circuit crushing circuit to a fine ore (30% + % ”). Open circuit rod and ball mills in closed circuit with cyclones grind the ore to 30% + 150 p (DamjanoVic 2000). The flotation circuit is shown in Figure 2 below. All flotation cells are 2.83 m3 (100 ft3) Denver cells. Soda ash is added to the rod mill feed to achieve a pH 10 in the roughers. This is the only treatment required to achieve the desired level of pyrrhotite rejection (55-60%). When soda ash is used for pH control, the rate of pentlandite flotation is increased relative to that of pyrrhotite compared to flotation at the natural pH of 8. A small amount of copper sulfate (0.01 gkg) is added to the roughers to enhance sulfide kinetics. Potassium amyl xanthate is added to the roughers with MIBC fiother. Rougher tails are reground and more xanthate and fiother are added to the scavenger cells. Rougher concentrate is cleaned and the rougha tails are combined with the scavenger concentrate to feed the scavenger cleaners. No recycle water is used in the Thompson Mill. Only fie& water &om the Burntwood River is used. The basic nature of the ores combined with the use of soda ash in the mill enables water to be discharged fiom the tailings area without treatment and still meet the environmental regulations for nickel in effluents (<0.5 ppm).

1146

RockTails

Pymhatii Tails

Figure 2. The Thompson Mill Flotation Circuit Copper nickel separation is readily effected with the addtion of lime to the rougher cleaner concentrate. Both pentlandte and pyrrhotite are depressed under condtions of saturated lime. Unlike in Sudbury area ores where cyanide is used to depress the pentlandite and pyrrhotite, the use of cyanide for CuNi separation in Thompson Mill results in the permanent depression of chalcopyrite (Agar 1991). The purpose of the copper cleaner circuit is to reject graphite to the cleaner tails that report to the nickel concentrate. The presence of graphite in the nickel concentrate helps to maintah the reducing potential in the electric fiunace bath so that the build up of magnetite in the b a c e can be controlled. The graphite rejection circuit consists of three stages of Denver 2.8 m3 cells operating in a counter current mode. Dextrin is added to the iir.9 cleaner concentrate to effect the depression of graphite. Overall pentlandite recovery fiom Thompson ore is 96% With 55% pyrrhotite rejection and 95% rock rejection.

Strathcona MiU The Strathcona Mill is located on the north- rim of the Sudbury Basin,some 400 km north of Toronto, Canada. The Sudbury Basin, an elliptical geological structure some 80 km long by 40 lan wide, is generally believed to be the result of a meteorite impact. The ore-bodies are located around the rim of the basin and in “off-set” deposits extendmg outward from the basin walls. The Strathcona Mill,built in 1968, processes 8,500 mtpd of nickel ore at 1.7% nickel and 1500 mtpd of copper ore at 6% copper. The valuable minerals in the ore are pentlaudhe and

1147

chalcopyrite. Pyrrhotite is the dominant sulfide in the nickel ore and occurs in a ratio of 5:1 relative to pentlandite. The pyrrhotite is predominantly monoclinic and contains between 0.6 and 1.2% nickel in solid solution. The ore contains small but economicallyrecoverable quantities of cobalt and PGE’s. The rock type is noritic with very amounts of minor floatable rock species such as talc and chlorite (Damjanovic2000). The mill makes a high grade copper concentrate assaying 31% Cu and 0.4% Ni for sale and a nickel concentrate grading 12% Ni and 3% Cu for processing at the Falconbridge smelter located some 70 km southeast of the mill. The nickel ore is ground to (55% - 74 p). Lime is added to the rod d l feed to maintain a pH of 9.2 in the rougher circuit. Potassium isobutyl xanthate is added to the rod mills and the feed boxes at the head of each bank of cells. Dowltoth 25OC is used as fiother. For scavenger flotation, copper sulfate and sulfiuic acid to pH 8 are added to activate the pyrrhotite. The pyrrhotite rejection circuit relies of fine grinding to 80% - 37 pm at pH 10 with lime to recover the pentlandite. The Strathcona flowsheet is shown in Figure 3.

Lime +

cyanide

Nickel Conc Copper Conc

Figure 3 The Strathcona Mill Flotation Circuit

The design of the Slrathcona flowsheet is based on maximizing copper and nickel recoveries and rejecting a high proportion of the pyrrhotite with m h h a l loses of nickel. The following points are key to achievingthis objective:

1148

0 Until recently the high copper ore was floated in a separate circuit. The tails reported to the

nickel circuit. This maximized copper recovery and m h h k e d the copper content in the nickel circuit. With decreasing tonnages being treated, the copper ore is now blended with nickel ore as it fed fom the 6ne ore bins to the rod mills. The rate of addition is controlled by the resulting head grade and copper recovery has been maintained. 0 Pentlandite is floated in the rougher cells in the presence of all the pyrrhotite. The pyrrhotite

acts as an oxygen sink and inhibits the oxidation of the pentlandite. Furthermore, according to a recent study, there is an interaction between patlandite and pyrrhotite whereby dixanthogen is depleted fom the surface of the pyrrhotite and accumulates on the surface of the pentlandite when the reagentized minerals are mixed together (Bozkurt, Xu and Finch 1998). High pentlandite selectivity is thereby achieved with 67% of the pentlandite and only 7% of the pyrrhotite reporting to the rougher concentrate in these roughers. 0

The secondary rougher concentrate is cleaned in magnetic separators. The non magnetics report to the nickel concentrateand the magnetic faction reports to the pyrrhotite regrind mill

O

The rougher tails are scavenged at pH 8 with sulfiuic acid and copper sulfate as pyrrhotite activators. A long retention time (60minutes) is used in the scavenger banks.

0

High selectivity is achieved in the pyrrhotite rejection circuit by using a combination of a high circulating load to achieve a low redox environment in conjunction with two stages of cleaning. The pH is controlled at 10.5 in this circuit With lime.

0

Copper*ickel separation is accomplished using a saturated lime circuit and cyanide. The lime is sufficient to depress the pentlandite but cyanide is required to remove activating copper ions fiom the pyrrhotite. The primary rougher concentrate is conditioned about 20 minutes with saturated lime @H 12) and cyanide. The pulp then reports to a primary separation stage consisting of a bank of conventional cells (OK 8 m3). The tailings fiom this bank report to 6nal nickel concentrate while the primary separator concentrate reports to a second conditioning step where more lime and cyanide are added. The conditioned feed then reports to one of three flotation columns. The column concentrate is the final copper concenrnte grading 31% Cu and 0.4% Ni. The column tails report back to a bank of conventional flotation cells adjacent to the primary separators.

The overall pentlandite Ipyrrhotite selectivity at Strathcona is high with recoveries of 95 percent of the pentlandite while 75 percent of the pyrrhotite and 97% of the rock is rejected (Wells etal. 1997). The Clarabelle Mill INCO’s Clarabelle Mill is located on the southern rim of the Sudbury basin - about 40 km south of Falconbridge’s S t r a t h c o ~Mill. The mill processes about 8 million mt/year of a copper nickel ore grading 1.6% Cu and 1.5% Ni, with a pyrrhotite to pentlandite ratio of 6.3:1. The ore contains a small amount of PGE’s. Both platinum and palladium occur in equal proportions (about 0.6 ppm each). The mill produces about 3200 mtpd of a bulk copper-nickel concentrate grading about 21% Cu+Ni which is pumped to a smelter feed preparation plant about 2 km away. Ruck and pyrrhotite tailings are pumped separatelyto a tailings area located 5 km west of the mill. The pyrrhotite tails are impounded under water in the center of the tailings area and the rock tails are used to build the peripheral dams.

The circuit has undergone a number of changes since it was built in 1971 with a major expansion completed in 1991 by the addition of a 9.75m (32ft) diameter SAG mill and 48 x 38m3

1149

flotation cells to increase the capacity fiom 29,OOOmtpd to 36,OOOmtpd. This expansion enabled INCO to maintah production while shutting down two other mills (Frood Stobie and Copper Cliff>.Two new flash fiunaces were built to smelt the bulk copper nickel concentrate. In 2001, a number of flowsheet enhancements were made to improve nickel recoveries while at the same t h e , increasing the level of pyrrhotite rejection. At the time of Writing, some of these changes are still under construCtion. The new flowsheet is shown in Figure 4 below.

The current Clarabelle Mill processes about 8 million mt/year of ore fiom which about lO0,OOO mtlyear of nickel is produced. The ore is ground to 55% passing 74pm a pulp in which lime is added to pH 9.2. The ground pulp reports to 52 one-metre diameter by two-metre long, wet drum magnetic separators. The nonmagnetic &action (about 7580% of the cyclone overflow) reports to six parallel lines of eight 38 m3 flotation cells. Potassium amyi xanthate and Unifroth 250 CM are the respective collector and fiother used. The concentrate fiom the fkst pair of cells h each line (“A” bank) reports directly to final product. The concentrate from the second pair of cells ( “ B bank) report to a regnnd mill. The pulp is reground to 85% passing 3 8 p and floated in two pardel h e s of 8 m3 flotation cells with the concentrate reporting to final product. The concentrate from the last two pairs of cells in each mainflotation line (“CD scavenger banks) are thickened, reground to 85% passing 38 pm and treated in a two-stage, scavenger cleaner circuit equipped with 2.3 m3 Denver flotation cells. The magnetic fiaction is ground to 85% - 74pn and reports to a two stage cleaning circuit also equipped with 2.3 m3Denver flotation cells.

Magnetic Discard

Figure 4 The Clarabelle Mill Flotation Circuit A novel feature of the Clarabelle Mill is the use of sodium sulfite and triethylenetetramine (TETA) to depress pyrrhotite. TETA is an extremely powerful chelating agent with high selectivity towards copper and nickel ions. When added to a pulp, it removes these ions fiom the surface of the pyrrhotite, effitively rendering the pyrrhotite un-floatable (Kerret al., 1989). The

1150

depressing effect of TETA is further enhanced with the use of sodium sulfite (Kelebek et al., 1995). The exact role that sulfite plays in enhancing the effectiveness of TETA has been the subject of numerous papers. To this day though, its role remains unclear. Its effectiveness however, is indisputable. One of the attractive features of the use of TETA is the simplicity of its use. It is added, along with sulfite in the grinding circuit. Xanthate is added to the flotatim cells. Reagent dosages are typically 0.1 to 0.5 g/kg of both TETA and sulfite. The simplicity of the TETNsulfite process is in marked contrast to other pyrrhotite rejection schemes that have been proposed and extensively tested on INCO’s Sudbwy area ores. The cyanide process (Wells et al., 1981) used large quantities of cyanide (-0.9 g/k) to depress pyrrhotite. Pentlandite floatabaity could be maintained with large doses of xanthate (-0.5 &). Selectivity was maintained by controlling the redox in the -400 mV range with the cyanide, Pentlandite pyrrhotite selectivity was excellent. The process was not used, primarily because of environmental concerns about the effits of large quantities of cyanide in the tailings area Another, more complex process, used a two-stage conditioning step to depress pentlandite. The stage used SO2and air while the second stage used a lime aeration step (Agar 1984). In the subsequent “reverse” flotation step, pentlandite reported to tails while pyrrhotite and chalcopyrite reported to the concentrate. These two minerals were separated in a subsequent flotation step with cyanide. Although this process was extensively tested in a 500 mtpd test circuit, it was never implemented - in part because of the complexity of the process but also because of the discovery of TETA as a highly effective and simple pyrrhotite depressing agent (Marticarena et al., 1994). h t

The TETNsulfite chemistry is used in a l l three pyrrhotite rejection circuits at Clarabelle Mill. In the magnetic pyrrhotite circuit, the new “ B cleaner and scavenger cleaner circuits which both reject hexagonal pyrrhotite. An example of the effectivenessof sulfite and “ETA on the selective flotation of pentlandite and TETA is shown in Figure 5 below (Xuet al., 2000).

Po Recovery (%)

Figure 5 Example of Effect of Solfite and TETA on PentlanditeRyrrhotite Separation Other features of the Clarabellecircuit are: 0

Xanthate is added to the grinding mills to enhance the recovery of plalinum.

1151

Copper sulfate and sulfuric acid are to be added to the scavenger cells to enhance P pyrrhotite flotation P The mill uses 100% recycle water fiom the tailings empoundment area As precipitation substantially exceeds evaporation in the Sudbury area of northern Ontario, a quantity of water equal to the net precipitation is discharged fiom the tailings area into the environmtmt after being treated with lime to pH 10.5 in a wastewater treatment plant.

Overall pentlandte recovery at Clarabelle is 91% while rejecting 82% of the pyrrhotite and 99% of the rock to tailings.

Noril’sk Mill No discussion on nickel flotation practice would be complete Without mention of the operations of Noril’sk Nickel RAO. Noril’sk Nickel has two principal subsidiaries. The Noril’sk Mining Company based in the Taymyr Peninsula in eastern Siberia and the Kola Metals and Mining Company based in the Kola Peninsular at the northwestern tip of Europe.However, it is estimated that 90% of the nickel output form Noril’sk Nickel comes fiom the T a p y r Peninsula. In 1998, Norilsk Nickel produced 19%of the worlds refined nickel production, 20% of the platinum and 50-66% of the worlds palladium (Bond 2001) Regrettably, despite the importance played by Noril’sk Nickel in terms of volume of nickel output, very little information about current plant practice is available in the western literature. The Noril’sk concentrator treats disseminated ore from the Noril’sk-1 deposit, copper ores fiom the Komsomolsky and Octyabrsky mines and “rich ore” from the Tainakh mines. What follows is a brief descriptionof the treatment of dissemiuated ore at the Norilsk concentrator. The disseminated ore contains about 0.5% Ni, 0.7% Cu and 7 ppm of PGMs. This ore is clearly mined primarily for its PGM content- The ore is ground to 4530% &us 74 microns. A recent innovation is the use of Knelson-48 concentrators to recover PGMs form the coarsely ground feed prior to flotation. A high degree of concentration of the PGMs (up to 4 kg/t) is achieved in these units. Platinum recoveries in the gravity concentrate are about 45% and losses to tails are reduced by almost 66% (Blagodath, 1998).

The Knelson concentrator tailings report to rougher flotation where iso-amyl xanthate is used. The rougher concentrate reports to a copper nickel separation stage in which a combination of lime and steam (60 - 70 “C) is used to depress the nickel (Gladishev, 2001) ROCK REJECTION The following section describes milling practice in nickel sulfide mills where the primary focus is the rejection of ultramafic rock species while attempting to maximize the recovery of nickel. Thompson Mill - Birchtree Ore

As mentioned above, the Thompson mill alternatively batches Birchtree ore With Thompson Ore. The flotation circuit fir Birchtree ore is the same as shown in Figure 2. The only difference is that a finer grind is used for the Birchtree ore (12% + 150 pn). The head grade of Birchtree ore is 1.7% Ni, 0.1% Cu With a Po/Pn ratio of 4.4. Problematic rock species include talc, chlorite and serpentine.

1152

Soda ash is the only reagent used for pynhotite rejection. Maintaining the pH at 10 in the rougher scavenga circuit achieves an amazing 85% pyrrhotite rejection while overall pe&miite recovery is 86%. Control of the problematic rock species is achieved by adding 0.5 gkg carboxy-methylcellulose (CMC). The preferred type has a mid-range viscosity. A novel aspect in the mill is that the CMC is added as a dry powder to the rod mill discharge without any loss in effectiveness. The MgO content of the concentrate averages 14%, which requires that it be blended with the low MgO bearing concentrate fiom Thompson ore before smelting. The effects of serpentine on pentlandite flotation were demonstratedby Edwards in pioneering work using electrophoreticmobility on Pipe Lake (Thompson) ore where he was able to show that the positively charged serpentine adsorbed on and depressed the negatively charged pentlandite (Edwards, Kipkie and Agar 1980). A further dramatic illustration of the negative influence of serpentine on pentlandite recovery was reported by Mani (1997). The dosage of as little as 30% by weight of serpentine to a ptlandite ore in a laboratory mill reduced the pentlandite recovery fiom above 80% to below 30% (Figure 6). The effect was shown to be caused by the serpatine slime coating the pentlandite. The effect was reversible by using much st~allowerfioths and 4 times the retention time in the cell, as well as with the use of CMC.

Pf

K

Serpentine Content of Feed (K)

Figure 6 Effect of serpentine on pentlandite recovery (Mani 1997) Raglan Mill Falconbridge’s, Raglan concentrator -is located on the northan edge of the Ungava Peninsula in Northern Quebec. It was commissioned in 1997 and was designed to treat 800,OOO mt/year of ore. However, recent improvements in the AG circuit have increased throughput levels to 1 million mt/year. (Langlois and Holmes, 2001). The mill processes ore fiom both an open pit and underground operation. The ore is ultramafic with significant concentrations of serpenthe and smaller but still significant quautities of talc and chlorite. The ore grade is 3.0% Ni, 0.76% Cu with a favourable PoPn ratio of 1.5. The mill produces 27,000 mt/year of nickel in concentrate

1153

The ore is ground in a 7.3 m (24-ft) AG mill. The oversize fiom the screened discharge reports to a crusher before returning to the AG mill. The screen undersize reports to a 4.3 x 6.4 m (14 x 2 1 4 ) ball mill operahg closed circuit with cyclones. The Ps0for the grinding circuit is 65 p. The rougher flotation circuit consists of 8 OK28 cells. Initially, all concentrate fiom the roughers reported to two stages of column cleauhg. However, because the concentrate fiom the first rougher cell was shown to be at or above final grade, it was directed to final concentrate. The remaining rougher concenlrateis cleaned in rectangular MinnovEx column cell (2m x 6m x 13m). Cleaner tails are scavenged in a bank of 4 OK28 cells. Initially, the roughers outperformed expectations and produced a much higher mnc grade than design (9%Ni vs 4%). The regrind however, was found to worsen the performance of the cleaner cells - likely due to overgrindingand was consequentlyleft off-line. With the introduction of a new ore source in 1998, the talc content of the mill feed increased significantly. This required the addition of CMC to depress the talc. Initially it was added to the cleaners only but talc flotation in the roughers overwhelmed the cleaners. So addition points fbr CMC were added to the rougher feed conditioning tank and to several points down the bank This improved the performance significantly although results were still not as good as with the non-talc ore. (Note, the benefits of staged addition of CMC were also noted at INCO's now decommissioned, Shebandowan Mill in northern Ontario (Agar et al., 1987). To reduce shipping charges, the mill increased the concentrate grade fiom 16 to 18% Ni in 1998. Plant tests had demonstrated that a high grade concentrate 16 -18% Ni could be pulled off the first cell by maximizing fioth depth and using lower airflows. Further enhancements of the concentrate fiom the first rougher cell were made by adding fioth washing. The second move was to divert the high grade 2"dcleaner tails to the 1" cleaner feed, thereby bypassing the cleaner scavengers where it was being downgraded (Figure 7).

Figure 7. The Raglan Mill Flotation Circuit

1154

Together these changes achieved the objective of 18%Ni grade with minimal loss in recovery. Overall pentlandite recovery at Raglan is 87% while rejecting 62% of the pyrrhotite and 97% of the rock. Mt Keith Mill The Mt. Keith Mill is located 630 km north-east of Perth in the Goldfields region of Western Australia. The mill, which is owned and operated by Western Mining Corporation, was commissioned in 1994. The orebody is a low-grade (0.58% Ni) disseminated nickel sulfide resource hosted in dunite, a serpentinized ultramafic rock. Mineralization consists of 3 -5% sulfides, predominantly pentlandite with lesser millerite, violarite and chalcopyrite. The copper content of the ore is less than 0.1%. The mill was originally designed to process 6.5 million muyear of ore and produce 28,000 muyear of nickel in concentrate. It was expanded in 1997 to a capacity of 10.5 million muyear of ore generating 42,000 muyear of nickel in concentrate (Francis and George 1998).

Ore fiom the open-pit is processed through two parallel grinding and primary flotation circuits or modules with common equipment used for cleaner and recleauer flotation and concentrate dewatering. The high Viscosity of the slurry is due to a cbrysotile component in the ore and requires low slurry densities in the grinding circuit (about 50%). After SAG and ball milling to a PW of 150 pn, the slurry passes through a unique two-stage deslimhg operation prior to flotation. Desliming rejects the -6pm fiaction (about 15% of the weight of mill feed) resulting in a major saving in flotation reagents and a halving of the downstream flotation requirement. Desliming involves two-stage cycloning using 300 x 4-inch and 848 x 2-inch Mozley cyclones. The small spigots (6.5 mm) on the polyurethane cyclones are protected fiom plugging by tramp material by inter-stage screening and fiom wear by ceramic inserts (Figure 8). The deslime cyclone overflow is subjected to flotation in a 5m by 4m by 14m high flotation column. Column tails are rejected to final tails. The cyclone underflow passes to the main flotation circuit. The rougher circuit in each module consists of 6x 100 Outokumpu tank cells as roughers and nine x 92 m3Wemco induced draft cells for scavenging. Each module also has two x 92 m Wemco cells and six x 42 m cells for cleaner scavenging. The rougher concentrate fiom the two parallel circuits are combined and report to two stages of cleaning using OK 38 and OK 8 cells. The reagents used in the circuit are sodium ethyl xanthate and H405 fiother. Copper sulfate is used as an activator. Guar Gum and CMC are used intermitteotly for talc and gangue control. Flotation is carried out in highly saline water (up to 80 gpl TDS) without any apparent negative effect on flotation recovery. Overall nickel recovery is between 60-70% depending on ore type, in a concentrate grading 19% Ni with 7.5%MgO.

1155

From Primary Grinding

Rash Rotaiim cell

Final Nickel Concentrate

Figure 8 Mt Keith Flotation Circuit A new feature in the Mt Keith circuit is the incorporation of a flash flotation cell in the grinding circuit. The OK-50 flash cell has had a very positive effect on reducing losses of fine pentlandite in the tails by capturing them before they become over-ground and oxidized. The use of flash flotation cells has become commonplace in the South African PGM mills where they have had a very positive effect on PGM recoveries.

SUMMARY Nickel flotation circuits in milling operations in Canada, Australia and Russia have been described. The need to achieve high concentrate grades and recoveries has challenged the creativity of nickel mineral processors around the world In recent years, there have been major advances in the efficiency of both pyrrhotite and rock rejection to achieve these objectives. There is a relationship between the ease of pyrrhotite rejection and the copper content of the ore. In ores with low copper heads, pyrrhotite can be effectively rejected using soda ash alone. As the copper content increases,more aggressive strategies are required to strip the activating copper ions off the surface of the pyrrhotite. Cyanide or the combination of TETA and sulfitehave been found to be effective strategies. Carboxymethyl cellulose is widely used in ultramafic ores to control the flotation of the hydrophobic rock species. There is a growing momtmtum to use flash flotation cells in the grinding circuits to maximize the recovery of pentlandite and PGE's.

1156

ACKNOWLEDGEMENTS

The author Wishes to thank Inco Ltd for permission to publish this paper. I would further like to thank Dr. Peter Wells fbr reviewing this paper and sharing his knowledge and expertise with me.

REF’ERENCES: Agar, G. 1991. Flotation of Chalcopyrite, Pentlandite and Pyrrhotite Ores, Int. Journal of Mineral Processing, 33, 1-19, Agar, G., Canadian Patent No. 1,238,430, Dec 19,1984 Agar, G., G.H. Styles, B. Lyons, W.B. IGpkk, 1987, Selection of Rock Depressants based on Laboratory Kinetic Studies, CIM Bul., vol. 80 (No. 907): 45-51 A l m k R, 1988, The Character and Occurrace of Primary Resources Available to the Nickel Industry, Extractive Metallurgy of Nickel and Cobalt, TMS The Metallurgical Society, ed. By Tyroler, G.P. and Landolt, C.A. Bacon, W., A. Dalvi and M. Parker. 2000. Nickel Outlook - 2000 to 2010, Mining Millenium 2000. Blagodath, Y.V., AA. Tatsenko, B.A. Zaltharov, and G.RPogosyants, 1998, Development of Gravity Concentration Technology at Norilsk Combine, Russian Journal of Non-Ferrous Metals, V0139, NO.10,5-8 Bond, A.R. and R.M. Levine, 2001, Noril’sk Nickel and Russian Platinum-Group Metals Production, Post Soviet Geography and Economics, 42, No. 2,77-104 Bozkurt, V Z. Xu, J.A. Finch. 1998. Pentlandite / Pyrrhotite Interaction and Xanthate Adsorption, Int. Journal of Mineral Processing 52, 203 - 214 Damjanovk, B., 2000, Canadian Milling Practice, Special Volume 49, Canadian Institute of Mining Metallurgy and Petroleum, Edwards, C. R, W.B. IGpkk, G.E Agar, 1980, The Effect of Slime Coatings of the Serpentine Minerals, Chrysotile and Lizardite, on Pentlandite Flotation. International Journal of Mineral Processing, Vol7,33-42 Francis, R, C. George 1998, The Design, Expansion and Operating Features of the Mt. Keith Nickel Concentrator, Mineral Processing and Hydrometallurgv Plant Design:World’s Best Practice, ed. by Jackson, N,; Dunlop, G.; Cameron, P. Gladishev, A., Y. Salaikh and J. Rubinstein, 2001, Trends in Flotation Equipment Developing, Froth FlotatiodDissolved Air Flotation bridging the Gap, United Engineering Foundation Inc. Conference. Kelebek, S., P.F. Wells and G.W. Heinrich 1996, Metal Ion Activation of Monoclinic Pyrrhotite &om Nickel - Copper Ores using EDTA Extraction, Changing Scopes in Mineral Processing Kelebek, S., P.F. Wells, S. Fekete, M.J. Burrows and D.F. Suarez 1995, Canadian Patent No. 2,151,316 June 8, 1995 Kerr, A.N., D. Liechti, D. Pelland and M.A. Mdcorena 1989, Canadian Patent No. 1,330,125 September 28,1989 Langlois, P., J. Holmes, 2001. Process Development at the Raglan Concentrator, Canadian Mineral Processing Conference.

1157

Mani, M., M. Xu, P. Q u h , R Stratton-Qawley, 1997, The Effect of Ultramafic Mineralogy on Pentlandite Flotation, Processing of Complex &s, 63 -76 Marticarena, M., G. Hill, AN. Ken, D. Liechti, D.A. Pelland. 1994. INCO Develops New Pyrrhotite Depressant, Innovations in Minerul Processing 5-21 Wells, P.F., S. Kelebek, M.J. Burrows. D.F. Suarez. 1997. pyrrhotite Rejection at Falconbridge’s Strathcona Mill. Processing ofCompltx Ores, 51 -62 Wells, P.F., G.E. Agar, KO. Reynolds 1981, Canadian Patent. No. 1,156,384 March 16, 1981 Xu, M, P. Quinn, G. Robertson and S. Wilson 2000, Development of a Two-Stage Middlings c i t at INCOS Clambelle MU,cunudiun Minerul Processors 3.TdConference,

1158

Nonsulfide Flotation Technology and Plant Practice J. Miller,’ B. Tippin,2and R. Pruet?

INTRODUCTION Minerals are often placed in categories according to mineral characteristics rather than chemical composition. Common definitions used in the industry are “metallic minerals,” “nonmetallic minerals,” and “mineral fuels.” The common metallic minerals are the mineral sulfides, such as pyrite and chalcocite, and the native metals, gold and copper. The mineral fuels include coal and oil sands. The nonmetallic mineral term is essentially synonymouswith the term “industrial minerals.” Probably the best definition of industrial minerals is “any rock, mineral, or other naturally occurring substance of economic value, exclusive of metallic ores, mineral fuels, and gemstones.” For a comprehensive overview of industrial minerals, the reader is referred to the SME book Industrial Minerals and Rocks, the 5th and 6th editions. In this regard, the category of nonsulfide minerals includes industrial minerals, energy minerals, and nonsulfide metallic minerals. There are hundreds of nonsulfide mineral flotation plants throughout the United States, and most have a plant design that depends upon the specific ore characteristics and the market specifications for their product. There is no common thread of technology between the design and processing of different nonsulfide minerals, or even between processing minerals of the same kind. Although most nonsulfide processing plants utilize flotation, other mineral separation techniques are often necessary to yield a marketable product. Typical beneficiation techniques used in conjunction with flotation include gravity separation, magnetic separation and chemical leaching. Sometimes several different flotation systems are used in the same processing plant, such as the flotation of mica, quartz and feldspar from pegmatite ores. Usually there is a primary material produced at a nonsulfide processing plant with other mineral products sold as by-product material. Often the economics of a facility is dependent upon the by-products and sometimes the by-products become the most profitable commodity produced at a plant. The stone, sand and gravel industry is the exception and is not included in t h s chapter. This nonsulfide mining industry is huge, exceeding all the metals industry and fuel industry tonnage combined. These mining operations do not beneficiate their material by flotationnor do they normally use any other mineral separation process. This industry mines a one-component deposit that only requires removal of fines and clays to produce a saleable product. The field of nonsulfide minerals is so varied and so large that to encompass the topic in this chapter is impractical, if not impossible. Some nonsulfide minerals are processed in very large tonnages in numerous plants, and some nonsulfide minerals are processed in only one or two small tonnage plants. These latter plants are so unique and so specialized that including them in this chapter would be of little use to most mineral engineers. Therefore the best this chapter can accomplish is to show the great diversity of nonsulfide flotation technology and provide several examples of actual nonsulfide plant practice.

1 University of Utah, Salt Lake City, Utah 2 North Carolina State University, Asheville, North Carolina 3 Imerys Pigments and Additives Group, Sandersville, Georgia

1159

General Aspects of Nonsulfide Flotation The dimensions of nonsulfide flotation technology extend in many different directions, as might be expected from the diversity of the mineral classes, which include soluble salt minerals (potash, borax, and trona), semi-soluble salt minerals (phosphate minerals, fluorite, calcite, and barite), insoluble oxides/silicate minerals (mica, quartz, and feldspar), and energy minerals (coal and oil sands). As a consequence, some flotation separations are accomplished from saturated brine, while other separations are achieved in solutions of rather low ionic strength. Also it is worth noting that certain nonsulfide minerals are naturally hydrophobic, such as talc, graphite, and coal. But in general, the nonsulfide minerals are hydrophilic and require relatively lugh levels of collector addition on the order of 1 lb/ton to establish a hydrophobic surface state. Further, the nonsulfide collectors generally are anionic or cationic surfactants, having hydrocarbon chains of ten carbon atoms or greater. In many instances, the collector is sufficiently insoluble that a distinct collector phase is present in the system, existing as a dispersion or as a collector colloid. This situation further complicates the analysis of nonsulfide flotation chemistry, and in view of the foregoing, it is evident that the flotation chemistry is distinctly different from the flotation chemistry of sulfide minerals. Nonsulfide flotation technology also differs from sulfide flotation technology with respect to particle size. In many nonsulfide systems, but not all, flotation is accomplished with deslimed feed and a particle size that extends up to several millimeters in diameter (coal, phosphate, and potash). On the other hand, some nonsulfide flotation systems involve flotation of micron-size particles (taconite and kaolin). With respect to flotation rate, again, the variation is great, with long retention times (even one hour) required for the flotation of impurities (anatase, etc.) fromkaolin, whereas rather rapid flotation (a few minutes of retention time) is required for phosphate flotation. Advances in flotation technology, including nonsulfide flotation, have been documented in a recent publication (Parekh and Miller, 1999). The Nonsulfide Mineral Industry Before the design and operation of a nonsulfide plant, it is imperative to understand the industry. In almost all cases, the plant design parameters and processes used for beneficiation are dependent upon a specific marketable product and the quality of the mineral in the deposit. The importance of this fact cannot be under estimated. Quartz is a prime example of this fact. In a sand (quartz) deposit there are several potential economic markets. If the deposit is located near an urban area, the sand quality will probably be acceptable for general construction, such as concrete, and find a market within 10 to 15 miles. Processing (crushing and sizing) cost would be minimal and the selling price of the sand would be in the range of 3 to 5 $/ton plus about 5 $/ton delivery by truck to a local market. If the sand could be processed by flotation and magnetic separation to yield glass sand quality, then the sand could sell for about 15 to 20 $/ton if there was a glass manufacturer within 100 to 150 miles. Yet, if the sand was of exceptional quality and could be processed to yield high-purity quartz, the product could be shipped anywhere in the world. Of course, other factors are involved in determining if the deposit is economic, but usually the overriding factors are (a) product quality, (b) market conditions and (c) location. Product Quality. The nonsulfide minerals industry usually produces a commercial product that remains a “mineral”, whereas the final product of the metals industry normally is a “metal”, produced from a mineral. Further, it should be noted that nonsulfide minerals have highly variable natural characteristics. Not all mica found in nature has equal economic value. The “aspect ratio” (diameter to thickness of the particles) of a mica crystal structure cannot be altered by processing, yet ths mineral characteristic dictates its value, Two mica conceneates can be exactly equal with 98%pure mica, yet a concentrate with a bulk density of 10 lb/cubic foot would make an excellent filler material at a very good price but the other mica concentrate may have a bulk density of 25 Iblcubic foot with limited, if any value, in the marketplace.

1160

Other mineral characteristics influencing the value and marketability of nonsulfide process plant design include the following.

* Particle Size (not necessarily for liberation and beneficiation but for market specifications):

* *

*

*

If the ore requires a 100-mesh grind for liberation during processing, but the size specification is 28x 100 mesh for the marketable product, then the material cannot be sold. If an ore is coarse liberating, after beneficiation the concentrate may have to be fine ground for sale. Particle Shape: The shape (round or angular, platy, fibrous, etc.) of a mineral can define a market and the mineral value. This is shown in the previous example of mica. Color: Color becomes a critical factor in some nonsulfide mineral markets. Kaolin, marble, limestone, mica, and quartz become more valuable with greater whiteness. Whiteness may depend not on the mineral concentration in a product but on the impurities contained withn the crystal structure, and processing cannot modify these minerals. Chemical Composition: Feldspar and other nonsulfide minerals can have a variable chemical composition. Feldspar contains calcium, sodium and potassium in the crystal. A feldspar from a deposit can be classified as a soda (Na) feldspar or a potash(K) feldspar depending upon the amount of these elements in the structure. This will define what commercial market is potentially available and at what price. Concentrate Grade: The grade of a specific mineral in a concentrate from a nonsulfide processing plant can define the market potential and the sales price. A concentrate with a grade of 30% tantalum can be sold at 80 $/lb but if it is upgraded further, the sales price may be much higher. The minimum potash grade for the market is 60% K,O, but if the grade drops to 58.8% K,O there is a substantial reduction in the sales price - if the material can be sold at all!

Market Demands. The size of nonsulfide mineral markets tends to be more restrictive and limited than those in the metals industry. This is because the markets are more specialized and directly connected with the consumer. The use of fillers (nonsulfide minerals such as calcite, mica or kaolinite) is expanding with the increased use of plastic in automobiles and with an expanding economy. The market for fertilizers (potash and phosphate) is cyclic because of weather and world economics. Barite is used for drilling mud in exploration and its demand is dependent upon the price and supply of world oil. The design and operation of nonsulfide plants must be flexible to adjust to the variations in market demands. These variable market demands may be tonnage (such as happens in the case of the demand for fertilizer in a wet weather in the planting season.) or quality (such as when more restrictive particle size specifications as specified for construction sand used in the “Super Highway” concrete). New industrial technologies can alter an industrial mineral market significantly and rapidly. Location. Location of the deposit is important because most nonsulfide products are high-volume, low-price material; therefore, transportation cost will have a major impact in the economics of the deposit. For example, even an excellent high-grade ore deposit of feldspar using a proven processing technique may not be economic because of its location. The ore deposit is likely to be discovered in a remote geographical area, but the glass manufacturing plant is probably located near an urban location. Transportation of bulk products is costly and often exceeds the cost ofmining and processing the ore. Salt produced from the Utah Great Salt Lake can be loaded on a unit train for less than $5/ton and yet the cost of shipping to a caustic chlorine producer in Nevada may be greater than $15lton. Consistency. An important factor in marketing most industrial minerals is to maintain product consistency. This is especially critical when attempting to enter a new market with an untried product. For example, glass manufacturers produce large batches using a specific mix of materials, (quartz, feldspar, limestone, soda ash and gypsum) and companies are very resistant to changing the mix. Specifications for feed mix limits are usually not the same for all glass manufacturers. Variations in the quality of the glass components can cause serious and expensive problems in their operations. Suppliers of industrial minerals must establish reliability with a client before sale contracts are made.

1161

Evaluation of Deposits and Products. The design, construction and operation of an industrial mineral facility are considerably different from a metallic processing plant. This is because the mineral products are market driven with specialized specifications and usually result in long-term contracts with the client. The metallic industry (copper, iron, gold, zinc, etc.) products are usually sold at published prices on the open market plus the processing technology is relatively uniform across the entire industry. A critical market study and product evaluationmust be carried out before spending capital money to develop a nonsulfide property. The mportance of a comprehensive market study cannot be over emphasized. The study should be one of the first things to be considered in developing a new property and should include (a) a survey of the potential use of the mineral, (b) size of the market, (c) price of the commodity, (d) transportation, (e) estimating the profitability or margin on sales, (0 potential by-products and (g) competition, not only competition from other companies but also competition from other types of material. In determining the capital and operating costs of a new industrial mineral plant, marketing is again important and the operating cost must include a sales staff or group. Process technology and marketing cannot be separated in the industrial minerals industry. Product quality and consistency are important in the evaluation of a new deposit and in the development of a new product from an existing plant. A market evaluation of the product is usually necessary in order to obtain a sales contract, and usually this requires a substantial quantity ofmaterial to test. Laboratory or batch testing to design the industrial mineral plant is always required but a large quantity of the material usually is needed to provide to the potential client for testing in their operation. This can be as little as several hundred pounds but often is 20 to 50 tons of test material. Continuous pilot plant testwork is normal when starting an industrial minerals operation not just to obtain engineering design and process criteria for constructing the plant but also to obtain acceptance of the material to the clients.

MICA Mica is a complex hydrous aluminosilicate with a varying chemical composition. There are several types of mica. Muscovite mica is commercially mined because of its unique characteristics, especially its color, laminated structure and insulating power. Phologopite mica is dark stained and is used in large volumes for sound deadening and fillers where color is not an issue. Biotite mica has a high bulk density and contains significant quantities of iron in the crystal (can be almost black in color), therefore, it has very limited use and generally are not considered of economic value. Mica uses include paint fillers, plastic extender,joint compound, surface coating, insulation board and numerous other industrial applications. Important mineral characteristics of the mica that dictate the market potential and selling price are (a) aspect ratio or bulk density, (b) color or whiteness, and (c) particle size. Most commercial applications require that the product be finely ground. Product grinding can be either wet or dry. Analytically determining mica is difficult because it is not a true mineral but a family of minerals with varying chemical composition. Potassium is the predominate element in muscovite but because of the variability of its composition, chemical analysis cannot be used to determine the mica content of samples. Usually commercial mica is not sold on chemical analysis but on specifications of its physical properties. Methods used in industry to evaluate mica flotation efficiency are vanning, magnetic separation and bulk density. Mica is easily floated because of its flake like crystal structure and can be floated in two systems: (a) an acid system using a cationic collector and (b) a basic system using an anionic collector. Both systems are used industrially but the acid flotation system is more common. Flotation selectivity in both systems is quite good and the flotation concentrates usually contain 90% mica or higher. Mica flotation is fast and produces a thick heavy froth. Further information on mica and the processing thereof can be found in the literature (Chapman 1983; Tanner, 1994; Tippin et al., 1999).

1162

Feed Preparation After crushing, either wet ball and/or rod mills are used to grind the ore in preparation for flotation. Grinding conditions should minimize the formation of fines whch are a detriment to the flotation process. Most ores are coarse liberating. Both the acid and the basic mica flotation system require desliming prior to flotation. One or two stages of desliming (at about 150 mesh) are generally required. The first stage of desliming follows grinding and removes primary slimes in the ore. The second stage of desliming follows scrubbing and removes the secondary slimes. Scrubbing and desliming is not absolutely necessary for mica flotation but the fines (clays) left in the flotation feed can cause a significant increase in reagent consumption and adversely effect selectivity. It is usually advantageous to precede flotation with a scrubbing step to clean the mineral surfaces and remove slimes from the surface of the mineral. The intensity, retention time and pulp density of scrubbing is not critical to flotation. Typically the ore is scrubbed for 3 to 8 minutes at 55 to 70% solids. Depending on the weathered characteristics of the feed, scrubbing can immediately follow grinding or after the first stage of desliming. Flotation in both the acid system (using amine) and the basic system (using petroleum sulfonate) have a conditioning step. The basic flotation system usually requires more conditioning time that the acid system. Here the pH is set, the collector added and the pulp density adjusted for flotation. Consumption of reagents for pH adjustment is dependent upon the minerals in the ore. High solids-high intensity and a long retention time during conditioning are not required and can actually be detrimental. Neither system is highly sensitive to the conditioning step. However, there is a saying that flotation is easy if the feed preparation is correct. In the acid system, conditioning time is very short, less than two minutes are needed. Since the cationic bond to the mica surface is not strong and the amine is slime sensitive, too much conditioning in this system can be detrimental. Typical amine collector usage is 0.3 to .06 lb/ton. The collector can be added as a chloride or an acetate amine. The pH for flotation ranges between 3.0 to 4.0 using sulfiu-ic acid. Conditioning time in the basic system is five to ten minutes in two stages at a pH between 8.0 and 9.0. Usually caustic is used to adjust the pH but other chemicals have been used. Flotation at the ore’s natural pH may be possible but this is usually less efficient and not worth the small savings of the caustic needed to adjust the pH. Typical reagent consumption is 1.O to 1.5 lb/ton of petroleum sulfonate. Mica Flotation The acidic and basic flotation systems are quite similar. The major difference is that the basic system has a tendency to float more of the gangue minerals, especially in the fine sues and the concentrate is more difficult to clean. The acid circuit has a more tenacious froth that has a tendency to entrap fines. The decision as to which system is used, sometimes is a matter of preference and experience. A typical system consists of a rougher (5 to 8 minutes), one cleaner (3 to 5 minutes) and a scavenger (3 to 5 minutes). Depending upon the ore and the market specifications, sometimes a second cleaner is used and the scavenger may not be necessary. The middling from the cleaner circuit are recycled back into the rougher circuit. Pulp density in the flotation feed (conditioner discharge) ranges between 20% and 30% solids. A frother, such a MIBC or a commercial fi-other reagent is normally added (0.1 to 0.2 lblton) to the rougher cell to loosen up the froth. Small frother dosages may also be added to the cleaner cells. Plant Practice Many industrial minerals are produced in multi-component plants, including mica. Mica is often a co-product or a by-product in an ore processing plant. Few mica-bearing deposits have the content and quality to be economic by selling only mica unless there are unique market conditions. It is common to have mica, quartz and feldspar in a deposit and to have a plant that beneficiates all three minerals commercially.

1163

Mica is beneficiated both by gravity and by flotation techniques in plant practice. Coarse mica can be concentrated by flotation but it is more economic to use gravity methods. Spirals, tables and screens are most commonly used. Because of the unique shape of the particles, pre-concentration of mica from an ore is sometimes done using a hydraulic classifier. Several stages of gravity concentration steps may be used to yield the grade of mica desired. Gravity techniques can be effective down to about 40 mesh. Very often the tailings andor middlings from the gravity circuit are ground to minus 30-mesh and sent to the flotation circuit. Gravity methods also have an advantage in that the mica product does not have any residual reagent on the surface. Some markets require a chemical free mica product. Typically the ore is ground and sized to 30 x 150 mesh before flotation. Due to the unique shape of the mineral, selective grinding naturally occurs during grinding. Because of this, mica plants often have spirals installed ahead or in the grinding circuit to recover the coarse mica. Gravity concentration is more effective and less costly than flotation for plus 10 mesh mica. Below 150 mesh, flotation selectivity is reduced as the platy nature of the mineral is reduced. Mica flotation is effective down to about 150 mesh but below this size the concentrate grade diminishes. Flotation selectivity in the finer sizes is usually better in an acid system than a basic system. Standard mechanical flotation cells are more common than column cells in mica flotation. Depending on the beneficiation process and the sales market, sometimes the dry mica flotation concentrate is further upgraded by magnetic separation or screening. To meet size specifications, the final mica concentrate from the beneficiationplant is often wet or dry ground to ultra-fine sizes before shipment. Wet ground mica is a higher value product than the dry ground product. The coarse mica concentrate may be combined with the fine mica concentrate or each can be sold separately depending on the commercial market. Typically mica products sold from a beneficiation plant are 95% mica or higher.

Q U ARTZIFELDSPAR Quartz and feldspar are among the most abundant minerals on earth, occurring almost everywhere. Depending on the chemical composition, physical characterization and ore location, these minerals have similar commercial usage. They also generally occur in the same deposit and are often recovered jointly as a feldspathic sand (a mixture of quartz and feldspar). Therefore, this section of the chapter considers quartz/feldspar flotation technology together for the production of (a) feldspathic sand and (b) the quartz/feldspar concentrates. The market for feldspar and quartz minerals is quite large, and includes glass, glass fiber, whiteware, tile, dinnerware, paint, adhesives andplastics. Glass and glass fiber are the largest tonnage market (68%). The quartz and feldspar products are sold under private contracts to individual users and there is no published national price. In general the quartz products can be sold at a price between $5 and $120 per ton and the feldspar products between $40 and $200 per ton. Feldspathic sand prices vary between $20 and $40 per ton. The wide range of prices for various usage results in transportation becoming an important factor in the economics of these two minerals The important characteristics for commercial feldspathic sand and quartzlfeldspar concentrates are (a) chemical purity, (b) whiteness or color, and (c) particle size. Also, the sodium to potassium ratio is an important factor in selling feldspar products. Flotation is the primary recovery process for these two minerals, excluding construction material usage. However, there is no known flotation plant in the United States that recovers only feldspar. Feldspar is recovered as a feldspathic sand (a mixture of quartz and feldspar) or as a result of a feldspadquartz separation that yields both a feldspar concentrate and a quartz concentrate. Both flotation systems use a first stage of flotation to remove impurities. The option of producing a feldspathic sand or quadfeldspar concentrates is both technical and economic. The technical factors are mineral liberation size between the quartz and feldspar crystals and the potassium:sodium ratio of the feldspar. The economic factors are product use, market tonnage, and end user location. If a plant is designed to produce quartz and feldspar concentrates, it is a simple matter to change the process to produce feldspathic sand. However, converting a

1164

feldspathic sand plant to one that produces individual concentrates is more difficult and requires capital investment for an additional flotation circuit. Further information on quartz and feldspar and the processing thereof can be found in the literature (Rogers and Neal, 1983; Kauffman and Van Dyk, 1994; Zdanczyk and Linkous, 1994; Ibrahim, 2002; Tippin et al., 1999). Feed Preparation After crushing, either wet ball andlor rod mills are used to grind the ore in preparation for flotation. Typically the ore is ground to 30-mesh to be applicable for most markets. Most ores are coarse liberating, up to 20-mesh, but the glass industry does not want any material over 30 mesh. Even though many applications for the minerals have a maximum iron specification grinding is done in a standard rodhall mill with metal liners. The grinding mills are in closed circuit with cyclones, screens or hydraulic classifiers. After grinding, one or two stages ofcyclone desliming (at about 150 mesh) are required. The first stage of desliming follows grinding and removes primary slimes in the ore. The second stage of desliming follows scrubbing and removes the secondary slimes. Desliming is not absolutely necessary but the fines (clays) left in the flotation feed can cause a significant increase in reagent consumption and adversely effect selectivity. It is usually advantageous to precede flotation with a scrubbing step to clean the mineral surfaces and remove slimes from the mineral surface, This improves flotation selectivity. The intensity, retention time and pulp density of scrubbing is not critical to flotation. Typically the ore is scrubbed for 3 to 8 minutes at 55 to 70% solids. Depending on the weathered characteristics of the feed, scrubbing can immediately follow grinding or after the first stage of desliming. Flotation Often the mined ore contains varying amounts of mica. Then the first beneficiation step is the removal of the mica by gravity methods andlor flotation. (see mica section of t h s chapter for details). Depending on the quantity and quality of the mineral, the mica will be sold as a by-product or rejected as a tailing waste. If a mica flotation is used, the tailings from the cells are dewatered to remove any excess reagents and increase the pulp density for the next flotation step (iron removal). To remove the contaminating iron minerals from the ore, a separate flotation system is used. A conditioning step precedes iron flotation where the pH is adjusted to about 3.5 with sulfuric acid, the collector added and the pulp density adjusted. High solids-hgh intensity and a long retention time during conditioning are not required as the iron flotation is rapid and effective. Conditioning time is short, less than two minutes at 60% to 65% solids. Typical collector usage is 0.3 to 0.6 lb/ton. Consumption of acid is dependent upon the minerals in the ore. Although the flotation time for iron flotation may be as much as five minutes, very little material (less the 2%) is actually removed in this process step. Sometimes it appears that the froth doesn’t contain any solids. However, this flotation step is very important to remove as much of the iron mineral as possible. Because of the small amount of material being floated, a frother is added to form a manageable froth. The iron concentrate is discarded as tailings waste and the underflow from the iron flot is dewatered. If a feldspathic sand is the desired product, the flotation cell underflow becomes a final product. If a feldspar product and a quartz product are desired, then another flotation step is necessary to separate the feldspar away from the quartz. A conditioning step precedes quartz-feldspar flotation step where the pH is adjusted to about 1.5 with hydrofluoric acid, an amine collector is added and the pulp density adjusted. Conditioning time is short, less than two minutes at 50% to 65% solids. Typical collector usage is 0.2 to 0.5 lblton. If needed, a frother is added to the feed immediately ahead of flotation. The feldspadquartz separation requires the use of the fluoride ion to activate the feldspar and to depress the quartz. The fluoride ion is a very effective depressant, and an excellent separation of quartz (cell underflow) from feldspar (cell froth) is obtained. Flotation time is three to five minutes. Sometimes the initial feldspar concentrate is cleaned once to recover any entrained quartz in the froth and to improve recovery. The middlings are usually recycled back to the rougher feed. A scavenger flotation circuit is almost never required.

1165

Plant Practice Hydrofluoric acid is a hazardous and dangerous acid that requires special safety conditions in the plant. Fluoride discharge fromprocess water is strictly controlled and limited. Several research studies have been made to find a replacement system. These have shown limited success but none have been used in actual plant operation. Flotation is the preferred method to remove the major amount of iron-contaminatingmaterial from feldspathic sand, quartz, and feldspar products. Dry magnetic separation is used as a final iron removal step on the flotation concentrates before shipment. However, one plant in Egypt has successfully replaced the iron flotation circuit with magnetic separation in a quartz-feldspar operation. This technique is not practiced in the United States. Many industrial minerals are produced in multi-component plants, including quartz and feldspar. It is common to have mica, quartz, and feldspar contained in a deposit and to have a plant that beneficiates all these minerals commercially. The flowsheet in Figure 1 shows the processing strategy for such a multi-component plant. Other saleable products from industrial minerals plants producing quartz/feldspar include crushed stone, sand, clay, spodumene and beryl. In the multi-component plants, it is often difficult to determine the primary product and the by-products. The design and operation of multi-component plants must remain flexible to alter the flowsheets to meet changing market situations. The feldspathic sand and feldspar concentrates from various mineral processing plants differ in chemical composition and physical characteristics. The sodium and potassium content make some flotation concentrates better suited for certain markets. After flotation, the feldspathic sand, quartz and/or feldspar are filtered, dried and stored for shipment. Most operators use hi-intensity dry magnetic separators as the final clean-up step before bulk shipment or bagging. Depending upon the market and specifications, most of the quartzlfeldspar produced is sold as ground material. Grinding is usually below 325 mesh and accomplished in ceramic ball mills to prevent iron contamination. The quartz products are sold with at least 99.5% silica. Typical iron specifications for quartz, feldspar and feldspathic sand are Insulating Fiberglass Glass Industry Ceramics Industry

Can be greater than 0.080% Fe,O, Typically between 0.080 and 0.065% Fe,O, Typically less then 0.05% Fe,O,

High Purity Quartz Technology. Quartz and/or silica sand (SiO,) has enjoyed a multitude of industrial uses. Today’s technology has enabled the mass production of quartz to various levels of purity, depending on the end use. With the growth of the modem electronics and computer industry there has arisen a demand for very high purity natural quartz. Very few deposits have the purity of the natural quartz to meet the market specifications in these industries. The Spruce Pine area of North Carolina produces over 90% of the high purity quartz used in the country. High Pure quartz (Quintas Grade) is produced primarily from physical beneficiation techniques, including grinding, scrubbing, flotation, and magnetic separation. Ultra High Pure quartz (Iota Grade), typically produced from the High Pure products, receives additional beneficiation, commonly in the form of intense acid leaching and chlorinationprocessing. Details of the technology are proprietary, and very little data is available. Because of the advances in many of the high tech industries, the use and price of the high pure and ultrahigh pure grades have grown accordingly, thus creating a higher level of interest in finding suitable raw material deposits of quartz. Therefore, the evaluation process, although highly objective in the initial phases, can become rather complex and time consuming. Once the material has been initially qualified and performance determined, h a c e trials must be conducted with potential customers to certify the raw material for their use.

1166

Figure 1. Schematic flowsheet - Typical Alaskite flotation. Mica, quartz, and feldspar flotation.

Because of the variety of end uses for the quartz products, it is difficult to assign exact product specifications,and quality is ultimatelydefined by performance. However, some typical guidelines can be applied to initially evaluate the quartz and/or to control the process. Typical limits on contaminants are as follows. High Pure Quartz (PPM) Fe Na Mg

20 ppm max. 100ppmmax. 30ppmmax.

Ultra High Pure Quartz (PPM) Fe 1.0 ppmmax. Na 2.0ppmmax. Mg 0.5 ppm max.

A1 Ca Ti

200ppmmax. 50ppmmax. 5 ppm max.

K Li

50-80 ppm max. 2 ppm max.

A1 Ca Ti

15 ppmmax. 2.0ppmmax. 1.0 ppmmax.

K Li

2.0ppmmax. 0.5 ppmmax

It should be noted that the above do not include all of the elements that could possibly disqualify a quartz, but are instead listed as a basis for further evaluation. It should also be noted that while many describe the purity of quartz in terms of % SiO,, this could be somewhat misleading, especially in the Ultra High Pure grades. A quartz product could be described as containing 99.999% SiO,, but still contain trace elements on a ppm level that could hinder the performance of the quartz for end use in electronics or computer applications. Once chemistry has been determined, actual performance must be evaluated to fully define the materials commercial viability, and this phase of the evaluation is almost always market specific. A common method is to produce a fused quartz product (e.g. crucibles) and examine the product for any flaws. This requires the production of bulk concentrates, sometimes ten tons or more, of the final quartz product on a pilot-plant scale under very controlled and noncontaminating conditions. The High Pure processing plants are costly to construct, operate, and maintain due to the special considerations required to prevent contamination. Because the High Pure markets are relatively low volume operations, unit costs for processing can be hgh. A High Pure quartz plant almost always produces both High Pure quartz and Ultra-High Pure quartz. Development of markets for the High Pure grades is almost essential for the economic survival of a high-grade quartz facility. The production of a High Pure grade quartz material with a price of $500 to $1500 per ton can create a steady cash flow while the Ultra High Pure markets are being developed. High Pure quartz is also a much larger market as compared to the Ultra-High Pure quartz, consisting of many low-volume niche uses. The Ultra-High Pure grade price ($2000 to over $10,000 per ton) can then create a nice profitability once fully developed and proven. Most plants producing High Pure quartz material also produce the lower grade quartz products also.

KAOLIN Kaolin is a white-to-near-white clay that contains kaolin minerals. The majority of kaolin particles have an equivalent spherical diameter less than two microns in products sold for coating pigments. Kaolin products sold as fimctional fillers or extenders have particle diameters less than 45 microns. Kaolin minerals are hydrous aluminumsilicates that include the minerals kaolinite, dickite, nacrite and halloysite. Kaolinite, Al,Si,O,(OH),, is the kaolin mineral of most importance for application as a white pigment in paper coating, paper filling, paint, plastics, rubber and a long list of other applications. Iron substitution for aluminumup to 2% on a number basis causes defects in the kaolinite crystal structure that is measured by the Hinckley Index. Kaolin deposits are classified as primary and secondary. Primary deposits are formed through the in situ kaolinization of rock-forming minerals, namely feldspar and mica. Secondary deposits are formed through the erosion, transport and deposition of kaolin mineral-bearing sediments. Sedimentary kaolin is a kaolin deposit hosted in sedimentary rock that contains kaolin minerals that

1168

are detrital, authigenic or both. Kaolin ore bodies show a continuum of kaolin mineral content that ranges from 10% to >95% by weight. The continuum of kaolin ore types with increasing kaolin mineral contents includes hydrothermal kaolin, residual kaolin, kaolinitic sandstone, sandy kaolin, and kaolin clay. The kaolinite particles in these ores have a particle size distribution that is different from deposit to deposit and show a particle size range from two millimeters to sub-micron. Flotation may be used to concentrate kaolin from ores that have high gangue content such as hydrothermal or residual kaolin. Quartz, mica, feldspar and tourmaline are common gangue minerals in primary kaolins having low kaolin mineral content. Flotation can be used to concentrate kaolin minerals when the size of gangue minerals is forty-five to five microns and separation from kaolin minerals by gravity settling or hydrocylones is not practical. Kaolin products are divided into several grades based on particle size, brightness or value-added processing. Standard grades are specified by percentage of particles by weight below two microns equivalent spherical diameter (ESD) as measured by a gravitational settling method for particle sizing. A #2 coating clay has 80 wt. % of particles below two microns, a #1 coating clay has 90 wt. % of particles below two microns, and a fine #1 coating clay has 95 wt. % of particles below two microns. Regular kaolin grades have GE brightness between 84 and 89, and high brightness grades have GE brightness greater than 89. GE Brightness is a measure of visible light reflectance made at 457 nm relative to a MgO standard on dried and pulverized kaolin compressed into a cake with 30 psi. Highly processed grades include delaminated,calcined, aggregated or engineered. Other important properties for kaolin pigments are shade, residue content measured on a 325-mesh screen, hard and dark particle counts, abrasion, moisture content, pH of slurry, Brookfield (low-shear) viscosity of clay-water slurry, and Hercules (high-shear ) viscosity of clay-water slurry. Flotation or selective flocculation may be used to beneficiate any kaolin grade that has a brightness or shade specification. Common minerals that harm kaolin brightness are iron oxides (hematite, goethite), iron sulfides (pyrite and marcasite), titania minerals (anatase, pseudorutile, rutile, brookite) and mica. Most kaolin operations located in Georgia use flotation or selective flocculation to remove titania minerals, which are difficult to remove by size fractionation or by high-intensity, wet magnetic separation (HIMS). Feed Preparation Kaolin used for pigments and additives is wet processed. Unless the kaolin is extracted from the wall of an open pit using a monitor, blunging is used, where kaolin ore is made-down into slurry with water in a range between 35% and 70% solids (Murray, 1980). The kaolin blunger is a tank with an impeller located near the bottom to provide mechanical work input. Chemicals are introduced during the blunging process to disperse particles liberated during blunging. There are two facets to dispersion of kaolin particles: pH adjustment to induce a negative charge on both the edge and face of kaolinite particles to enable deflocculation, and the use of a dispersant chemical to increase the negative charge on mineral surfaces and thereby increase particle-particle repulsion. The choice of pH used to disperse the kaolin slurry is partly dependent on downstream processes utilized for beneficiation. Generally pH is adjusted to around 6.5 or 7.0 using sodium carbonate (soda ash) for clay slips transported by pipeline and beneficiated using a HIMS or selective flocculation. When flotation is used for beneficiation, the pH may be adjusted to between 8.0 and 9.5 using ammonia or sodium hydroxide. Primary dispersants used during blunging may include sodium silicate, sodium hexametaphosphate, sodium polyacrylate or a combination of dispersants. Dispersant dose depends on several parameters such as particle size, surface area, organic matter content and mineral content. Dispersant doses typically range from 0.1 wt. % to 0.5 wt. % on a dry basis. Care must be taken to select a dispersant chemistry that does not interfere with selection or extraction of impurities.

1169

Flotation Froth flotation is used to recover kaolinite in rejects from kaolin-refining operations. The rejects from hydro separators used for processing primary kaolins in Cornwall and Devon may contain upwards of 60% kaolinite (Pemberton, 1989). The kaolinite from these rejects can be recovered using flotation with chemistry selective for kaolinite that pennits a separation from quartz, feldspar, tounnaline and other nonclay minerals. Reverse froth flotation was introduced to kaolin processing in the early 1960s (Greene and Duke, 1962; Grounds, 1964). Flotation of kaolin can operate with or without carriers. Greene and Duke (1962) used a ground limestone carrier. Their approach to kaolin flotation comprised adding the carrier to a dispersed kaolin slip, conditioning the kaolin-ground limestone slurry with a collector and frothing agent, and floating out the ground limestone with attached anatase. Cundy (1976) avoided the use of carriers by dispersing the kaolin slip at pH 9, introducing divalent ions such as calcium or barium to aggregate anatase particles, introducing a fatty acid such as oleic acid to coat the anatase particle surfaces, and aggregating the conditioned slip in the presence of air bubbles to separate of the anatase. Yoon and Hilderbrand (1986) developed a noncarrier flotation process using hydroxamate collectors instead of fatty acids. Selective Flocculation The basis of selective flocculation is to flocculate one or more minerals and leave the other minerals in a dispersed state to permit separation by gravitational settling in a thickener . Selective flocculation of kaolin can entail flocculation of kaolinite and removal of impurities in the overflow from a thickener or entail flocculation of specific impurities such as anatase and collecting a kaolin product from the thickener overflow. A common selective flocculation process used by Georgia producers flocculates anatase out of a dispersed kaolin slip (Maynard et al., 1969). The process developed by Maynard (1968, 1974) requires that the blunged kaolin slip at pH 6.5 and dosed with sodium hexametaphosphate and sodium silicate,be conditioned with sodium chloride, and aged to aggregate anatase. A water-soluble, strongly anionic, high molecular weight (>lo6) polyacrylamide polymer is then added which acts to form anatase flocs that settle in a thickener. Other developments in selective flocculation were made by Shi (1 986), who used an ammonium salt conditioning agent, and Behl et al. (1996) who used calcium salt and fatty acid to accomplish the titania aggregation. Garforth et al. (2000) further improved the process by adding sodium polyacrylate, a dispersant, in conjunction with addition of the high molecular weight polyacrylamide polymer flocculating agent. Although counter-intuitive for a flocculation process, this patented technology promotes the formation and separation of anatase- and mica-bearing flocs from the dispersed kaolin slip to increase floc settling rate and increase overall process recovery. Plant Practice Kaolin flotation and selective floccuIation are practiced at wet process plants that produce white pigments. These plants are typically very high volume and have multiple process steps. After the kaolin has been extracted from the ore body and made down into slurry, the kaolin is transported by pipeline to a degritting operation. Grit is defined as particles >45 microns in size. Degritting is the removal of grit particles from the kaolin slip using dragboxes, spiral classifiers, hydrocyclones, and screens. Flotation and selective flocculation typically follow degritting or HIMS. HIMS is used to remove iron-bearingparamagnetic impurities that are harmful to brightness, HIMS may be used before or after flotation or selective flocculation in a kaolin plant. Some kaolin operations using gray kaolin containing organic matter have an ozone process before HIMS. Ozone gas is reacted with the organic matter in the kaolin in contact towers. The ozone breaks complex organic molecules into simpler and lower molecular weight organic compounds that can be removed with fbrther processing and that do not absorb visible light, thereby increasing kaolin brightness.

1170

Fractionation is the size classification of particles using solid bowl decanter centrifuges to make particle size cuts in the clay that are appropriate to grade. Fractionation may occur after titania impurities and other discoloring impurities have been removed. The coarse centrifuge rejects from fractionation that contain kaolinite stacks are typically ground using an attrition grinder. The attrition grinder delaminates the coarse kaolinite stacks into component platelets. Following fractionation, kaolin is reduce-acid leached. The kaolin slip is acidified with sulfuric acid to pH in the range pH 2 to 4; alum may be added to help flocculate the kaolinite particles, and sodium hyposulfite is added to reduce ferric to ferrous iron. Ferrous iron absorbs less visible light in the blue wavelengths, thereby increasing brightness. The flocculated kaolin slip is then dewatered using a rotary vacuum filter to remove soluble salts that are inherent in the crude or were added during processing. The filter cake from the rotary vacuum filter is then dispersed again in a repulper with pH increased to a range between pH 6.5 and 7.5 using soda ash and a dispersant such as sodium polyacrylate. The dispersed kaolin slip typically gets some chemical treatment with biocide to prevent bacteria and mold growth that may spoil the product. Higher biocide dosages are needed where process chemicals such as fatty acids are used. The clay slurry is firther dewatered to a slurry shipping solids near 70% or is dried for shipment as bulk or bagged product. Drying is typically done using a spray dryer that yields a kaolin product form of small beads.

COAL The processing and/or disposal of fine coal is of significant importance to the coal industry. In the US about half of the more than 1 billion tons of coal produced annually is processed in preparation plants. More than 300 plants in 20 states process as much as 3700 tph in one case, average plant capacity about 700 tph (Dorsett, 1995). Most of the clean coal production comes from coarse and intermediate particle sizes by gravity separation at a low cost ($1.50-2.50/ton). It is reported that 65% of the plants use heavy media. In almost one third of the preparation plants in the US, froth flotation is used for fine coal recovery (-0.5 mm), which amounts to about 10% of the total clean coal production. In this regard, it is estimated that 30 to 50 million tons of fine coal are discarded each year. Fine coal recovery is impeded by higher preparation costs ($4.50-7.00/ton) and dewatering costs ($2.006.OO/ton). Flotation Coal, which consists of complex hydrocarbon molecules derived from the anaerobic alteration of plant matter, tends to be naturally hydrophobic and is associated with other inorganic mineral matter, including clay, quartz, and gypsum. Sulfur occurs as sulfate sulfur, pyrite sulfur, and organic sulfur chemically incorporated into the complex hydrocarbon molecules. Coal is ranked as anthracite, bituminous, sub-bituminous, and lignite, according to its carbon content, volatile matter, and energy value. Not unexpectedly, the coal hydrophobicity as revealed by contact angle and flotation response varies with coal rank and its petrographic composition,maceral type (Aplan, 1976; Arnold and Aplan, 1989; Laskowski, 2001). In fact, on this basis, resin macerals can be separated from other coal macerals by flotation (Miller and Ye, 1989; Miller et al., 1993). Of course the natural hydrophobic surface state of coal can be altered by oxidation processes which lead to an increase in polar oxygen groups at the surface, making flotation more difficult. Also, the naturally hydrophobic surface state is influenced by solution chemistry phenomena. For example, the adsorption of water-soluble macromolecules (organic colloids) such as dextrin can cause the coal surface to become hydrophilic (Haung et al., 1978; Miller et al., 1984). Such phenomena are the basis for the reverse flotation process in which pyrite is floated from coal (Miller, 1973; Miller and Lin, 1986). Plant Practice The general strategy is to float coal from the mineral matter, relying on the natural hydrophobicity of coal. Typically the feed, minus 0.5 mm, is conditioned with an oily collector, fuel oil, at a dosage of about 1 .O lb/ton as necessary and pH 6 to 8. For unoxidized high-rank bituminous coal, collector

1171

addition may not be necessary, and flotation can be achieved with just a frother. In most other cases, an oily collector is required. The critical issue then is distribution of the collector and the selective wetting of the coal particle surfaces. In this regard the use of surfactants for dispersiodemulsification facilitates the distribution of the fuel oil collector and enhances its wetting of the coal particle surfaces (Yu et al., 1990). As a result, even difficult-to-float subbituminous coals can be floated if the oil emulsion is prepared properly. Of course frother is required to develop a froth phase to carry the coal, and this is most commonly accomplished with MIBC (methyl isobutyl carbinol) at a level of addition which varies from 0.1 to 0.5 lblton (Aplan, 1976). Although a number of dispersantsldepressants can be used to depress the mineral matter, generally this practice is not necessary. However, for feed material containing a high content of clay, contamination of the clean coal product can be a problem, and, in this regard, a dispersant might be added. Alternately, the feed can be deslimed or a split feed strategy can be practiced, in which case the slimes (minus 100mesh) are separated by classifying cyclones and floated in a separate flotation circuit. The use of column flotation cells has become quite popular in the coal industry. With the use of wash water, a clean coal product containing 5 to 6% ash can be achieved, as compared to 10 to 12% ash as might be achieved in single-stage flotation with mechanical flotation cells. In this way, improved coarse coal yield can be achieved with columns and the yield optimized to meet ash specifications. The key is to maximize the incremental ash. A most recent review of coal flotation technology and practice is given by the set of five papers published in Advances in Flotation Technology (Parekh and Miller, 1999).

PHOSPHATE World fertilizer production from phosphate continues to be a crucial factor for the efficient growth of crops to feed the peoples of the earth. Phosphate is an essential ingredient of fertilizer, and the world production of phosphate rock is around 140 million tons per annum with reserves of 12,000 million tons. It should be noted that the U.S. is the world’s largest producer ofphosphate with most of its -40 million tons per annum coming from the vast sedimentary deposits of Central Florida. In the US. phosphate rock was mined by 10 firms in 4 states, and upgraded into an estimated 39.7 million tons of marketable product valued at $1 billion, f.0.b. mine site in 2000. Florida and North Carolina accounted for 85% of the total domestic output, with the remainder produced in southeastern Idaho and northeastern Utah. However high-grade phosphate ores, particularly those containing few contaminants, are being progressively depleted and production costs are increasing. As discussed at the International Engineering Foundation Conferences on Phosphate Beneficiation, December 1993 and 1998, Palm Coast, Florida, and more recently at the Engineering Foundation Conference Beneficiation of Phosphate 111, December 2001, St. Pete Beach, Florida, many technological problems must be solved if we are to continue to produce phosphate rock at our current rate of consumption. Proceedings of the first two conferences have been published (El-Shall et al., 1993; Zhang et al., 1999). The 2001 conference is scheduled to be published in 2002. There are two main types of phosphate deposits, igneous and sedimentary, whlch have widely differing mineralogical, textural and chemical characteristics. Igneous rock is often associated with carbonates and/or alkalic intrusions and is generally quite crystalline and low in grade. Most phosphate rock production comes from sedimentarymarinephosphates. Sedimentary deposits account for the majority (87%) of the phosphate reserves and resources. Consequently it is not surprising that 80% of the world production of phosphate rock comes from the mining and concentration of sedimentary phosphate ores. Most of the sedimentary phosphates are earthy in appearance, cryptocrystalline, and associated with gangue minerals which may include appreciable amounts of dolomite, quartz, hematite, aluminosilicates and clay minerals. The phosphate content in currently mined rocks can range from over 40% to below 5% P,O,. The biggest technical problem facing the Florida operations is lower grade resources with an increasing content of MgO. As the mining operations progress to the south, the overburden and clay slimes increase significantly. Disposal of the slimes is a serious environmental problem that has yet to be solved.

1172

Flotation The chemistry fundamentals of fatty acid flotation of phosphate minerals have been studied extensively, and this research provides a basis for the understanding of industrial practice. However, there is still much unknown regarding the collector adsorption and oil wetting phenomena which occur in the actual flotation of phosphate rock at plant sites. Important issues include water quality as well as structure and composition of phosphate minerals. For example, the significance of composition is revealed in oleate adsorption studies for different apatites as shown in Table 1 (Yehia et al., 1993). Table 1. Comparison of Adsorption Density, Contact Angle, and Flotation Recovery for Apatite Minerals with Oleate as Collector at 5x10“ M and pH 10. Mineral Fluorapatite Carbonatapatite Hydroxylapatite Chlorapatite

Adsorption Density, mg/m2

Surface Coverage

Contact Angle, degrees

Flotation Recovery, %

1.08 0.86 0.52 0.34

0.49 0.39 0.24 0.20

55 43 38 No Attachment

88.7 71.3 62.5 53.2

It is evident that the adsorption density, hydrophobic surface state, and the flotation recovery decrease in the order fluorapatite, carbonatapatite, hydroxylapatite, and chlorapatite. Certainly it is to be expected that similar variation would be expected for the flotation of phosphate rock in plant operations. In any event, it has been well-established that oleate will chemisorb at the apatite surface, forming a calcium oleate surface species (Lu et al., 1998). Unlike other semi-soluble salt minerals, the adsorption generally is limited to monolayer coverage, or less, apparently due to solubility considerations and the calcium site density at the apatite surface. Nevertheless, it seems that this chemisorbed fatty acid at the apatite surface provides sufficient hydrophobicity to allow for the spreading of fuel oil typically used in plant operations. Plant Practice Conventional plant practice for the siliceous sedimentary phosphate resources involves the double float process, also known as the “Crago process”, as shown in Figure 2. This flotation process is used around the world, and has been practiced for at least 60 years. A number of useful references exist which discuss the flotation technology for phosphate (Wiegel, 1999a and 1999b; Zhang, 1993; Davis and Hood, 1993; MaHeesh et al., 1996; Lu et al., 1997and 1998). The phosphate ore is classified into four size fractions, the pebble (+16 mesh), the coarse (16x35 mesh), the fine (35x150 mesh), and the slime (-150 mesh). The pebble product, containing over 28% P,O,, is one of the final products. The slime, containing a significant amount ofultra fine clay mineral particles, is discharged to tailings due to difficulties in the separation of phosphate minerals from the very fine clay particles. Typically the coarse and fine size fractions, containing substantial quartz, are fed to the plant for beneficiation by froth flotation with an insoluble oily collector consisting of a mixture of fatty acid and fuel oil. The coarse and fine feeds are conditioned separately at 70% solids with fatty acid and fuel oil at alkaline pH. After conditioning the phosphate minerals (such as francolite, a cryptocrystalline carbonate fluorapatite) are generally floated in mechanical cells. The flotation concentrates are finther upgraded by reverse amine flotation to remove the entrained quartz particles after being scrubbed with sulfuric acid to remove the adsorbed fatty acid and fuel oil. Both mechanical and column flotation cells are used in the Florida operations. Interestingly the columns have been used for rougher flotation of coarse feed. In most applications in other industries, columns have been used for fine particle flotation. High solids conditioning (>70% by weight) with fatty acidlfuel oil mixtures is used extensively in the double float process as practiced for sedimentary ores to improve the phosphate recovery and reduce collector consumption. Some research has been carried out to investigate the phenomena involved and to optimize the conditioning parameters. It has been found that the phosphate recovery

1173

Phosphate R o c k R u n of M i n e O r e

Class ification

Pebble ( + I 6 mesh) (final product)

High Solids (>70%) Conditioning

Coarse

Fine

( 1 6 x 3 5 mesh)

(35x150 mesh)

High Solids (170%) Conditioning

4

Coarse Fatty Acid Flotation

Slimes (-150 mesh) (to tailing pond)

’

I Acid Scrub

Acid Scrub

Coarse Amine Flotation

Fine A m i n e Flotation

A

I

*

Coarse Co n c e ntr at e

T ailing s

Figure 2. Flowsheet for the processing of Florida phosphate rock.

1174

Fine Concentrate

increases as the energy input for conditioning increases up to a critical value and then decreases as the energy is increased further. At a higher level of energy input, recovery decreases due to the generation of slimes under extremely strong agitation. In view ofthe mechanisms of anionic collector adsorption, reducing the water content during conditioning has the dual effect of increasing the concentration of collector in the aqueous phase and reducing the quantity of activating ions. As is the case for most systems that involve insoluble oily collectors, the distribution of the collector is an important consideration. In addition to high solids conditioning, the use of certain nonionic surfactants significantly increase the recovery of coarse phosphate apparently due to improved froth stability and improved dispersiodattachmenthpreading of the fatty acidfuel oil collector. Recent research relates the improved flotation with high solids conditioning to the selective wetting of phosphate minerals by the fatty acidfuel oil mixture (Lu et al., 1997). Results from an experimental study with a high-speed video system clearly show the preferential transfer of oil to a francolite surface (carbonate fluorapatite) and helps to explain the high solids conditioning mechanism. When a quartz particle with an attached oil drop interacts with a francolite particle, the oil drop will spread at the francolite interface. When these two particles are forced apart, a major portion of the oil drop is transferred to the francolite surface. A small portion of the oil drop may remain at the quartz surface. This procedure happens again and again during high solids conditioning, and the fatty acidfuel oil collector which initially may be at quartz surfaces will ultimately be transferred to the surfaces of francolite particles. In this way, flotation selectivity is improved. Therefore, both recovery and selectivity in phosphate flotation with fatty acidfuel oil are improved by high solids conditioning.

POTASH The world production of potash, which exceeds 30 million tons of equivalent K 2 0 annually, comes primarily from Canada, Russia, and the Ukraine. U.S. production of potash from New Mexico and Utah has diminished significantly because of low grades and high clay content. The important potash ores typically contain sylvite, halite, and insoluble gangue minerals, including clays and carbonates. Other common potash minerals are listed in Table 2. Table 2. Common Potash Minerals (Zandon, 1985) Mineral

Formula

Sylvite Camallite Kainite Langbeinite Leonite Schoenite Polyhalite

KCl KCl.MgC12.6H,0 KC1.MgSO,.3H2O K,SO,.MgSO, K2S0,.MgS0,-4H20 K2S0,.MgS0,.6H20 K2S04~MgS0,.2CaS04.4H20

Specific Gravity

% K,O

1.99 1.60 2.13 2.83 2.25 2.15 2.78

63.17 16.95 18.92 22.70 25.69 23.39 15.62

Flotation is used to separate the potash minerals, particularly sylvite from halite, in about 70% of the operations. Alternately, evaporation and crystallization are used in certain instances. The unique feature of potash flotation is that the separation is accomplished from a saturated brine, typically using long-chain alkyl amines for flotation of the sylvite. Almost all potash flotation plants have a crystallization step to maintain a brine balance and reduce potash losses. In addition to sylvite flotation, other potash minerals that are recovered by flotation include schoenite, carnallite, and langbenite. Flotation In general, soluble salt flotation with both cationic (alkyl amines) and anionic (alkyl sulfates) collectors can be explained based on hydration phenomena at the salt surface (Hancer et al., 2001).

1175

Some salts tend to promote water structure, and these salts, whch are extensively wetted by their saturated brines, cannot be floated with either alkyl amines or alkyl sulfates. Such is the case for halite, NaCl. On the other hand, some salts tend to be water-structure breakers, and these salts, which are not wetted completely by their saturated brine, can be floated with either alkyl amines or with alkyl sulfates. Such is the case for sylvite, KC1. For example, the brine contact angles for sylvite and halite are compared in Table 3 in the absence of collector. In view of the foregoing, it is not unexpected that sylvite flotation will be sensitive to brine composition, and, for example, the presence of magnesium (a structure maker) in the brine depresses the amine flotation of sylvite. Table 3. Advancing Contact Angle Measurements for Saturated Brine at the Surface of KCI and NaCl (NoCollector Present) Salt

Contact Angle, degree

KCl NaCl

7.9 f 0.5 (12.0 f 1.4*)

O*

*Measured on the natural cleavage plane of a single crystal. In addition to the salt properties and the brine composition, the presence of clay slimes is an important factor which can impede flotation (Laskowski, 1994). This is particularly true in the case of sylvite flotation. Generally in potash flotation, the slimes must be blinded or the feed deslimed prior to sylvite flotation with amines. It is well established that amine flotation of sylvite occurs when the amine precipitates from the brine. In this regard, is appears that such a similar event occurs at the sylvite surface, leading to the hydrophobic surface state. Finally, the effect of temperature in sylvite flotation with amine is worth noting. It has been established that higher temperatures result in a decrease in sylvite flotation with amine collectors. Such an effect is even seen in plant operations where the potash recovery decreases during the summer season. This temperature effect might be due to the increased solubility of the amine at higher temperatures. Also, the behavior of KC1 may change with temperature, from a water-structurebreaking salt at lower temperatures to a structure-makmg salt at temperatures above 30°C. Plant Practice One of the major concerns in flotation operations is the presence of slimes and their elimination prior to sylvite flotation from halite. A number of process strategies are used, and these include:

* *

mechanical desliming with classification cyclones flocculation-flotation desliming

In the case of desliming by classification, intensive scrubbing is required, and this treatment may contribute to sylvite loss to the slimes product. To avoid this loss of sylvite by abrasion during scrubbing some operations flocculate the clay slimes and remove them by flotation with low levels of coliector addition prior to the flotation of sylvite (Tippin et al., 1998). For feed containing a low level of clay slimes or for treatment of feed with residual slimes, slime depressants, polymers, can be used to prevent amine adsorption by clay. These polyelectrolytes include both anionic and cationic polymers as well as nonionic polymers, and carboxy methyl cellulose. Considerable savings in collector consumption can be achieved with the use of such polyelectrolytes with improved selectivity in the flotation of sylvite from halite and clay slimes. Finally, it should be noted that frequently potash flotation plants have both fine and coarse flotation circuits, split feed, not unlike the practice in phosphate plants and in some coal plants. In the case of the flotation of coarse sylvite, oil is frequently used to promote the flotation response (Laskowski, 1997). Again, the use of oil is similar to the strategy used in phosphate and coal flotation.

1176

ACKNOWLEDGMENTS Funding of research in the area of nonsulfide flotation chemistry at the University of Utah has been provided by DOE Grant No. DE-FG-03-93-ER143 15.

REFERENCES F. Aplan, 1976. Coal flotation. Flotation, A. M. Gaudin Memorial Volume. SME, 1235-1264. B. J. Arnold and F. F. Aplan, 1989. The hydrophobicity of coal macerals. Fuel, 68: 651-658. S . Behl, M. J. Willis, and R. H. Young, 1996. Method for separating mixture of finely divided minerals. U.S. Patent 5,535,890. G. P. Chapman, 1983. Mica. IndustrialMinerals andRocks, 5thed., Vol. 2. S. J. Leford, ed. AIME, New York, 915-924. E. K. Cundy, 1976. Flotation of fine-grained materials. U S . Patent 3,979,282. B. Davis and G. D. Hood, 1993. Improved recovery of coarse florida phosphate. Mining Engineering. 7: 596. J. Dorsett, 1995. The 1995 prep plant survey adds more sites. Coal. October: 30-41. J . Drelich, J. S. Laskowski, M. Pawli, and S. Veeramasuneni, 1997. Preparation of a coal surface for contact angle measurements. Journal of Adhesion Science. 11 (1 1): 1399-1431. H. El-Shall, R. Moudgil, and R. Wiegel, eds., 1993. Beneficiation of Phosphate - Theory and Practice. SME, Littleton, Colorado. W. L. Garforth, R. J. Pruett, D. L. Archer, J. Yuan, J. J. Garska, and H. V. Brown, 2000. Method for separating mixture of finely divided minerals and product thereof. U.S. Patent 6,068,693. E. W. Greene, and J. B. Duke, 1962. Selective froth flotation of ultrafine materials or slimes. Mining Engineering, 14:51-55. Grounds, A., 1964. Fine-particle treatment by ultraflotation. Mine & Quarry Engineering, March, 128-133. M. Hancer, M. S. Celik, and J. D. Miller, 2001. The significance of interfacial water structure in soluble salt flotation systems. Journal of Colloid and Interface Science. 235: 150-161. H. H. Haung, J. V. Calara. D. L. Bauer, and J. D. Miller, 1978. Adsorption reactions in the depression of coal by organic colloids. Recent Developments in Separation Science, CRC Press Inc., Volume IV, 115-133. S. S. Ibrahim, 2002. Egyptian silica. Industrial Minerals. London, Issue 412 (Jan.): 34-37. R. A. Kauffman and D. Van Dyk, 1994. Feldspars. Industrial Minerals and Rocks, 6th ed., D. S. Carr, ed. AIME, New York, 473-48 1. J. S. Laskowski, 1994. Flotation of potash ores. Reagents for Better Metallurgy. Mulukutla, ed. SME, Littleton, Colorado. J. S. Laskowski, 2001. Coal floation and fine coal utilization. Developments in Mineral Processing, Elsevier, Vol. 14, 368 p. J. S. Laskowski and Q. Wang, 1997. Amine-containing oils as extenders in the flotation of sylvinite ores. Proceedings of the XX IMPC, Aachen, 21-26 September 1997,605-616. Y. Lu, J. Drelich, and J. D. Miller, 1997. Wetting of fiancolite and quartz and its significance in flotation of phosphate. Mineral Engineering. lO(10): 1219-1231. Y. Lu, J. Drelich, and J. D. Miller, 1998. Oleate adsorption at an apatite surface studied by ex-situ FTIR internal reflection spectroscopy. Journal of Colloid and Interface Science. 202: 462-476. Y. Lu, N. Liu, X. Wang, and J. D. Miller, 1998. Improvedphosphate flotation withnonionic polymer. Engineering Foundation Conference, Beneficiation of Phosphate: Second International Conference, Palm Coast, Florida, December 6-1 1. C. Malteesh, P. Somasundaran, and G. A. Gruber, 1996. Fundamentals of oleic acid adsorption on phosphate flotation feed during anionic conditioning. Minerals Metallurgical Processing. 17:17. R. N. Maynard, 1974. Method of rapid differential flocculation of TiO, from kaolin slurries. U.S. Patent 3,857,78 1. R. N. Maynard, N. Millman, and J. Iannicelli, 1969. A method for removing titanium dioxide impurities from kaolin. Clays & Clay Minerals, 17, 59-62.

1177

R. N. Maynard, B. R. Skipper, and N. Millman, 1968. Method of beneficiating clay by removal of titanium impurities. U.S. Patent 3,371,988. J. D. Miller and C. L. Lin, 1986. Characterization of pyrite in products from the reverse flotation of coal. Coal Preparation, 2: 243-261. J. D. Miller, C. L. Lin, and S. S. Chang, 1984. Coadsorption phenomena in the separation of pyrite from coal by reverse flotation. Coal Preparation, 1: 2 1-38. J. D. Miller, Q. Yu, K. Bukka, and Y. Ye, 1993. Analysis of selective resinite flotation from Wasatch Plateau coal by pH control. Coal Preparation, 13: 3 1-51, J. D. Miller and Y. Ye, 1989. Selective flotation of fossil resin from Wasatch Plateau high volatile bituminous coal. Minerals and Metallurgical Processing, SME, May: 87-93. K. J. Miller, 1973. Flotation of pyrite from coal: Pilot plant study. USBM RI 7822. H. H. Murray, 1980. Major kaolin processing developments. International Journal of Mineral Processing, 7,263-274. B. K. Parekh and J. D. Miller, eds. 1999. Advances in Flotation Technology. SME, Littleton, Colorado, Sections 3 and 4, 197-341. M. Pemberton, 1989. Developments in china clay processing. Mine & Quarry, 40-43. C. P. Rogers, Jr., and J. P. Neal, 1983. Feldspar. IndustrialMinerals and Rocks, 5thed., Vol. 1. S. J. Leford, ed. AIME, New York, 709-722. I. C. S. Shi, 1986. Method of beneficiating kaolin clay utilizing ammonium salts. U.S. Patent 4,604,369. J. T. Tanner, Jr., 1994. Mica. Industrial Minerals and Rocks, 6th ed. D. S. Carr, ed. AIME, New York, 693-710. R. B. Tippin, H. L. Huiatt, and D. Butts, 1999. Silicate mineral and potash flotation. Advances in Flotation Technology. B. K. Parekh and J. D. Miller, eds. SME, Littleton, Colorado, 199-211. U.S. Geological Survey, 2001. Mineral Commodity Summaries. US.Government Printing Office, Washington, DC, 190 pp. R. L. Wiegel, 1999a. Phosphate rock beneficiation practice. Advances in Flotation Technology. B. K. Parekh and J. D. Miller, ed. SME, Littleton, Colorado, 213-218. R. Wiegel, 1999b. Phosphate rock beneficiation practice in Florida. Beneficiation of PhosphatesAdvances in Research and Practice. P. Zhang, H. El-Shall, and R. Wiegel, eds. SME, Littleton, Colorado, 271-275. A. Yehia, J. D. Miller, and B. T. Ateya, 1993. Analysis of the adsorption behavior of oleate on some synthetic apatites. Minerals Engineering, Pergamon Press, Great Britain, vol. 6, no. 1, pp. 79-86. R.-H. Yoon, and T. M. Hilderbrand, 1986. Purification of kaolin clay by froth flotation using hydroxamate collectors. U.S. Patent 4,629,556. Q. Yu, Y. Ye, and J. D. Miller, 1990. A study of surfactant/oil emulsions for fine coal flotation. Advances in Fine Particle Processing, Elsevier, 345-355. V. A. Zandon, 1985. Potash. SME Mineral Processing Handbook. Weiss, ed. SME, New York, Section 2 2 . M. J. Zdunczyk and M. Linkous, 1994. Silica. Industrial Minerals and Rocks, 6th ed. D. S. Carr, ed. AIME, New York, 879-891. J. P. Zhang, 1993. Phosphate beneficiation - Trends of the 90s. Beneficiation of Phosphate: Theory and Practice. H. El-Shall, B. M. Moudgil, and R. Wiegel, eds. Society of Mining Engineers, Littleton, Colorado, 426-441. P. Zhang, H. El-Shall, and R. Wiegel, eds., 1999. Beneficiation of Phosphates - Advances in Research and Practice. SME, Littleton, Colorado.

1178

Design of Mechanical Flotation Machines Michael G. Nelson', Frank P. Traczy,?, and Dariusz Lelinski2

ABSTRACT Economics have driven the development of large, mechanical flotation cells with cylindrical tanks. Proper design and scale-up of these larger machines has been based in part on years of experience with successful scale-up of smaller rectangular cells, using geometric criteria. Current scale-up procedures also incorporate hydrodynamic analysis. Improved hydrodynamic analysis is based on several recent technical advancements. These include (1) better understandkg of the influence of power input and mixing on metallurgical performance, (2) use of computational fluid dynamics (CFD) to validate designs before going to the field, and (3) use of new flotation models, based on first principles and empirical data, such as the bubble-surface-area flux (Sb) parameter to model flotation response. This chapter discusses the design of mechanical flotation equipment, including the techmques used for froth removal and adaptations for process control. ACKNOWLEDGEMENTS The authors gratefully acknowledge the assistance of the following individuals in the preparation of this chapter: Jouko Kallionen and Don Foreman, GL&V/Dorr-Oliver; Richard Peaker, Metso Minerals; and Heikki Oravainen, Outokumpu Mintec OY. INTRODUCTION Mechanical flotation machines have been designed in a wide variety of configurations [Srewis, 19961. Machines used for most mineral flotation applications may be designated in one of three categories: 1. Mechanically-agitated,self-aerating tanks, with the rotor near the top of the tank; Mechanically-agitated,externally-aerated tanks, with the rotor near the bottom of the tank; or Non-agitated and externally-aerated columns.

2. 3.

Other types of machines are used in some specialized applications, but those listed above comprise the vast majority of machines in present operation. The first two types of machines are called mechanical cells and are the workhorses of mineral processing. They are often categorized together and called tank cells, circular cells, or some other general name to distinguish them from flotation columns. This chapter discusses in detail the design of such machines. Self-aerating machines are manufactured only by Eimco Process Equipment Co., under the name WEMCO. Externally-aerated machines are manufactured by GL&V/Dorr-Oliver, Inc., Outokumpu Mintec OY, Metso Mmerals (formerly Svedala Industry), and others. The theoretical basis for design of flotation machines was described in detail by Harris, in the A.M. Gaudin Memorial Volume Flotation, published by SME in 1976 manis, 19761. Harris gave an exhaustive discussion of design theory, and provided design details for machines made by 13 I 2

College of Mines and Earth Sciences, University of Utah, Salt Lake City, Utah, U.S.A. EIMCO Process Equipment Co., Salt Lake City, Utah, U.S.A.

1179

manufacturers. In 1999, SME sponsored a symposium entitled “Advances in Flotation Technology,” which included a paper summarizing machine design by Arbiter, as well as papers by representatives of Eimco and Outokumpu [Arbiter, 1999; Weber, et al., 1999; and Jonaitis, 19991. This chapter draws on those papers, and on correspondencewith representatives of Eimco, GL&V/Dorr-Oliver, Outokumpu, and Metso Minerals. The detailed discussion of flotation machine characteristics is based directly on information provided by the respective manufacturers. Not surprisingly, each manufacturer believes its machine to be superior. The statements made by each manufacturer about its own equipment are thus included here without qualification.

T h ~ schapter does not attempt to distinguish one design as superior to others. Rather, it presents methods currently used by major manufacturers, with the intent of offering the reader an understanding of design practice among those who now put flotation machines into operation in concentrators around the world. Each mechanical flotation machine must provide the following functions: 0 0 0 0

0 0

Good adpulp contact, Adequate solids suspension, Good mixing without stagnant zones or short-circuiting, A quiescent zone for froth separation, Adequate froth removal, and Adequate residence time to allow the desired recovery of the valuable constituent.

The residence time is influenced by a machine’s hydrodynamic characteristics, but is also a function of the its volume and the rate at which it is fed. HYDRODYNAMIC CONCEPTS Dimensionless Numbers Hydrodynamic analysis plays an important role in the design of modem flotation machines. For that reason, it is discussed briefly here. Hydrodynamic analysis is a particular application of the technique known as dimensional analysis. Dimensional analysis was developed by a variety of practitioners to allow the design of full-scale machmes or devices on the basis of tests using prototypes of a much smaller size. Zlokarnik [ 19911 gives a brief history and excellent description of the methods. Of the many applications of dimensional analysis, those most familiar to most mineral processing engineers are probably the dimensionless numbers used in the solutions of certain engineering problems. These include the Reynolds number, the Mach number, the Prandtl number, etc. Zlokarnik [ 19911 lists 24 such numbers. Dimensionless numbers are widely used in fluid mechanics. For example, the Reynolds number, usually designated R or NRe,is defined as

where p is the fluid specific gravity, V is the linear flow velocity, L is a characteristic length (often the diameter of the pipe), and p is the fluid viscosity. If the units for the component variables are consistently expressed, the Reynolds number is dimensionless. The Reynolds number is used to characterize the degree of turbulence in a fluid system, with a higher number indicating increased turbulence. The Reynolds number can also be used to perform analysis of scale-up problems. For example, one might ask, “A 1:15 model of a submarine is to be tested in a towing tank containing salt water. If the submarine moves at 12.0 mph, at what velocity should the model be towed for dynamic similarity?' [Giles, 19621 This problem is solved by simply equating the Reynolds number for the model with that for the full-scale machine, thus:

1180

(12Lp)/p = pV(L/lS)/p.

(2)

Because the specific gravity and the viscosity are the same in each case, V is easily determined as 180 mph. The dimensionless numbers used in fluid mechanics may be understood as relating a force that acts on a fluid to the inertia of the fluid. Table 1 summarizes those numbers and the forces they respectively characterize.

Table 1 Dimensionless numbers used in fluid mechanics

Euler Number Reynolds Number Froude Number Cauchy Number Weber Number

Expression in Basic Quantities Ma/PA MdzA M g MafEA Ma/oL

Standard Expression 0V2/D I

1

P W 4

v2fu pv2/E OLV2/O

Expression for Flotation Cells

Quantities Related Inertia - Pressure Force (ND2P)/cLp Inertia - Viscous Force (N2D)/g Inertia - Gravity Force Inertia - Elasticity Force ( N ~ D ~ Inertia ~ ) ~ -~Surface Tension

In the above table, the symbols are defined as follows: A D g M P V p

=area, = impeller diameter, = acceleration due to gravity, =mass, =force, = linear velocity, = absolute viscosity of flotation pulp, = specific gravity, and

a = acceleration, E = bulk modulus of elasticity, L =length, N = impeller speed, p =pressure, p = absolute viscosity, y = surface tension at air/pulp interface, CT = surface tension.

Residence Time The concept of residence time, or retention time, arose in the design of chemical reactors, and is widely applied to sizing of flotation machines. Only a brief theoretical discussion is given here. The flow pattern through a vessel depends on the configuration of the vessel, and the conditions therein. The simplest flow pattern occurs when each particle (molecule) in the fluid spends the same time in the vessel. This is called plug flow. There is no mixing in the vessel, in the flow direction, although lateral mixing may occur. Every particle will exit the vessel at a time z after it enters. The residence time z is given simply by z

where V Iv

=V&,

(3)

= volume of the vessel, and = volumetric flow rate.

The conceptual opposite of plug flow is perfect mixing, in which any input is immediately dispersed uniformly throughout the vessel. These concepts may be understood by considering a near-instantaneous, impulse input to the vessel of a tracer material, whose concentration is then measured over time. In a perfectly-mixed vessel with a nominal residence time z defined as in Equation 3, the concentration C of the tracer measured at time t is

c = cIe(-"), where CI = tracer concentration at time t = 0.

(4)

1181

Figure 1 shows residence time distribution plots for the cases of plug flow (the vertical line) and perfect mixing (the curve). The cases just described are ideal. In practice, stirred tanks depart from perfect mixing for a variety of reasons. Levenspiel [19961 describes three typical examples. Figure 2 shows residence time distribution of a tracer in a vessel that has short-circuiting, as indicated by the sharp, early peak. Figure 3 shows residence time distribution for a vessel with stagnant fluid, or dead space, as indcated by the early curve. Figure 4 shows reasonably good mixing, with exponential decay and the mean residence time approximately equal to the theoretical residence time, z. Figures 1-4 are all based on Levenspiel [ 19961, with permission. Residence time distribution tests are usually conducted by adding a tracer material to the operating vessel. The tracer may be a chemical that will not react in the process, a radioactive compound, or

1

c

0 1

08 06

c/q

0)

2

04

3

02 0

0

2

1

3

zobs

7

Time

tlz I

Figure 1 Plug flow and perfect-mixing

Figure 2 Short-circuiting

Figure 3 Stagnant fluid or dead space

Figure 4 Reasonably good flow

some other material that can be detected in samples taken. In any tracer test, care must be taken to gather and analyze the data correctly. Furthermore, because of the difficulties often encountered in making these measurements, it is tempting to draw conclusions on the basis of only one test. Multiple measurements will increase the validity of the conclusions drawn. In a typical test, the tracer enters at the machine with the feed. The best location for sampling the machine contents is at the exit point of the machine. Samples are taken periodically, over a period equal to at least twice the theoretical residence time, z, of the vessel. The samples are analyzed for tracer concentration, and a residence time distribution is plotted, normalized as shown in Figure 1. The actual or observed residence time is the time at whch half the tracer material has passed through the machine. This can be determined by integrating the area under the residence time distribution curve. The analysis of residence-time data for flotation applications is lscussed in detail by Nesset [ 19881.

1182

The effective volume of the machine is determined by comparing the observed residence time with the theoretical residence time. An observed residence time that is sigmficantly less than the theoretical time indicates the presence of sanding or stagnant areas. In mechanical flotation machines 8-15% of the machine volume is normally acceptable for air hold up, froth layer, and quiescent zone. This volume is subtracted from the machine volume during initial calculation of design and scale-up criteria. If tracer tests show an apparent stagnant volume greater than this, the vessel may be sanded or inadequately mixed. The determination of retention time by tracer tests has received considerable attention, but its relevance is not universally accepted. RTD tests without accompanying determination of flotation rates have questionable value, because flotation kinetics are not analogous to chemical reaction kinetics. The first-order equations generally used are simpllfications of the more complex phenomena known to occur in flotation. They also assume that reaction rates are equal throughout the tank volume, which is not the case with flotation. The main concern is to ensure sufficient efficient machine volume for each application. For example, Outokumpu does not use tracer tests as the sole criterion for retention time determination, although these tests are made to ensure the characteristic mixing properties are maintained in scale-up. HYDRODYNAMIC ANALYSIS OF FLOTATION MACHINES Characteristic Numbers Hydrodynamic analysis of flotation machines was introduced by Arbiter, Harris, and Yap [1969] A detailed, dimensional analysis of standard flotation machines was performed, and the results were verified with tests in laboratory-size machines, using the well-known aflinity laws for turbomachinery, Q -ND3, and P

-N3D5,

where Q is the volumetric flow rate, N is the rotor rotational speed, D is the rotor diameter, and P is the net power consumption [Simon and Korom, 19971. Using dimensional analysis, Arbiter listed seven “performance parameters” for hydraulic characterization of flotation machines. These parameters are shown in Table 2, where 4is the cell surface area, & is the draft tube diameter, g is the acceleration due to gravity, Qr is the liquid circulation flow, Qa is the air ingestion rate, Vc is the cell volume, Vt is the rotor tip speed, and p is the specfic gravity of the liquid or slurry. P, D, and N are defined as above. Table 2 Flotation cell hydraulic performance parameters (Arbiter, et al., 1969) Definition

Parameter Specific Air Flow Circulation Intensity Liquid Rise Velocity Specific Power

Qafvc

QJVC QJAat

Air Flow Number, N, I

I

I

A later analysis by Zlokarnik [1973] confirmed the importance of the air capaciy number, C,, in the scale-up of flotation machines and argued that the Froude number, F, = (N D)/g should be used in place of the power number.

1183

The relationships of these parameters to the critical functions of flotation machines listed above are clear. Circulation Intensity, Specific Power, Power Number, and Froude number relate to pulp suspension, interface stability, and the solids/& contact. Specific Power also may be referred to as Power Density and Power Intensity. These parameters also relate to residence time, because they influence pulp suspension and the amount of short-circuiting in the machine. Specific Air Flow, Au Flow Number, and Air Capacity Number relate to interface stability and soliddair contact. Figure 5 shows data fkom laboratory tests conducted by Rodrigues, Filho, and Masini [2001], in which the correlation of flotation performance with basic hydrodynamic quantities was clearly demonstrated. These tests were conducted in Hallimond tubes under carefully controlled conditions, using very uniform particles. Thus identical results may not be found in industrial-scale equipment. Nonetheless, the correlations found are instructive, and show the value of examining fundamental, hydrodynamic parameters.

0

0

0

0 Glass spheres

0 Quartz particles

Figure 5 Correlation of recovery with Reynolds number (courtesy Minerals Engineering) Losses in flotation are usually in the fine or coarse fractions. Often, the largest losses are in the coarse material, so it is important to provide the proper levels of mixing and suspension, giving sufficient energy for bubble-particle collision and suspension of coarse particles. However, excessive mixing can result in lower recovery, so a balance must be sought. The specific energy used in flotation may be different for different minerals and even different size distributions of the same mineral. For example, in copper flotation power averages typically range from 1.2 to 1.5 kilowatts/effective cubic meter, while power in platinum flotation may run as high as 3 kilowatts/effectivecubic meter. Residence Time Conventional practice has been to size flotation cells based on residence time. The required residence time for a given application is determined from laboratory or pilot-plant testing. The laboratoly data are then used to calculate the parameters for a kinetic model, such as the one proposed by Klimpel [1980], where recovery, R, is given by R

=

K [ 1 - (l-e-Kt)/Kt1,

(8)

1184

where

& t K

= recovery at time zero, = time, and = flotation rate constant.

When the required residence time is known, the flotation capacity is determined by a simple calculation poling, 19801. This procedure may be summarized in the following equation:

C = Qpulp' R, (9) where C = total required flotation capacity, m3,and R = required residence time, min. In t h ~ scalculation, the solids feed rate to the mill, the solids specific gravity, and the pulp density are used to calculate the volumetric flow rate of the pulp. When the required total tank volume is determined, the number of machines of a given size is found from Equation 10: 7

= VedQpdp.

where = theoretical residence time, min, Ven = effective cell volume, m3, and Qpulp = volumetric flow rate of pulp, m3/min.

z

The effective tank volume must subtract the volumes of the froth layer and the internal tank components. It must also account for the air holdup, which is the amount of air suspended in the pulp during machine operation [Levenspiel, 19961. Calculation of residence time using Equation 10 assumes that the machine is operating with no imperfections or deviations in its flow pattern. In practice, all stirred tanks experience some combination of dead space, short circuiting, back mixing, and other imperfections [Levenspiel, 19961. Measurements of actual residence times in operating flotation cells can be made, using techniques suggested by Levenspiel. Table 3, provided by Metso Minerals, shows relationships between test-bench and plant retention times for various materials. It may be used as a guide where no other information is available. It is clear that bench times need to be increased to obtain correct plant design retention times for plant design. Outokumpu reports using scale-up factors somewhat higher than those shown. GL&V/Dorr Oliver further caution that care should be taken during laboratory testing to ensure that the froth removal rate results in the same concentrate grade as that anticipated in the plant design.

1185

Table 3 Residence time guidelines for various applications

Material' Copper Lead Molybdenum Nickel Tungsten Zinc Barite Coal Feldspar Fluorsr>ar Phosphate Potash Sand (Impurity Float) Silica (Iron Ore) Silica (Phosphate) Effluents Oil

Typical solids Typical residence concentrations for times for roughmg applications, roughmg applications, minutes3 percent? 32-42 13-16 25-35 6-8 35-45 14-20 28-32 10-14 25-32 8-12 25-32 8-12 30-40 8-10 4-8 3 -5 25-35 8-10 25-32 8-10 30-35 4-6 25-35 4-6 30-40 7-9 40-50 8-10 30-35 4-6 As received 7-12 As received 4-6

Typical laboratory flotation times, minutes 6-8 3 -5 6-7 6-7 5 -6 5 -6 4-5 2-3 3 -4 4-5 2-3 2-3 3 -4 3 -5 2-3 4-5 2-3

Scale-up factor 2.1 2.0 2.6 1.8 1.8 1.8 2.0 1.6 2.6 2.0 2.0 2.0 2.3 2.6 2.0 2.0 2.0

Notes: 1. Material must be in floatable form. 2. For cleaning applications, use 60% of the solids concentrationfor roughmg. 3. For cleaning applications, required residence time is approximately 65% of that required for roughmg. Superficial Gas Velocity3 Superficial gas velocity, J, is defined as the volume of air,Qa, which rises through the froth area A of the flotation cell within a time unit, thus J,

= QaIA

The superficial gas velocity may be used as an index to facilitate comparison of air feed volumes between different cell sizes and especially their froth areas. Depending on ore type and circuit design, limits may be defined for Jg values that will produce successful flotation. Low gas rates often result in low recoveries, due to i n s a c i e n t froth production and removal rate, while high rates have the same effect because air rises too rapidly through the froth bed, producing boiling or geysering. In the extreme, either case can result in eventual froth collapse and zero recovery. Aeration rates must also be optimized for particle size. Figure 6, provided by GL&V/Dorr-Oliver, shows recommended superficial gas velocity vs. particle size. Superficial gas rate alone does not determine the optimum operational air feed rate because in otherwise similar conditions froth depth can be used to manipulate this. As shown in Equation 6, the froth residence-time index, q,combines variables of air feed rate Qa and pulp level (or froth depth) h in one parameter. Froth residence time expresses the flotation air mean residence time in the froth bed. 3

The discussion in this section was contributed by Heikki Oravainen of Outokumpu Mintec OY.

1186

Froth residence time may be used as a parameter to optimize metallurgical flotation results in different applications, Although not always directly applicable, often an optimum time range for froth residence can be defined. In such cases, this time can be used to control flotation selectivity by manipulatingfroth depth.

I Figure 6 Recommended superficial gas velocities in Dorr-Oliver machines (courtesy GL&V/Dorr-Oliver) The effects of slurry pumping rate, aeration rate, and froth residence time combine to control recovery and grade in a flotation machine. Low aeration rates generally necessitate shallow froth beds with low selectivity and vice versa As froth depth is increased to bring up concentrate grade, the accompanying recovery loss can to a certain extent be restored by increasing air flow, to maintain optimum froth residence time. However, when air flow is increased without changing pumping characteristics, a point will be reached at which recovery does not improved. This is primarily because bubble size increases, and bubble surface-area flux decreases. In addition, overaeration can cause excessive turbulence, and lead to geysering at the froth surface, as previously mentioned.

Bubble Surface Area Flux Recent research has suggested an addtional characteristic number called the bubble-surface-area flux,denoted s b [Gorain, Franzidis, and Manlapig, 19971. This number is total surface area of all bubbles passing through an i m a g m y plane in a given time, as shown in Figure 7. It may also be thought of as the volumetric air-flow rate divided by the effective cross-sectional area of the cell. Bubble-surface-area flux is defined as sb

where J, d,

= 6J,/ds, = superficial gas velocity, and = Sauter mean bubble diameter.

1187

bble Surface Area

o

w

Figure 7 Bubble surface-area flux Table 4 shows measured values of J, and Sbr along with rotor speed, N, for several commercial flotation cells. The data are summarized from Gorain, Franzidis, and Manlapig, 1999, and Deglon, Egya-Mensah, and Franzidis, [2000]. Table 4 Bubble surface-areaflux in commercial flotation cells

Tests were conducted in a 2.8 m3 flotation cell, treating zinc cleaner feed in an Austdian concentrator. Sixty-four tests were conducted using four different impeller types, four impeller speeds, and four air-flow rates. It was found that bubble-surface-area flux correlated reasonably well with the flotation rate constant for each set of conditions - better than with bubble size, gas holdup, or superficial gas velocity [Gorain, Franzidis, and Manlapig, 19971. From these tests, the bubble-surface-area flux showed promise as a single parameter that could be used to characterize flotation performance. Unfortunately, measurement of the parameter proved difficult to accomplish in the field [Oblad, 2002, and Oravainen, 20021 and the parameter has not been widely used outside academic studies. Further research meiskanen, et al., 20001 has shown that bubble surface-area flux does not adequately characterize performance in the flotation of coarse particles. POWER CONSUMPTION Power consumption is an important consideration in the design of mineral processing plants. This is especially true in remote areas, where power must be generated on site. Figure 8 shows how specific power decreases as cell size increases. This is another factor that has driven the design and installation of increasingly larger cells.

1188

It is important to consider total power consumption when analyzing power requirements for a flotation installation. Extemally-aerated cells usually have smaller motors driving the cell mechanisms, because the mechanism pumps only slurry. Self-aerated WEMCO cells have larger motors dnving their mechanisms, but do not require additional power to drive an air blower. The trend in total power requirement is similar for both types of cells, but actual values are higher for WEMCO machines. WEMCO states these higher specific power levels may contribute to higher values of bubble surface-area flux,but this is difficult to quantify. Cell Volume, m3 (approximate)

-

". 3

A

0.10

3 &i

;

0.09

2.63 2.37

0.08

2.11

0.07

1.84

4

E ._

3

0.06

1.58

0.05

1.32 , "

8

3

a

1.05

3 y. 2

E

&

a

rA

0

5 0.03

0.790

0.02

&

0.526 'CI 0

0

E:

0

N 0

0 0

m

0 0 0 0 d m

0

0,

0 0 0 N

8 8 0

0 8

0

m

d v ,

Cell Volume, ff3

Figure 8 Specific power in flotation cells COMPUTATIONALFLUID DYNAMICS Computational Fluid Dynamics (CFD) is a now commonly used by cell manufacturers for initial cell design. CFD is applied widely to modelling mixing tanks in the chemical process industries [JSrawczyk, 19971, and was thus naturally extended to modelling flotation tanks. However, the flotation models are more complex, because three phases are present.

In a CFD model, the tank volume is divided into numerous, small elements - usually from 50,000 to 200,000. The equations of classical fluid mechanics are then solved in approximation for the flow of material in each element, using partial differential equations. Solving the large number of equations involved requires many floating-point calculations, so workstations or parallel computing systems are usually used [Studt, 19971. After the model is solved, visualization allows users to understand the results of the model. Using color graphics, contour or vector plots of pressure, flow, and other variables can be prepared. Two examples are included here to show the tremendous utility of CFD models. Figure 9 shows the predicted velocity vectors in a vertical plane for a Metso Mmerals RCSTMmachine. Figure 10 shows a contour plot of the predicted bubble surface-area flux in the same machine. Of course, the figures are much clearer in color, but their utility can still be seen in the black-and-white versions.

1189

-.-

... .

. .....

.

Figure 10 Bubble surface-area flux Figure 9 Velocity vectors in a Metso contours in a Metso Minerals cell Minerals cell (Both figures courtesy AMIRA and Metso Minerals) SPECIFIC DESIGN CRITERIA Dorr-Oliver Machines Dorr-Oliver machine design is based on the approach of Arbiter, previously described. Specifically, a general formula relates various parameters as shown Kallionen, 19991:

WT= K x (TD)"x (C/D)b x (P/W)" x Xd x C,", (14) where H = dispersion height, T =tank diameter, D = rotor diameter, K = mixing factor, C = bottom clearance of rotor, C, = Air Capacity Number, P = power consumption, W = particle weight X = particle size, and a, b, c, d, and e = constants. Units are not specified for the variables in the above formula. It is assumed that they must be consistent. Dorr-Oliver machines are designed to provide Froude numbers greater than 0.8 for smaller cells, and greater than 0.4 for large cells. However, hydrodynamic analysis is not the only consideration in cell design and scale-up; past experience and operating conditions also play an important role. Dorr-Oliver tanks have conical bottoms, as shown in Figure 11, to reduce sanding in the corners. In th~scase, the value of T, tank diameter, used in Equation 14 is T', the virtual tank diameter. Tank aspect ratio (WT) is vaned depending on the size of particles to be floated. In general, WT is much less than one for coarse particles, and approximately equal to one for finer particles. Dorr-Oliver provides two impeller designs, as shown in Figure 12. The Verti-MixTMimpeller is for coarse and intermediate particle sizes. It generates a vertical mixing pattern with low peripheral speed. The conventional impeller is for fmer particle sizes. It generates a horizontal mixing pattern, with a peripheral speed 25 to 30 percent higher than that of the Verti-MixTM design. This results in a higher-energy environment, preferred for fine-particle flotation. These

1190

designs resemble the Outokumpu designs, described below. In both cases, the rotor is higher above the cell floor for coarse particle flotation.

T

I

Figure 11 Dorr-Oliver tank dimensions

Conventional

verti-Mixm

Figure 12 Dorr-Oliver impeller designs Outokumpu Machines Outokumpu flotation cell development began in the late 1960s primarily to satisfy the company’s own needs, improving on currently-available, commercial equipment for treating complex-sulfide ores at Outokumpu’s mines in Finland. The original design goals were to develop a machine with the following properties: 0 0

0

0

near ideal mixing, sufficiently turbulent conditions in the contact zone, maximum capability for air dispersion, energy efficiency, ability to start up under full load (sanded conditions), and h g h metallurgical selectivity.

The original rotorlstator configuration resulted from scientific testing and hydrodynamic calculations. The rotor profile was calculated to balance the hydrodynamic and static pressures,

1191

allowing a uniform air flow over the major part (approximately the upper two-thirds) of the blades, thereby increasing the air volume dispersible by the impeller. The blade design provides separate slots for air distribution and slurry pumping, to maintain independent action on these two functions. Details of the rotor design are given by Fallenius [1976]. Figure 13 shows the two current Outokumpu mechanism types, which differ only with regard to the stator design. The Multi-Mix mechanism is the general-purpose design for all-round duty with enhanced capability for floating fines. The Free Flow (FF) mechanism has less turbulent action with more pumping capacity designed for coarser flotation applications. These two designs resemble two Dorr-Oliver designs, described above. In both cases, the rotor is positioned higher above the cell floor for coarse particle flotation.

MM design

FF design

Figure 14 Outokumpu cell design

Figure 13 Outokumpu rotor designs

Since hydrodynamics is an integral part of Outokumpu flotation cell design, the use of dimensionless numbers in flotation machine scale-up is standard practice. However, it has been found that variations arising from different flotation applications are magnified as cell size increases. Similarly, use of large scale-up ratios requires the application of correction factors or exponents to practically all dimensionless numbers. In view of the above, for example, Outokumpu does not use the power number as a sole design parameter for power scale-up. Neither is the Froude number considered as a constant in flotation cell scale-up by the principles of hydrodynamics. Outokumpu uses an extensive database from different applications acquired through on-site measurements from actual installations worldwide. This has allowed the indexing of different process and ore types to be used for power calculations. The power numbers for the two Outoknmpu mechanism types are approximately four for the MM and six for the FF mechanism.

Outokumpu determines specific power figures for different characteristic volumes inside the flotation cells to identify the overall Power Density and its gradlent in the machine. The ratio between specific power per cell volume and the power per slurry volume inside the rotodstator free space varies from 23.6 to 35.6 depending on machme size. Since the initial design in 1970, various tank configurations have been devised. The rectangular OK-R and OK-U maclunes represent the traditional models with one or more shafts per tank in the lower volume (below 50 m3 per shaft) range. A prototype of the oripnal,

1192

cylindrical Tankcell@was tested in 1986 at the Pyhasalmi Mine, with the first commercial sale in 1991. The Tank Cell@is offered in standard sizes from 5 to 200 m3 effective volume. A typical design is shown in Figure 14, above.

Metso Minerals Machines The RCS Machine was developed in the mid-1990s. The primary design goal was to create the two classic zones within a flotation cell, an active lower zone to ensure effective particle suspension and transportation and a relatively quiescent upper zone to minimize bubble-particle separation. A circular cell design was adopted as being the most suitable tank shape to provide symmetrical hydraulic flow patterns with minimum upper-zone turbulence. The RCS mechanism was also specifically designed to maximize sluny circulation through the high-energy zone withm the mechanism to promote ultra fine particle recovery. The RCS mechanism is said to have strong, primary bottom flow and secondary, top-flow slurry recirculation to ensure effective particle suspension. The symmetrical flow patterns across the cell floor towards the mechanism are said to effectively minimize short-circuiting, an effect that is further enhanced by the careful design of the location and dimensions of the slurry entry and exit ports. Aeration rate is a function of the total sluny flow through the impeller, which has been designed with an integral air shelf to ensure that air can be effectively mixed into the sluny flow stream. The effective radial distribution of the air across the complete cell cross section is the result of balancing the radial slurry pumping rate and bubble rise velocity. The balance between particle capture and particle drop-off governs the overall metallurgical performance of all flotation cells. Particle capture within the flotation cell is thought to be by impingement for the majority of particle sizes and possibly by high-energy bubble nucleation at the very fine sizes. The upper zone recirculation of the RCS mechanism is a key feature of the design, for two reasons: First, it leads to additional particle bubblekontacts as the recirculation flow crosses the bubble rise path, and second, it maximizes slurry circulation through the highenergy impeller zone to increase ultra-fine particle recovery. The quiescent upper zone in the RCS cell is created by the use of a single horizontal baffle, located on the cell wall. The baffle promotes recirculation to the mechanism as well as allowing development of a quiescent upper zone, which is said to minimize potential particle-bubble separation and lead to a very stable frotWslurry interface. A typical RCS machine is shown in Figure 15. Machines are provided with two internal, cross-flow launders, with froth discharge to one side of the cell bank. This is said to simple froth removal and minimize transport distance. Cell sizes range from 5 to 200 m3.

1193

Figure 15 Metso Minerals RCSm flotation machine Table 5 shows characteristic numbers for some Metso Minerals machines, calculated from company literature [Svedala, 20011. Table 5 Characteristic numbers for Metso Minerals machines Machine Size, m3 Number Specitic Power, b i m 3 Air Rise, d m i n

Expression P I v, QaJ&

5 3.0 0.95

30 1.5 0.93

100 1.1 0.89

200 1.25 0.90

WEMCO Machines Historical background for use of hydrodynamic methods in the design of WEMCO machined is given in Nelson and Lelinski [2000]. Current practice for WEMCO machmes includes the calculation or measurement of characteristic numbers, shown in Table 6 for a range of machine sizes. These numbers are geneally maintained in the ranges shown for all machine sizes. Air and water delivery numbers are calculated at standard engagement and submergence for the rotor. Table 6 Characteristic numbers for WEMCO machines Machine Size. m3

1194

Figure 16 shows a WEMCO Smartcell@machme.

Hood Disperser

Rotor

Draft tube

Figure 16 WEMCO SmartCell@flotation machine The openings and clearances within the machine are sized to provide the requisite minimum linear flow rate, based on the design flow of slurry through the machine. SCALE-UP CRITERIA Flotation machine scale-up is based primarily on maintaining geometric similarity as machine size is increased. Modifications are made to maintain the minimum flow rates required to prevent sanding, and to accommodate special conditions in certain applications. In addition, as cells have become larger, modifications have been made to froth launder designs, to allow removal of the increased fioth quantities produced by the large machines.

Dorr-Oliver Machines Don--Oliver machines are sized for application based on Equation 14, above, with additional guidance provided by the company’s data from previous installations. In general, geometric similarity is maintained, but this may vary, depending on application. Hydrodynamic criteria are not usually evaluated for each application. Tank aspect ratio (HIT) is varied depending on the size of particles to be floated. In general, H/T is much less than one for coarse particles, and approximately equal to one for finer particles. Outokumpu Machines In scale-up, rotor submergence in Outokumpu cells is maintained at a geometrically similar location, so that the original turbine type mixing flow pattern remains virtually unchanged regardless of scale. The intent of this practice is that the tanks and components of the Outokumpu machines maintain geometric and hydrodynamic similarity as machine size is increased.

1195

Outokumpu determines the average turnover of slurry for each application. Because the flow circulation inside the tank involves several identifiable flows, these are also calculated separately. The average turnover is the sum of a combination of at least four main flows, the fastest of which is in the bottom of the tank,where probability of initial bubble/particle attachment is most critical. Turnover time is defined as the average number of times the cell volume is pumped through the mechanism, and depends on the residence time of the pulp in the tank. With practical residence times, Outokumpu machines have an average turnover from about 3.5 to 8.5. Outokumpu machines are scaled up based on the criteria described in the two preceding sections. The preliminary design for a new machine size is based on geometric similarity to preceding models. The basic calculation is therefore quite straightforward. The geometricallyscaled design is then adjusted to ensure sufficient flow rates for mixing and suspension as well as specific power in the rotorhtator region. The design of the open froth area and the concentrate launders are the parameters that change the most, dependent on application.

In Outokumpu machines, rotor submergence has little effect on aeration capability. Further, change in ambient air density or pulp back-pressure (hydrostatic head) can be compensated by choosing a blower of the correct size. Because the mechanism is in the bottom of the tank,pulp suspension and mixing properties can be maintained without having to alter the flow pattern with modlfcations to the cell internals to enhance pulp circulation. Thus scale-up to larger sizes does not introduce unwanted pulp surface turbulence, so additional dampeners or deflectors are not employed. Rotor speed and size are selected to maintain the necessary solids suspension and pulp circulation patterns inside the tank and to provide the necessaty shear forces in the primary rotor/stator contacting zone. These parameters are checked separately for each application. If necessary, rotor speed may be altered by changing belt position, on a multi-sheave drive, or through use of an electronic, variable-speed dnve. Rotor tip speeds exceeding 7.6 d s e c are generally not recommended to avoid increased wear. The Outokumpu designs have been analyzed by CFD since 1992. The models are continuously validated in industrial scale by pressure and flow measurements in close association with complementing power, stress and vibration analyses. CFD analysis shows that the flow pattern in the Outokumpu Tankcell@machines maintains turbulence at the comer of the cell, effectively minimizing sanding. This is one reason why Outokumpu do not use tank-bottom beveling in any of the tank sizes. Figure 17 shows velocity vectors in an Outokumpu cell, calculated by CFD. The velocity vectors are gray-scale coded, with values in meters per second. Metso Minerals Machines Scale up in the full range of RCS flotation machine sizes is based on both hydrodynamic principles and experience from a wide range of installations. The RCS machines in the standard range are geometrically similar, all having a substantially constant aspect mtio of tank diameter to froth lip height. Constant power and pumping numbers are used to ensure that the RCS mechanisms are hydrodynamically similar over the complete range of cell sizes. However, Metso Minerals do not use dimensionless numbers such as the Froude number to fully define scale-up of the complete cell. This is because there are several hydrodynamic zones, each Merent, within a flotation machine, which precludes the application of these general functions. Metso Minerals examine localized slurry velocities, localized specific power, air dispersion, and other factors to ensure similarity in performance in the scale-up of the RCS machine. CFD simulation is used to give additional insights into the overall hydrodynamic behavior of the machme.

1196

Figure 17 Velocity vectors in an Outokumpu cell Other than the general hydrodynamic scale-up, the ratio of new feed volume flow rate to mechanism pumping rate is also considered to ensure that the symmetrical flow patterns developed by the RCS mechanisms are not significantly distorted by the entry and exit slurry volume flow rates. CFD simulation is also being used to further refine the historically-assumed values of feed flow rates. Table 7 shows Metso Minerals’s guidelines for the number of cells per bank the slurry flow rate, as functions of cell size. These numbers are typical for all manufacturers.

I

Cell Size, m3

Maximum Cells er Section

Maximum Slurry Flow, m3/min

5

4

150

4

450

15 20 30 40

160 200

1

3

1

600

1

3 3

1 1

900 1200

4800

6000

WEMCO Machines W M C O machines are scaled up based on the criteria described in the two preceding sections. The preliminary design for a new machine size is based on geometric similarity to preceding models. Th~sis easy to calculate, and provides a convenient starting point. The geometrically-

1197

scaled design is then modified to maintain the hydrodynamic parameters in the desired ranges. The most important characteristic numbers are: 0 0 0 0 0

specific power (power per rotor volume and power per cell volume), residence time of the air and pulp in the disperser region, superficial air velocity into the froth, air flow number, circulation intensity in the flotation cell, and, minimum slurry flow rates at constricted points.

For the WEMCO machine, the disperser volume is the capacity contained within the disperser cavity. Specific power input gives the f i s t estimation of the size of the mechanism (rotor diameter, motor size, etc.) for a new cell size. The Specific Power necessary for expected metallurgcal results is dependent on the particular application. The residence time of the air and pulp in the disperser region is also considered. Thls is an indication of the intensity of bubbleparticle contact and is defined by the total volume (air plus pulp) contacted within the volume of the disperser, per unit time. Air flow number relates to the necessary balance between air ingestion and pulp suspension as controlled by rotor speed, once the rotor diameter has been established. Either insufficient or excessive air can result in lower recovery; the former results in a reduction in the flotation rate constant, while the latter results in surface geysering and particle detachment. Circulation intensity represents the number of times the pulp will pass through the mechanism prior to exiting the vessel. The higher the circulation intensity, the greater the number of times the mineral-laden pulp will come in contact with the induced air within the disperser region. Pulp velocity is a superficial value, which is determined by the liquid flow rate per cross sectional area. The ability to increase the pulp velocity with machine size is a design feature used to increase the suspension characteristics of the larger flotation machines. As such, this parameter is used to quantify the solids suspension characteristicsof the machine. In conjunction with these hydrodynamic parameters, other features are included as part of the scale-up criteria. These features serve such functions as reducing pulp/f'roth interface turbulence and providing adequate froth-removal capacity. For example, the froth-removal capacity of the machine is characterized by the lip loading, which is defined as the weir or launder length per cell volume. In W M C O machines, lip loading can be easily adjusted by changing the number of radial launders. In the last five years, WEMCO designs have also been analyzed by CFD models. These models have included all three phases present (solid, liquid, and gas) in flotation applications. The models have been used to eliminate spots where solids may tend to settle. Figures 18 and 19 respectively show CFD models of flow in rectangular and cylindrical cells. Again, these images are best shown in color, but their value is still evident here . However, in Figure 18, the lighter regons show areas of higher energy, where particle-bubble contact is intense, while the darker regions show the quiescent zone at the top of the tank. In Figure 19, the light-colored lines show individual flow paths through the machine, with a typical path highhghted in white dots.

1198

Figure 18 CFD model of WEMCO l+l@ machine (courtesy of Gert Van der Linde and Foskor Ltd.)

Figure 19 CFD model of WEMCO SmartCell" machine

OPERATING VARIABLES Dorr-Oliver, Outokumpu, and Metso Minerals Machines Operating variables in the externally-aerated cells are aeration rate, rotor speed, and froth depth. Aeration rate is controlled by adjusting valves in the air delivery line, or by changing the blower speed. If this is varied independently of other variables, bubble size distribution in the cell will change. Aeration rate is to a certain extent independent of the site elevation and slurry density. However, the configuration of the blower and air delivery system may limit the ability to adjust aeration rate. It is also important to realize that when the external aeration rate is changed, other parameters associated with aeration, such as superficial gas velocity, gas holdup, and bubblesurface-area flux, will also change. The type of air blower selected is important. Multi-stage, centrifugal air blowers are nonnally recommended, as they have the best characteristic relationship of discharge pressure vs. delivered air volume. Other blower types can be used but they generally require variable speed drives or other means of modifying the pressure/volume relationship to ensure successful operation at reduced aeration rates. Rotor speed is easily changed when variable-speed drives are fitted. However, variable-speed dnves add to the costs of the installation. Froth depth is changed by altering the height of the froth-pulp interface, which is accomplished by controlling the flow through the cell, using the discharge valves. WEMCO Machines Because WEMCO machines are self-aerating, air flow is controlled differently than in other machines, where air is provided by blowers or compressors. The air intake for each WEMCO machine can be fitted with manual or automatic control valves. Either of these may be used to decrease the air flow to the machme. Pumping characteristics of WEMCO machines may also be changed by changing the size of the rotor. This however is less common because it requires that the machine be shut down and the mechanism removed. The pumping of air and slurry by a WEMCO machine are controlled by changing the position or speed of the rotor. Rotor position may be changed without removing the mechanism, by raising or lowering the mechanism, or changing the length of the draft tube. Rotor speed may be changed by changing belt position, on a multi-sheave drive, or through use of a variable-speed drive. Again, use of variable-speed drives increases costs.

1199

Rotor position is described in terms of submergence and engagement, as shown in Figure 20. Submergence is the distance from the slurry surface to the top of the rotor blades; engagement is the depth to which the rotor engages the draft tube. WEMCO can provide detailed performance curves, showing air and water pumping rates in relation to rotor speed, submergence, and engagement.

n

Froth surface ,.

Launder

I./

i

4.

. ............... . . ................ . ............/... . ........ . ,..,........... .,. ..... _....,. ................

Submergence Hood Dispe Rotor

False bottom

1

I

Figure 20 Operating variables in the WEMCO cell In some applications, high power densities require the installation of specially-designed hoods and Ispersers. In other applications, spray water may be applied to the froth, to remove entrained gangue particles. This can be done in all types of cells.

MECHANICAL COMPONENTS In addition to the tank and rotor mechanism, each flotation machine, or row of machines, has additional components. These include launders for discharge and conveyance of froth, boxes or connections for feeding and discharging slurry, valves for controlling slurry flow rates, and in some cases, baffles for modification of flow patterns within the cell. All manufacturers use automated, computer-based dmftmg programs to prepare design drawings, and perform finiteelement analyses to determine stresses on cell walls and other components. Because design of these components is similar among all manufacturers, the discussion here will be geneml, with some references to specific machines for illustration. Froth Launders Launder designs have changed with cell design. Early, rectangular cells had launders along one or both sides, as required by the volume of froth generated. The first cylindrical cells had circumferential launders, sloped to one, two, or more discharge points around the cell. Current cylindrical cells have combinations of radial, circumferential, and cross-flow launders, to provide additional capacity for froth removal. Launders are sized to provide the correct lip length for a given cell volume and froth area. Launder layout is based on the total lip length, and on maintaining a maximum distance of travel from any point on the froth surface to the nearest launder. For example, the WEMCO design, with patented radial launders, maintains a maximum froth travel distance of 0.5 meters.

1200

Feed and Discharge Connections Feed and discharge connections were originally made in rectangular boxes. These have generally been discontinued because of their high cost. Current designs for feed boxes are usually semicircular. They are intended to minimize the required floor space and costs of materials and fabrication. In selecting the feed box design for a given installatioq the respective costs and space requirements must be considered. There are various designs for discharge connections. For example, WEMCO uses three designs. The semi-circular dwharge, with dart valves for control, allows the closest spacing, and the easiest maintenance of the dart valves. It is also the most expensive. The conventional design, a modfied box, is less expensive, but requires greater spacing between cells. The hinged dart valve design allows for close spacing, but requires that adjacent cells be shut down and vacated for valve maintenance or repair. In some discharge connections, Outokumpu can spec@ pinch valves instead of the customary dart valves. GL&V/Dorr Oliver use a replaceable “drop-box design” junction and discharge box in their large tank cells to reduce row length and facilitate maintenance. The size and number of the Qscharge valves is determined by comparing the required range of slurry flow rates to the performance curves (percent open vs. flow) for available valves. The valves are sized to function near the middle of their performance curves during normal cell operation. MACHINE SIZE AND CONFIGURATION The recent trend among all manufacturers has been to increase the size of individual machines. Large machines are generally preferred by operators because they allow for the same total volume, with lower installation and maintenance costs, and less required floor space. In addition, when there are fewer cells, the extent of back-mixing and short-circuiting between successive cells is decreased. When each cell is separated by a connection box with a control valve, its operation is virtually independent fiom the cells on either side. The cost advantages of larger machines are shown in Figure 21, where cost-per-cubic-meter is shown vs. cell size. The costs shown are approximate, but are typical for machines at this time. It is also important to consider that,when there are fewer cells, and fewer rows of cells, the down-time for each cell or row becomes more important, as a greater share of production capacity is lost. This is often negated in modern mills through the use of scheduled maintenance shutdowns, in which one or more rows of cells may be inspected and maintained. For many years, flotation cells were joined in the “hog-trough” configurationtoform banks or rows of cells. As cell size increased, rectangular cells were configured in shorter and shorter banks, sometimes comprising only two or three cells. Both these designs - the hog trough and short banks of rectangular cells - are still used in some applications, but large, cylindncal cells are favored in most large, new concentrators.

1201

1800

5000

1600

4M0

1200

2mo

1000

1000 SO

75

100

ID

IS0

175

2W

0

10

20

ca "OIIPIO.

CeIVobm5 m'

30

40

m1

Figure 21 Comparative Costs for flotation cells REFERENCES Arbiter, N., 1999. Development and Scale-Up of Large Flotation Cells, in Advances in Flotation Technology, ed. B. K. Parekh and J. D. Miller, 345-352. Littleton, CO: SME. Arbiter, N., C. C. Harris, and R. F. Yap, , 1969. Hydrodynamics of Flotation Cells. Transactions, Society ofMiningEngineers of A I . . 2441134-148. Brewis, T., 1996. Flotation Cells. Mining Magazine. 160:7, 18-24. Deglon, D. A., Egya-Mensah, D., and J. P. Franzidis, 2000. Review of Hydrodynamics and Gas Dispersion in Flotation Cells on South African Platinum Concentrators. Minerals Engineering, 13:3,235-244. Fallenius, K., 1976. Outokumpu Flotation Machines. In Flotation, Volume 2, ed. M.C. Furstenau, 838-861. New York: AIME. Gorain, B. K., J. P. Franzidis, and E. V. Manlapig, 1999. Empirical Prediction of Bubble Surface Area Flux in Mechanical Flotation Cells from Cell and Operating Data. Minerals Engineering, 12:3,309-322. Gorain, B. K., J. P. Franzidis, and E. V. Manlapig, 1997. Studies on Impeller Type, Impeller Speed and Air Flow Rate in an Industrial Scale Flotation Cell. Part 4: Effect of Bubble Surface Area Flux on Flotation Performance. Minerals Engineering, 10:4,367-379. Harris, C. C., 1976. Flotation Machines, in Flotation, Volume II, ed. M. C. Furstenau, 753-815. New York: AIME. Heiskanen, K., K. Fob, S . Junnikkala, J. Aho, P. Lampinen, and K. Honkivaara, 2000. Flotation of Different Particle Sizes as a Function of Bubble Surface Area Flux. In Proceedings. 32"d Annual Meeting of Canadian Mineral Processors, chapter 12, 197-211. Ottawa, ON, CIM. Jonaitis, A. J., 1999. Design, Development, Application, and Operating Benefits of 100-m3+ Outokumpu Tank-Cell Flotation Cells. In Advances in Flotation Technology, ed. B. K. Parekh and J. D. mller, 371-380. Littleton, CO: SME. Kallionen, J., 1999. Advances in Application Driven Design of Flotation Cells. In Proceedings of Comer 99/Cobre 99 International Environmental Conference, Volume 11- Mineral ProcessingEnvironment, Health and Safety, ed. B. A. Hancock and M. R. L. Pon, 29-39. Warrendale, PA: TMS,.

1202

50

Klimpel, R. R., 1980. Selection of Chemical Reagents for Flotation, in Mineral Processing Plant Design, ed. A. L. Mular and R. B. Bhappy 907-934. New York: SME-AIME. Krawczyk, J. R., 1997. Applications and Limitations of CFD in the KPI. Processing, July, 53-57.

Hydrocarbon

Levenspiel, O., 1996. The Chemical Reactor Omnibook, 61.1-61.18. Cowallis, OR: OSU Book Stores, Inc. Nelson, M. G., and D. Lelinski, 2000. Hydrodynamic Design of Self-Aerating Flotation Machines. Minerals Engineering, 13: 10-11,991-998. Nesset, J. E., 1988. The Application of Residence Time Distributions to Flotation and Mixing Circuits. CIM Bulletin, November, 75-83. Oblad, H. B. @ h c o Process Equipment Co.), personal communication, January 28,2002. Oravainen, H. (Outokumpu Mintec OY), personal communication, January 25,2002. Poling, G. A., 1980. Selection and Sizing of Flotation Machines, in Mineral Processing Plant Design, ed. A. L. Mular and R. B. Bhappy 887-906. New York: SME-AIME. Rodrigues, W. J., L. S. Leal Filho, and E. A. Masini, 2001. Hydrodynamic Dimensionless Parameters and their Infiuence on Flotation Performance of Coarse Particles. Minerals Engineering, 14:9, 1047-1054. Simon, A. L., and S. F. Korom, 1997. Hvdraulics, 181-182, Upper Saddle River, NJ: PrenticeHall. Studt, T., 1997. Flow Dynamics Models Take New Turns. R&D Magazine, April, 26-32. Svedala Industri AB, 2001. Flotation Machines, Reactor Cell System RCS. Technical Data. Malm): Svedala Industri AB (now available from Metso Minerals, Colorado Springs, CO). Weber, A., C. Walker, L. Redden, D. Lelinski, and S. Ware, 1999. Scale-Up and Design of LargeScale Flotation Equipment, in Advances in Flotation Technology, ed. B. K. Parekh and J. D. Miller, 353-369. Littleton, Colorado: SME. Zlokarnik, M., 1973. Ahnlichkeitstheoretische Kriterien zur dimernsionierung von Flotationszellen. Sonderdruck und Zeitschrrt “Erzmetall”, 26:3, 107-113. Zlokamik, M., 199 1. Dimensional Analysis and Scale-up in Chemical Engineering. Heidelberg: Springer-Verlag.

1203

Flotation Equipment Selection and Plant Layout Kenneth R. Wood

ABSTRACT The key factors in flotation equipment selection and sizing are the particular flowsheet chosen to treat the ore and the retention time required at each stage. Many other factors have to be considered when calculating the total cell volume required. This paper also discusses flotation plant layout and design considerations, as they relate to the flotation equipment and associated ancillary systems INTRODUCTION It is assumed for purposes of this chapter that the flowsheet has been fixed and flotation times to achieve target metallurgical results for the various stages of treatment have already been determined. Methods for these tasks are covered in another Section of this volume. This chapter discusses factors that influence the translation of required flotation times to particular equipment sizing calculations. Column flotation cells are not dealt with specifically here except for some layout considerations, since they are covered in detail in another chapter. EQUIPMENT SELECTION: SIZING INPUTS The design criteria are key to the whole design process in that they set the desired throughput rate, operating time and schedule, expected operating conditions, as well as describing expected ore characteristics.They provide a framework within which the sizing and selection of equipment fits. The first step in flotation equipment sizing is to convert theoretical retention time to a plant value. Once the operating retention time has been set, other factors which affect the volume and flow rate of slurry at any point in the circuit must be accounted for. A detailed mass balance for solid, liquid and slurry flows needs to be prepared based on the chosen flowsheet and metallurgical balance and considering expected operating conditions and design factors. Detailed aspects to consider are discussed below. Flotation kinetics and scale-up The type of test information available affects the translation of test data to actual equipment specifications. As the quality and scale of the testing improve, confidence in the results increases. Optimum retention time as determined by testing may need a scale-up factor applied to determine target plant retention time. Comparison to other ores. If no actual test results are available, as might be the case at a very early stage of evaluation of a deposit or for a hypothetical ore, the design values are based on typical industry practices or on comparison with existing plants treating deposits with similar mineralization. In this case, the referenced plant is scaled according to throughput and grade ratios. Laboratory tests. Once laboratory test results are available, they provide a better indication of needed treatment conditions for the particular ore. Kinetic data from timed batch tests for the different stages of treatment provides a basis for determination of flotation retention time for each stage to achieve a target metallurgical result. Locked cycle test results, if available, can provide an estimate of the amount of circulating load to be expected at each stage of the flowsheet and its

1204

effect on grade and recovery. However, whether for batch or lock cycle test data, a scale-up factor must be applied to arrive at actual or equivalent retention time required in the plant. Laboratory batch tests allow the possibility of reasonably accurate measurement of conditions at the various stages through the test since the entire product fraction can be weighed and analyzed. However there is the possibility of greater scale up errors due to the differences between laboratory conditions and commercial cell conditions. This relates to different performance characteristics of laboratory size cells and production models, in such areas as energy intensity, air rates, surface area to volume ratios, effective cell volume and short circuiting in plant installations, among others. The foregoing applies to both mechanical and column cell tests. The scale-up factor varies for different mineral types as well as for different cell designs. Pilot plant testing. Pilot plant testing provides more accurate values for retention time and circulating load at each stage. Results from a well executed pilot plant test should not require any scale-up factor to estimate the retention time for larger commercial cells. (Arbiter 1978; Ounpuu 2002) For pilot plant work it is very important to work out a detailed mass balance to get the most accurate results. For both pilot plant and actual plant scale testing, there is more room for error in the measurements since it is more difficult or impossible to make precise measurements of some of the process streams. There is also the possibility that the circuit is not at steady state operating conditions. While the operation may be more representative of what could be expected from new cells, it is more difficult to ensure the accuracy of the data. Programs are available and widely used for rationalizing test data to minimize the potential errors in plant data and can help improve the usefulness of test results. (Richardson 1986) Actual plant results. Actual results from plant tests or plant operations can provide reliable information, although as noted above it can be difficult to make accurate measurements of some of the process streams. Even with plant testing there can be scale-up effects when going from one type of equipment to another. Early work with large tank cells showed that overall required retention time was about 65%-85% of what would be needed with “conventional” trough-type mechanical cells (Greene 1996). Current estimates for some tank-type cells indicate that as little as 30%-70% of the retention time of conventional cells may be adequate, depending on the specific application (Kujawa 2002). Scale-up factors. Simulators based on various flotation models are available to help scale up to plant requirements from laboratory batch test data as well as from pilot plant or plant data. Two such programs are described in the chapters on Models and Simulators. If such a simulator is available it can assist in estimating plant equipment needs and possible flowsheet configurations. These programs try to be general enough to apply to cells from different manufacturers as well as different types and sizes of cells within the product range. Some correction factor may be needed to try to adapt the model output to a specific cell type or size. (Kujawa 2002). Many cell manufacturers have developed proprietary simulators based on their own cell designs, which use kinetic data generated from laboratory, pilot or plant tests and apply fundamental analysis of flow within the cells and in the froth to calculate actual requirements for cell size, number of cells and froth area. In these cases, no further factor is required to adjust for retention time scale-up. This is more applicable to the newer cylindrical tank-type design of mechanical cells, which have greater hydrodynamic efficiency and are easier to model than more traditional designs. It also applies to column suppliers and manufacturers of contact-type cells. If no simulator is readily available for a first pass estimate of plant requirements from laboratory batch test data, historical plant data can be used. As shown in Table 1 average scale-up factors for the range of Denver conventional flow-through type cells varied from 1.6-2.6 with an average of 2.1. Another unspecified manufacturer suggested values of 1.5-3.0 (Weiss 1985). Degner indicated a “kinetic factor ratio” of 2.5 to scale from laboratory batch test flotation rates to commercial cell rates (Degner 1986) although there is some question as to the validity of scaling the rate factor itself on the basis of laboratory results (Barbary 1986). However, whether applied directly to the retention time or to the kinetic transient factor in the flotation rate equation, the end

1205

effect is that of a multiplier on laboratory flotation time to estimate plant retention time. Note that some simulators may use this or similar empirical data for their batch test scale-up factors (Blot 2002) As a guide, a reasonable scale-up factor on laboratory test retention time results for rougher scavenger applications can be taken from Table 1 for conventional open flow cells and about 75% of those values for tank-type cells, in consultation with specific cell suppliers. For pilot plant results, the default number is 1. Of course, if more specific information is available for a particular ore type, say from a cell supplier, test facility or reliable modeling program that value should be used. Given the uncertainty of the scale-up factor for laboratory test results, the value of pilot testing the flotation process before proceeding to final selection of cells and detailed engineering is evident. For any major project, piloting should be considered as a key input to the design process. A number of equipment suppliers and certainly independent testing organizations have pilot facilities that in many cases can be easily set up at site for testing. These can also provide comparisons of different flowsheet arrangements and equipment types. Function of cell type. As indicated above, the scale-up factor is different for conventional cells and tank-type cells. For contact-type cells, the manufacturers indicate that no scale-up factor is required between their specific small pilot scale results and any of their cell sizes since capacity is based on flow through the contactor and froth removal capacity rather than retention time. This only applies to their own testing equipment and not to “conventional” laboratory batch flotation test cells. While they may be able to scale up from laboratory test data on the basis of their own database they get a much more accurate scale-up from small pilot-scale tests (Bulled 2002; Murphy et a1 199?). In all cases with these cells it is necessary to work with the supplier for testing the ore and to determine suitability of the process, expected metallurgical results and sizing and arrangement of the equipment. Factors Affecting Design Distinct from the scale-up of kinetic test data to determine a target operating retention time, there are a number of variables that affect actual retention time in the plant and that have to be considered when setting total flotation volume or capacity. Grindability variations. Grindability variations are probably the most important factor to consider in setting design factors for flotation capacity (Weiss 1985). This is particularly so for plants with SAG mills in the grinding circuit ahead of flotation. Variations in grindability can cause changes to plant feed rate, flotation circuit feed size or both. In most cases, grinding circuit throughput is the constraint on production and the flotation circuit must be able to handle what it gets. Plant throughput variability must therefore be accounted for in determining equipment volume and capacity. Of particular importance are the grindability values in the initial year or two of operation as these are what will be experienced before any significant changes or additions to the plant can be made. Often the ore from open pit mines increases in hardness with depth, resulting in lower throughput and/or coarser grind as the operation progresses. Conversely, initial throughput rate can be significantly higher than the overall average. Design flotation volume must provide the required retention time for what will be fed to the plant in significant amounts. There will be extreme situations where retention will be reduced by excessive flow rates and where the amount of this ore does not justify the extra capital to install full retention capacity. The amount of variation that can be covered has to be traded off against the higher capital and operating cost of installing extra cell volume. In some instances the flotation circuit capacity may be limited by the need to make a specific separation or meet a required product specification. This should not be a factor at the design stage, unless cell requirements get very excessive at high throughputs or there are constraints on total cell volume. The upstream grinding equipment can influence control over throughput rates. If the grinding circuit contains a fixed speed SAG mill, the mill must be fed at its full grinding capacity in order to protect the mill from grindouts. With a variable speed drive on a SAG mill, grinding

1206

rate can be limited if necessary to control extreme flow rates to the downstream process. For rod mill-ball mill or single stage ball mill circuits, feed rate can also be adjusted somewhat independently of grindability. In these cases, the allowances for grindability variations can be smaller than for fixed speed SAG circuits. The possibility of a coarser grind as a result of harder ore or higher throughput has to be considered for its effect on the flotation circuit. Expected particle size affects slopes on launders as well as particle suspension in the cells. Particularly when the grinding circuit includes a large SAG mill and has a relatively coarse product to flotation, such as is common in porphyry copper circuits, it is often useful to include some form of safety screen ahead of the flotation feed distributor. This can prevent buildup of coarse particles in the cells that would otherwise result from cyclone blockages in the grinding circuit. An alternate approach is positive and reliable control of cyclones to prevent blockage of the underflows that would send coarse particles to the overflow. Feed grades and variations. Variations in feed grades will have been considered at the test stage when selecting the target retention times. While feed grades won’t necessarily change rougher-scavenger cell volumes, they do impact on concentrate handling considerations. If extreme variations in grade are expected, concentrate launder capacity must be adequate to handle the extreme flow, particularly in the first few cells. At very high grades froth carrying capacity may be exceeded, which indicates that more cell froth area is required. Extra launders or launder lip length may be required to reduce froth travel distance to the launder. This can be determined from the specific scaled up flotation rate to estimate expected concentrate volumes at each cell. Launder or piping size must allow for the maximum expected concentrate flow. The expected range of feed grades has a significant effect on total cell volume at the cleaning stages. While cleaner volume may not allow for the absolute maximum feed grade expected, it has to be able to handle what is expected to occur for a significant fraction of the plant feed, especially in the first year or two of operation. Pump capacity is also determined by the maximum concentrate production rate allowed for in the design. Concentrate grades. When considering how much of the variation in grindability and feed grade to allow for in setting maximum design throughput rates, the specific application has to be kept in sight. What is the impact of reduced concentrate grade or recovery if retention time is reduced? How acceptable is it to make a less than optimum product in consideration of product sales? How is separation of multiple products affected if cleaning capacity is exceeded? What effect does excessive flotation time have on product grade or on premature recovery of next-stage product in sequential flotation of multiple product ores? Many of the answers to these questions relate to flotation kinetics and have to be evaluated on an economic basis to see what the best tradeoffs are on a case-by-case basis. Air Holdup. An allowance for the volume of the cell taken up by air in the pulp has to be added to the cell volume. For mechanical cells this is typically 15% so the design factor will be divided by 0.85. For tank-type cells the holdup is likely to be 10% or less. If cell requirements have been produced by the use of a simulator, determine how much, if any, allowance for air holdup has been included in the simulation. Rheology. Presence in the ore of certain minerals that increase pulp viscosity may affect cell volume requirements. As an example, clay minerals can have a significant impact on suitable pulp density to be able to make a clean separation. This should have been evaluated at the ore testing stage. If there is a significant presence of such interfering material in the ore, it may be useful to treat the slimes fraction separately from the coarser fraction. In this case, the slimes are treated at a lower density to achieve lower viscosity in the pulp and reduce entrainment of gangue slimes in the concentrate. The lower density may also reduce overall reagent consumption. At the lower density cell volume has to be increased to provide the required retention time. The coarser fraction on the other hand can be treated at a higher density, reducing the total cell volume for that fraction.

1207

The coarse and fine fractions might also have different flotation rates, which would change target retention times for each stream. Depending on the particular cell manufacturer, the impeller/stator configuration may be different for coarse and fine materials. Slurry density. Flotation density can have an impact on recovery and product grade. It is one of the conditions that are normally evaluated in a test program since it has a major effect on cell volume requirements. For fine particle flotation, lower density typically provides better product grade from less entrained gangue. Coarser particles tend to work better at higher densities. Whatever slurry density is decided on it affects the cell volume needed to provide the chosen retention time. Higher density also results in higher power draw for cell agitation and pumping. The effect of density on volume is shown in the Table below. Reduction of density from 40% to 30% solids increases the required number of cells by 45%, or conversely for a given set of cells reduces retention time by over 30%. Flotation should be done at the highest density that allows efficient flotation in order to minimize cell requirements. However in some cases upstream processes determine the density. Typical operating densities for different minerals are shown in Table 1. Effect of slurry density on cell volume requirements and retention time Retention Time Relative to Weight % Solids in Slurry Slurry Volume Relative to 40% Solids 40% Solids 0.42 20 2.42 0.69 30 1.45

40 45

1.oo 0.85

1.oo 1.18

Assumes solids specific gravity = 3.0 Recycle streams. Where recycle streams are required, they must be included in the total flow for calculation of cell volume. Routing of recycle flows will have been established at the testing stage of flowsheet development. Return point(s) should provide an escape route for impurities or middlings that could otherwise keep building up in the circuit, until overloading of the system results in loss of clean product along with the unwanted or untreatable fraction. Recycle streams often have a different density than the main stream and the dilution (usual) must be taken into account when calculating total slurry flow volume. Possible bypass routings. It is often useful to provide bypasses for higher-grade concentrate from the initial cell or cells of a stage to go direct to final product or to a later subsequent stage, if the material characteristics allow. This can reduce the allowance for extra cleaner cell capacity to account for wide swings in throughput and/or feed grade. Typically this is liberated fast-floating mineral from the first cells of the roughers which could go direct to final product, reducing the load on regrind and cleaning stages. This is a similar principle to the use of flash flotation cells within a grinding circuit to remove early-liberated particles, often directly to final concentrate. In fact with judicious selection of equipment it may be practicable to design in a flash flotation stage at the head of the roughing circuit to produce a finished concentrate stream on a normal basis. Provision should also be made for return of lower grade streams to previous stages for further upgrading. This might be required if grade of concentrate from the final few cells of a bank is below acceptable level as a result of low heads or low throughput, for example.

1208

Table 1 Typical retention time and bank lengths for rougher flotation in mechanical cells.

Mineral

W

Feed % solids

Laboratory Average flotation time, plantflab min. time Denver cell scaleup Retention time min.

Open flow square cells

Usual no. cells per bank

RCS round cells'

Minimum Maximum no. per . no. per bank bank

Metallic minerals 32-42 13-16 6-8 2.1 12-18 12 Copper Lead 25-35 6-8 3-5 1.8 4-8 4 Molybdenum 35-45 14-20 6-7 2.6 12-18 12 Nickel 28-32 10-14 6-7 1.8 10-16 10 Tungsten 25-32 8-12 5-6 1.8 6-10 6 Zinc 25-32 8-12 5-6 1.8 6-10 6 Non-metallics 30-40 8-10 4-5 2 4-8 4 Barite Coal 4-12 4-6 2-3 2 3-6 3 Feldspar 25-35 8-10 3-4 2.6 4-8 4 Fluorspar 25-32 8-10 4-5 2 5-10 5 Phosphate 30-35 3 2 3-6 4-6 2-3 4 Potash 25-35 4-6 2-3 2 4-8 4 4-8 2.3 30-40 7-9 3-4 Sand (impurity) - _. Silica (iron ore) 40-50 8-10 3-5 2.3 8-14 8 Silica (phosphate) 30-35 4-6 2-3 2 4-6 4 Effluents as received 6- 12 4-5 2 4-8 4 Oil as received 4 For cleaning applications density i s typically 50-65% of rougher density. (Use 60% for RCS')

Retention time in cleaning is about 60-75% of roughing time (Use 65% for RCS')

20 10 20 20 12 12 10 6 10 12 6 8 8 14 6 10 8

Usual no. cells per bank

Minimum Maximum no. per no. per bank bank

8-12 6-8 10-14 8-14 7-10 6-8

6 5 8 6 6 5

16 12 10

6-8

4 3 4 5 3 4 4 6 4 4 4

10 6 10 10 6 8 8 12 6 8 8

4-5

6-8 6-8 4-5 4-6 6-8 8-10 4-6 4-6 4-6

16 10 16

Design Factor The above-mentioned inputs influence the magnitude of the design factor. Within reason, a larger design factor provides for greater operating flexibility, but also contributes to a higher capital cost. Excessively high design factor can result in equipment being oversized and having to operate at unsuitable conditions for significant periods of time, although this is more applicable to pumps than flotation cells. The greater the variability in the ore and especially when combined with a fixed speed SAG mill in the grinding circuit, the more desirable it is to be generous with the design factor. Note that this is not the same as the scale-up factor for converting laboratory/pilot plandplant test flotation time to new plant retention time. It does include such factors as feed rate variability through the grinding circuit, grade variations, and the like. Close attention should be paid to the mine plan and ore characteristics to determine how hardness and ore grade are related. If soft ores tend to be higher grade than harder ores, then a greater design factor will be required than for the opposite situation. Typically, the design factor for flotation is likely to be 1.25 or more although different parts of the circuit may require different factors. It should be evaluated in detail for each case based on the above variables. Once the design factors to apply to theoretical plant retention times have been established, determine the overall mass balance for solid, liquid and slurry at each stage of the flowsheet, including recycle and bypass streams. Each stream should be calculated for nominal and design values. Plant Capacity Terminology. In this discussion cell refers to the individual flotation machine, whatever type it may be. Mechanical cells usually have a single mechanism, although some have two or four. A unit or block of cells refers to one or more cells at the same horizontal level connected in series, with a feed arrangement at the head end and a discharge control at the outlet end. The discharge control function may be served by a junction or connection box joining two blocks of cells, in which case it is a discharge box for the upstream block and a feed box for the downstream block. The junctioddischarge box provides control of flow from the block of cells, which determines pulp level in those cells, and also provides the necessary hydraulic drop at the end of the unit to ensure controllable flow through that block. A bank of cells is dedicated to one common function, such as rougher, or rougher-scavenger, or cleaner, etc. Banks are made up of one or more blocks of cells connected in series, starting with a discrete feed box and ending with a final discharge box or control. Figure 1 illustrates, from left to right, feed, junction and discharge boxes as well as the vertical step after each unit or block of cells.

1210

Figure 1 Example of cell arrangement in bank Courtesy of Outokumpu Mintec Number of cells vs. capacity. Flotation banks have a functional minimum length, which is determined by the acceptable amount of recovery loss resulting from short-circuiting within and between the cells. This is more important for banks which discharge a finished waste product (tailings) from the circuit, such as scavenger tails, where there is no further opportunity to recover lost values from the stream. Where the tailing product goes for retreatment, the number of cells in the bank is less critical since the losses can be recovered again, although the increased recycle itself may require extra cell volume. There is also a practical maximum length limited by flow rates through the cells. The maximum and minimum lengths are different for different types of cells and for different minerals. In older open trough style cells with no baffles between cells, more cells were needed as uncontrolled flow between cells down the bank resulted in significant short-circuiting within the block of cells. As the flow connection between cells becomes more tightly controlled and mixing in the cell improves the cells act more like unit reactors and there is less short-circuiting. As the number of cells in a bank increases the overall flow approaches a plug flow condition and there is a high probability that any given particle will remain within the system for the mean residence time (Weiss 1985). Increasing the retention time per cell can also reduce short-circuiting, as long as overall retention time is sufficient (Lindsberg 1988). As cell size increased and cell design changed over the years, banks tended to fewer cells, particularly for slow floating minerals such as those of Cu, Ni and Mo. Working minimum lengths of 8 to 10 cells (for copper type applications) became more common although somewhat longer rows were still preferred (McAllister 1980). For large cells with significant retention time per cell, even shorter rows can still provide acceptable recovery. In this case, 6 cell rows would work for a retention of 3 to 4 minutes per cell. However, for short retention times (say less than 1 to 1% minutes per cell) the longer banks are needed to limit short-circuiting (Lindsberg 1988). Typical industrial values for minimum, usual and maximum number of cells per bank are shown in Table I.

1211

For contact-type cells the number of cells in a bank is determined by how many “stages” are needed to get to the target recovery. Retention time is determined by froth removal rate from the separation vessel and is typically about 2-3 minutes per cell. Size trends. Plant practice tends toward fewer and larger units to minimize capital and operating costs. Flotation equipment size can be maximized to the limits of either available equipment size or minimum row length needed to keep potential short-circuiting to an acceptable level. Larger and fewer cells require less floor area so building or supporting structure size can be minimized. With fewer units, it is more viable to provide advanced controls for each cell. Operator requirements are also minimized with fewer equipment units and increased instrumentation and control.

Cell volume calculation Once the target retention time has been established, the calculation of actual cell volume is relatively straightforward. For any given stage of mechanical flotation cells, total volume can be calculated as follows: 1 W e f f = -QQTEPF

60

Where: = Effective cell volume. If not known, assume Veff = 0.95 V Total. Veff N = Number of cells in the bank. = Dry ore feed rate, metric tonnes per hour Q - Dlytld 100 24 %availability Where availability = actual average hours operated per day124 hours per day

T E P

F

= Plant retention time [= test time x scale-up factor] 100 = Pulp expansion factor = = -(default) lOO-%Vol.~irinpulp 0.85 = Pulp volume per dry tonne solids 100 - 1 for water slurries. l + solids s g . pulp density wt% solids = Design Factor

“

1

For small cells, the usual range of cells per bank can be used to select a bank length and corresponding cell size. For large tank-type calls, a minimum bank length of 6-8 cells divided into total volume gives an indication of maximum cell size. Back calculation with a standard cell size gives the actual number of cells required. The design object is usually to use the least number of the largest cells that meet minimum bank length requirements. Repeat this procedure for each stage of flotation. Where specific conditioning time is needed at some point in the circuit, calculate the conditioner volume by multiplying flowrate to the conditioner by required retention time by design factor, in consistent units. Choose a standard conditioner size that provides the volume.

Cell Type Current practice tends toward use of mechanical cells for roughing and scavenging operations, with either mechanical or column cells as cleaners. Some more recent installations have also included contact-type cells in various parts of the circuit, although these are not yet as widespread

1212

as mechanical or column cells. Some applications have used all columns or all contact cells although these are uncommon outside the coal industry. Mechanical cells. Mechanical cells are available in a wide variety of sizes and configurations. They range from rectangular cells with flat or U-shaped bottoms connected in blocks, to individual unit cells with conical bottoms for use in grinding circuit applications, to the more recently develo ed tank-type cells which typically are cylindrical in shape and range in size to more than 200 m . They can be naturally aspirated or provided with forced air. Their common feature is that an agitation mechanism provides both air dispersion and solids suspension within the pulp. Tank configurations are designed to improve mixing and air dispersion characteristics or reduce solids sanding or both. Impeller and stator designs are varied, all with the intent of optimizing performance while reducing energy consumption and minimizing maintenance. In some cases, different mechanism configurations are used to treat different types of feed, such as coarse sands or very fine material. Mechanical cells often give a higher recovery over a broader size range than columns, particularly towards the coarse end of the distribution and as a result are more commonly used in applications where the tailing leaves that circuit, such as roughers or scavengers. There has been a strong move towards the recently developed round tank-type cells. These have allowed much larger sizes to be built. With the round shape and tightly controlled flows in and out, short-circuiting within and between the cells has been reduced compared to the more traditional square cells with end-to-end connections. The design also allows much more accurate modeling and scale-up of the cells for different sizes. This type of cell is illustrated in Figure 1. The shape lends itself to relatively low cost construction in large sizes since the cylindrical form is self-supporting without a lot of structural reinforcement. The design is lower cost than conventional square cells above about 40 m3 capacity and competitive down to a much smaller size. With the large cells, power consumption per unit volume decreases, as shown in Figure 2.

P

Specific Power Consumption {Typirsl for noiniital35%solids of 2.7 6.g.) 0.1

0.W c UI @.E8

0

a I 0 0 7'

2

0.06

0

I-

%

0.05

g

0.04

J

Y L 2

.=

0.03

0

0

10 a 0.01

Cell Volume (cft)

Figure 2 Effect of cell volume on specific power consumption Courtesy of GL&V/Dorr Oliver With the large surface area of the big cells, various approaches are used to improve froth removal. This includes froth crowding, which when adjustable allows fine-tuning of each cell to suit the conditions at its position in the circuit. Internal cross-launders and radial launders also reduce froth travel distance to aid concentrate removal and reduce froth overloading.

1213

While the main applications of these cells is in rougher-scavenger service, there has been a recent move to include them in cleaning circuits as well. Different impeller/stator configuration can be used to adapt to the particle size (Colwell 2001). Flash Flotation. Unit cells for flotation of cyclone underflow streams in the grinding circuit have the objective of recovering coarsely liberated fast-floating particles of heavy minerals early in the process. Typically the concentrate is coarse and at a high grade, often suitable to send to final concentrate, This can provide benefits that include increased recovery of values, reduced overgrinding, reduced load on regrind and cleaning circuits, improved dewatering from coarser overall concentrate. An example of this type of cell is shown in Figure 3.

Figure 3 Example of unit cell for Flash Flotation Courtesy of Outokumpu Mintec Retention time in the cell is usually about 2-4 minutes, with grade increasing at shorter times. Longer flotation times result in higher recovery but to a lower grade concentrate. This can be upgraded in a dedicated cleaner cell to produce final concentrate. Because of the high densities and presence of very coarse particles these cells can be difficult to control and, especially in smaller sizes, can be subject to plugging unless the control valve is carefully sized for the flow. Several approaches have been implemented to get around these problems. A dual outlet arrangement allows a very dense, coarse stream to return rapidly to the mill feed while a less dense stream with particle sizes more suited to flotation is directed to the impeller and flotation zone. This lower density stream recombines with the coarser stream to maintain the original density to mill feed, but having a lower density for the actual flotation.

1214

Another approach passes the cyclone underflow through a DSM screen ahead of the unit cell, which removes oversize and effectively reduces the density of feed to the unit cell. Again, the bypassed material and unit cell tailings are recombined to maintain mill feed density. Note that insertion of a unit cell to treat cyclone underflow normally requires increased height of the cyclones above the grinding mill feed chute, unless only a bleed stream is treated and tails are returned to mill discharge rather than mill feed. Higher cyclones mean increased pumping costs. Some dimensions and data on unit cells are shown in Table 10. An additional complication that requires careful consideration is the effect that removing a significant amount of high-grade material can have on sampling. If head samples are based on cyclone overflow, the bypassed concentrate will not show up in the feed sample. Sampling the concentrate itself can be complicated by the relatively coarse particle size, particularly where gold is present. The problems are not insurmountable but should be well thought out and allowed for in advance. Columns. Columns are good at producing high-grade concentrates at reasonably high recoveries. The best application for columns is recovery of slow floating fine particles where entrainment of gangue limits the separation (Finch 1995). Feed slurry enters the column part way up its height and flows down toward the bottom discharge. Air is usually injected at the bottom of the column and flows up through the slurry, collecting floatable particles as it rises. Some columns have the air injected into the feed or recycled tailings slurry outside the column. Wash water distributed onto the upper froth layer helps remove entrained gangue from the froth as well as limiting froth viscosity. Froth overflows the top of the column to a collection launder. Surface area of the column is an important factor when the concentrate load approaches the maximum carrying capacity of the column. These details are all discussed in detail in the chapter on column design, The ore should be tested with a laboratory or pilot column to ensure that it is suitable for treatment with columns and to develop scale-up data. With their large retention time and good froth rejection of entrained gangue they can often provide the required upgrading in fewer stages of cleaning than would be needed for mechanical cells. For example, molybdenite has traditionally needed many stages of cleaning with mechanical cells to produce an acceptable product, often up to 8 to 12 stages. However, where suitable the same upgrading can often be achieved in as few as 3 stages of columns if the circuit is properly set up. Columns have relatively low operating costs as the main energy inputs are for pumping and compressed air. Capital cost for a given capacity is less than for mechanical cells through lower equipment cost and smaller building or structure size. However, they require advanced instrumentation and controls to derive maximum benefit from them (Murdock 1988). Because of their height relative to the rest of the plant, they are often installed outside the concentrator building itself, where climate and conditions allow. Contact cells. More recently developed contact-type cells are a development from column flotation. There is still a contact or collection zone where very fine air bubbles are contacted with the feed slurry and a separation zone where the mineral bearing bubbles separate into the froth phase for collection to concentrate. They differ from columns in that the contact zone is relatively small and subjected to intense shear for maximum bubble-particle contact and is in a physically separate component of the cell from the separation zone. The two types are distinguished by whether the air is self-aspirated or forced into the slurry under pressure. In each case, the suppliers have established specific test equipment and procedures to provide design information for evaluating suitability of the ore for treatment with the cells and subsequent sizing of the equipment. Figure 4 shows an example of the self-aspirated Jameson cell in a rectangular configuration. Figure 5 shows a schematic of the forced air Minnovex Contact Cell.

1215

Figure 4 Example of self-aspiratedJameson contact-type cell Courtesy of MIM Process technology

Both types provide a high capacity from a relatively small size unit so provide potential capital cost benefits over conventional cells, particularly for retrofits. The overall number of installed units is still relatively small compared to mechanical cells but is growing. The selfaspirated Jameson cells have been widely installed in coal operations, particularly in Australia, where they have been successful at treating coal slimes streams. A number of cells have also been installed in metal applications at various stages of the circuit. Organic recovery from aqueous phases in SX/EW plants is another application where contact-type cells have made inroads. Jameson cells are available in rectangular or round tank configurations. Tank diameter ranges from 1 m up to 6.5 m. The number and diameter of the downcomers in which the air and slurry are contacted determine the volumetric flow rate to the cells. The area of the separation tank is sized for the expected solids loading on the froth. Downcomer diameter goes up to 500 mm which can handle 250-300 m3/h of slurry. Most of the smaller downcomers have flow capacities in the range of 15-80 m3/h (Jameson 2001). The forced air Minnovex Contact Cells are a more recent development, with applications more in roughing or flash flotation in grinding circuits although with the froth-washing capability they are also suitable for cleaning. These cells are mainly in the range of 1.5 - 4.5 m diameter, with height on the order of 1.5 times diameter. Diameter is sized for expected froth carrying capacity while height is set to provide the retention time to achieve the separation. The self-aspirated Pneuflot cells by Humbolt Wedag have been applied mainly to coal, industrial minerals and wastewater, with most installations in Euro e. These cells range in size from 0.8 to 6.0 m while pulp flow capacities range from about 5 m /h for the smallest unit up to

P

1216

Next Page

1200-1500 m3/h for the largest. They generally have a single reactor or downcomer per cell (KHD 1997). Other cells are also available which represent variations of detail on the same basic concept. In general, scale-up of the cells from the cell-specific test results to full scale production models is reported to be very reproducible and more accurate than mechanical cells as it is not dependent on retention time but rather on entrained air rates, air rise velocity and concentrate loading o n froth (Gray 1998).

Air

Wash Water (Optional) Spargers

Feed Slurry Froth Launder

Contactor

Concentrate

Figure 5 Schematic of Minnovex forced air contact cell Courtesy of MinnovEX EQUIPMENT SELECTION: TYPICAL CELL DATA For current information on specific cells it is recommended to contact the manufacturers for their latest brochures. Specific information may also be available on the websites of the various

1217

Column Flotation Glenn Dobby’

ABSTRACT Industrial application 0, ~olumnflotation began in earnest in the early D’s, anc is now acceptel as a conventional application. For example, most of the world’s major copper producers-use column flotation as the final stage of cleaning. In addition, the technology is common for final cleaning of zinc, lead and molybdenum sulfides, and for processing of iron ores and phosphates. This article includes a description of the key features and concepts of column flotation, and then reviews recent industrial applications of column flotation, covering pilot testing and scale-up, circuit design of selected applications, and instrumentation and control. INTRODUCTION The invention of the flotation column by Pierre Boutin and Remi Tremblay (patented in 1962) came about from earlier work conducted on solvent extraction at the research laboratory of Eldorado Mining, in which they applied the column as it is known today to solvent-in-pulp processing of uranium ore. Slurried ore was fed near the top of a laboratory column and solvent droplets were generated at the bottom of the column. The density difference caused a countercurrent flow of solvent droplets and slurry, and a solvent phase was created at the top of the column, with a distinct solvent-pulp interface. An aqueous diluent was introduced near the interface in order to minimize contamination of the solvent with ore particles. Sluny was drawn out the bottom of the column and solvent overflowed the top lip. (Boutin 2002)

Boutin and Tremblay then applied the concept to ore flotation, replacing the solvent droplets with air bubbles and replacing the diluent with water. The first successful applications were on amine flotation of silica from iron ore samples. The first notable industrial success of flotation columns was by Column Flotation Company of Canada on moly cleaning at Noranda’s Les Mines Gasp6 (Cienski and Coffin, 1981; Coffin and Miszczak, 1982). Today column flotation has become an accepted means of froth flotation for a fairly broad range of applications, in particular the cleaning of sulfides (copper, zinc, lead and molybdenite) and the flotation of iron ore, phosphate and coal.

0

0 0 0

Flotation columns differ dramatically from mechanical flotation machines in several ways: there is no mechanical agitatiodshear the cell is relatively tall and narrow gas bubbles are generated by sparging froths typically are deeper, and wash water typically is liberally applied to the surface of the froth.

There has been a wealth of research and fundamental studies on aspects of column flotation during the past 15 years or so (some of which has been described by Finch and Dobby, 1990 and by Finch, Uribe-Salas and Xu, 1994.). However, the fundamental aspects is not the focus here; rather, this article will review recent industrial applications of column flotation, covering pilot testing and scale-up, circuit design of selected applications, and instrumentation and control. We begin with a description of the key features and concepts of column flotation. 1

MinnovEX Technologies Inc., Canada

1239

KEY FEATURES AND CONCEPTS A schematic of a flotation column is shown in Figure 1. Industrial flotation columns are 6-14 m in height (from the bottom discharge to the top lip), and range in diameter from 0.5 to 5 m. Recent large scale applications have typically been 4 to 4.5 m diameter and approximately 12 m tall. [There are many installations of columns that are rectangular. The fabrication cost for rectangular columns is significantly higher than for circular columns, as approximately twice the mass of steel is required; however, there is no evidence that rectangular columns will provide better metallurgy than from circular columns.]

As with mechanical cells, two distinct zones are evident: the collection zone (extending from the spargers to the frotkpulp interface) and the froth zone. Interaction between these two zones, especially the recycle of collected particles from the froth to the pulp, is particularly important in understanding design and operation of a flotation column; this is addressed in the Scale-up section. The froth of a flotation column is usually water washed, with approximately as much wash water as there is water reporting to the froth. The water is most commonly added via pans, perforated with 4-8 mm diameter holes, located 20-30 cm above the froth,thus generating a “rain“ of water onto the surface of the froth. The key objective of water washing the froth is to minimize recovery of hydrophilic gangue particles into the concentrate2via hydraulic entrainment; the wash water replaces feed water that would have otherwise reported to the concentrate, carrying gangue particles with it. The effect of the water addition is to create a froth where the gas bubbles do not coalesce to the same degree as in non-water washed froths (Yianatos, Finch and Laplante, 1986 and 1987). Hence, a column froth is usually very stable even when deep; an industrial column froth is typically 0.6-1.0 m deep. The wash water streaddroplets must be large enough to penetrate the top layer of the froth, because the washing action takes place primarily at the froth/pulp interface (Yianatos, Finch and Laplante, 1987). If the water stream is too light then there will be a tendency for the water to simply bypass the froth directly into the concentrate. For heavy froths, some operators implement wash water delivery via perforated pipes immmersed several centimeters below the top of the froth. Wash Water

I

1

I

.............................. Froth Zone

’Concentrate

!

.............. \I /\

4

Feed

Collection Zone

.............

4

Air

The froth product will be referred to in the article as the concentrate, unless otherwise stated

1240

Some key terminology and concepts follows. Supe&ial velocity Phase flowrates can be expressed as a superficial velocity by dividing the flowrate by the column cross-sectional area. Hence, gas flowrate is often expressed as a gas velocity J,, with usual units of c d s . The range of J, observed in industrial flotation columns is typically between 1 and 2 c d s (this is the flowrate at the top of the column; a static head of typically 1 atm means that the true gas rate at the bottom of the column is about half that at the top.). Likewise, feed rate of slurry can be expressed as a velocity. Bias The difference between wash water flow rate and concentrate water flow rate is referred to as the “bias”. When wash water flow exceeds concentrate water flow, the bias is positive; the bias is negative when the reverse occurs. A zero bias means that the two flows are the same; as stated earlier, a common approach with column flotation is to operate in the range of a zero bias, perhaps a bit higher or a bit lower. As with other flows, bias can be expressed as a superficial velocity, JB (again, expressed in c d s ) .

Another method to describe the relationship between wash water flow and concentrate water flow is to quantify the wash water in terms of displacement washes. One displacement wash means that the wash water flow equals the concentrate water flow (one displacement wash is the same as a zero bias). Carrying Rate and Carrying Capacity The available surface area of a flotation machine is an important consideration in most cleaning circuits and, because of its small diameter:height ratio, it is particularly so with flotation columns. The concentrate solids flux is referred to as the carrying rate C,, described in units of tph/m2. Industrial columns typically operate at between 1 and 3 tph/m2, depending on the level of wash water addition and the particle size of the concentrate (smaller particles will result in lower carrying rates). The maximum carrying rate is referred to as the carrying capacity CM.Effects of particle size and water addition on carrying rate and carrying capacity are discussed further in the following section. Gas Holdup and Bubble Size The volume of the collection zone occupied by gas bubbles, referred to as the gas holdup and expressed as a percentage, can be used as a diagnostic feature. This is because the gas holdup is a b c t i o n of both gas rate and bubble size; hence, knowing the gas rate and the gas holdup can lead to an inferred calculation of average bubble size. Bubble size, as with other flotation machines, clearly plays an important role in column operation. Smaller gas bubbles are generally preferred. However, in primary cleaning of mineral systems with high kinetics, very small gas bubbles will tend to hurt froth mobility, and hence are undesirable.

Bubble generation in flotation columns was originally achieved via sparging through pierced rubber or through woven fabric. That approach is no longer very common, as the column must be shut down and drained to assess or replace sparger units. Another early method, developed by the US Bureau of Mines, was injection of an air-water mixture through multiple 1 mm diameter orifices. Today there are two other principle bubble generation technologies. The most common method of bubble generation, developed by MinnovEX in the early 1990’s, is jet sparging of air through orifices. Air is forced under high pressure (30-100 psig) through annular or circular orifices, creating a jet of air through the slurry. The high shear that exists at the gas-slurry interface results in generation of gas bubbles. Bubble diameter is affected by the jet velocity; the higher the jet velocity the smaller the resulting bubbles. The spargers are fit into the column through a nipple and valve assembly, extending only several centimeters into the pulp. With this configuration, the spargers can be removed and inserted while the column is full of slurry and under operation. A large column will use 15-20 spargers. Variations of the original approach are now supplied by a few manufacturers, with atomatic shutoff of air on the loss of air supply a recent feature.

1241

The second principle method of bubble generation employed today is through pumping of a portion of the underflow through static mixers back into the pulp, employed by the Microcel technology (Yoon and Luttrell, 1994). Air is injected at the static mixer, and the resultant high shear through the mixer generates very small gas bubbles; hence, the bubble generation is external to the column. A typical column will have several static mixers fed by a common pump. An advantage of the static mixer approach is that it can produce smaller gas bubbles. However, this is at the expense of pumping a very high slurry flow through the mixers. The Microcel approach has been applied primarily to coal flotation, with a large and successfid installation at BHP Coal’s Peak Downs coal preparation plant (Brake, 1998). Whatever method of air addition employed, the pressure drop of a column will require that the air is supplied via a compressor (not a blower). Selection of the correct compressor capacity dearly is important. PILOT TESTING AND SCALE-UP Metallurgical testing of column flotation has generally been through continuous pilot plant operation. For retrofitting at existing plants, the approach has been to conduct pilot testing on site, feeding a bleed stream from the plant circuit to the pilot plant circuit (Kosick et. al. 1991) Typical pilot columns are 10 cm diameter by 6 m tall, although pilot testing has been conducted with columns as small as 5 cm and as large as 30 cm. There have been some attempts to conduct pilot testing and scale-up though operation of a batch laboratory column (Flint, 2002).

The ideal approach for design and sizing of columns for cleaning on a green-fields design is via pilot testing. However, if the column application is for final cleaning and the head grade is low (and rougher concentrate mass recovery consequently is low), the quantity of rougher feed required to generate sufficient rougher concentrate for column pilot testing is generally too large to justify a pilot plant. Kinetic data obtained from bench mechanical flotation machines can be employed, but only with specific precautions and the application of recently developed advanced modelling tools. While it is difficult to apply the bench data directly to column sizing, the selectivity that is obtained from the bench testing can be used for helping design the number of stages of cleaning. The new modelling tools have been developed to assist in the design of circuits using flotation columns (Dobby, Kosick and Amelunxen, 2002), and now the design of sulfide cleaner circuits for green-field plants is entirely feasible without the need for pilot testing (aside from some complex ores). Scale-up of flotation columns has been well documented (Finch and Dobby, 1990). The general approach is to apply a first-order kinetic model to particle collection, and select froth zone recoveries appropriate for the size of the columns and the froth mobility. Particle retention time may be significantly different from that of the water, especially for coarse particles, so the retention time of particles needs to be calculated and applied. Hydraulic entrainment of fine particles is accounted for by quantifying the recovery of water to the froth, for columns operating with a bias close to zero, the effect of entrainment is usually minor. A key factor in column design is the froth recovery, which directly controls the extent of recirculation within the column. As a general observation, the froth recovery of pilot columns is considerably higher than in large industrial columns, due to the stability imparted to the froth by the walls in a pilot column. Therefore, in sizing large columns from pilot plant data the designed retention time will be longer for two reasons: (a) short-circuiting Since the large column has a fluid flow pattern that is closer to a stirred tank than a plug flow vessel (opposite to that of a pilot column), a longer retention time is required to achieve the same collection zone recovery as in the pilot column

1242

(b) lowerfi.oth recovery Since the plant column will have a lower froth recovery (i.e. higher internal circulating load) it is necessary to have a higher collection zone recovery than the pilot column in order to attain the same overall column recovery as with the pilot column. Three issues deserve further elaboration: carrying capacity, bias flow and column height. Carrying Capacity Operation of full scale columns has shown that carrying capacity (CM)for a large column can be substantially lower than that of a pilot column. For example, Espinosa and Johnson (1991) reported a CMfor 2 m and 2.5 m diameter plant columns on lead and zinc cleaning that was about 50% of the CMfor a 5 cm pilot column. However, the trend of higher carrying capacity for higher bois still valid for plant columns.

An important and often neglected issue when quoting CM,whether for a pilot column or a full scale column, is the value for the operating bias. Figure 2 is an example of measured concentrate solids flux Ca for a copper cleaning pilot column, plotted as a function of bias. In this case feed grade to the column is high, so Ca is reasonably close to CM,and CMwould be expected to follow the same trend as in Figure 2. A full scale column circuit was installed subsequent to the pilot column work, and the plant operating region for a 3.0 m diameter column is also shown in Figure 2; it is at a significantly lower level than that obtained in the pilot column operation and is in reasonably good agreement with the observation reported by Espinosa and Johnson (note that the plant columns did use internal launders in addition to the external launder). Required Bias Flow Figure 2 clearly indicates that carrying capacity will increase as operating bias is decreased. Bias can be decreased by decreasing the wash water rate or by increasing the gas rate. An example of the latter is shown in Figure 3, which was obtained from column cleaning on copper rougher concentrate with a 10 cm diameter pilot column, upgrading from about 14 %Cu to 34 %Cu. The data in Figure 3 was derived with two sparger fabrics, with Fabric 2 being considerably more permeable than Fabric 1 and thereby generating larger gas bubbles than Fabric 1 (the generation of larger gas bubbles could be inferred from gas holdup measurements that were made for each test). The bias is expected to decrease when smaller gas bubbles are generated (and gas rate remains the same) because the larger interfacial area will carry more water from the collection zone into the froth zone. 6

"

I

-0.1

-0.05

0

0.05 0.1 Bias (cmls)

0.15

0.2

0.25

Figure 2 Concentrate solids carrying rate versus bias with a pilot column used for copper cleaning, and for the resulting plant column operation (Dobby and Kosick, 1995).

1243

0.2 Wash water at 0.08 to 0.16 cmls I

t

0.1 A

2 0.0

I

-

0

v

u) (D

g -0.1 -0.2 -0.3 0.4

0.9

1.4 Gas Rate ( c d s )

1.9

2.4

Figure 3 Effect of gas rate and bubble size on bias in a pilot column.

_ zie

-z?

80

36

70

35 Grade (Yo CU)

Q)

>

0 0

2

60

34

5 50

33

kn n

40 -0.15

32 -0.1

0 Bias (cmls)

-0.05

0.05

0.1

Figure 4 Example of the effect of bias on grade and recovery for copper cleaning in a pilot column. It has been demonstrated several times (e.g. Furey, 1990, Espinosa and Johnson, 1991) that operation of a column at zero to slightly positive bias will usually maximize the concentrate grade. A further increase in bias, through wash water addition, to a point significantly above a zero bias will generally result in minimal grade increase but a substantial recovery loss. An example of this is shown in Figure 4, for copper cleaning in a 10 cm diameter pilot column. In general, a good guideline is to provide sufficient wash water such that the column operates at close to zero bias. More definitive guidelines for a given installation must take into account the operating grade-recovery curve and economic efficiencies specific to that operation; in some cases

1244

it will be more profitable to run with a slightly negative bias. Such a situation arises when the column circuit must temporarily process high grade feed. As another example, if insufficient column capacity has been installed, it will be difficult to operate the columns with a positive bias while still attaining target recovery. Column height The height of a flotation column is generally determined by required retention time, accounting for both short-circuiting and a significant degree of froth dropback. That being said, it has been reasonably common to adjust the designed column height slightly to better fit within a plant layout. There are situations where tall columns are undesirable. This arises when the feed grade is high and the floatable mineral has very fast flotation kinetics, which will cause the froth to become hlly loaded with solids with a short retention time. Any further retention time, i.e. more collection zone height, is ineffective, as the gas bubbles become fully loaded before reaching the froth-pulp interface. An installation that takes advantage of this is the zinc cleaning circuit at the Doe Run Company’s Fletcher concentrator in Missouri. The cleaner circuit consists of two columns in a counter-current cleaning configuration, with 1st cleaner tailings recycled to rougher feed. Rougher concentrate is typically 15% Zn and final concentrate is typically 60 %Zn. Both columns are 1.7 m diameter. The 1st cleaner column is 12 m tall (from floor to lip) and the 2nd cleaner column is 8 m tall. This allows gravity flow of 1st cleaner concentrate to the 2nd cleaner, via a head tank, and saves a pumping stage. The overflow of the 2nd cleaner column is setup at the same floor level as the existing mechanical cells. Another installation that has successfully applied a combination of tall and short columns is the iron ore flotation plant at Samitri, Brazil (described further in the following section). CIRCUIT DESIGN AND APPLICATIONS Most installations today use a combination of columns and mechanical cells within distinct circuits. This has arisen partially from the retrofit application of columns, where stages that produce an intermediate concentrate can be implemented with existing mechanical cell equipment. However, the application of columns together with mechanical cells has been driven primarily by the desire to match optimum cell properties to the duty required. This has resulted in the final stage(s) of cleaning by column flotation and cleaner scavenging in mechanical cells.

With the advent of flotation columns has come the option of arranging cleaning circuits in a scavenger configuration in addition to the conventional counter-current cleaning approach. The selection of cleaner circuit configuration in either a cleaner mode or a scavenger mode is determined from the level of upgrading required in conjunction with the selectivity of the particular process, and the total quantity of solids to be removed as froth product. In applications where there is a high level of solids recovery to the froth, there will be significant capital cost savings in being able to use a scavenger configuration, with which the solids are quickly removed from the circuit. Further circuit design issues will be highlighted in the section as specific applications are described. Examples that follow are for sulfide cleaning, iron ore flotation and phosphate flotation Examples of Copper and Zinc Cleaning The most accepted application of column flotation is for final cleaning in copper sulfide flotation; a majority of large copper plants today utilize columns. Two of the most common cleaner circuit flowsheets are shown in Figures 5a and 5b. Both the Candelaria concentrator and the Collahausi concentrator in Chile use 5a, while Antarnina (in Peru) and Escondida Phase 3.5 (in Chile) employ the flowsheet in Figure 5b. The Antamina copper installation uses eight 4.3 m diameter columns, four for each stage, while Escondida has a total of 14 columns (usually 12 are in use), each 4 m square in cross-section. 1245

For zinc cleaning, the column flowsheet employed depends on the rate of zinc flotation and the selectivity between sphalerite and the gangue minerals (usually pyrite). When the sphalerite flotation rate is fast, then a reasonably high stage recovery can be obtained in the first column, allowing the use of the flowsheet shown in Figure 5b. When the sphalerite is slower floating and selectivity is not high, as typically occurs with complex sulfides, then the stage recovery of the first column is not very high and it is common to employ two stages of columns in series, Examples of this are from the Laronde Division of Agnico Eagle (Figure 6 ) and BHP-Billiton’s Les Mines Selbaie (Figure 7). Rougher Concentrate Regrind 1 71 - 1L

L Final Concentrate

tI b i

-,----------,-,J

I

I

Cleaner Scavenger Tails (usually final tails)

Figure 5a A simple column cleaning circuit, applied in several porphyry copper concentrators.

Final Concentrate

t

Reground Rougher Concentrate

b

4

SECOND STAGE COLUMNS

FIRST STAGE COLUMNS - 1

r

v

SCAVENGERS

.

Scavenger

Figure 5b A three stage column-mechanical cell circuit applied on some copper and zinc ores.

1246

3-stage Conditioning

-

Roughing

Primary Cleaning Column cell A

-

Secondary

b

Secondary Cleaning

L-L \

3

-c

Regrind

4

Figure 6 Zinc flotation circuit at Agnico Eagle’s Laronde concentrator ( Werniuk, 2000)

Feed (Cu circuit tailings)

Zn Column Recleanen

I

I

Figure 7 Zinc flotation circuit at the BHP-Billiton’s Les Mines Selbaie (Wright, 1995).

1247

For many column applications there is a choice between implementing either two or three stages of columns/mechanical cells. This is particularly so in copper and zinc cleaning. Using three stages will increase the overall project capital cost (though not necessarily proportionately) with the benefit of improved selectivity. A clear example of this is shown in Table 1, which summarizes results from pilot column cleaning on zinc 1st cleaner concentrate at Les Mines Selbaie in Canada, and compares results from the main plant mechanical cleaners, a CUscavenger-open pilot column circuit and a CCC/scavenger-closed pilot column circuit. Both column circuits performed significantly better than the plant 2nd and 3rd cleaners (as would be expected, because the plant cleaners were undersized for the application). The three-column circuit produced a concentrate about 2 % higher in zinc than the two column circuit (even though feed grade was about 3% lower). The feed rate to both circuits was similar, so the three column circuit had a higher overall residence time. However, only part of the improvement of the CCC circuit over the CC circuit can be attributed to the increased residence time; a significant portion of the improvement was due to the better selectivity attained by increasing the number of circuit stages. A three stage (CCWscavenger) circuit was installed at Selbaie, treating 1st cleaner concentrate (Chevalier and Dobby, 1996.). [This example was from a complex sulfide application, where zinc flotation rate and selectivity were not high. A faster floating zinc ore with better selectivity may not see as significant a difference between two and three stages of cleaning.] Table 1 Example of column and plant performance in zinc cleaning, showing the difference between two stage and three stage column circuits. Circuit Feed Rate Feed %Zn Conc %Zn Zn Recovery (L/min) Plant mechanical cleaners 42.6 55.5 57 Two Columns 3.3 41.4 57.2 82 Three Columns 3.3 38.2 59.1 84 Molybdenite Cleaning As mentioned in the introduction, the first notable industrial application of column flotation was on molybdenite cleaning at Les Mines GaspC, where three columns replaced approximately 10 stages of mechanical cells. Since then, column flotation has become the standard approach for molybdenite cleaning circuits. Typically three sequential column stages conduct the final cleaning, treating either rougher concentrate or 1st cleaner concentrate generated in mechanical cells. An approach to determining the number of stages required has been described by Amelunxen (1990). At the Endako concentrator in British Columbia, Canada five columns are used, two at 2.7 m diameter, two at 1.5 m diameter and one at 1.1 m diameter, all 10 m tall. An additional advantage of using columns for molybdenum cleaning from bulk copper-moly concentrate is that the columns are well suited to the use of nitrogen in place of air (which is sometimes practiced in order to significantly reduce the consumption of sodium hydrosulfide, used for copper sulfide depression). Iron Ore Flotation The selectivity of silica flotation from hematite can be extremely high, and pilot cotumn flotation of silica has shown extremely impressive separation performance, especially on applications of Brazilian iron ore. This is because column wash water is very effective at rejecting from the froth the high level of hematite fines that occur in the feed. However, there have been several failures in industrial applications, primarily due to errors in scale-up. Scale-up must be done with care, for two reasons: the target level of silica in the underflow is usually <1%, so when the feed grade is high (1525 %silica) the recovery of silica to the froth product must be higher than 90 %, at which point short-circuiting of slurry in the collection zone plays a dramatic role; and

1248

as discussed earlier, the plant column will provide a considerably lower degree of stability to the froth,which means that there will much higher dropback from the froth to the pulp Hence, the increased short-circuiting that occurs in the industrial column and the significantly higher degree of internal circulation combine to prevent successful separation in one (or even two) column stages. In addition, the lower froth recovery in the industrial column is usually magnified for coarse silica, often resulting in difficulties with removal of coarse silica. The solution to the problem described above is to use several stages of flotation. A good example of this is from CVRD’s Samitri circuit in Brazil, Figure 8. All of the four columns are 4.6 m diameter. The first (rougher) rougher and the cleaner-scavenger column are both water washed and only 8 m tall, while the cleaner and re-cleaner columns are not water washed and are 15 m tall. Froth from the two tall columns feeds the scavenger column by gravity flow. The feed grade to this circuit is typically 20 % silica or higher. Most of the silica is removed in the front end column circuit, and the mechanical secondary circuit ensures removal of coarse silica. Phosphate Flotation Flotation columns have been applied successfully for phosphate flotation in both Brazil and Florida. A good example is the Serrana concentrator (Brazil), where six columns are in operation (Guimariles and Peres 2000, GuimarZLes et al. 1999). Each (rectangular) column is 3 m by 4.5 m by 14.5 m tall. The columns were all installed as retrofits to the existing mechanical cells; ultimately 66 8.4 m3 cells were replaced by the six columns. The flowsheet for the plant is shown in Figure 9, and typical operating conditions for each column are summarized in Table 2.

CLEANER WASH

RECLEANER

J

WASH

J

4-

FEED

\ b

I‘.

CLEANER SCAV

ROUGHER \

v

T I

v FINAL TAILINGS

FINAL CONCENTRATE

Figure 8 Flowsheet for silica flotation from iron ore at CVRD’s Samitri (Alegria) concentrator.

1249

Table 2 Typical operating conditions for the phosphate flotation columns at Serrana

Item

Units

I

Air Wash water Bias Residence Time Carrying Rate Lip loading Froth depth

Apatite: Natural Fines 0.90

I cds

I

Apatite: Fines from Grinding 0.90

I

I

Apatite: Coarse 0.97

I I

Apatite: Reground 0.82

+0.05

630

I

Barite: Coarse

+0.17

1000

440

510

ROD MILL

MAGNETIC SEPARATION (LOW INTENSITY)

DESLIMING

CLASSIFICATION

ov

ov 'I BALL MILL

APATITE FLOTATION *.W)

Fines from grinding DESLIMING

I

DESLIMING

.c

ov

CONDITIONING

1 BARITE FLOTATION

PTNE CONCENTTUTZ

(ahk UP)

1

+ L I DESLIMING

CONDITIONING

REGRINDING

l l I

CONDITIONING

1

I

& COARSE CONCENTRATE

Figure 9 Flowsheet for phosphate flotation at the Araxh concentrator of Serrana (GuirnarHes et al, 1999). 1250

INSTRUMENTATION AND CONTROL The degree of instrumention on an industrial flotation column can range from very basic to reasonablely complex. The minimum control requirement is that for pulp level. However, a typical industrial column will have automatic control of level, air flow rate and wash water flow.

The low ratio of surface area to volume inherent with a column means that the pulp level (which directly affects froth depth of course) will change quickly and significantly with changes in feed flow rate. Hence, this puts emphasis on good tuning of level controllers, but more importantly on ensuring that a reasonably stable feed flow is provided to the column. Feed to a column may be directly from a pump; however, multiple columns in parallel are usually fed via a slurry distributor. Typically, pulp level sensing is through a floadplate assembly that acts as a target for an ultrasonic sensor. The level control element normally is a pinch valve, often raised a few meters from the bottom of the column in order to reduce pressure drop (and hence wear) across the valve. In some applications a variable speed pumps is used used in place of the pinch valve. Other sensors and control elements may include pressure sensors for detecting bulk density of the column, and automatic shutdown and startup knife-gate valves for slurry.

Various strategies exist for control of a flotation column. Level control, as stated earlier, is clearly the most important. There has been much discussion on control of bias; the dificulty with this is that an industrial sensor for bias is presently unavailable. In fact, for many applications the bias does not need to be controlled. Consider the following example from sulfide cleaning. Assume that wash water has been set for typical operating conditions to provide approximately a zero bias. When concentrate production will be higher because of higher feed grade, the bias will shift to negative; however, this is probably where the operation should be in order to ensure good recovery across the column under conditions where concentrate grade should be less of an issue than recovery. And on the other hand, when concentrate production will be lower, the bias will shift to positive. Again, this is not necessarily bad, as there is more capacity available and the higher level of water washing will be beneficial. Advanced control that manipulates wash water, air and level set points needs to be built into an overall circuit control strategy, which will be unique for each application. LIMITATIONS/CAUTIONS Development of column flotation in the minerals industry has not been without it’s failures. There are several aspects of application and implementation which require particularly close attention, in either pilot testing or scale-up. Two of these follow. Feed conditioning Some applications will require fairly intense feed conditioning and, unlike mechanical cells, a flotation column will provide very little conditioning. Hence, when conditioning is necessary it must be entirely complete before being fed to the column. If conditioning is applied in pilot testing then it must also be applied in the full design, to the same extent as used in the pilot testing. In some cases this is impractical, in which case the need for conditioning, and the degree of conditioning applied, needs to be closely evaluated at the pilot stage. Fine particle flotation Depending on the nature of the surface chemicals employed, flotation of very fine particles in flotation columns often is not as effective as in mechanical cells. This could be related to poorer bubble collection efficiency for fine particles in the absence of high shear, or the need for “surface cleaning” of slimes via a high shear environment. The size at which performance deteriorates significantly is a function of the nature of the mineral system and surface chemicals employed, and must be determined through pilot testing.

1251

CONCLUDING COMMENTS The engineering of flotation columns has advanced significantly over the past 20 years. 0 Through the application of well defined scale-up methodologies, flotation columns can be reliably designed and sized. 0 Sparger technology is reliable and robust. 0 Methods for instrumenting and controlling columns are well defined and routinely implemented. The next challenge will be advanced process control of flotation circuits employing flotation columns. ACKNOWLEDGEMENTS The author would like to recognize efforts of the team responsible for development of flotation column technology at MinnovEX during the past 14 years, especially Glenn Kosick and Dale Coupland. In addition, support of staff from the many operating mines where MinnovEX has applied column flotation technology is gratefully acknowledged. REFERENCES Amelunxen, R.L. 1990. Moly upgrading profiles with flotation columns, CMP Proc. 22nd Annual Meeting, p.258 Boutin, P. 2002. Private communication to the author Brake, I.R. 1998. The development and commissioning of a new Microcel column flotation circuit for BHP Coal’s Peak Downs coal preparation plant, XIIIth Coal Preparation Congress, Brisbane, Australia Chevalier, G. and Dobby, G. 1996. Flotation column installations for zinc and copper cleaning at Les Mines Selbaie, at Copper ‘96, CIM Conf of Metallurgists, Montreal Cienski, T. and Coffin, V.L. 1981, CMP Proc. 13th Annual Meeting, p.240 Coffin, V.L. and Miszczak, J. 1982. Proc. 14* Int. Mineral Processing Congress, Toronto, paper 421 Dobby, G. and Kosick, G. 1995. Case studies on circuit design using flotation columns, CMP Proc. 27th Annual Meeting, p. 171 Dobby, G., Kosick, G. and Amelunxen, R. 2002. A focus on variability within the orebody for improved design of flotation plants, CMP Proc. 34th Annual Meeting, p.77 Espinosa-Gomez, R. and Johnson, N.W. 1991. Technical experiences with conventional columns at Mount Isa Mines Limited, Proc. Column ‘91 (Agar, Huls and Hyma, eds.), Sudbury, CIM Flint, I. 2002. Batch flotation modeling, PhD thesis, UBC, Canada Finch, J.A. and Dobby, G.S. 1990. Column Flotation, Pergamon Press Finch, J.A., Uribe-Salas, A. and Xu, M. 1994. Column flotation: a selected review Part 111, in “Flotation- Science and Engineering” (ed. K.A. Matis), Marcel Dekker Furey, J.T. 1990. Rougher column flotation of gold tellurides, CMP Proc. 22nd Annual Meeting Guimaraes, R.C. and Peres, A.E.C. 2000. Industrial practice of phosphate ore flotation at SerranaAraxa, Proc. 21“ Int. Mineral Processing Congress, p.B9-17 Guimaraes, R.C., Takata, L.A., Peres, A.E.C. and Wyslouzil, H. 1999. Column flotation applied to the production of phosphate rock in Araxa, Brazil, Proc. 31“ Annual Canadian Mineral Processors Operators Conf. Kosick, G., Dobby, G. and Shimizu, P. 1991. The use of column flotation pilot plants for column flotation circuit design, Proc. Column ’91, Canadian Inst. Min. & Metall., Vol2, p.359 Werniuk, J. 2000. Getting down and dirty at Laronde, Canadian Mining J., August, p.13 Wright, A. 1995. Flotation at Les Mines Selbaie, CMP Proc. 27th Annual Meeting, p.120 Yianatos, J.B., Finch, J.A. and Laplante, A.R. 1986. Holdup profile and bubble size determination of flotation column froths, Canadian Metall. Quarterly, 25( l), p.23 Yianatos, J.B., Finch, J.A. and Laplante, A.R. 1987. Cleaning action in column flotation froths, Trans. Inst. Min. & Metall., Section C, 96, p.199 Yoon, R.H. and Luttrell, G.H. 1994. Microcel column flotation scale-up and plant practice, CMP Proc. 26th Annual Meeting, paper 12

1252

Characterization of Process Objectives and (General) Approach to Equipment Selection Charles E. Silverblatt’, Jeflery H.Easton’

ABSTRACT There are many processes in the Minerals Industry that require some type of solid-liquid separation. At the same time, there are many different equipment designs championed by a variety of manufacturers, each seeking to apply their own proprietary design. It is in the manufacturer’s best interest to provide the most economic solution to a separation application, and every effort is made to do so. But providing the most economic design requires a cooperative effort between the process design engineer and the manufacturer. The design engineer must provide the manufacturer with a thorough understanding of the process requirements, and the manufacturer must in turn provide the design engineer with the capabilities of the separation equipment. INTRODUCTION The design and equipment selection process becomes much easier when the design engineer has a thorough understanding of the characteristicsof the solid-liquid suspension to be separated and how these slurry characteristicsaffect equipment design. The simple fact is this - The slurry dictates the equipment design and selection, not the application engineer. The application engineer assists by providing a design that satisfies the demand of the slurry separation objectives. This chapter provides the reader with an introduction to separation equipment options, an approach that the application engineer can use in choosing which option(s) best fits a given process, and the primary factors involved in solid-liquid separation. Later chapters will provide detailed discussions of sedimentation and filtration unit operations and the specific equipment designs that provide a suitable application for each of unit operation. Once an understanding of the various unit operations has been established, selection of the correct separation device can be confidently made. The discussions in this section are based on non-biological solids. CRITICAL PLANT SEPARATIONS Solid-liquid separation is an important, and many times critical, step in a mineral processing plant. Some examples of typical critical separations are as follows: Operation of a seven to eight stage countercurrent decantation (CCD) washinghhickening circuit to effect high recovery of the caustic and aluminum values in an alumina plant. This must be accomplished in a closed circuit system with highly scaling slurries at starting temperatures of 100OC. 2. Operation of a vacuum filter to produce an iron concentrate cake with a moisture content required for pelletizing in order to minimize bentonite addition and maximize pellet quality. A good quality filtrate must be maintained to prevent abrasion of the internal parts of the filter. 3. Operation of a closed water system in a coal washing plant so that reclaimed water containing less than one percent suspended solids is produced for reuse to maintain beneficiation efficiency. Refbe solids must also be thickened so that they can be effectively dewatered to meet EPA requirements. 1.

WesTech Engineering, Inc., Salt Lake City, Utah

1255

4. Recovery of phosphoric acid values from slurry containing gypsum and other gangue solids at an elevated temperature. This must be done while minimizing wash water usage, maximizing PIOs recovery and minimizing scale problems. 5 . Neutralization of acid wastes from mining or metallurgical processes to produce water suitable for discharge or recycle to the process and solids suitable for landfill.

While solid-liquid separations are important to the overall performance of the plant, they also represent a substantial capital investment and significant operating and maintenance expense. Once the equipment selection process has reduced the number of choices to a few suitable options, comparative costs become a very important part of the selection process. AVAILABLE EQUIPMENT DESIGNS

Equipment designs intended to separate solids from liquids or concentrate slurries vary widely depending upon the specific process requirements and the characteristics of the feed slurry. There are generally two basic objectives in any solid-liquid separation process: The production of a clear liquid and a properly washed and dewatered solid. Dilute slurries frequently require some type of pretreatment before final dewatering, while concentrated slurries can usually be handled directly by the final dewatering machine. Gravity sedimentation is frequently used to clarify relatively large volumes of liquid or to concentrate dilute slurries when preparing them for dewatering. A few of the available solid-liquid separation designs are: Standard Clarifiers Disc Filters Clarifiers With Solids Recirculation Standard Drum Filters - Internal & External Drum Belt Filters Clarifiers With Internal Flocculation Horizontal Belt Filters Chambers Vacuum Precoat Filters Pulp Blanket Clarifiers Plate & Frame Pressure Filters Standard Thickeners Recess Plate Pressure Filters Thickeners With Internal Flocculation Tower Pressure Filters Chambers Centrifuges Thickeners With Self Diluting Feed Systems Dissolved Air/Gas Flotation Granular Media Filters - Fixed & Moving Bed How does one make a selection from such varied equipment types, particularly when there are a number of variations within each type? By determining the: 1. Objective ofthe separation step 2. Effect of the separation step on the overall flowsheet 3. Flow rate of the slurry 4. Type of operation: Continuous or batch. If batch, the time available to process a batch. 5 . Characteristics of the slurry involved: a. Size and density of the solids: coarse or fine, heavy or light b. Concentration of the slurry: dilute or concentrated 6 . Need for flocculation: a. Are coagulants and polymers permitted? b. Are the flocs fragile or tough? 7. More valuable phase: Solid or liquid 8. Required moisture and solute content of the final slurry or filter cake 9. Required clarity of the liquid 10. Need & suitability of filter aids such as DE, etc. I 1. Availability of representative sample and arrangements required for testing. This information will eliminate many possibilities, point directly to a few possibilities and help in designing a suitable testing program. The results from a test program will allow a detailed evaluation of the available equipment options. In order to best understand how the various equipment options can fit into the proposed flowsheet, one must understand the primary factors that influence equipment and process design.

1256

PRIMARY FACTORS IN SOLID-LIQUID SEPARATION The discussions that follow address only the more important of the many factors affecting separation. Required Capacity It is not surprising to find that the choice of equipment type is frequently a function of required capacity. Relatively small to medium capacity requirements may, for example, best be met by using filters, either batch or continuous, while the most economic solution for large capacity requirements frequently involves thickeners. There have been cases where the design engineer had a strong preference for filters in the washing circuit based on historical usage, but when testing showed that the filtration rate was lower than expected, a CCD circuit was found to be the more economic solution. With very large capacity requirements, thickeners, when they are applicable, are generally the more economic solution, even though they require more space than filters, as they operate with less operator attention, less maintenance, and less overall cost than filters. Particle Size Distribution Particle size is one of the most important factors influencing performance and cost. Information such as the percent minus 200,325 or 400 mesh will usually provide the application engineer with an idea of the unit’s capacity and expected performance. When extra fine solids are involved, specific surface measurements may be more meaningful. One thing is certain, the finer the solids, the lower the unit’s capacity, the poorer it’s performance and the more restrictive the equipment choices. However, fine particles with a very narrow distribution, such as may be produced by a crystallizer, are much easier to handle than a wide distribution range with the same average particle size. Sometimes the particle size distribution or specific surface area is purposely controlled and kept within a narrow range to improve the quality of the final product. In the production of alumina trihydrate, a once through precipitation would produce a very fine product that would be difficult to dewater and wash. However, classification procedures after precipitation along with fine particle recycle and other crystal growth controls produce a relatively coarse, but narrow particle size range product that dewaters rapidly and washes efficiently. Poor design can destroy a good size distribution. A three stage, countercurrent cyclone classification pilot plant circuit was installed to eliminate extreme fines in a plant handling soft phosphate rock. Unfortunately, greatly oversized centrifugal pumps with high shear closed impellers were employed which required excessive throttling. The resulting attrition produced a product size distribution that was essentially the same as the initial feed and product values were lost. Particle Shape Particle shape starts to influence performance when it varies greatly from a general spherical form. Platelets and long needles represent two extremes. Platelets act as multiple flapper valves within a filter cake and severely restrict cake formation rate, particularly at higher vacuums. Long needles can cause severe blinding of filter media by imbedding in the pores of the cloth. In this case, continuous cloth washing on a belt type filter is required to maintain cloth porosity. Heavy (high specific gravity) particles of irregular shape, such as iron ore particles, can pack so tightly on the bottom of a thickener that they are very difficult to move. The problem is overcome by operating at low solids inventory, high rake speed and special rake blade designs. Feed Suspended Solids Concentration Feed solids concentration strongly influences equipment selection. Generally, dilute slurries must be concentrated, usually in a gravity thickener, before going to the final dewatering device. Since the final dewatering device almost always operates most efficiently on more concentrated slurries, it behooves the operator to control the thickener so that it consistently produces concentrated slurry. If flocculation is required, thickener operations can be influenced in the opposite way. Dilute slurries not only flocculate more easily than concentrated slurries but yield a different floc structure that settles and concentrates more rapidly, If a relatively concentrated thickener feed must be flocculated, it should be diluted for best results. The need for thickener feed dilution was clearly demonstrated many years ago at a coal washing plant when very poor flocculation and settling resulted in a dirty overflow, a low underflow solids concentration, and a solids build up in the circulating water

1257

system. The problem was corrected by simply diluting the feed with a portion ofthe overflow. A ruleof-thumb developed in the coal industry is that the solids concentration ofthe minus 200 mesh fraction of thickener feed should be no greater than 1%. Recent experience has shown that most required thickener feed dilution can be conveniently carried out in a self-diluting feed system. -10 Micron Solids Content The finest particle sizes, particularly those finer than 2 microns, have a tremendous affect on solidliquid separation processes because of the associated very large surface areas. If fine particles are dispersed as true colloids, they exhibit Brownian movement and do not settle. Therefore, charge neutralization (coagulation), and sometimes polymer addition, are required for their removal. Even then, these finer sizes adversely affect settling rate, filtration rate, and cake moisture content. I f flocculation is not employed, these finer solids can build up within the system and adversely affect the whole operation.

Flocculation The use of long chain polymers for flocculation, frequently following charge neutralization, is an excellent and important tool in all types of sedimentation and filtration, and their use has turned many difficult separations into easy ones. When long chain polymers are used in thickening, the operator must be very careful not to overuse or mis-use them. Over-flocculation not only leads to the creation of a gelatinous mass that is difficult or impossible to transport to the discharge point, but is an unnecessary expense. It is easy to understand how over-flocculation occurs, particularly when there is significant variation in the mass flow rate and settling ability. The easy, potentially troublesome and more costly, approach to flocculation is to set the polymer addition rate high enough to cover all circumstances, otherwise, someone must check the degree of flocculation each time there is a change in feed rate or quality. The correct approach to polymer control is to vary the polymer rate to maintain a constant settling rate, while at the same time, checking to make sure that solids do not form a gelatinous mass. The best approach to polymer control is the use of one of the computer control systems that are now becoming available. Viscosity & Feed Temperature Increases in fluid viscosity almost always decrease solid-liquid separation rates. This is true in sedimentation as well as in final dewatering. In the case of dewatering, the effect is usually an increase in the moisture content of the discharged solids. Thus, while production rate of the dewatering device might be maintained, moisture content of the product would increase, possibly causing downstream problems. Viscosity is affected primarily by temperature. For example, the viscosity of water is 0.98 centipoise at 2 1"C, but increases to 1.7 centipoise at 2"C, a 73.5% increase. Obviously, this increase in viscosity will affect both settling rate and moisture reduction in the final dewatering step. In the filtration of fine coal, a study some years ago showed that viscosity had a substantial affect on cake moisture content, while surface tension, which is also affected by temperature, made little difference. An examination of the operating data of that same plant, which used a lot of make-up water, showed that the moisture content of the total coal plant product was a function of the temperature of the river water used in the plant. Dissolved Solids Build Up In Feed Liquor Overall plant economics usually dictate that a minimum of water be discharged to waste. This also means that even minor contaminates have an opportunity to build up to significant, and sometimes harmful, concentrations. While it happens infrequently, long-term solute build up can change water phase ionic strength sufficiently to require a complete re-investigation of flocculation because the original polymers and procedure failed completely. Build up of other constituents, such a chlorides, may result in unexpected corrosion.

1258

Slurry Scaling Whenever lime addition is involved, there is the risk of incomplete reaction such as in the precipitation of gypsum and calcium carbonate or other compounds that form supersaturated solutions. These reactions will go-to-completion and scale on equipment surfaces unless care is taken to carry out these precipitations in the presence of previously precipitated solids and with ample reaction time. Care in the design of the precipitation reaction will not only reduce scaling tendencies, but can produce larger, easier to handle particles. The various high-density sludge acid mine drainage precipitation processes and the solids recirculation units used for water softening are good examples of the application of these principles. DETERMINATION OF SLURRY CHARACTERISTICS Operating Experience When considering a plant expansion, an operating plant is the ideal source for information on slurry characteristics. Even then, design engineers frequently make changes in the flow sheet that will significantly impact feed slurry characteristics, Nevertheless, data from a plant operating on a similar slurry is very valuable information that should be used to amplify laboratory or pilot plant testing results, as an operating installation will often bring to light important factors that may not be evident during a short term testing program. When these types of data are not available, a well-constructed and conducted test program will provide sound sizing and design information. Unit Operations There are a number of design variables in both sedimentation and filtration unit operations that must be investigated and defined in the process of selecting the proper type of equipment. Test results are usually defined in terms of these design variables. The lists that follow, although not exhaustive, contain most of the variables of concern. Gravity Sedimentation Chemical flocculation. Current practice almost always makes use of chemical flocculation to reduce the size of the sedimentation vessel, enhance settling rate, and increase underflow concentration. Chemicalflocculant selection. A wide selection of chemical flocculants is available from a number of manufacturers. Screening tests on small slurry samples can quickly reduce the number of chemical candidates to less than three. Cliemicalflocculant mixing. The method and intensity of mixing influences flocculation. Properly conducted screening tests can define the best mixing method. Solids recirculation. The minimum suspended solids concentration required for flocculation to produce a practical settling rate usually occurs within the range of 300 to 700 mg/L. Previously concentrated solids from the clarifier underflow are recirculated to the feed when the feed solids must be increased. Feed dilution. When the solids concentration is too high, flocculation is inefficient. It is almost always best to dilute the feed to less than 20 wt.% suspended solids, and frequently to as low as 10 wt.% solids.. Dilution may be by external pumping or by natural internal dilution into the flocculation zone. Mechanical floc growth. Mechanical floc growth is the slow mixing of a polymer solution with the feed slurry that is required to form flocs large enough to settle at a useful rate. This mixing is usually carried out within the feedwell to prevent the floc breakdown that occurs when flocs are transferred from one vessel to another. The nominal design time is usually three times the observed bench test time. Settling rate. Settling rate is the rate at which the solids settle out of the liquid. The observed rate obtain during a bench test must be derated to account for the short circuiting that exists in every gravity sedimentation machine. One historically accepted scale-up factor is 50% of the observed settling rate.

1259

Clarification time. Clarification time is the time required for the fine particles left in suspension after the interface or bulk of the solids have settled out to agglomerate and settle out. The time observed in a bench test must be derated to account for the short-circuiting that exists in every gravity sedimentation machine. The derating factor is variable, based on the diameter to depth ratio of the clarifier. Solids handling. Time and depth are required for the solids to settle to some desired concentration. This requirement is frequently expressed as Unit Area, the area required for a ton per day of solids settle to the required concentration. This area increases at a rapidly increasing rate as the solids concentration approaches maximum. Filtration Chemical flocculation. Flocculation immediately ahead of filter may or may not be required. If the filter follows a thickener that used flocculation, additional polymer will usually be beneficial. Cliemicalflocculant selection. A wide selection of chemical flocculants is available from a number of manufacturers. Flocculation and filtration tests on small slurry samples are used to reduce the number of chemical candidates to one or two. Cliemicalflocculant mking. Casual mixing is seldom sufficient for slurries concentrated enough for efficient filtration. A power mixer is usually the most effective. The mixing system should be located adjacent to or directly above the filter. Cake formation rate. Rate at which the cake is formed. This rate is strongly influenced by feed solids concentration and by the cake thickness required for discharge. Correlating techniques are readily available in the literature. Cake dewatering rate. Rate at which residual liquid is removed from the cake during a period when only air or gas passes through the cake. Cake washing rate. Rate at which wash liquid passes through the cake. Wash efficiency. Cake wash displaces the residual liquid in the unwashed cake, and the efficiency of this displacement is the wash efficiency. Correlating techniques are available in the literature. Airflow rate. Rate at which air or gas passes through the cake during a dewatering period. Steam drying. Application of steam to a filter cake. Steam not only displaces residual liquid, but heats both the residual liquid and the cake solids. The resulting elevation in temperature of the residual liquid increases the rate of dewatering and thus reduces the final cake liquid content. TESTING PROTOCOL

Representative Sample Before deciding upon any type of testing program, it must be possible to obtain a representative sample, either for laboratory testing or for pilot plant testing. Laboratory tests are much less expensive than continuous pilot tests and can also be carried out in a much shorter time period. Any sample used for laboratory testing represents only a snapshot of the process, and one must be certain that a single or several snapshot samples will truly represent the process variations, or are the best representation that can be obtained. If process development is based solely on core samples, laboratory tests are the only choice, and several samples are usually required to cover the expected range ofoperations. Even then, allowances must be made for the fact that the commercial product will not be as clean as the product produced from core samples. Today’s laboratory test scale up techniques are sufficiently dependable to use for full-scale design, particularly when the equipment supplier has experience in similar operations. The weakest link in the sizing chain is usually the sample. Test Objective The objective or objectives of any test program must be clearly defined, or the test program is likely to be a waste of time. A test program frequently includes several separation steps, and specific collateral data may be required to assist in sizing pumps, mixers, etc. Care must be taken to ensure that the persons carrying out the test program are familiar with the overall process and the affect of each step upon the rest of the process. Careful planning is essential.

1260

Data Collection & Analysis It is essential that each sample be characterized, as a bare minimum, with respect to suspended solids concentration, particle size distribution, liquid phase TDS, pH and temperature. Other important tests include a detailed chemical analysis of the liquid phase, a chemical analysis of the solids, and any other information that may be specific to the process. This information allows the results of the test program to be evaluated against the expected feed quality or the actual feed obtained after plant start UP. The data required for such standard separation steps as clarification, thickening, filtration, centrifugation, etc. are well defined by the companies supplying those items of equipment. Nevertheless, there are frequently special analytical requirements or other unusual considerations that require special planning. Details of the tests including all data sheets and all support assistance must be prearranged to ensure that all information is obtained in a timely fashion. Equipment Selection The results of the test program will detail equipment selection and sizing options required to meet solids-liquid separation objectives. Capital and operating costs then become the primary equipment selectors. The following chapters will detail specific equipment characteristics and the sizing tests required for each type of equipment.

REFERENCES D.A. Dahlstrom. 19-. ASME

Chapter 6 , Fundamentals of Solid-Liquid Separation. Solid-Liquid Separation, , Edited by Harma, R. 0. & Degner, V. R.

R.H. Perry & D. W. Green (Eds.). 1997. Liquid-Solid Operations and Equipment. Perry’s Chemical Engineers’ Handbook, 71h Ed. D. Purchas (Ed.). 1977. SolidLiquid Separation Equipment Scale-Up, Uplands Press, Croydon, England.

1261

Centrifugal Sedimentation and Filtration for Mineral Processing Wallace Leung, Bird Machine Company

ABSTRACT Centrifugal separation has been commonly practiced in many key separation steps in mineral processing, some of which include separation, clarification, classification, degritting, dewatering, and purification. Literally thousands of sedimenting and filtering centrifuges are employed today in processes such as coal, tar sand, kaolin, potash, soda ash, calcium carbonate, and drill mud etc. In the past decade, significant advancement has been made to develop effective technology and know-how to improve the process separation and to get higher-grade product(s) with improved throughput and lower energy consumption. In this paper, the theory, types, process functions, applications, and new developments of centrifuges for mineral processing are presented. THEORY OF CENTRIFUGAL SEPARATION Regimes of Sedimentation The sedimentation behavior of a suspension may be classified into four categories in accordance with the solids concentration in suspension and the degree of aggregation of solids. This is illustrated in Figure 1, which in essence is a modified Fitch diagram. For dilute concentration and low degree of solids aggregation, solid particles settle independent of each other and they follow the Stokes’ law of sedimentation, which was developed for spherical particles settling under earth gravity, one g (9.8 m/s2). As solids concentration increases, the sedimentation rate of particles is affected hydrodynamically by neighboring particles despite there is no physical contact between them. Under this condition the settling rate may be less, or even higher, than the Stokes’ settling velocity. For a given solids concentration, as the particles tend to agglomerate due to weak or negligible electric repulsion they form aggregate and settle as a large floc, which can be modeled by fractal analysis. This allows both small and large particles to settle at the same speed, also known as zone settling, without discriminating the size of individual particles. Addition of coagulant and flocculant (polymer) may further promote formation of agglomerates and flocs leading to zone settling; while introduction of dispersant extends the discrete particle settling condition well into the concentrated solids region in which hindered and zone settling normally prevail, see Figure 1. The former finds applications such as clarification of waste slurries, while the latter finds applications such as classification of valuable fine-particle slurries for coating and pigment market. Dense thick slurry forms networking as particle concentration and the degree of aggregation both increase in a suspension. Under gravitational body force the solid network or matrix compresses downward (compaction) while liquid expresses counter-currently upward (expression). This is delineated as the region to the upper right in Figure 1.

1262

Aggregate

Degree of Aggregation

Discrete Particles

Discrete Particle Settling Dilute

Dispersant Adde Solids Concentration

Concentrated

Figure 1 Different regimes of sedimentation

Stokes’ Law The sedimentation of discrete particles in a viscous fluid under centrifugal gravity, G, can be described by Stokes’ law,

Ap=ps-p~is the density difference between solids and suspension, x the particle size (equivalent diameter), $ the solids volume fraction, ?+I the hindered settling function, p the viscosity of suspension. For dilute suspension where $<<1, h ($)=1 and the above reduces to the Stokes’ form of sedimentation for single particles if centrifugal gravity G is replaced by earth gravity g. For concentrated slurry h($)
Boundary/Critical-Layer Model The flow inside a centrifuge with a cylindrical clarifier is very complex especially for decanter centrifuge where the conveyor scroll is rotating at angular speed, either faster or slower, relative to the bowl. There are several models to quantify the separation capacity based on the geometry and the operating parameters including the Sigma theory, G-surface and G-volume approaches. None of these satisfactorily explains and quantifies the effect of flow in the complicated geometry and operation. The most useful model is the improved boundarylayerlcritical-layer model, which is a generalization of the boundary-layer model 121. This model is still somewhat inadequate. Nevertheless this is the only model, among all, that accounts for particle size distribution (PSD), which is vital for separation, classification and clarification. In the improved boundary layer model, a critical moving layer in the vicinity below the pool surface at a mean radius Rp flows across the pool. The model assumes that once a solid particle settles across the layer, it transverses a stagnant pool below and ultimately settles in the sediment or cake. A major difference between the improved model with the earlier boundary-layer model is that feed

1263

solids of all sizes are assumed to be uniformly distributed across the entire critical layer thickness instead of concentrating at the surface of the layer. Depending on the operating differential and the geometry of the conveyor the layer thickness can be very thin or as thick as the annular pool depth. When the critical layer is indeed thin in comparison with the radius, the actual thickness of the layer does not enter directly into the result. The G's for separation is determined from the critical radius Rp wherein G=R2R,. For some geometry with an axial-flow design, the critical radius is located at the pool surface. Whereas for other geometry and operating conditions where there is substantial mixing across the pool due to secondary flow caused by the differential speed between the conveyor and bowl, an effective radius can be used instead, such as the geometric mean of the bowl radius and the pool radius. This can account for effect due to variation in pool depth.

Solids Recovery The solids recovery, respectively, in the cake R, or in the centrate Re, for the improved boundary layer model, are expressed as R, ( L e ) = 1- F, ( x )+ I ( x , ) / x,' R, (Le) = 1- R, X

I ( x ) = J x ' f , (x)dx 0

x , I X" =

yfi

Le

particle size % cumulative undersize x in feed slurry frequency of a given size x in feed slurry cut size, i.e., maximum size of particles in the overflow reference particle diameter, convenient taken as 1 pm flow rate cylindrical clarifier length viscosity of slurry density difference between solid particle and suspension angular rotation speed of bowl critical layer radius average feed acceleration efficiency for the feed stream in clarifier hindered settling function due to hydrodynamic interaction applicable for dense slurry From a practical standpoint, it is difficult to determine the viscosity of dense slurry especially with fast settling particles and the hindered settling function. The ratio p/h is combined as a

1264

“viscosity factor” p’ given h is dimensionless and is mostly less than unity. Often p’ is used as a matching parameter for a given process separation from available test data with projection to performance on other operating conditions which are unknown. For that matter, if the average clarifier acceleration efficiency, nominally 0.7-0.9, is not determined accurately, any discrepancy with the actual performance could also have been absorbed in the viscosity factor when matching the actual performance data with theoretical prediction. Le is the dimensionless Leung number [2] governing centrifugal sedimentation. Note the solids recovery in the cake R,, for application of separation and clarification, as well as solids recovery in the centrate %, for classification application are both functions of Le. As evident from Equation 3e, Le is directly proportional to the cut size in the centrate. In fact, the cut size x, is simply 1.69Le pm given x, is taken conveniently as 1 pm. Suppose Le for the geometry and operating condition of the centrifuge is such that it assumes a value of 3, the maximum particle size in the centrate is thus 5.2 pm. Table 1 enlists the typical cut sizes, the corresponding operating Le, and the typical applications.

Cut Size Tyler Mesh Microns 0.1 0.5 1 2 5 10 25

Le

Typical Applications

Dl

0.06 0.3 0.6 1.2 3 6 15

325

45

27

200

75

44

150 100 150

100 150 212

59 89 125

Ultra high-G classification of valued ultrafines, waste slimes, and colloids .

Classification in coatings, pigments (e.g., kaolin, calcium carbonate, silica, mica, etc.) and drill mud with high-speed centrifuges Classificatioddegritting of oversize particles above 25 pm Classificatioddegritting of oversize particles above 45 pm Sedimenting fine particles for which continuous filtering centrifuge does not separate well

Relatively coarse separation with lowerspeed sedimenting centrifuges

Note Le has taken into account, among many variables, the feed rate and speed (or G). For a given application, the feed rate and speed may be very different for different size machines, however for the same application with the intent to get the same cut size, Le has to be identical for the machines. This becomes a very important scale-up criterion for separation, classification and degritting. One can use the Le scale-up approach to design and scale-up a machine to perform with a specified feed rate and attain a specified performance such as recovery etc. Alternatively, it can be used to predict the performance of a given machine if say the feed rate or speed is changed such as in process optimization. Most PSD has a polydispersed distribution with the population of particle size spreads out from very fine to very coarse sizes as characterized by the cumulative undersize function Fr (x) or the frequency distribution function ff (x). On the other hand, monodispersed distribution results with the particle size concentrates near a relatively narrow range as with precipitated minerals (e.g., precipitated calcium carbonate) from dissolution and crystallization processes. Also particles may be populated in two seemingly distinct size ranges, i.e., bimodal. It can be found in slurry with single species or commonly found when two species are mixed (such as in weighted drill mud), each with their respective characteristic size distributions and densities. The PSD of the slurry is critical to separation, classification and clarification.

1265

Particle Size Distribution of Overflow The particle size x k in the product centrate (i.e., overflow of centrifuge) from the improved boundary-layer model is given by,

(4) The centrate PSD F, is therefore a function of the size x k the size cut x, which depends on Le, and the PSD of the feed Ff. Cumulative Size Recovery The cumulative recovery of particles with size below x k in the product overflow from the centrifuge can be determined using the improved boundary-layer model as,

Again SR is a function of xk and Le through Ff and I. It is clear that the PSD of the feed slurry plays a vital role in determining the size distribution of the centrate product, size recovery and total solids recovery.

Filtering Centrifuges The final dewatering or deliquoring of cake in a filtering centrifuge depends on the dimensionless time tD [2] which incorporates several important operating variables in deliquoring - the G-force, time duration t, hydraulic diameter of particles xh, cake height h, viscosity pL,and density of liquid pL. It is defined as,

Granted the article characteristic size stays constant, the cake permeability which is proportional to xh is also constant. The final moisture of the cake depends on the G-force, retention time, inversely related to the cake height (which in turns relate to solids throughput), and also inversely related to the liquid viscosity. The latter usually decreases with increasing operating temperature. The shape of the moisture drainage curve depends on the surface properties of the cake and the kinetics of liquid film drainage from the cake surface.

P

PROCESS FUNCTIONS AND MACHINES Separation (without and with added chemicals) - In separation feed slurry is introduced into the centrifuge where a concentrated stream of solids and a liquid stream primary with low concentration of solids are desired. Polymer can be added for dewatering and separating waste stream when majority of the particles is too fine in the micron to sub-micron size range. For mineral applications, the polymer dosage is typically less than 0.1%kg polymer per kg of dry solids.

1266

Clarification (liquid phase product) - When liquid phase should be free of solids for reuse or discharge, solid concentration needs to be minimized and typically a tolerable limit is set on the maximum tolerable solids concentration in the centrate. Classification - Finer solids fraction is the desired product with coarser fraction discarded; vice versa coarser fraction is the product discarding the finer fractions, e.g., slimes with particles less than 0.5 pm. Degritting - This concerns removal of oversized particles, such as 25 pm, 45 pm, etc. and/or foreign particles with a different density compared with the solids in the process stream. Deliquoring (dewatering) - Cake moisture needs to be minimized to meet process requirements (final end product, further downstream processing such as thermal drying, purification, etc.) Washing - Cake is required to be washed with appropriate solvent or wash liquid with minimal contaminants when cake solid needs to meet purity requirement. Wash liquid should be minimized while achieving good cake purity. Deliquoring followed by reslurrying - When particles in the slurry are in microns, say 5-10 pm, and cake washing is desired to achieve high purity, the slurry is first deliquored by centrifugation followed by reslurrying. This can be done in stages, with say two stages being very common. Alternatively a basket centrifuge can also be used if feed volume is limited and storage tanks are required for silo.

TYPES OF CENTRIFUGES There are two types of centrifuges - continuous-feed and batch basket centrifuges. Within each group there are many different designs and variations. Most applications use continuous-feed Centrifuges. Solid bowl or decanter is the most versatile among all. Solid Bowl or Decanter A schematic of the solid bowl or decanter is shown in Figure 2. Feed slurry, after accelerated in the rotating feed compartment or accelerator, is introduced to the annular pool. Under high centrifugal force, the heavier solids migrate radially outward toward the bowl to form cake displacing the lighter liquid toward the center. Solids are compacted against the bowl wall by the centrifugal force and conveyed by the screw conveyor, rotating at differential speed relative to the bowl, to the small diameter of the conical beach for discharge. As the cake is lifted above the annular pool in the dry beach, liquid further drains back to the pool leaving a drier cake for discharge. The gear unit and/or backdrive control the differential speed between the bowl and conveyor changing the solids retention time as needed. The clarified liquid overflows the weirs located at the opposite end of the machine. The pool is controlled by the discharge diameter of the weirs. The performance of the centrifuge depends on the various operating variables such as feed rate, pool depth, rotation speed or G-force, and differential speed and should be optimized for a given process.

Feed Rate. Liquid residence (settling) time of the slurry in the bowl may directly affect the degree of centrate clarity that can be obtained. Decreasing the feed rate will increase the liquid residence time and permit more efficient settling of suspended solids. With very dilute suspensions (solids concentration much less than l%),gravity or cyclonic thickening upstream of the centrifuge is recommended to concentrate and reduce the total volume

1267

of feed sluny or liquid handled. Hydraulic loading affects the main-drive motor requirement from perspective of accelerating the feed stream while solids loading affect the conveyor torque load. Backdrive, or Gear Unit

Bowl

Housing

Conveyor Adjustable Weirs Feed Inlet

Effluent

Cake Discharge

Figure 2 Solid-bowl or decanter centrifuge schematic

Pool Depth. The proper pool depth depends on the settling characteristics of the solids in the feed slurry. By reducing the pool depth, a drier cake is normally obtained because a longer dry beach is available for cake drainage before discharge. Pool level should not be lowered to a point where centrate clarity suffers or solids conveyability is hindered. When the pool depth is increased the length of the drying beach is reduced. This generally results in higher cake moisture. A deeper pool usually improves centrate clarity, since liquid retention time is increased giving lighter and smaller particles more time to settle. Deepening the pool also eases movement of the cake due to liquid buoyancy, resulting in improved cake convey ability. Rotation Speed and G-Force. Higher rotation speed produces higher centrifugal force acting on the solids in the cake and improved settling rate. The consequence is lower cake moisture and/or a clear centrate. However, this does not necessarily always hold. Some solids especially the finer size fraction have density very close to that of the suspension (i.e., somewhat neutrally buoyant) due to adhesion of contaminants or bubbles to the solid surfaces. They do not settle regardless of the magnitude of the centrifugal force. Also different solids tend to pack tightly (i.e., compactible cake) under high centrifugal force and some drain more readily under lower speeds where larger voids exist with higher cake permeability. For compactible cake, increasing G beyond a certain point does not warrant increase in cake dryness due to increasing cake resistance to deliquoring [ 2 ] .In general, for maximum clarity, cake dryness and least power consumption, operate at the “lowest possible” speed compatible with the material characteristics and performance requirements. It is always good practice during the initial start-up period to compare cake dryness and centrate clarity at different rotation speeds and G’s. This allows selection of the optimum centrifugal forces for the specific application. Differential Speed. By lowering the speed differential between the conveyor and the bowl, the solids residence time is increased. This causes an increase in cake depth against the bowl wall with greater compacting stress and higher cake dryness in most cases. This is often accompanied by increasing conveyance torque as well. Lower conveyor differential provides less turbulence and less re-suspension of solids. However, low conveyor differential may have the opposite effect if the incoming feed solids rate is higher than the capability of the conveyor to remove them in which case solids un-transported buildup in the cylinder and eventually overflow at the liquid discharge ports. The latter is often accompanied with increasing conveyance torque over time. There must be a balance between solids input and solids removal to prevent plugging of the machine and loss of the centrate clarity.

1268

Differential speed may be changed by changing the gear ratio for the same box if available or changing the gear box with a different gear ratio. The differential speed is related to the bowl speed Qb and the pinion speed Qp by the following kinematic relationship:

AQ = QtJ

-Qp

(7)

Suppose Qb=2600 rpm and the pinion shaft is locked stationary with Qp=O, with a gear ratio r=80:1 it gives a differential speed of 32.5 rpm. An alternative is to provide an electric backdrive where the pinion is driven by a DC motor, or an AC motor, which can be controlled by a variable frequency drive (VFD). By tuning the frequency, the pinion speed is adjusted thus changing the differential while the machine is running. For example, the VFD is tuned with pinion shaft rotating at speed respectively 1000 rpm and 2440 rpm, the differential speed becomes 20 rpm and 2 rpm. The small differential speed allows longer retention time, which facilitates deliquoring cake to high dryness. Flocculant andor coagulant are used to agglomerate fine particles improving centrate clarity. This is especially on the waste application in which polymer dissolved in liquid stream is of lesser concern 1131. Also a hydraulic pump/motor is used as backdrive for centrifuge. The pump pressure is used to control the torque while the oil flow rate in the pump is used to control the differential speed between the scroll and the bowl. The centrifuge can be over-torqued due to plugging with unconveyed solids accumulating in the bowl. If temporary stopping of feed to the machine while continuously maintaining the differential speed between the conveyor and bowl does not clear the machine, the rotation speed and thus the counteracting centrifugal force is reduced to facilitate cake transport. Unfortunately the differential speed also reduces in lieu of Equation 7. Fortunately, a centrifuge equipped with an electric or hydraulic backdrive allows maximum differential speed to transport cake out of the machine despite the reduced bowl speed or when the bowl stops rotating.

Feed Acceleration. In a decanter the feed stream needs to be accelerated to match the tangential speed of the rotating pool to generate centrifugal force to effect separation. Also the feed should be distributed uniformly on the pool with minimal radial velocity to avoid disturbance and turbulence. Unfortunately, conventional feed-accelerator designs by-and-large poorly accelerate the feed and frequently the latter is introduced in concentrated streams jetting into the pool causing turbulence, resuspension of sediment, and wear on rotating surfaces. This is especially significant for a high volumetric rate application. Under high flow rate, feed slurry that has difficulty swallowing through the feed ports floods the feed chamber; and when serious feed may leak back along the outer diameter of the feed pipe. A comprehensive patented feed acceleration system [2, 4,5, 6, 71 has been developed for continuous-feed centrifuges including sedimenting and filtering centrifuges. To visualize the effect of feed acceleration, two identical rotating bowls of 250-mm diameter, cantilever outward from the support, were set up side-by-side to demonstrate the visual effect of feed acceleration [7]. Water was introduced in one centrifuge bowl via an improved feed accelerator and for the other at the same flow rate with a conventional accelerator design. The surface of the pool was observed with a strobe light tuned at the frequency of the rotating bowl at 1000 rpm. A pool meter in the form of a water wheel which has a pair of paddles dipped into the rotating pool by 3 mm was set up under free wheeling condition and the pool meter was driven by the rotating pool where the feed entered. The rotational velocity of the feed at the pool based on the rotating speed of the pool meter was measured using a strobe light and compared with the bowl speed at 1000 rpm. Figure 3a is a strobe picture showing the pool region where feed was introduced via a conventional feed-accelerator design. The surface of the pool appeared like a rough sea with turbulence. A close-up of the pool in Figure 3b reveals the entire pool surface was wavy and chaotic. In fact, the pool where feed was introduced was rotating at a speed significantly

1269

less than that of the bowl. This “slip” or mismatch in tangential velocity between the incoming feed and the remaining pool is the cause of turbulence. On the other hand, Figure 3c shows the same rate of feed being introduced to the pool after flowing through an improved XL*PLUSB feed accelerator system. The pool was quiescent reflecting the image and color of the pool meter. Also the pool meter appeared stationary rotating at the same pool speed as the bowl. This fosters the “calm” condition for sedimentation where heavier particles under the driving centrifugal force settle to the bowl wall. An application of this technology is demonstrated later in Figurel9. Figure 4 shows the ratio of measured pool speed to that of the bowl for different feed rates. The speed ratio, which measures the feed acceleration efficiency, stays constant at 100% independent of feed rate for the XL*PLUSB feed system while the efficiency for the conventional design drops off sharply to an disappointing level approximately 50% with high feed rate (above 7 m3/h for the 250-mm bowl without a conveyor). This is a common experience with all conventional designs. The square of the velocity ratio, which is an important measure of the actual G-force (a measure of separation ability) versus the rated G-force - i.e., G-efficiency, decreases even more sharply. Despite this, under-accelerated feed would eventually get accelerated in a decanter by contact with the pool and the rotating surfaces but not without skidding, wear and turbulence. This is at the expense of consuming clarifier and pool volume for feed acceleration leading to lower throughput, lower solids capture and possibly feed leakage. For a decanter with a short clarifier (such as a screen bowl) this is seriously disadvantageous. Improved feed acceleration systems and related technologies have been presented in much greater detail elsewhere [2,4, 5, 6, 71.

Figure 3a Strobe picture of the pool in a 250-mm rotating bowl fed with water after accelerated by conventional feed accelerator design (courtesy of Bird Machine Company)

Figure 3b Close-up strobe picture of the pool for configuration of Figure 3a showing turbulence and waves on pool surface (courtesy of Bird Machine Company)

1270

Figure 3c Strobe picture of a pool in a 250-mm bowl fed with water after accelerated by XL.PLUSB feed accelerator design (courtesy of Bird Machine Company)

Conventional

0 % )

0

I

1

I

I

I

I

1

1

I

I

I

I

2 3 4 5 6 7,8 9 1 0 1 1 Flow Rate, m /h

Figure 4 Acceleration efficiency (pool speed at feed entrancehowl speed) and G-efficiency (square of acceleration efficiency) for, respectively, conventional and XL*PLUSB accelerators shown in Figures 3a-3c Sizes. A variety of different diameters are available for decanters from as small as 150-mm (6-in) diameter to over 1400-mm (55-in) diameter. The length-to-diameter (aspect) ratio is as large as 4.2 to 4.5, and as small as 1.3. Generally, the larger machines operate at lower speed (1OOOg-3OOOg) and accept higher throughput when compared with small diameter machines operating at higher speed thus higher centrifugal gravity (3000g-4000g) processing more difficult-to-separate materials. Materials of Construction. Duplex steel with higher yield strength is employed for high-speed decanters. For processing abrasive materials, all of the high-wear areas such as feed zone, blade tips, possibly bowl wall, and cake discharge are protected with a wear resistant coating such as ceramic, tungsten carbide tiles or spray, or other specially prepared sacrificial coatings. Screen Bowl Centrifuge With an added cylindrical-screen section attached to the small cake discharge diameter, a screenbowl centrifuge can be viewed as a modified decanter centrifuge, see Figure 5. Given the distance between the two bearings are the same for a centrifuge with a given diameter, therefore the length of the larger-diameter cylindrical section of the solid bowl is reduced to accommodate for the screen length of the machine.

1271

The cylindrical screen section has openings to accept soap-dish shaped screens, which are installed externally of the bowl held by a big band clamped on the outer diameter of the bowl, or screen panels, which are installed internally. The screen can be made of stainless steel, tungsten carbide ligaments, and ceramics. The screen is mostly profiled (such as wedge wire, see Figure 6) with the smallest opening of the cross section at the screen surface facing the cake and the largest opening at the exit outer diameter to prevent trapping solids in the screen causing blinding of the screen to flow of filtrate. A ceramic conveyor is shown in Figure 7. Cake is washed at the initial screen region while the remaining screen section toward the cake discharge is used for deliquoring. It is important to wash the cake while it is still fully saturated with the mother liquor otherwise air bubbles trapped in the cake may block the wash liquid to the cake pores when the cake is partially desaturated. Gear Unit

Housing

Screens

~~~l

Adjustable Weirs Feed Inlet

Screened

Effluent Proiuct

Conveyor

Main

Effluent

Figure 5 Screen-bowl centrifuge schematic

200-300 pm (typical)

filtrate Figure 6 Self-cleaning wedge-wire profile used for screens for filtering centrifuge

Figure 7 Screen-bowl conveyor blade tips lined with ceramic tile protection (courtesy of Bird Machine Company)

1272

Pusher Centrifuge The pusher centrifuge is cantilevered supported and has single (single-stage pusher), or multiple cylindrical screens (multiple-stages with two- and four-stages common) with a progressively larger screen diameter. A schematic of the two-stage pusher is shown in Figure 8. A thickened feed slurry of 40%-60% after properly accelerated by the feed accelerator is introduced to the first-stage basket. An annular push plate with an outer diameter closely fitted to the inner diameter of the first-stage basket pushes the cake as the first-stage basket reciprocates along the axis. For the case of a two-stage pusher, the cake at the end of the first-stage basket drops onto the secondstage basket as the first-stage basket retracts. In the second-stage basket the cake is further deliquored under a higher G-force. For a four-stage pusher, the first two stages are equipped with an end ring (acting as a push plate) at the exit of the screen section for pushing the cake. Both the first stage as well as the third stage reciprocates. In the retrieved stroke, the cake is dropped from both the first and third stage, respectively, on the second and fourth stage. In the forward stroke, the cake is pushed from the second and fourth stage to the third stage and discharge, respectively. Deliquoring is accomplished as the cake gets thinner with capillaries between solids opened up as it spreads in successive stages at increasing diameter and higher G-force. Cake washing is often carried out for a multiple-stage pusher at the transition between the first and second stage to remove cake impurities. The wash liquid needs to be delivered and pre-accelerated through a wash pipe equipped with a nozzle so that the velocity of the wash liquid approximately matches the tangential speed of the cake in the rotating basket to establish the same centrifugal gravity. Otherwise, the wash liquid would appear to have a “lighter” gravity and would not be able to penetrate the cake displacing the mother liquor saturated with impurities. In applications requiring high performance, the cylindrical basket of the last stage is replaced with a conical basket. This reduces the load on the push mechanism at high solids throughput because the longitudinal component of the centrifugal gravity is assisting cake transport. There are additional enhancements on cake deliquoring with this design. As the cake is conveyed to the large diameter of the cone, the cake height is reduced with more surface area and concurrently the G-force is increased.

Second Sago Basket

First Stage Basket

Pusher Plate

,Accelerating Cone Feed Pipe

Hydraulic Pushing

Mechanlsm

C’ake Discharge

cornp&tmental Effluent Chamber

Figure 8 Two-stage pusher centrifuge schematic

1273

Next Page Conical Screen Centrifuge Vibrating Screen Centrifuge. The inertia generated by vibrating two eccentric rotating masses facilitates cake transport along the conical basket. Commercial sizes with 1100-mm (44-in), 1200mm (48-in), and 1300-mm (56-in) diameter are available. The centrifugal gravity is typical under 1oog. Screen-Scroll Centrifuge. The screen-scroll centrifuge has a rotating conical basket mounted either horizontally (see Figure 9) or vertically. Feed slurry is accelerated in a conical section of the hub, which forms a feed compartment. Accelerated feed is introduced into a rotating truncated conical screen at the small diameter. After filtration, the cake is conveyed to the large diameter by a scroll with complete “wrap-around” helical blades, or a discrete number (4 or 8) of blades profiled with a small helix angle. Cake washing at the small diameter of the conical screen is possible, in which wash liquid is introduced into a separate compartment in the conveyor hub, where wash liquid is sprayed through nozzles mounted on the hub outer diameter. As the cake is conveyed to the large diameter it spreads to a thinner pile and under higher centrifugal force the cake attains low cake moisture before discharge at the large diameter of the basket. Common sizes are 250 mm (10 in), 400 mm (16 in), 700 mm (28 in), 900 mm (36 in) and 1000 mm (40 in). Vertical centrifuge operates at lower G, 230+g, while the horizontal centrifuge operates between

300-8OOg. Wedge-wire profiled screen and perforated plates are commonly used. The opening of the screen is nominally 300 pm and a more open screen with 700-800 pm is employed for processing coarse materials. Laser-cut screen with uniform sieve size over the entire screen can also be used for screen-scroll centrifuge. The loss of fine solids in the filtrate can be regulated by selecting the appropriate sieve size. For example, a 60-pm sieve has an 8% opened area and provides good filtering capacity as well as screening capability. The laser-cut screens are coated with a chromium layer for wear protection. Rotating Basket

Single Input -Dual Output Planetary Gear Speed Reducer

Scroll 10’ Feed Pipe

Dewatered Product Discharge Centrate Discharge

Figure 9 Screen-scroll dryer schematic

1274

Characterization of Equipment Based on Filtration Principals and Theory Glenn D. Welch'

ABSTRACT The characteristics of the fluid/particle mixture to be filtered, its intrinsic filtration properties, and the objective of the solid-liquid separation dictate the equipment selection that would be technically and economically feasible for a prospective application. Definitions of the physical and chemical characteristics of the slurry to be filtered and correlations obtained from testing representative feed samples by applying principals of filtration theory are used as a basis for filter design. A systematic approach is used initially to reduce the number of options considered feasible for a given filtration application. A brief discussion of important considerations relating to equipment design is provided as a further guide in the selection process. Testing to develop design criteria is the final requirement to complete the selection process. INTRODUCTION The selection of filtration equipment is an integral part of overall process design. The main focus of this paper is to provide a general guide for the engineer to reduce filtration options for a specific application and to develop sufficient information to properly select the optimum design. The economic selection of filtration equipment begins with choosing a design basis, and narrowing down filtration equipment options through systematic consideration of solid-liquid separation system operations prior to the filter, post-treatment requirements, overall process goals, and characterization and nature of the feed material. A test program based on the filtration options selected is then the final step necessary for further evaluation. The test program is focused on the goals of the specific application and serves to define the important aspects of the filtration cycle time (i.e. form time, wash time, and dry time) required for filter design. Once the data has been obtained and correlated, final selection by evaluation of engineering requirements and ecnnnmic analysis can commence. Evaluation of the rate of cake formation and/or the rate of filtrate production achieved for an applied pressure drop (driving force) dictates the type and size of the filter that will be required based on the tonnage to be processed. The important properties of the feed material affecting filtration equipment design can be accurately defined by performing tests on representative samples in a suitable laboratory scale testing apparatus through correlation of test data by applying standard filtration principals. A complete discussion of correlation methods for filtration data can be found in the cited references (Smith and Townsend 2002). CHOOSING A BASIS FOR FILTRATION DESIGN The design of filtration systems will either be based on:

0

Cake formation (removal of bulk solids from liquid), or Clarification (removal of solids from bulk liquid).

' Pocock Industrial, Inc., Salt Lake City, Utah 1289

Which of these general categories an application falls into has an affect on the test work to be done, and the equipment type and choices to consider. Cake Formation Cake formation processes involve the removal of solids that are present in bulk from the liquid phase. Either the solids or the liquid may be the important recovered product. Definition of whether the solids or the liquid is the important phase has bearing on the method of treatment and the filtration equipment selection. In either case, filtrate clarity, washing of the filter cake, and final cake moisture content are important design considerations. Common types of filtration equipment used for cake formation applications include: Continuous vacuum filters (horizontal belt, disc, and drum), pressure filters (recessed plate, plate and frame, and tower presses), belt presses, and filtering type centrifuges. Gravity thickening is often used to increase the solids concentration of the filter feed. Clarification Clarification processes involve the removal of low concentrations of solids from the liquid phase present in bulk. The liquid phase is usually the important recoverable quantity. Solids are removed to meet specifications for downstream liquid processing, and/or to meet environmental regulations. Clarification filtration equipment may include: Precoat drum filters, pressure filters (recessed plate, plate and frame, and tower presses), deep-bed granular filters, cartridges, pressure leaf or bag type filters, and filtering type centrifuges. For many applications involving vacuum and pressure filters, some type of filteraid is used to improve filtration rates and filtrate quality.

SOLID-LIQUID SEPARATION SYSTEM OPERATIONS In the design of any filtration system, there are four SLS system operations to consider: 0 0 0

Pretreatment Solids Concentration Solids Separation Post Treatment

Pretreatment Some method of chemical or physical pretreatment may be required to improve the properties of the stream for SLS, or to make it amenable to separation. Pretreatment of a stream generally involves increasing the particle size of the solids present, and/or reducing the viscosity of the liquid. Hence, pretreatment of the feed makes the SLS separation easier by improving the sedimentation, and/or filtration properties. Some common forms of chemical pretreatment include the addition of coagulants or flocculants, extenders such as bentonite clay, and pH adjustment. Common forms of physical pretreatment may include temperature changes, crystallization, aging, freezing, and the addition of admix materials such as diatomaceous earth to modify slurry properties. Concentrating Solids Gravity thickeners or clarifiers are the most common types of SLS equipment used to concentrate solids in a dilute stream prior to filtration. Gravity sedimentation can be used alone, or in combination with a variety of filtration options depending on the extent of liquid removal required for the application. Control of solids concentration fed to filters will improve separation rates and process operation as well as reduce the size of the equipment required. Washing in counter-current thickening or clarification circuits (CCD) provides an efficient means to removehecover soluble components. A CCD circuit may also be used in combination with a filter or centrifuge to provide a more efficient and economical means of meeting process

1290

goals when large tonnages are involved. To achieve the same wash efficiency as a CCD circuit by using vacuum or pressure filtration alone, staged filtration steps with intermittent repulps would be required. Cross-flow filters may also be used to concentrate solids in some applications.

Solids Separation Selection of the type of equipment used for the final separation step depends on the characterization of the feed, the process goals, and the post treatment required. When filtration is to be considered as a final SLS step, the most important aspect related to equipment selection is the rate of cake formation. The rate of cake formation can be used as an initial guide in selecting filtration equipment. The following list shows a breakdown of equipment that may be considered based on cake formation rates achieved (Tiller 1974): Rapidly filtering materials (>5cm/s): Gravity pans, screens, horizontal belt or top-feed drum filters, or filtering centrifuges. Medium filtering materials (0.05 to 5 c d s ) : Vacuum drum, disc, horizontal belt or pan filters. Slow filtering materials (cdhr): Pressure filters, or disk and tubular centrifuges. Clarification (negligible cake): Cartridges filters, deep granular bed filters, precoat drums, and admix filters. Post-treatment Frequent review of the post-treatment goals to be accomplished by the separation is an important step in the selection of filtration equipment. Important factors affecting the design of filtration operations include:

0

Filtrate clarity Soluble material in the filter cake Liquor or moisture content of the filter cake End use of solids

The type, size, and cost of filtration equipment depends upon process requirements. Equipment required to produce a small change in any listed item could significantly increase the cost of equipment. Economics demand that only minimum requirements are met. Specification of equipment to yield results in excess of the basic requirements could be very costly and of little value to overall process goals. Filtrate clarity. Post-treatment of filtrate produced from a filter may be somewhat relaxed if clear filtrate is not a requirement and a thickener or clarifier is provided upstream of the filter as solids can be recovered by recycling the filtrate to the thickener feed. If a thickener or clarifier is not included in the flowsheet, the filtrate can be collected and a polishing filter used to clarify the primary filtrate. This will eliminate the necessity of using the primary filter as a clarifier. Specific equipment types of gravity media, pressure, and vacuum filters are used to clarify streams containing low solids concentrations. Soluble material in the filter cake. To remove/recover soluble components in a filter cake washing is required. The amount of wash solution needed to achieve a specific goal depends on the efficiency of the washing method, and the characteristics of the material in the filter cake. For filtration systems wash volume is defined by the number of displacements of liquor in the cake with wash solution (wash ratio). Wash efficiency is determined during testing, and is usually displayed as a function of wash ratio. The efficiency of washing on filters depends on the properties of the solids, the porosity of the filter cake (porosity is a function of particle size), and the method of washing used. There are two

1291

basic methods of washing on a filter, flood washing and spray washing. In flood washing, the wash solution is applied and forms a pool on top of the cake until forced through by the driving force of the filter. Flood washing is the most efficient method of washing. Pressure filters all use some form of flood washing to wash filter cakes. In spray washing, the wash solution is sprayed on the cake at a rate that equals the rate of passage through the cake. This type of washing is used for rotary drum applications. Filter cake washing can be accomplished in vacuum and pressure filtration equipment. The extent of washing required dictates equipment choices. For vacuum filtration equipment involving washing, horizontal belt, pan, and rotary drum are the best equipment options. These filter types allow for separate collection of strong and weak filtrates and facilitate counter-current washing. Most pressure filters are well suited for cake washing but counter-current washing is more complex. Counter-current washing can be accomplished in a pressure filter circuit by using cycled wash solutions from surge tanks or by successive re-pulping and re-filtering. For processes where extensive washing is required, inclusion of a CCD circuit prior to the pressure filter is common practice. Cake moisture content. Cake moisture content is a function of particle size, shape, and composition, and is also affected by temperature, liquid viscosity, and cake porosity. Minimum cake moisture is obtained by maximizing dewatering time and driving forces that control the rate of airflow through thin filter cake. Steam and heated air are also used to reduce cake moisture content. Mechanical expression or squeezing cake to remove excess moisture is also used in some types of recessed plate and tower (vertical) presses. Whether a cake can be successfully dewatered by air or mechanical means depends on capillary backpressure, which is a function of pore size. For air blowing alone to be effective, solids particles are usually larger than 10 microns diameter. Pressure filtration is used for smaller particle sizes, and for slower filtering materials when low cake moisture content is required. Vacuum filtration is limited by low driving forces and airflow rates during cake dewatering. End use of solids. Definition of filter cake end use impacts filter selections. Cake handling characteristics and downstream process requirements will often dictate the type of filter utilized. For example, if the filter is required to produce a constant supply of solids to downstream processes, continuous vacuum filtration may be applicable. Pressure filtration cakes are produced batch-wise, but could be stored for supplying continuous systems. Pressure filter cakes are also typically drier, and may require breaking or pulverizing to meet further processing demands.

FILTRATION EQUIPMENT DESIGN CONSIDERATIONS Overall Process Goals Process and logistic factors affecting the overall flowsheet require consideration throughout the equipment selection process. There are many types of filtration equipment available to accomplish a particular separation goal, some types may be more effective than others, or involve different techniques or mechanisms. The selection process should focus on choosing a generic equipment type that can provide required performance reliably and economically. Characterization of the Feed Characterization of the feed slurry is an important initial step. Aspects of the feed to be defined include: Particle Size Distribution Suspended solids concentration Dissolved solids concentration pH requirements Temperature and Volatility

1292

0

Tonnage to be processed Special properties and nature of the solids or liquid present

Particle size distribution. Has a significant impact on cake formation rates and subsequent filtration equipment selection. The amount of material present below 44 microns can have an especially significant impact on cake formation rates, and be a deciding factor in equipment selection. For this reason, samples used for testing must be prepared very carefully or they will bias results for equipment design. Over-grinding or mixing during preparation of a sample may also severely bias test results. Softer materials tend to slime depending on the material(s) present, posing a significant problem. For harder materials or precipitates sliming may not be an issue. However, the impact of attrition on the sample to be tested should be carefully considered. Suspended solids concentration. Plays a key role in equipment selection. Feeds containing high solids concentrations (usually 5% or greater) are typically concerned with cake filtration (removing bulk solids from the liquid). Cake accumulation rates for high solids feeds usually vary from several seconds ( c d s ) to several minutes (cdmin). Feeds containing low solids concentrations (usually less than 5%) involve the removal of small amounts of solids from the bulk liquid (clarification). Selecting equipment for clarification processes often involves equipment combinations. By their nature, clarification processes typically involve removal of ultra fine colloidal solids particles that are difficult to filter unless pre-concentration is practiced. Dissolved solids concentration. Impacts solute viscosity and is a function of the degree of saturation for the components present. Dissolved solids can potentially cause scaling in the internals of the equipment or blind the filter cloth due to precipitation. The latter can be initiated by pressure drop or temperature changes during the separation process. This may be an important consideration depending on the degree of solute saturation, and nature of the dissolved solids present in the treated stream. The pH of the feed stream. The initial pH of the feed stream, and the flexibility of making changes to the pH in the process has potential impact on filtration performance. Changes in pH can provide some flexibility in pretreatment of the slurry as a result of its affect on the behavior of particle surface chemistry. Use of coagulants and flocculants are affected by pH. Construction materials and their selection is impacted by pH and solute corrosion activity. Corrosivity may limit the selection of certain types of equipment. Temperature and volatility. Of the feed liquid present have impact on equipment design and selection. For operating temperatures above 5 0 T , and/or for volatile liquids, vacuum filtration equipment will need to provide for the increased vapor load to be treated as flashing occurs. Vapors produced could also present various explosion or health hazards, or may cause damage to the vacuum pump. For high temperature, and/or for volatile liquids, pressure filtration options should be considered. Process tonnage. The tonnage to be processed has an impact on the economics of equipment selection. Continuous filtration equipment may not be economically justified for low tonnages. For processes involving low tonnages and/or solids concentrations, there may be a definite economic advantage in manual or semi-automated batch operations, such as pressure filtration. Some pressure filters are fully automated and specifically designed to handle large tonnages. Pressure filters of this type approach the economics of continuous vacuum filtration. The economics of continuous versus batch filtration equipment is governed by the filtration rates achieved, tonnage processed, and cake moisture attained. Special properties and nature of the solids and liquid. May impact filtration equipment selection. Problems such as scale formation and cloth blinding can usually be overcome or prevented if recognized early in the design process.

1293

SLS TESTING: THE FINAL PHASE I N EQUIPMENT SELECTION After the above aspects of the type of process and separation have been examined, the engineer should now be at a point where some basic equipment type(s) can be selected to achieve the specific requirements for the application at hand. Testing is the next important phase required to complete the process of equipment selection by providing the needed design criteria. The equipment choices should reflect the complexity of the separation and the importance of the specific process requirements to be met. To make the selection choice easier, testing relevant to all the types of equipment chosen may be conducted. Once testing is complete and design criteria provided, an economic analysis can be completed, and final judgments for equipment selection can be made. Final equipment selection must also take into account process reliability and robustness, which are significant considerations in the overall economic picture. A typical SLS testing program will provide the following information: An in-depth analysis of pretreatment options, including dosage requirements for all chemicals and filteraid added to the system, and details of any necessary physical treatment. Settling rates and sizing basis required for thickener and/or clarifier design. Suspended solids concentrations achievable for the type of clarification method desired. Selection of suitable filter media and dischargeable cake thickness. Filtration cycle time needed for filter design (form time, wash time, and dry time). Production rate, washing efficiency, cake moisture, and air flow data required for the design of continuous vacuum filtration equipment (horizontal belt, disc, or drum type vacuum filters). Area or volume requirement, washing efficiency, and cake moisture data required for the design of batch pressure filtration equipment (manual, semi-automatic, or fully automatic recess plate, plate and frame, leaf or bag type, horizontal and/or vertical pressure filters). Filtration test results may also include; steam, ambient, or heated air drying data, compression drying data, data indicating scaling or cloth blinding tendencies, and cake discharge recommendations and properties. Viscosity data needed for slurry pump and pipeline design. REFERENCES R. C. Emmet, and C. E. Silverblatt. 1974. When to Use Continuous Filtration. Chemical Engineering Progress (Vol. 70, No. 12). B. Fitch. Choosing a Separation Technique. Chemical Engineering Progress (Vol. 70, No. 12). C. E. Silverblatt, Hemant Risbud, and Frank M. Tiller. Batch, Continuous Processes For Cake Filtration. Chemical Engineering (April 29). Cory B. Smith, and Ian G. Townsend. 2002. Testing, Sizing and Specification of Filtration Equipment. Article is included with this SME publication. Frank M. Tiller, and Joseph Wilensky. 1974. Pretreatment of Slurries. Chemical Engineering (April 29). Frank M. Tiller. 1974. Bench-Scale Design Of SLS Systems. Chemical Engineering (April 29).

1294

Testing, Sizing, and Specifying Sedimentation Equipment Tim Laros’, Steve Slottee’, and Frank Baczek]

ABSTRACT Various methods of testing have been developed over the years to study the sedimentation characteristics of solids for the purpose of sizing thickening or clarification equipment. Proven testing and data correlation procedures are presented for developing design specifications for the most common types of sedimentation equipment. The paper covers a brief review of sedimentation theory, feed sample characterization, flocculant selection, flocculation conditions, solids flux optimization, thickening rate, detention time, and thickened solids rheology. The paper reviews data evaluation methods to specify equipment features and predict performance. INTRODUCTION Sedimentation is the partial separation of suspended solid particles fiom a liquid by gravity settling. There are two primary sedimentation operations: thickening and clarification. Thickeners increase the concentration of solids in a stream and maximize liquid removal. Clarifiers remove relatively small quantities of suspended particles to produce a clear effluent. Sedimentation equipment designed for either function typically looks the same, having similar features; however, the approach to design is different and feed conditioning features are typically based on achieving either underflow slurry density or overflow liquor clarity. The basic principles and testing procedures for determining the size, performance, and basic design guidelines of thickeners and clarifiers are the subjects of this chapter. In recent years, new designs of high efficiency ultra high rate and high density underflow slurry thickeners have been introduced to the market. Testing techniques for sizing these newer technology units is in the hands of the specific equipment vendors as proprietary information. However, the testing techniques discussed in the following sections apply in principle to the majority of proven sedimentation equipment designs. A BRIEF REVIEW OF SEDIMENTATION THEORY

The primary forces present in sedimentation separations are gravity, buoyancy, and friction. The factors influencing these forces are: 0 0 0 0 0

Liquid density Particle density Particle size and shape Liquid viscosity Temperature

0 0

0 0 0

Particle flocculation Particle concentration Thickened slurry rheology Distance of settling Horizontal and vertical motion

The solids settling and rate of separation can be theoretically related to many of these factors, and therefore, influence thickener and clarifier design within the specified process requirements. While it is important for the designer of a thickener or clarifier to understand this, there is a point where theory becomes incomplete and empirical testing begins. The sizing of a thickener and clarifier is a combination of applied theory and testing. 1 EIMCO Process Equipment Company, Salt Lake City, Utah.

1295

The relative settling characteristics of particles can be separated into three basic regimes: 0

0

Free independent particle settling systems - settling rates are independent of particle concentration. Hindered settling - settling rate steadily decreases as the particle concentration increases. Compression - particles settling rate is restricted by the mechanical support of particles below, causing compaction and deformation.

All three regimes normally exist in each sedimentation operation, with one typically having the primary influence on the size and design of the equipment. Each regime is a function of solids concentration and how “flocculent” a particle is. “Flocculent” refers to the tendency of particles to cohere or stick together. Flocculent particles will generally be less than 20 microns in diameter, metal hydroxides, chemical precipitates, and most organic particles. Free, independent particles are discrete and include many mineral solids, salt crystals, and particles will little tendency to cohere. Figure 1 illustrates how the settling regimes are related to concentration and flocculent nature. Understanding the particle settling regimes is required to determine which tests to conduct for a specific thickening or clarification application. LOW CONCENTRATlOh

PARTICULATE SElTLlNG REGIME

COMPRESSION REGIME - .

CONCENTRATIOL

DEGREE OF PARTICLE COHERENCE TOTALLY DISCRETE

EXTREMELY FLOCCULENT PARTICLES

PARTICLES

Figure 1 Effect of particle coherence and solids concentration on the settling characteristics of a suspension. Sizing Nomenclature Thickener or clarifier area must be sufficient to allow the slowest settling particle to reach the bed of compressing solids or bottom of the tank before its associated liquor overflows. There is a critical concentration for which the settling rate, or upward liquor flow rate will limit the solids throughput rate. The area requirements for thickeners are frequently based on the solids settling rates measured in the hindered or zone settling regimes, where the settling rate steadily decreases as the particle concentration increases. Since the surface area of the thickener is the main variable, the mass flow through the system is considered a flux (the product of concentration and velocity), such as tonnes solids/m2/day. Frequently the inverse of this flux is used, called “Unit Area”, m2/tonnessoliddday. This allows system properties to be described in a general way for any size unit. The same nomenclature applies to clarifiers, where “rise rate”, usually expressed as d m i n or m3/min/m2,limits or sets the criteria for overflow clarity.

1296

WHEN IS TESTING RECOMMENDED? Data fiom full-scale sedimentation equipment, operating in the application under consideration, is always a first choice for sizing new equipment. However, care must be taken in evaluating the applicability of these data. The characteristics of the feed stream for the new application (i.e., mineral characteristics, particle size, viscosities, pH, etc.) must be identical to the existing application. It is also necessary to know whether the existing equipment is operating at its “capacity”, and what factors might be influencing how close to “capacity” it is. To the extent that the characteristics and operating conditions are different, bench or pilot scale testing may be required. Also, various sedimentation equipment designs may be offered by equipment vendors, with the potential of more efficient operation, and the specific design or features may require special testing to verify sizing or benefits. Any significant difference between the existing and new feed streams, or new equipment design features, will most likely require testing. OVERALL APPROACH TO TESTING SEDIMENTATION EQUIPMENT The following data is required to design either a thickener or clarifier, and testing must be designed to produce this information:

0 0

Feed stream characteristics Feed flow rate Expected underflow slurry density and/or overflow liquor clarity Flocculant type, solution concentration, and dose (if used) Conditions for flocculation (solids concentration, mixing time and energy) Vessel area and depth Settled solids rheology (for raking mechanism design and drive torque specification) Site-specific requirements: seismic zone, weather related specifications, local mechanical design codes, and the user’s preferred design specifications. Local operating practices

There are three basic approaches to testing for sedimentationequipment:

Continuous piloting - a small diameter thickener or clarifier of same configuration as the full-scale equipment. Semi-continuous bench scale tests - using laboratory pumps which pump feed slurry and flocculant solution into settling cylinders from which overflow liquor and underflow slurry are continuously collected. Batch bench scale settling tests - the conventional procedure requiring a relatively small amount of sample and graduated cylinders in which a sample is placed and allowed to settle. Generally, batch bench testing is sufficient for scale up if the testing engineer has previous experience in the application being evaluated, and proven scale-up methods are used. Larger scale testing is dictated by the following factors: Sample size - small amounts of sample will limit testing to batch bench scale Test costs - semi-continuousbench scale tests are more expensive than batch tests and pilot tests are significantly more expensive than both. Accuracy - although bench scale testing is relatively well developed and accurate, scale-up factors are still required. Pilot scale testing may be required for certain applications where the scale-up factor is uncertain or function ofthe equipment is otherwise uncertain (i.e. Will a thickener be able to discharge underflow with the rheology produced in the batch test? Are the feed characteristics expected to vary frequently, and will batch testing be able to quantify the effects on performance?).

1297

Demonstration - observing a clear overflow or a thick underflow from pilot equipment of the same design as the full-scale equipment significantly reduces uncertainty. Sample Characterization Whether clarification or thickening tests are conducted, sample characterization is common and necessary for both. Any sizing from testing is based on the characteristics of the sample tested. Without this data included in the basis of design, the sizing and predicted performance cannot be validated for the specified feed stream. Characterization requires the following measurements as a minimum: 0

0 0 0

0

Particle size distribution - include coarse (+lo0 micron) and fine (minus 20 micron) particle diameters Particle Surface Area Particle specific gravity Liquid specific gravity Dissolved materials, if any Temperature PH Feed solids concentration

Coagulant and Flocculant Screening Coagulants and flocculants are widely used to enhance settling rate which reduces thickener and clarifier diameter and improves overflow clarity and/or underflow slurry density. The terms “coagulation” and “flocculation” are sometimes used interchangeably, however, each term describes separate functions in the agglomeration process. Coagulation is a preconditioning step that may be required to destabilize the solids suspension to allow complete flocculation to occur in clarification applications. Flocculation is the bridging and binding of destabilized solids into larger particles. As particle size increases, settling rate generally increases. The science of flocculation is not discussed here but can be found in numerous texts and literature which is readily available from flocculant vendors. Both coagulation and flocculation are typically considered in designing clarifiers, whereas, flocculation is normally the only step in designing thickeners. Coagulants may be either organic such as polyelectrolytes or inorganic such as alum. Coagulants can be used alone or in conjunction with flocculants to improve the performance of the flocculant or reduce the quantity of the flocculant required. In some systems, where a flocculant has been used in an upstream process, a coagulant may be needed to allow additional flocculant to be effective. There are two primary types of flocculants: 0

Natural flocculants - Starch, guar, and other natural materials have historically been used for sedimentation flocculation, but have been replaced by more effective synthetic polymers. Synthetic polymeric flocculants - There are hundreds of synthetic polymers available developed for specific mineral types and applications

Because of the many available flocculants, a screening program is necessary to choose an effective flocculant. Testing time and expense usually limits the number of flocculants screened to less than ten. The choice of flocculant can be narrowed by considering the following: 0

0

0

Prior experience with flocculants on the feed stream under evaluation is always a good source of data. Experience of the test engineer or end user with a particular type of flocculant for the application (e.g., copper tailings, alumina red mud, etc) Test one each of the major types of flocculant charge: anionic, nonionic, and cationic.

1298

0

Test one each of the synthetic polymer length: long chain, short chain.

The purpose of the screening tests is to select a coagulant or flocculant whose generic type will most likely be effective in plant operation, and therefore, suitable for clarifier or thickener testing. Although a thickener or clarifier may be started up on the flocculant selected in the testing, it is very common to conduct furfher tests on the full-scale machine to firther optimize dosage or flocculant type. The flocculant manufacturer can be a source of great assistance both in the testing and the full-scale optimization of flocculant use. Equipment: Several 100 - 250 ml beakers for testing slurries for thickening, or, laboratory gang stirrer apparatus for testing liquors for clarification Several flocculants of different charge, polymer length A stirrer or hand-held laboratory spatula Syringe or burette for measuring the dosage Procedure: Coagulant or flocculant solutions should be made up according to the manufactures instructions and used within the shelf life recommended. The solution concentration recommended for testing is typically more dilute than the recommended “neat” concentration so that the viscosity is lower to make dispersion more rapid during testing. In the screen tests, each coagulant or flocculant is added to the beaker samples of representative of slurry or liquor in a drop-wise fashion, while the sample is mixed with a spatula or stirrer. The amount of coagulant or flocculant required to initiate floc particle formation is noted along with relevant notes as to the size of the floc, capture of fines, resultant liquor clarity, and stability of the floc structure. The dosage is typically noted in ghonne solids it the sample is primarily solids (thickener design), or in mg/l liquor if the sample is primarily for clarification and the solids concentration is low. Once a coagulant andor flocculant is chosen, and a dosage estimated, optimization testing is conducted to further quantify the conditions for the larger volume design tests.

Optimization of Thickening Test Conditions After a flocculant type is selected, the next step is to conduct a range of tests on larger sample volumes, using the selected flocculant, to gather data on the effects of feed slurry solids concentrations on flocculant dosage and settling rate. There is a range of feed solids for which flocculation effectiveness is maximized, resulting in improved settling characteristics. Operating within this feed solids range, results in smaller diameters, higher underflow slurry densities, better overflow liquor clarity, and lower flocculant dosages. Equipment: 500 ml beaker Laboratory spatula Syringe or burette for measuring flocculant solution Stop watch Procedure: Prepare a series of equal volume feed slurry samples at different solids concentrations. A good method is to start with the design feed slurry solids concentration, and then prepare a series of samples, decreasing in increments of 5% solids. For some very fine solids samples (e.g., alumina red mud, clays, leached nickel laterites, etc), it is recommended to also check a sample diluted to 2-3% solids. The final volumes if each sample should be around 350400 ml so there is room in the beaker for adding the flocculant and mixing. Begin adding the flocculant solution drop-wise and make notes on the dosage at which flocculation begins, and the settling velocity. Continue adding flocculant incrementally and noting the floc structure, fines capture, liquor clarity, and settling velocity. Once the settling velocity remains constant for a few tests, sample testing can be stopped, and then move on to the next sample. From the above tests, the plot shown in Figure 2 can be drawn and the results used to set conditions for the larger and final tests for sizing the thickening equipment. The test procedure for the design tests should be structured to span both the optimum solids concentration

1299

and concentration points higher and lower. The flocculant dosage should be checked at the optimum and at dosages slightly higher and slightly lower than that determined in the above tests. RANGE OF SOLIDS CONCENTRATION8 FLOCCULANT DOSE FOR DESIGN TESTS I

I

I

SOLIDS CONCENTRATION, wt%

Figure 2 Effect of solids concentration on flocculant dose and settling flux. Optimization of Clarification Test Conditions As discussed later in this chapter, clarification is typically controlled by the time necessary for the dilute suspension of solids to coagulate, flocculate, and then settle. In many instances, the rate of clarification is enhanced by increasing the solids concentration in the flocculation zone of the clarifier. This is done in a full scale operation, by internally or externally recycling previously settled solids into the flocculation zone where they are mixed with fresh, coagulated feed. The higher population of solids improves the flocculation efficiency and settling rate. To conduct these tests, a large sample of feed liquor (typically enough to contain ten grams of suspended solids) is first coagulated at the dosage and mixing intensity determined in the screening tests and flocculated according to the screening test. The solids are allowed to settle and the supernatant carefully decanted. The settled solids are then transferred into a smaller beaker and saved for dosing samples of freshly coagulated feed samples. The final, optimizing tests are conducted by preparing approximately 750 ml samples and coagulating as previously determined. Recycle solids are added, as a slurry, to give suspended solids concentrations of 1, 2, 3, and 5 g/l. Flocculant is then added and the mixing stopped. Settling rate is measured and clarity observed. The recycle solids concentration that gives the best clarity is selected for the final, larger design tests. Settled solids from these tests should be saved for the final tests since they are representative of recycle stream solids. In some leach solutions, very fine colloidal solids are present and are very difficult to coagulate. In these cases, it is typically necessary to test for mixing intensity and mixing time to obtain coagulated solids that are more amenable to flocculation. TESTING STRUCTURE AND EQUIPMENT TO SIZE THICKENERS Thickener sizing requires determination of two values: the area necessary to prevent the formation of a critical concentration zone of solids which can rise to completely fill the thickener (unit area), and the bed depth needed to attain the desired underflow concentration. Sizing of thickeners normally requires evaluating settling data to identify the critical solids settling flux to establish the minimum unit area for liquor release and separation from settling solids. For applications of slow thickening suspensions, such as clays, or where a high underflow density approaches paste consistency, the testing should also study the effect of the compression zone to determine the settled mud bed unit volume for thickening. Which ever is the limiting condition will have an effect on sizing and design of the thickener. In some instances, the calculated compression depth may be too great to be practical, and the area must be increased as necessary to provide the needed retention time with a lesser depth.

1300

Method of Coe Clevenger for Non-Flocculated Pulps In 1916, Coe and Clevenger proposed a method for sizing thickeners using a zone sedimentation or compression subsidence model, which has been proven to be valid in applications where the slurries do not require flocculation, or use only reagents such as lime which cause little variation in the floc size or dewatering characteristics. The basis for this sizing approach is that the settling rate of the pulp is a function only of the solids concentration, as represented by the zone settling regime of Figure 1. By definition, the critical point will be that solids concentration just before, to the beginning of, the compression zone. To test the initial settling rate, of a number of pulp samples ranging in solids concentration from feed slurry to underflow slurry is measured using conventional laboratory settling tests. For each concentration, an area value can be calculated from the observed settling rate and the volume of liquor which would report to the overflow, based on a pre-selected underflow concentration. A limiting maximum area per unit weight of solids per day can be determined. The equation used is: Unit Area, m2/tpd Where,

=

(1/C - l/C,) / v

C =test solids concentration, kg/L C, = underflow solids concentration, kg/L v = initial settling rate at test conditions, m/d

At the end of the series of tests, the solids are dried and weighed and values substituted in Equation (1) to calculate the corresponding unit area, A maximum value will be obtained, representing the limiting or size-determining conditions, as illustrated in Figure 3.

log (FEED - UNDERFLOW), kg liquidlkg solids

Figure 3 Coe Clevenger method of determining thickener unit area. Methods Derived from Kynch Theory for Flocculated Pulps When polymers are used to flocculate a suspension, dilute suspensions will produce much larger flocs, compared to higher concentrations as discussed earlier, and it is preferable to run a single test at the expected feed solids concentration. Procedures derived form Kynch theory can be used to determine the unit area, m2/tpd. One of these methods is the Talmage and Fitch method which uses Equation (2): Unit Area, m2/tpd = t, / C,H Where,

,

(2)

t, = settling time, days C, =test of feed solids concentration, kg/L H, = initial height of pulp in the test, m The value oft, is determined from the settling curve by any of various methods, the selection of which depends on the particular testing organization and its experience in scaleup from this

1301

approach and organization's proprietary methods proven in actual practice. One commonly employed system which produces conservative results is illustrated in Figure 4, and is based on the bisection of the angle formed by two tangents to the straight line portions of the curve, the intersection of the bisection and the settling curve defining the critical point. A tangent to the curve drawn at the critical point intersects a line representing the height of the pulp at underflow concentration, giving the value oft,. Another approach, called the Oltmann method employs a straight line drawn from the start of the settling curve to a discernible point on the curve at which compression is believed to begin. The extension of this line to the underflow line, as also shown in Figure 4, gives the value oft, to be used in the equation.

5

-bE W-

0 t, determined by Oltmann Lu I-

s A W 2

PULP LEVEL AT SELECTED UNDERFLOW CONCENTRATION -----------

d-

-I

n -I

J

3

n

I

I

t, ! t,

t, determined by Kynch

ELAPSED SETTLING TIME, min

Figure 4 Sizing a thickener from batch data using Talmage and Fitch equation, and selecting time, t, by Kynch o r Oltmann methods Both methods described above usually apply a scale up factor which accounts for the mud bed depth in an actual thickener. Various other sizing methods have been developed over the years. The Warren Spring Laboratory published an extensive review on 1977. one such "flux" type model is the WilhelmNaide model which is an adaptation of the Yosioka & Hassett models of the 1950's and 60's. The flux type models were developed to overcome some of the over and under sizing problems that were found to occur with Kynch and Talmage & Fitch, especially with modern flocculants. The Wilhelm-Naide model uses the total thickener flux, settling flux in addition to withdrawal flux, to define unit area as:

UnitArea= Where,

I(b- 1)/bib-' ab

C" b-'

(3)

a and b = coefficients developed by the equation vi = aCib where Ci is measured as weight of dry solids per unit volume, and vi is settling velocity of a layer with suspended solids concentration Ci, lengtwtime. C,

= underflow

solids concentration, in weight dry solids per unit volume.

As proved by Kynch, tangents to the test settling curve extrapolated to the vertical axis permits determination of the solids concentration or Ci. The slope of the tangent line yields the settling rate. Thus, by plotting velocity as a fhnction of solids concentration on log-log paper, the values

1302

of a and b are determined. It is interesting to note that normally 3 straight line relationships are obtained and at least two in any case. Accordingly, the appropriate values of a and b are used in Equation 3 to calculate the unit area value at the desired underflow solids. The reader is referred to the literature for the complete development of the method. Determining Effects of Compression on Sizing As pulp enters the compression regime, different factors come into play. Compression bed depth will typically have an effect on the overall thickening rate and higher bed depths will reduce the unit area. Predicting the effect of increased bed depth is not possible from theoretical considerations alone, as too many other factors have an influence, including the metallurgical nature, particle size distribution, and particle of the solids. The flocculant dosage and the characteristic floc structure affects the unit area, as well as the mechanical action of the rake and the particular rake design. Ideally it would be desirable to carry out this test in a cylinder in which the pulp could be maintained at a depth approaching that of a full-scale thickener. Usually, this is impractical and appropriate scale-up methods are needed. If cornpression is expected to be a critical factor, additional tests should be carried out in deep cylinders, and the values plotted so as to permit extrapolation to grater depths, using the log-log relationship shown in Figure 5. Generally, this extrapolation should not be extended beyond a pulp depth in compression of l m unless full-scale data indicate the benefit of greater depths. The mechanical action in the compression zone and the depth of the pulp in actual compression will influence the rate of thickening. In evaluating the compression zone requirements from a batch test, one can calculate the unit volume (cubic meters of compression zone volume per tonne per day) from tests conducted under identical conditions except in cylinders of deeper pulp depths. It will be found that the calculated volume increases as the cylinder depth increases. Using the unit volume calculated from tests at a normal cylinder depth, as a design basis, can result in an undersized unit if the compression zone determined the size of the thickener and if only the detention time were considered. This is generally manifested as a lower unit density than expected or the need for a greater amount of flocculant than indicated in the original test work. In determining the compression zone unit volume, the following equation can be applied: Unit volume, V, m3/tpd Where,

=

t, (ps - PI) /

[Ps(Pp

- Pi)]

(4)

t, = compression time required to reach a particular underflow concentration, days. ps, PI , and pp = densities of solids, liquid, and pulp (average), respectively, tonne/m3.

The log-log plot of unit volume versus average pulp depth (Figure 5) will usually produce a straight line having a slope less than 1, decreasing in value as underflow concentration increases.

1303

2 m Cylinder

2 Liter Cylinder

-m 0

log AVERAGE PULP DEPTH, rn

Figure 5 Log-log plot of compression zone unit volume vs average pulp depth Batch Settling Test Tests are conducted using the apparatus illustrated in Figure 6 . Figure 7 shows a typical batch test data sheet. Equipment: 2-liter cylinders with markings to measure height, settling rate, and volumes Picket rakes with motors (rakes should turn at 6 rph) Flocculant addition and mixing apparatus Stop watch, or suitable timing device Balance for weighing cylinders and samples Apparatus to filter, wash (if required), and dry solids Flocculant solutions Procedure: Record the tare weight of all cylinders and beakers used for gathering sample weights, and prepare flocculant solutions. Add the slurry sample to the cylinder in a manner so that the slurry sample is representative of the characterized feed sample, and prepare the dosage of flocculant determined for the test from previous screening tests and the planned testing program. Measure the weights and volumes as required by the data sheet. When ready to begin the test, mix the cylinder of slurry using a suitable apparatus and add the flocculant solution while mixing. Figure 6 shows a plunger type mixer made with a rubber stopper mounted to a hollow tube with a syringe attached to the top for delivering a pre-measured amount of flocculant solution while mixing. Once the flocculant is mixed into the slurry, the mixing apparatus is withdrawn, rakes are quickly inserted, and the settling rate of the slurry interface is recorded. The time for each test depends on how long it will take for the settled slurry to reach its final level and density. When the test sample has reached terminal density, the rakes are carefully removed from the cylinder in a manner to minimize re-suspension of the settled slurry. Allow the cylinder and sample to set for a while after the rakes are removed so any disturbed slurry resettles, and take the following data: Total volume of sample before decanting Total weight, plus tare, of the cylinder and sample Total slurry volume after decanting supernatant Settled slurry weight after decanting Settled slurry height

1304

MOTORIZED PICKET RAKES

FLOCCULANT ADDITION 8 MIXING DEVICE: SYRINGE, HOLLOW TUBE, RUBBER STOPPER WITH BOTTOM FLOW DEFLECTOR

2 LITER GRADUATED CYLINDER

Figure 6 Batch settling test apparatus. After the supernatant has been decanted and the above measurements made, repulp the settled slurry with the mixing apparatus and pour off as much as will flow into a beaker. With a tarred syringe, sample 25mls of the slurry and weight the sample to obtain the density, and then dry the sample to obtain the per cent solids concentration by weight. If dissolved solids are present in the liquor, the dried sample weight must be corrected for the dissolved salt concentration. Complete the data sheet and correlate the data using one of the methods described earlier. Most testing laboratories that conduct thickening tests on a regular basis, typically have a computer programmed that will plot the settling curve and correlate the data to determine a unit area for a specified underflow slurry density.

1305

THICKENER TEST DATA SHEET COMPANY.

TEST No.: DATE: LOCATION:

CONTACT:

BY:-

MATERIAL: OBJECT OF TEST: TEST I.D.: THICKENER TYPE: PICKET RAKE SPEED, RPH: FLOCCULANT MIX INTENSITY:

FLOCCULANT CONCENTRATION, g/l: VOLUME ADDED, ml: DOSAGE, glt: SElTLING VESSEL SIZE, mllft: UNDECANTED SLURRY VOLUME, ml: __ SLURRY WEIGHT, gm: TARE: DECANTED SLURRY VOLUME, ml: SLURRY WEIGHT, gm: DRY SOLIDS WEIGHT, gm: -TARE: ULTIMATE INTERFACE HEIGHT, ml: ___

FEED SLURRY INFO: TEMPERATURE, 'C: pH: DESCRIPTION:

-

-

PARTICLE SIZE DISTRIBUTION-------Mesh wt% Mesh wt%

-

-

-

-

-

-

-

SUPERNATANT DESCRIPTION-------

THICKENED PULP DESCRIPTION---.

SP. GRAVITY LIQUOR: SP. GRAVITY SOLIDS: INITIAL % SOLIDS: FINAL % SOLIDS: REMARKS:

TIME, min.:

INTERFACE HEIGHT, ml:

2. 4.

5. 6.

-, 9. 10. 11. 12. 14. 4c

... 16.

18. 19. 20. 21. 22. 23. 24. 25. 26.

Figure 7 Batch thickening test data sheet. Semi-Continuous Settling Tests A semi-continuous settling test can be effective is determining the initial settling velocity for various feed solids concentrations and flocculant dosages if enough sample is available for conducting the test through all the variables desired. The test is usually representative of the initial slurry free settling zone seen on the batch test settling curve (Figure 4). To obtain compression zone settling data, the flocculated sample is withdrawn from the bottom of the test apparatus and the settling rate is measured in a separate 2-liter cylinder fitted with picket rakes. Compression zone data is collected in the same manner as that described above for the batch settling tests. Some thickener vendors use a semi-continuous testing apparatus with a deep cylinder that is raked so that the test can be interrupted and the flocculated slurry allowed to settle to obtain the compression zone settling data. The same data is typically collected as in the batch tests. Figure 8 shows one type of semi-continuous testing apparatus. A well mixed, homogeneous, feed slurry sample is pumped into the test apparatus. Dilution water can also be pumped into, and mixed with the feed slurry, prior to entering the apparatus to adjust the feed slurry to the different feed solids concentrations determined for the test. Flocculant solution is then pumped and mixed with the feed slurry in a mixing device that is integral with the feedwell of the test apparatus. In addition to volumetric flow rates of the different feed streams, the volumetric flow rates and density of the overflow liquor and underflow slurry must be measured to obtain a mass balance for the particular test conditions.

1306

I

I

14d-J

I

FEEDSLURRY MIXTANK

FEED SLURRY PUMP

FLOCCULANT PUMP

FEEWVELL

SoLUT'oN FLOCCULANT 'INK

OVTRFLOWUOUOR HOLDINGTANK

OPERATING PULPLEVEL-

"\

CLEPiRTEST CYLINDER

iiB

1 L

UNDERFLOWSLURRY HOLDINGTANK

CONTROL VALVE

Figure 8 Semicontinuous test apparatus Rheology Tests and Drive Torque Selection The basic test methods described above will allow the design engineer to select a unit area for design of a thickener for a particular application. They will also set the feed solids concentration and flocculant type and dosage required for the particular unit area. Selection of a rake drive torque can be done fiom historical data of operating units in the same or similar applications. A table presented in the chapter "Design Features and Types of Sedimentation Equipment" gives guidelines on the use of drive torque K factors, based on experience across a range of applications, which are applied in Equation 5 for the selection of drive torque: Torque, Nm = K DZ Where,

(5)

K = torque factor D = thickener diameter, m or ft

The information given in the table relates to thickeners operating at underflow slurry densities with yield stress values in the range of 10 - 70 Pa. In recent years, thickening and thickener design technology has advanced to allow the production of much higher underflow slurry densities that exhibit much higher yield stresses and approach the limit of flowability. In these particular applications and designs, the rheology and yield stress of the underflow slurry becomes a critical design guideline €or rake mechanism design and drive torque selection. For paste and high density applications, the K factor is related to the yield stress of the thickened slurry. When a thickener is designed to produce a very high underflow slurry density, or paste, the yield stress of the underflow increases rapidly as the solids concentration increases and approaches the consistency if a formed, but un-dewatered, filter cake. Since this is a relatively

1307

new field, most vendors of thickener drives have developed company proprietary information on the capability of their respective drives. Figure 9 shows a typical curve of yield stress versus slurry solids concentration with the corresponding effect on the required torque capability of the thickener drive. The particular curve must be developed independently for each application since the rheology characteristics will vary depending on the particle size distribution, particle shape and surface area, slurry solids concentration and mineral characteristics.

PASTE THICKENER UNDERFLOW SLURRY

1 I

TYPICAL THICKENER UNDERFLOW SLURRY I I

REF: FORMED

I

SLURRY SOLIDS CONCENTRTION, wt%

-

Figure 9 Typical curve showing how required drive torque increases relative to slurry yield stress. Sizing Rakeless Thickeners The new ultra high rate rakeless thickeners are typically sized from pilot plant operating data obtained with a pilot scale unit of the rakeless thickener. The unit’s design incorporates feed dilution to optimize flocculation and maximize settling rate, and also internal dewatering cones that enhance liquor removal and the rate thickening. The particular effects of these internal features are difficult to simulate on the bench, and the most accurate sizing is obtained from the pilot plant testing. The unit design incorporates a 60 degree cone which facilitates thick underflow slurry movement to the discharge cone without rake action. TESTING STRUCTURE AND EQUIPMENT TO SIZE CLARIFIERS There are a variety of clarifier designs applied in minerals processing, and most are designed to use coagulants and flocculants to improve the efficiency and rate of clarification. These are covered in the chapter “Design Features and Types of Sedimentation Equipment” of this manual. The particular stream to be clarified, the concentration and characteristics of the solids to be removed, and the degree of clarification required will usually dictate the feed conditioning features of the clarifier. These feed conditioning features typically are incorporated into the feedwell system, where the flocculant and recycle solids are combined with the fresh feed, and the mixing jntensity and retention time designed to provide floc formation and solids capture. The current practice and equipment designs used in similar applications, and which are performing well, can be used to narrow the options, and a testing program can then be designed to optimize the feed preparation conditions, special feed conditioning features and equipment size. Clarifier testing involves an understanding of the free particle settling regime and how it is affected by the flocculent mature of particles and the imperfections in full-scale clarification

1308

resulting from thermal and density velocity gradients. If the process allows, the use of coagulants and flocculants greatly improves the efficiency and rate of clarification, as well as the size of the clarifier. Clarifiers are sized on a combination of the area requirements which are based on the free settling of particles, or overflow liquor rise rate, and the detention time required to achieve the desired overflow liquor clarity. These values define the clarifier diameter and depth. The area requirements are based on the free settling rate of the slurry in m/hr divided by an accepted scale up factor, either at 0.5 or 0.75, to give a rise rate. The rate of liquor overflowing the clarifier must be less than the free settling rate of the particles, hence the 0.5 or 0.75 scale up factor depending on the application. The quantity of liquor overflow divided by the rise rate determines the required clarifier area and hence diameter. The detention time requirements are based on a second order detention curve relating liquor clarity to settling time. The laboratory detention time is dsually multiplied by 4 to give a required dynamic detention time. This dynamic detention time then gives the volume of liquor that needs to be detained or stored in the clarifier, generally below the bottom of the feedwell. Together with the diameter, and feedwell dimensions, the clarifier height is then determined.

Jar Tests For clarification, the polymer, type of addition, and mixing is generally critical. Determination of the type of polymer addition (staged or single application) and the degree of mixing (flash or gentle) is usually accomplished with jar tests. These tests are carried our in a gang stirrer with variable speed, flat paddle stirrers in a standard testing apparatus as manufactured by the Phipps Bird Company. From these tests, polymer type, dose, and the degree of mixing and mix time are determined for the best overflow liquor clarity. Settling Tests Once the polymer and mixing requirements are determined, a standard two liter settling test is conducted, in which interface height versus time is recorded. Generally, picket rakes are not used in this test since the objective is not to define a thickener size, but a clarifier size. This test yields the free settling rate. Detention Test The detention test is carried out preferably in a relatively large vessel, 2 to 4 liters in volume, with a diameter-to-depth ratio of 0.5:1, thermally insulated and covered if at an elevated temperature. The sample is placed in the vessel, flocculated as determined in the screening tests, and allowed to settle for a length of time which will yield the desired clarity. Samples are withdrawn from a point near the middle of the vessel and analyzed by any suitable means. It will be observed that the clarity in this test vessel generally is about the same in a zone from 1 to 2 cm below the surface to just above the settled solids; thus, the depth of sampling is not important, provided it is near the center of this zone and care is taken not to stir up the settled material during sampling. Normally, five or six samples, collected at 5, 10, 20, 30, and 60 minutes, will be adequate to define the settling characteristics of the suspension. Data from the test are plotted as a log-log plot of clarity (suspended solids concentration) versus time as shown in Figure 10. Surface area is then selected to make sure the rise rate is not excessive and in a safe region. If interface settling occurs in the batch test, the interface settling or bulk settling rate can be used to determine a minimum area. The settling velocity or bulk steeling rate is measured and converted to I/min m2 to select the minimum area, and a scaleup factor applied based on experience in the application.

1309

i

P

iz I-

W

0

z 0

0

.........

log SETTLING TIMES, rnin

Figure 10 Log-log plot of suspended solids concentration vs time Inclined Plate Clarifiers Inclined plate clarifiers, commonly known as lamella clarifiers, are shallow depth sedimentation devices in which the inclined plates serve to increase the total settling area of the clarification zone; essentially creating a series of small clarifiers in parallel. The settling area of an inclined plate clarifier is the horizontal projected area of each plate multiplied by the number of plates. Sizing inclined plate clarifiers is similar to sizing conventional clarifiers. First, a jar test is conducted to determine polymer type, dose, and mixing conditions to achieve the desired clarity. A detention time test is then conducted to generate the second order detention curve. This test is slightly different from the conventional detention time test in that the sample of clarified liquor collected a prescribed distance below the surface of the liquor. This distance is normally the spacing of the plates in the clarifier. From the detention curve, a required time to achieve a desired liquor clarity is determined. From this time, a settling velocity is calculated according to Equation (6):

Vs = (Sp - Suf) / Tr Where,

(6)

Vs = Settling velocity, m/sec Sp = Plate spacing, m Suf = Scale up factor, usually 0.5 Tr = Required time, sec

The Projected Plate Area, PPA, in m2 is then determined from Equation (7): PPA = Q I Vs (3600) Where,

(7)

Q = Overflow from the clarifier, m3/hr Vs = Settling velocity, m/sec

The appropriate manufacturer’s model having the required projected plate area can then be selected. Counter Current Decantation (CCD) Design Methods Washing in a thickener consists of mixing solids and associated solution with wash water, settling the solids, decanting the clarified solution and then repeating the process as required until a targeted removal of dissolved material, which was present in the original slurry liquor, is achieved. The variables in the design of a CCD circuit to achieve a desired dissolved solids removal are:

1310

The number of washing stages. The ratio of overflow volume to the volume of liquor in the underflow, which is often referred to as “wash ratio.” The “efficiency” of each stage which refers to the completeness of mixing of the underflow and overflow slurries and liquors which enter each washing stage. The objective of CCD testing is to quantify each of these design parameters. This is usually done by using data collected from bench settling tests to determine the expected underflow slurry density of a generic stage. A mass balance around the CCD circuit, using the “LDC” method, is used to determine the number of stages required at varying design variables of wash ratio and inter-stage mixing efficiencies. Figure 11 shows the “LDC” method of calculating a mass balance around the CCD washing circuit to determine the overall circuit washing or recovery efficiency. In the method, L = kg liquor / kg solids, D = kg solute / kg solids, C = kg solute / kg liquor, C=D/L, and the calculations begin at the last stage and work towards the first stage. To begin the calculation, set D=l at the last stage to represent an unknown value. Once the mass balance is complete, the efficiency, or recovery of solute, is determined by dividing the value D for the last stage underflow (D=l) by the value of D for the feed and subtracting from 1, then converting the decimal result to a percent as shown in Equation 8: CCD Circuit Recovery, % = (1-(1/x)) 100 Where,

(8)

x = feed solute value of D, kg solute / kg solids, calculated from the mass balance.

L D=X C

WASH LIQUOR L D C

. I MIXED FEED L D C

Figure 11 Example of mass balance approach to determining CCD recovery efficiency. CONCLUSION The overview of testing and sizing sedimentation equipment should give the reader an understanding of the general approach to designing and defining a test program for sizing sedimentation equipment for different applications and desired results. Although testing techniques will vary depending upon the engineer’s experience and between vendors of sedimentation equipment, the same basic principles apply to all.

1311

REFERENCES Dahlstrom, D. A. 1986. Selection of Solid-Liquid Separation Equipment. In Advances in Solid-Liquid Separation, ed. H. S . Muralidihara, Chapter 9. Columbus: Battelle Press. Baczek, F. A., R. C. Emmett, E. G. Kominek. 1988. Sedimentation. In Handbook of Separation Techniquesfor Chemical Engineers, ed. P. A. Schweitzer, 2nded., Section 4.8. New York: McGraw Hill Book Company. Pearce, M. J. 1977. Gravity Thickening Theories-A Review. Warren Spring Laboratory, Department of Industry. Baczek, F. A. 1985. Equipment Design and Process Consideration in Flocculation and Sedimentation. Proceedings of the Engineering Foundation Conference. 26 1-272.

1312

Testing, Sizing and Specifying of Filtration Equipment Cory B. Smith', Ian G. Townsend2

ABSTRACT Specification of filtration equipment to meet process performance and economic objectives can be reliably determined by bench scale testing when representative samples are available. Through development of a series of generalized vacuum and pressure filtration relationships, filter selection and sizing can be obtained for specified feed samples. Impact of process variables on filtration performance can be assessed as an integral part of economic evaluation. Complex filtration systems that include counter-current filtration and various washing schemes can also be reliably evaluated using bench-scale techniques. Test sample representation is the key to obtaining results that can be confidently used to select and size filtration equipment. Pilot plant scale testing may be required if samples are affected by aging, particle size variation, temperature effects, and other process variables that would change filtration characteristics of the feed sample. Pilot scale operation is also required when bulk samples of filter cake or filtrate are required for downstream process testing. Pre-thickening of the sample and use of flocculants to enhance filtration rate requires careful evaluation to obtain reliable results.

INTRODUCTION The purpose of this paper is to give an overview of the steps in sizing and specifying filtration equipment. These steps include determining the ultimate objectives, obtaining representative samples of the materials to be filtered, conceiving and executing a testing program, and finally, interpreting test data to determine filter sizing. The first step should always be to determine what objectives must be accomplished. Selection of filtration equipment requires choosing the optimal type of equipment. For example, horizontal belt, rotary drum and disc types of vacuum filters, as well as horizontal or vertical filter presses may all be candidates for filtration of many types of metallurgical solids. Each type, however, requires a different testing approach. Furthermore, process requirements must be understood prior to initiating laboratory work. Will the process require cake washing to reduce or remove the liquor solute content of the cake? What moisture targets are there for the cake produced? These types of questions should be answered prior to starting a testing campaign to insure all data required to meet the process objectives are collected during testing. Once objectives are understood, a test program can be conceived. Vacuum and/or pressure filtration tests can be planned as needed to determine unit production rates for each relevant piece of equipment. Planning should include collection of all necessary data to describe various process options and their associated operational trade-offs. Cake washing rates and efficiency data should be collected if the filter is to be a washing application. Similarly, drying cycle data should adequately describe the trade-off between cake dryness and production rate. The effects of expected variations in feed conditions (such as feed slurry solids content, temperature, applied vacuum, mineralogy, precipitation conditions etc.) should be considered in the testing scope as well. Once a filtration testing program is properly conceived and outlined, the execution of the testing can be confidently undertaken. Data interpretation and correlation are relatively straightforward with an understanding of filtration theory basics. 1 Pocock Industrial, Inc., Salt Lake City, Utah. 2 Larox, Buckinghamshire, UK.

1313

BRIEF REVIEW OF FILTRATION THEORY Classical treatment of filtration theory nearly always begins with Equation (1) Poiseuille's equation (Carman, 1938; McCabe and Smith, 1976). This form of the equation contains a couple of cumbersome terms, V for the volume of filtrate, and c for solids concentration expressed as the weight of solids per unit volume of filtrate.

For more practical use, this can be converted to:

By solving the expression, and then plotting W as a h c t i o n of Of on log-log paper, a straight line of 0.5 slope should result (and it generally does). This equation will be the basis of the cake form rate correlation, which is explained later in this paper and is used for filter sizing. A graphical representation of this correlation is shown in Figure 2 later in this paper. The full mathematical derivation of equation (20) is shown for reference in the Appendix to this paper. The notation used above in equations (1) and (2), and hereinafter, is as follows: Bf

=

P A Ap u c R,

= =

P

=

v = =

= = =

w = s = s, =

Cake formation time. Volume of filtrate. Liquid viscosity. Filtration area. Pressure drop across the filter cake. Average specific (filtration) resistance (units of length/mass). feed slurry solids concentration in terms of mass of solids per unit volume of filtrate. Resistance of the filter medium (units of length'). Mass of dry solids in the filter cake per unit area. Liquid density. Weight fraction of solids in the filter feed slurry. Weight fraction of solids in a formed, but undewatered filter cake.

FILTRATION TEST PLANNING Define Objectives Process needs should drive the testing objectives; hence, it is crucial to understand these process requirements prior to testing. Examples of possible process requirements for filtration equipment include, but are not limited to: Feed slurry dewatering. 0 Cake washing. 0 Filtrate clarity requirements. Cake moisture content targets. Cake transportability. Filter production rate. All process requirements must be envisioned prior to planning the test strategy.

1314

Sample Selection, Characterization, and Preparation Selection of a representative samples is critical if test results are to apply to the actual processes. Sample availability may be plentiful when testing materials from an existing operation; however, sometimes sample quantity may be extremely scarce, such as during pre-feasibility research programs. In any case, the sample should be matched as closely as possible to the actual expected stream and conditions for which the filtration equipment is to be specified. Sample characteristics that should be considered when determining if a sample is representative include several physical and chemical parameters. Solids concentration, specific gravity, size distribution, slurry pH, ORP, and chemical dosage (such as flotation reagents or flocculants) should be matched as closely as possible. Variations in ore type or mineralogy in a mine can significantly affect filterability of flotation concentrates and should therefore be taken into account. Sample age can also be extremely important. Many metallurgical solids are reactive and change with time significantly enough to alter filtration characteristics. An example would be sulfide flotation concentrates, which will oxidize over time. For this reason, samples for filtration testing should be as fresh as possible. Ideally, testing should be conducted immediately after the sample material is produced. This could mean that testing must be conducted at a mine or plant site, or at a pilot plant where flow-sheet development work is underway. Samples should not be allowed to freeze, dry, or be dewatered and then re-pulped prior to testing, since the resulting sample will probably not behave the same as it would have originally. Finally, expected process parameters should be closely matched in preparation of the sample for filtration testing. Feed solids concentration, temperature, pH, etc. should be carefully matched to the process conditions. When testing metallurgical slurries, a fall in temperature can cause crystallization of salts, thereby changing cake-washing process to re-dissolution, with very different kinetics. Temperature changes also influence filtrate viscosity, and cloth blinding through salt deposition. Feed solids concentration is especially import to consider, since any de-watering process will be greatly affected by the quantity of water present. Often processes produce samples at solids concentrations that are not amenable to filtration. In these instances it is generally recommended to pre-thicken filtration feed. An example would be flotation concentrates, which generally are thickened prior to filtration. If thickening of the slurry is to be conducted, it should be carefully done so that the resulting filter feed is at a reasonable pulp density for thickener underflow, and contains a proper dose of the correct flocculant. It is often advantageous to conduct all solidsfliquid separation testing in concert, since (for example) thickening test underflow material could be used for rheology and filtration testing feeds. Equipment Needed Vacuum filtration testing. The primary equipment required for vacuum filtration test work consists of a grid of known area covered with an appropriate filter cloth and surrounded by a metal or plastic shim to contain the pulp sample. This drainage grid or filter leaf is supported vertically on a vacuum flask. Alternately, a similar vacuum leaf can be attached to a flexible tube that connects to a vacuum flask. This alternate arrangement would allow for the leaf to be inverted and immersed in a container of filter feed slurry to simulate drum type vacuum filter operation. The differential pressure is translated from the vacuum pump to the filter leaf surface through large bore tubing and fittings. The vacuum pump should be equipped with an internal bypass system to control the vacuum level without the introduction of bleed air. More information on testing equipment, as well as diagrams of possible testing apparatus can be found in Perry’s Chemical Engineering Handbook (Perry and Green, 1991). Pressure filtration testing. Bench-scale pressure filtration test work can be performed using a pressure bomb device. The apparatus consists of a 250 mm section of nominal 50 mm pipe, capped with two flanges. The upper flange contains fittings for air pressure connection and the sample feed port. The lower flange contains an integral drainage grid, which supports the filter media. The filtrate port is centered in the bottom flange, below the filter media, . More information on testing equipment, as well as diagrams of possible testing apparatus can be found in Perry’s Chemical Engineering Handbook (Perry and Green, 1991).

1315

DATA COLLECTION Vacuum filtration and washing tests should be conducted to collect a general set of filtration and wash efficiency data to design and size vacuum filters. Testing should (as applicable) examine the effect of applied vacuum level, feed solids concentrations, cake thickness, dry time, filter aid (flocculant or D.E. etc.) and the volume of applied wash solution on production rate, filter cake moisture and wash efficiency. Similarly, pressure filtration and washing tests should be conducted to collect ageneral set of filtration and wash efficiency data to design and size pressure filters. Tests should examine the effect of applied pressure, feed solids concentration, cake thickness, air blow time and the volume of applied wash solution on production rate, filter cake moisture and wash efficiency. General Procedures Vacuum filtration testing. To produce a test filter cake from a slurry sample, a given weight of pulp

at the proper temperature and known to yield an approximate cake thickness should be poured onto the upturned test leaf while the ball valve connecting the leaf to the vacuum flask is opened to apply the differential pressure. As the last of the liquid phase disappears through the surface of the formed cake, the form time ends and is noted, and a known volume of wash water is poured onto the surface of the newly formed cake (only applicable if wash is to be tested). Once the last of the liquid wash solution disappears through the surface of the cake, the wash time is ended and noted and the subsequent dry time begins. At the end of the dry time, the filter cake is discharged from the leaf, and the wet weight and cake thickness are determined and recorded. After drying, the dry cake weight is determined and recorded for cake moisture calculations. Additional cakes should be collected in the same manner, each at different test conditions, until the range of test variables selected has been adequately covered. Pressure filtration testing. To produce a test pressure filter cake from sample slurry, a given weight of pulp at the proper temperature and known to yield an approximate cake thickness is poured into the pressure chamber. The sample port is closed and air pressure applied above the feed slurry to facilitate initial cake formation and dewatering. As the last of the filtrate is produced, shown by rapid air flow through the drainage grid, the form time is noted and recorded. Immediately following the form time the filter is depressurized, and a known volume of wash solution is poured onto the surface of the newly formed cake (for washing cases). The sample port is then closed and air pressure applied to facilitate cake washing and dewatering. Once the last of the liquid wash solution is produced, the wash time is noted and the subsequent dry time begins (only applicable if wash was applied). At the end of the timed air blow, the filter cake is discharged from the filter, and the wet weight and cake thickness are determined and recorded. After drying, the dry cake weight is determined and recorded for cake moisture calculations. Additional cakes should be collected in the same manner, each at different test conditions, until the range of test variables selected has been adequately covered. Wash procedure discussion. Experience has shown that the construction of a washing curve from data is often made more difficult than necessary by the choice of methods employed to collect these data. Hence, when washing effectiveness data are collected, the method of testing should be varied as described in the following section, A washing stage in vacuum and pressure filtration consists of cake liquor displacement by both air and the applied wash fluid. Remaining cake liquor content will be a function of time of air displacement, cake thickness, pressure drop across the cake, particle characteristics, liquor viscosity and other factors. The wash stage begins the instant wash fluid is applied, continues as the fluid passes through the cake, and ends with a period of liquor displacement by air. The end of a wash stage may precede cake discharge from the filter, or the application of a subsequent wash (in cases with multiple wash stages).

1316

A correlation of washing data should fall on a single curve and should illustrate the relationship between the volume of wash applied and the degree of solute removal without being affected by variations in the liquor concentration throughout the washing stage. It is possible to achieve this type of correlation using a parameter R, which is plotted as a function of a second parameter N, where N and R are defined as follows: N is defined as the volume of wash applied divided by the volume of liquor in the cake at the end of the wash stage, which includes some liquor displacement by air. If L is defined as volume of applied wash and L 2 as volume of liquor in the cake at the end of the wash stage, then N is defined: N = k . L2

In situations where the wash fluid applied contains no soluble value, R is defined as the concentration of solute in the cake liquor at the end of the wash stage divided by the concentration of solute in the cake liquor at the beginning of the wash stage. In cases where the wash fluid contains soluble value, this concentration must be accounted for. Theoretically, with infinite wash applied, the concentration of the solute in the cake liquor would equal the concentration of solute in the wash liquor. Therefore if Cz is defined as the concentration of solute in the cake liquor at the end of the wash stage, C1 is defined as the concentration of solute in the cake liquor at the beginning of the wash stage, and C, is defined as the concentration of solute in the applied wash fluid then R is defined:

R =

cw c1- c w

c2

-

The method of determining cake liquor concentration at the end of a washing stage has been found to be critical in terms of the quality of the correlation achieved. Unsatisfactory results are obtained if either the wash filtrates or dry (washed) cake residues are analyzed for residual soluble value content. The suggested approach consists of repulping each filter cake of a wash series in a known volume of pH adjusted de-ionized water. After repulping, the resultant slurry is filtered using a new filter mat and clean and dry buchner funnel and vacuum flask for each cake; Filtrate reporting to the vacuum flask is submitted for assay while filter cake solids are dried to obtain dry suspended solids. Complete accountability for all cake solids must be maintained throughout this repulp procedure. When the analytical results are obtained from the cake repulp liquor, the actual concentration of soluble value in the liquor associated with the cake upon discharge from the filter can be calculated by solving the following expression:

c 1=

P

+ Vr WI P

Crl

Where: C1 = Cake liquor concentration at discharge (g/ml). W1 = Weight of cake liquor (grams). p = Density of cake liquor (g/ml). V, = Volume of repulp fluid used (ml). C,, = Concentration of solute in repulp liquor (analytical result).

GENERAL FILTRATION TESTING DATA INTERPRETATION AND CORRELATION Data collected during testing can be correlated to show several general relationships, which can be subsequently used to predict filter performance for given conditions (Dahlstrom, 1957; Silverblatt, 1974). The key correlations include: Cake weight versus cake thickness. Cake formation rate verses cake weight (or thickness). 0 Cake moisture content verses relative dry cycle length. 0 Cake wash penetration verses cake thickness and applied wash volume. 0 Solute removal verses applied wash volume.

1317

These correlations can each be used to describe filter operation during each portion of a complete cycle, and when used together will describe the total filter performance and cycle time for the conditionstested. An example data set is used below to illustrate the correlationsdescribed above. An example of each correlation will be shown in Figures 1 through 5. During testing, several cakes of various thickness should be produced so that data for the fist two correlations can be collected. The first correlation shown in Figure 1 demonstrates the relationship between wet filter cake thickness (in millimeters) and dry filter cake weight W (with units of dry kg/m2). For the purpose of design, a possible cake thickness of 15 mm could be used for horizontal belt filters. According to Figure 1 the unit weight of a 15 mm cake would then be 2.9 dry kg/m2when flocculant is added, while a 15 mm cake produced without flocculant would weight 3.9 dry kg/m2 5

4.5 4 3.5

3 2.5 2

1.5 1

oWdh Added Flomlanl

0.5 0.1

0

0

5

10

Cake Thickness (mm)

15

7 0.1

20

Fig. 1 Cake weight vs. cake thickness

1

10 Form Xme (minutes)

100

Fig. 2 Cake weight vs. form time

Figure 2 displays the logarithmic relationship of dry cake weight, W, with units of dry kg/m2,as a function of cake formation time, in minutes. As predicted by theory, the slope of each curve is ?h (Silverblatt, 1974). The correlation shown in Figure 2 indicates that, for the case where flocculant is used, a 15 mm cake, which weighs 2.9 dry kg/m2will form in approximately 2.1 minutes, while when flocculant is not used, a 15 mm cake, which weighs 3.9 dry kg/m2 will form in approximately 7.2 minutes. The relationship between filter cake moisture at discharge and the dry time factor ( e p , with units of min*m2/kg)is shown in Figure 3. The dry time factor is the dry time ( e d , in minutes) divided by the dry cake weight per unit area (W, with units of dry kg/m2). The dry time factor permits a correlation between cake moisture and dry time for all variations in cake thickness,by normalizingthe dry time for cake weight, which depends on cake thickness. The correlation indicates that, when flocculant is employed, approximately a 2.9-minute dry time ( e d m =1.0 min.m2/kg and W = 2.90 dry kg/m2)following the cake wash will yield filter cake with 79% moisture. Similarly,when flocculant is not used, approximately a 4.9-minute dry time ( e d m =1.27 min.m2/kg and W = 3.9 dry kg/m2) following the cake wash will yield filter cake with 75% moisture. Figure 4 shows the correlation of the wash time (0,) and the wash time factor (WoVw),plotted as 8, in (minutes) vs. W-Vw in (kg*liter/m4).Similar to the dry time factor, the wash time factor permits a correlation between wash time and specific wash volume Vw (liter/m2)for all variations of cake thickness. From this plot it can be seen that for a WoVw value of 27.82 kg*liter/m4(corresponding to a wash ratio of N = 0.5) a wash time of 8.98 minutes is required when no flocculant is added, similarly,

1318

when flocculant is used, a W*Vw value of 19.89 kg*liter/m4(corresponding to a wash ratio o f N = 0.5) a wash time of 1.37 minutes is required.

’‘k?iL)

12

__

10

8-

6

I.

0,:

0

’

72.0

0

0

Figure 5 shows the relationship of fraction of solute remaining (R), determined from assay values on the repulped filter cakes, with respect to wash ratio (N). The wash ratio “N’ is defined as the number of cake liquor displacements with wash solution when the filter cake is at the discharge moisture content. Fraction of Solute Remaining, R 1

0.1

0.01 0 No Added Flocculanl

0.001 0

0.5

1

1.5 2 2.5 Wash Ralio, N

3

3.5

4

Fig. 5 Solute remaining vs. wash ratio

Fig. 6 Cake washing material balance

The recovery data correlation displayed in the in Figure 5 (often referred to as a “wash curve”) allows a quick estimate of where increases in wash volume produce little effect, so that the optimal level of wash can be selected. The predicted “R’ values also allow for construction of accurate material balances for a variety of washing filters. An example of a material balance for a three-stage

1319

counter-current washing belt filter is shown in Figure 6. While this is a fairly complicated filter arrangement, data fiom the wash curve allows recovery prediction.

VACUUM FILTER CYCLE TIMES AND PRODUCTION RATE CALCULATIONS Horizontal Belt Vacuum Filters. Cycle time for horizontal belt filters is the summation of the various filter function times required, that is, the sum of cake formation time, cake washing time, and cake dry time is the cycle time. Production rate (with units of weight per filter unit area per time) for a horizontal belt vacuum filter is calculated as follows: Production Rate =

cxw CycleTime

An empirical scale-up factor is included in the C term as well as the required constants for engineering unit conversions. Minimum cake thickness is considered during selection of cycle times used to calculate production rates. Minimum design cake thickness for horizontal belt vacuum filters range from 1/8" to 3/16" (3.2 - 4.8 mm) depending upon material characteristics.

Rotary Drum Vacuum Filter Configuration, Cycle Time, and Production Rate. Standard data, such as that discussed for Horizontal Belt Vacuum Filters above, are utilized in the same manner for Rotary Vacuum Drum and Disc filters with a few modifications that account for the configuration of a rotary vacuum filter. Form time, wash time (when applicable) and dry time and thus total cycle time and production rate are generally based on vacuum drum filters apportioned as shown in Table 1. Table 1 Vacuum drum filter area apportionment Drum Filter Zone Form Zone Wash Zone Dry Zone DischargeJ+e-wash/Dead Total:

I

Drum Filter w/o Wash % of cycle Arc Length 30% 108' NA NA 180" 50% 72'1 20%1 loo%l 360'1

Drum Filter with Wash Arc Length % of cycle 108' 30% 90" 25% 65' 18% 27% 360'1 100%

PI

These fractions of the total cycle time result from the geometry of the drum filter itself. Refer to Figure 7 as an example of how a vacuum drum filter might be proportioned.

DRVZ0NE=~plo180'

I \

SLURRY PAN

2 /

Fig. 7 Vacuum drum filter zone configuration without wash

1320

Hence, for the general case, cycle time is given by the following equation: Cycle Time =

FormTime Dry Time Wash Time Form Zone Fraction Dry Zone Fraction Wash Zone Fraction

For a specific example, the cycle time for a washing drum filter is given by: Cycle Time = FormTime - DVTime - Wash Time 0.30 0.18 0.25 Filter operation can be considered to be form time limited, wash time limited, or dry time limited based on which portion of the cycle is shown to be limiting during testing. The cycle time used is always the largest of the fractions shown above, since this slowest portion of the cycle limits the overall filter operation. A form, wash, or dry time limited operation is possible within limits imposed by the maximum and minimum rotational speed of the drum. By determining the times needed for cake formation, washing, and drying for a given cake thickness in consideration of washing and moisture targets, it is possible to calculate the total required cycle time. It is possible for the form cycle time to be limited by the minimum cake thickness required for discharging. The minimum design cake thickness for various types of vacuum drum filter discharge methods are shown in Table 2. Table 2 Vacuum drum filter minimum design cake thickness. Cake Thickness m Inches 118 - 3/16 3.2 - 4.8 1/32 0.8 114 6.4 1/8 - 3/16 3.2 - 4.8 6.4 114

I

Discharge Method Belt Roll Scraper Coil

String

The limiting zone of the drum governs the cycle time for a complete circular rotation. Nonlimiting zones may not require the full available zone in order to meet individually required targets. The form zone may be limited to the specific required form time by placing rotary valve bridge block settings within this zone. The production rate (with units of weight per filter unit area per time) for a rotary drum vacuum filter is calculated using the following equation where W is the dry cake weight per unit filter area and C is a constant including an empirical scale-up factor as well as required constants for engineering unit conversions: Production Rate =

cxw Cycle Time

Rotary Disc Vacuum Filter Configuration, Cycle Time, and Production Rate. Form time and dry time, and thus total cycle time and production rate, are generally based on vacuum disc filters apportioned as shown in Table 3. Table 3 Vacuum disk filter area apportionment Disc Rlter Zone Form Zone Dry Zone DischargePre-wasNDead Total:

I I

Disc Hlter Arc Length % of cycle 108' 30% 135' 37.5% 1170 32.5% 360' 100%

I

I

1321

I

Hence, for the general case, cycle time is given by the following equation: Cycle Time =

Dry Time FormTime Form Zone Fraction Dry Zone Fraction

Specifically, the cycle time for a rotary disc vacuum filter is given by: Cycle Time = Form Time - Dry Time 0.30 0.375 Filter operation can be considered to be form time limited or dry time limited based on which portion of the cycle is shown to be limiting during testing. The cycle time used is always the largest of the fractions shown above, since this slowest portion of the cycle limits the overall filter operation. A form or dry time limited operation is possible within limits imposed by the maximum and minimum rotational speed of the disc. By determining the times needed for cake formation and drying for a given cake thickness and moisture target, it is possible to determine the total required cycle time. It is possible for the form cycle time to be limited by the minimum cake thickness required for discharging. The minimum design cake thickness for a rotary disc vacuum filter ranges from 3/8” to 1/2” (9.5 - 12.7 mm) depending upon material characteristics. The limiting zone of the disc governs the cycle time for a complete circular rotation. The nonlimiting zone may not require the full available zone in order to meet the design target. The form zone may be limited to the specific required form time by placing a rotary valve bridge block setting within this zone. The production rate (with units of weight per filter unit area per time) for a rotary disc vacuum filter is calculated using the following equation; where W is the dry cake weight per unit filter area and C is a constant including an empirical scale-up factor as well as required constants for engineering unit conversions: Production Rate =

cxw Cycle Time

Pressure Filter Cycle Time and Filter Press Sizing Pressure filter cycle time will simply be the sum of the individual portions of the cycle time for each operation, that is: cake formation time, cake washing time, dry time (including diaphragm squeeze time, if applicable), and miscellaneous time required to open and close the press, discharge the cake, and clean the cloth, etc. Therefore, the total cycle time is given by: CycleTime = FormTime + DryTime+ WashTime + Misc.Time Once a cake thickness is selected (based on filter press chamber thickness), the components of the cycle time can each be predicted by the correlations shown in Figures 1 through 4.Note that within a vertical recessed plate or plate and frame pressure filter chamber, two (2) filtration surfaces exist; hence, a half-cake forms from each filtration surface simultaneously. This half-cake thickness must be used for form time prediction from the correlation in Figure 2. In pressure filters with horizontal plates filtration is single-sided, so full cake thickness is used for form time calculation. Total cake thickness (chamber thickness) should be used for all other calculations. The correlation in Figure 1 relates cake thickness to cake weight per unit area (W), and the correlation in Figure 2 relates form time to cake weight W. The correlations in Figures 3 and 4 similarly predict washing and dry times based on selected wash volume and cake moisture. The miscellaneous time will be equipment specific, and should be supplied by the vendor of the selected equipment Once the pressure filter cycle time is known, the filter sizing can be undertaken. This is not as simple as determining the production rate based on test data. For example, if filtration rate alone determined pressure filter sizing, the production rate (with units of weight per filter unit area per time) would be given by:

1322

W CycleTime This equation is generally only valid for hydraulically limited filters, for which the production rate is relatively small. Most pressure filters are actually limited by the volume capacity of the filter, which limits the amount of cake that can be produced in a single cycle. Since pressure filters are closed chambered devices, the fixed volume of the chambers limits the volume of cake that can be produced in a single cycle. Hence the dry bulk cake density (dry kg/m3) is an extremely important experimental value. This value can be inferred from the correlation in Figure 1, which describes the dry cake weight per unit area per unit of cake thickness. With the dry bulk density known, the desired tonnage will set the total volume of cake to be produced in a given day. The cycle time can now be used to set filter sizing, since the total hours of filter operation per day will limit the number of cycles that can be run by the filter. Theoretical ProductionRate =

Hours Per Day Worked Cycle Time (hours) Therefore the filter press required cake volume (in cubic meters) must be: Cycles Per Day =

Filter Press Volume =

C x 1OOOx Daily Tonnage Cake Dry Bulk Density x Cycles Per Day

Where C is an empirical scale-up factor, tonnage is in metric tons, and cake dry bulk density has units of kg/m3. This volume specification can be considered to be the required filter sizing The effective production rate for a filter sized previously described would then be: Effective Production Rate =

Cake Weight Per Cycle Filter Press Areax Cycle Time

If this effective production rate (for volume limited filter presses) is less than the theoretical production rate previously described, then the filter design can be considered to be volume limited. Since this happens to be the case in many instances, this shows the danger inherent in sizing a pressure filter based on production rate alone.

PILOT SCALE TESTING Laboratory testing described in the previous sections will generate filter sizing data using standard, bench scale equipment. It is particularly useful when only small quantities of representative sample are available. The test data can be used to evaluate many vacuum and pressure filtration options as part of the flowsheet and equipment selection process. A short-list of preferred filter types can then be produced. While laboratory scale testing focuses on the filtration process, the scope of pilot scale testing is larger, and includes data collection for full-scale plant design including ancillaries. If sufficient representative sample is available, for example at a production plant considering an upgrade, testing can proceed to pilot scale. Pilot scale testing may not always be necessary, but is conducted when factors including the following are important: To evaluate a specific type of filter, and to demonstrate that it will work as predicted at laboratory scale. 0 To check filter operating characteristics that cannot be demonstrated at laboratory scale, such as cake discharge and cloth blinding. To produce large samples of filter cake or filtrate to develop downstream unit operations, or for marketing purposes. To collect engineering data for filter plant design and ancillary equipment selection. To compare existing and proposed filters in parallel over time, and over a wide range of operating conditions.

1323

To gain operating and maintenance experience on a filter as part of the selection process, or to train staff before the production unit is installed. Whereas standard laboratory equipment can be used to evaluate almost all types of vacuum and pressure filters, pilot scale units tend to be more specific. Pilot scale filters are often supplied by manufacturers, and are designed to replicate the operation of that company’s equipment. This should be clearly understood when analyzing test data and trying to apply it generally. “Pilot Scale” operation is not rigidly defined, and test units can have filtration areas rangingfrom 0.1 to 5m2. Pilot plants may be located permanently, for example at in a fixed location such as a research center. However, they are commonly mobile units that can be transported and operated wherever needed. Testing on-site often ensures that samples are both representative and fresh. Smaller pilot scale units are operated “off line” in batch tests, whereas larger units can be integrated into an existing process for continuous operation. In both cases, preparations for pilot scale testing must be made well in advance. Although some transportable pilot filters may be delivered with their own ancillaries such as feed pumps and compressed air supply, others may need these to be provided at the test site. In all cases, plans must be made to provide electricity and water, analytical services, and assistance with installation and maintenance. A single technician can conduct laboratory scale testing, but pilot scale operation is significantly more demanding on resources. Figure 8 shows a test pressure filter with a filtration area of 0.1m2. This unit is manually operated and simulates a single cycle of a membrane pressure filter including cake formation, cake compression, cake washing, second cake compression, and air-drying. This type of test filter can be pneumatically powered, although electricity may be required for slurry heating and agitation. The test unit includes all necessary ancillary equipment such as feed tank, feed pump, pressing pump and cake wash liquid tank. The feed tank holds 100 liters slurry, and filter cakes of several kilograms can be produced. Different chamber depths can be used, and flow rates and pressures can be adjusted. The test unit illustrated is 2 meters long, 1.5 meters high and 0.6 meters wide and weighs 400 kilograms. Such units can be easily moved through the plant to the test location on wheels, with a forklift truck, or by crane.

Fig. 8 Pressure filter (0.1m2)(courtesy of Larox) Similar test filters can be fully automated, with continuous recording of pressures, flowrates and filtrate volume direct to computer. Figure 9 shows a 1.6m2fully automatic pilot scale pressure filter built into a standard shipping container. This pilot unit includes all necessary ancillaries, and can be integrated into a continuous process.

1324

Fig. 9 Automatic pilot scale pressure filter (l.6m2)(courtesy of Larox) Data collection and analysis methods are similar to those used in bench-scale testing. Of course larger sample volumes are involved, and sample handling logistics more demanding. Power, air and water consumption are monitored during pilot scale testing. The aim is to optimize the filtration cycle for energy consumption, as well as process results, and also to avoid incorrect sizing of ancillaries. Pilot scale testing will generate large samples of filtrate and filter cake. Valuable information on materials handling can be obtained by observing ease of cake discharge, and transport characteristics. For example, filter cake samples can be used for angle of repose tests for downstream surge bin design. Extended pilot plant trials give an opportunity to test different materials of construction, especially in aggressive, high temperature and corrosive conditions. Above all, pilot-scale testing provides tangible evidence to decision makers that the filter provisionally selected during laboratory scale testing will, in fact, perform as expected at production scale.

1325

APPENDIX DERIVATION OF CAKE FORMATION RATE EQUATIONS Classical treatment of filtration theory nearly always begins with Equation (1) Poiseuille's equation (Carman, 1938; McCabe and Smith, 1976). This form of the equation contains a couple of cumbersome terms, V for the volume of filtrate, and c for solids concentration expressed as the weight of solids per unit volume of filtrate. More succinctly, most applications we encounter in either vacuum or pressure filtration deal with the solids filtration rate rather than the liquid filtrationrate. It follows that an expressionfor the rate of cake formation that avoids the V and c terms will be more useful in practical applications. The choice of an equation from which to start a revised derivation for the purpose stated above depends on the reference source selected. The same is generally true for the notation that is used. The starting equation is usually derived from Poiseuille's equation and is typically in the following form which shows the instantaneous rate of filtrate flow as the ratio of the driving force to the product of the viscosity and the sum of the filter cake and filter media resistances:

This can be converted to:

The notation used above in equations (1) and (2), and hereinafter, is as follows:

Of =

v = P

=

A Ap a c R,

= = = = =

P

=

w =

s = s, =

Cake formation time, Volume of filtrate. Liquid viscosity. Filtration area. Pressure drop across the filter cake. Average specific (filtration) resistance (units of length/mass). feed slurry solids concentration in terms of mass of solids per unit volume of filtrate. Resistance of the filter medium (units of length-'). Mass of dry solids in the filter cake per unit area. Liquid density. Weight fraction of solids in the filter feed slurry. Weight fraction of solids in a formed, but undewatered filter cake.

Integrating equation (2):

Eqn (3):

Eqn (4):

$f

ef

def =

-&[ Ji

VdV

P

= -

AP

1326

Rm j i dV

]

However, where: Eqn (5):

w="v A

and, it follows:

It should be clear that the rate of cake formation, as we usually express it, (in terms of weight of dry cake solids per unit area and per unit time), is equal to W&. Rearranging equation (7) yields:

Eqn (8):

-W-

of

-+

Rm

This is a rather simple equation, and it avoids the use of the V term. However, the c term is one that has an unusual definition, and this definition is probably responsible for more misunderstanding regarding filtration theory than any other factor. Hence, the term is ripe for change, and must be replaced with another term for feed slurry concentration that can readily be understood, i.e., weight % solids. Actually, the term that will be used is S, the weight fraction of solids in the feed slurry, which is almost the same thing and easier to manipulate. The definition of c is the weight of dry solids in the feed slurry per unit volume of filtrate. The liquid remaining in the formed (but yet undewatered) cake is, of course, not part of the filtrate, but it certainly was part of the feed slurry and it must be taken into account. Note the reference to the formed but undewatered filter cake. The mechanism of dewatering or draining of a filter cake by displacing the entrained liquid with air (as occurs in the cake drying portion of a pressure or vacuum filter) has nothing to do with the mechanism of cake formation, which is described by equation (8). Therefore, because there are two mechanisms involved, another term, Sc, denoting the weight fraction of solids in a formed but undewatered cake, must also be used. The definition of c could be expressed as:

Eqn (9):

c =

Weight of Solids Volume of Filtrate

By dividing c by p, the density of the liquid, it is obvious that:

Eqn (10):

-c _p

Weight of Solids Weight of Filtrate

1327

Also, S, can be defined as:

Eqn (11):

s,

=

Weight of Solids Weight of Formed (but undewatere d) Cake

Or, also as:

Eqn (12):

sc =

Weight of Solids Weight of Slurry - Weight of Filtrate

Therefore,

Eqn (13):

Weight of Filtrate = Weight of Slurry -

Weight of Solids

sc

and, Eqn (14):

-C _-

Weight of Solids Weight of Solids Weight of Slurry SC

By multiplying the term containing Sc by (weight of slurry/weight of slurry),

Eqn (15):

c -- S

P

I - -S

and,

Eqn (16):

c = - PS

S l - -

Substituting equation (16) into equation (8),

Note that the steps between equation (1) and equation (17) were made using only mathematics, with no assumptions at all, so if the validity of equation (1) is accepted, then equation (17) is rigorous. Equation (17) can be particularly useful in ascertaining what the effect of changing a variable on the right side of the equation has on the left side (the "form filtration rate"). For instance: If Ap is doubled, the form rate will double. Note that doubling the form rate can only be due to Of being halved as W was not changed on the on the right side, W cannot be changed on the left side (since they are the same). This illustration is filtration at constant cake thickness. On an actual vacuum filter, if the operator doubled Ap, he would have to increase the filter speed by a factor of 2, to get the same cake thickness (he had before Ap was doubled). On a pressure filter, if slurry feed or fill

1328

time to the chamber was held constant, but air-fill and diaphragm pressures were doubled, the form rate would be halved. If

S (the newly defined feed slurry concentration) were doubled, the form rate would I - -" 3 C

double. Once again, this is true only if the cake thickness were held constant by changing the filter cycle time (vacuum filter), or the fill time (feed time) were halved (pressure filter). A seemingly logical question follows: What is the effect of changing a variable if the appropriate filter function, such as cycle time or fill time, were not changed? In other words, consider the effect on form filtration rate, now at constant cycle or function time. To evaluate this possibility and at the same time avoid the use of a complicated equation, a simplifying assumption must be made, that is, the resistance due to the filter medium, R,, is small compared to the resistance due to the filter cake, Wd2, which is usually the case in most vacuum and pressure filter applications. After setting Rm equal to zero, we obtain:

r

1

Rearranging:

Eqn(19):

(W>z = O f

2APP -

S

-

and further,

I S

Eqn (20):

I--

S

sc

By solving the expression, and then plotting W as a h c t i o n of &on log-log paper, a straight line of 0.5 slope should result (and it generally does). By dividing W by Bf to calculate the form filtration rate, we obtain: I

r

1

By solving the expression, and then plotting the form filtration rate, W/ef, as a function of form time, 81, on log-log paper, a straight line of minus 0.5 slope should result (and it generally does).

1329

REFERENCES Carman, 1938. Trans. of the Inst. Chem. Eng. (London) 16: 174. McCabe and Smith, 1976. Unit Operations of Chemical Engineering, 3rdEdition. New York: McGraw Hill. p937. Dahlstrom D.A., and Nelson, P.A., 1957. Moisture-Content Correlation of Rotary Vacuum Filter Cakes. Chemical Engineering Progress. 7 : 320-327. Silverblatt, C.E., Hemant Risbud, and Tiller, F. M. 1974. Batch, Continuous Processes for Cake Filtration. Chemical Engineering. 4: 127-136. Perry, R.N., and Green, D.W., eds.1991. Chemical Engineering Handbook. 6” Edition. New York McGraw Hill.

1330

Design Features and Types of Sedimentation Equipment Fred Schoenbrunn’ and Tim Laros’

ABSTRACT Sedimentation equipment can be specialized to cover a wide range of process goals. Various equipment designs are available that have been optimized for different processes and process objectives. Some of the common designs used in the minerals industry include clarifiers, solids contact clarifiers, inclined plate clarifiers, conventional thickeners, high rate thickeners, high rate rakeless thickeners, high density thickeners, and Alcan deep thickeners. Within these designs, options include bridge and center column mounting, as well as a variety of tank construction options. Selection of the optimal design for a project depends on the process objectives and can be a function of capital versus operating costs, with project life, maintenance, materials selection, and site layout considerations also being factors.

INTRODUCTION There are a variety of sedimentation equipment designs available to the minerals industry. Selection and sizing of the proper piece of equipment depends on the process and objectives. Details of the design of the equipment can vary not only with process objectives, but also with site topography, material specific gravity and particle size, and availability of local building materials, as well as plant operating philosophy. The development of modern flocculants has lead to high rate and high capacity designs that are optimized for their use. Modem flocculants can decrease the required sizing for sedimentation equipment by an order of magnitude or more. The cost of flocculant is a significant operating cost and should be considered in an economic evaluation. There is usually a trade off between equipment size and required dosage, with smaller units requiring a higher flocculant dosage up to a point. It used to be that “conventional” meant that flocculant was not being used. Now it implies a mechanism that is not specifically designed for the optimal use of flocculant, and is usually used as a reference point regarding the amount of risk in a design. However, it should be noted that a well-designed high capacity thickener could operate more consistently and with lower risk than a larger well-designed conventional thickener using flocculant. In the minerals industry, sedimentation equipment is ubiquitous. The most common applications in base metals involve thickening of tailings and concentrates. In alumina, nickel laterite, uranium, gold, and some of the other leach circuits, counter current decantation (CCD) is a fundamental part of the process, using as many as 8 thickeners in series. Other applications include clarification of plant discharge water, process water treatment, leach liquor clarification, removal of precipitates, pre-leach, and grind thickening. There are three general types of thickener rake drive mechanisms; bridge mounted, center column mounted, and traction drives. Bridge mounted drives are centered on a bridge that spans the thickener, with a shaft attached to the rakes. Because of the bridge, there is an upper limit on the size of machine this design can be economically applied to, generally around 40 - 50 m diameter. Center column drives are used on larger thickeners (or small ones without lifts) and are mounted on a center column that typically also supports an access bridge that spans one half of the tank. The shorter bridge compensates for the addition ofthe column and the use of a cage around 1 EIMCO Process Equipment Company, Salt Lake City, Utah.

1331

the column to drive the rakes. These are economical starting at about 30 m diameter and are currently available up to a diameter of 130 m. Traction thickeners use a peripheral drive mounted on one (or two) of the rake arms. These units can develop very high torques, but do not have lift capabilities and are sensitive to environmental conditions regarding contact between the drive wheel and the rail or traction surface. They are usually only considered for large diameter applications. Conventional thickener design uses center underflow outlets with the floor sloped at 1:12 2:12 towards the center. Lighter duty applications such as clarifiers usually use a 1:12 slope. Large diameter machines often use a dual slope design, with an inner slope of 2:12 and an outer slope of %:12 or 1:12, to avoid making the machine excessively deep. Some applications such as uranium yellow cake and magnetite use steeper slopes such as 3:12. The tank sidewall depth is usually about 3 m and is determined by both process and mechanical considerations. A freeboard of 150 - 300 mm is usually used on both the tank and feedwell. The center outlet is typically a 45" cone on bridge type thickeners and a trench on center column thickeners. Either requires a scraper assembly attached to the rake structure to keep the material moving. A generalized thickener is shown in Figure 1. Figures 2 and 3 are photos of typical center column and bridge mounted thickeners respectively. Rake Drive

Feedwell

-Overflow Launder

-

\Drop Box and Outlet

Center Cone

Figure 1 Thickener schematic DESCRIPTION OF THICKENER COMPONENTS Feedwell Thickeners and clarifiers are typically fed in the top center and use a feedwell to still the feed stream. The feedwell is also the most common point for flocculant addition. Properly designing the feedwell and flocculant addition location to maximize the effectiveness of the flocculation can have great impact on the operation. The feed slurry concentration for optimum flocculation and most economic thickener size is not necessarily the slurry concentration reporting to the thickener from the upstream process. Quite often, the feed slurry requires dilution prior to addition of flocculant to achieve best flocculation and thickening performance. This effect is presented in Figure 4. This figure shows a maximum in the settling flux at a relatively low solids concentration. The settling flux is the amount of solids settling through a given area, which is the product of the solids settling rate times the concentration. The optimal concentration depends on the characteristics of the feed solids. Very fine solids may need to be diluted to 5 wt% where the maximum flux may occur at 15 wt% for a coarse grind tails. For clarification, the feed slurry may be too dilute and may benefit from an increase in concentration by recycling underflow slurry back to the feed slurry. This will be discussed under Solids Contact Clarifiers.

1332

Figure 2 Center column mounted thickener. Note the concrete on-ground tank, truss type rake arms, and feed launder mounted under the bridge. Photo courtesy of EIMCO Process Equipment Company

Figure 3 Bridge mounted thickener. The unit uses an elevated steel tank. Photo courtesy of EIMCO Process Equipment Company Many methods of feed dilution have been tried since the advent of synthetic flocculants. Pumping of overflow liquor into the feed slurry from the overflow launder or directly from the top of the thickener with submersible pumps is quite common. This method requires a pump, which can be quite large and expensive in cases where large quantities of dilution are required. Early dilution methods without external pumping used slots or holes cut into the feedwell to draw dilution liquor into the feed slurry due to the density differential between the feed stream and the clarified liquor. These types of feedwells are often referred to as Cross type feedwells, as Harry

1333

Cross developed one of the first units. The Outokumpo Floc-Miser is an example of this type of feedwell.

Settling Flux,

Flocculant Dose, glt

tlhlrn2

Solids Concentration, wt%

Figure 4 Solids concentration versus settling flux and flocculant dosage An alternative method of dilution is to place an eductor or jet pump in the feed line to dilute, flocculate, and mix the slurry prior to entering the feedwell. The degree of dilution can be designed into the eductor through the geometry of the eductor and the driving head of the feed slurry through the eductor. This type of feedwell is distinguished by the EIMCO E-DucB SelfDiluting Feedwell. An example of this is shown in Figure 5.

Polymer Undiluted Feed

Figure 5 E-DucB Self-Diluting Feedwell Rakes Most thickeners use a set of arms that move through the pulp to help thicken the pulp and move the thickened material to the underflow outlet. These typically have blades set at 30-45 degrees to the tangent of motion to push the solids towards the outlet. The rake arms must be strong enough to transmit the torque needed to push the solids towards the thickener discharge. Rakes are usually designed for specific process applications. Processes that produce a heavy scale build up on the rakes such as alumina refining require a rake containing a minimum of steel surface area. Rakes for magnetite thickening usually have spikes attached to the blades such that heavily thickened magnetite can be resuspended. Some rakes are streamlined to help reduce the torque on the rake structure. Rakes may have pickets or Thixoposts (posts to distance the blades

1334

from the rake arms) attached to them for processes in sticky, viscous materials. In general, rake design is process dependant. Rake Drives The rake drive provides driving force (torque) to move the rake arms and blades against the resistance of the thickened solids. The drive also provides the support for the rotating elements. Bridge type thickeners can use spur gear, worm gear, or planetary or other commercial reducer drives. Column mounted thickeners almost always use spur gear drives. Since the rake steel is usually designed based on the 100% torque rating of the drive, two or more levels of overload protection are often designed into the drive to prevent torque overloads and associated rake damage. Reliability is a critical issue with the drive, since failure frequently means digging out the thickener. Key drive elements are hardened steel gears, large precision bearings, oil bath lubrication, accurate torque measurement, and strong housings. There are a variety of torque descriptions such as; peak torque, design torque, normal operating torque, duty rated torque, AGMA 20 yr torque, cutout torque, etc. This is partly due to the relationship of torque to gear life, where gears last a very long time at low torque, but exponentially shorter as the torque increases. There are two methods of motive power for drives, either electric motors or hydraulic power packs. Using hydraulics offers features such as soft starts, variable speed, torque indication by hydraulic pressure, low speed hydraulic motors, excellent torque sharing on multiple pinion drives, and pressure relief as an overload protection. The downsides are low efficiency, higher cost, maintenance, and the added complexity of another system. Similar features are now available for electric drives using electric VFDs. Electric drive motors are relatively simple and can use mechanical, load cell type, or electronic load sensing torque measurement and protection. Rake drive sizing is dependant on the application with variables such as particle size distribution, flocculant use, solids loading, rake design, and design underflow concentration affecting the selection. Since the torque is related to the diameter by a square power fimction, a “K” factor is usually used to refer to a drive size independent of diameter, where Torque = K x Diamete?. A table of typical values for standard duties is shown below, covering clarifiers, conventional, and high rate thickeners. However, the selection of an appropriate K factor should also consider the variables listed above, and may need to be significantly different fiom those listed. For example, very high underflow densities may dictate a K factor an order of magnitude higher. Duty Light

I Standard Heavy

Extra Heavy

I

K Factor (N/m) Examples River, or lake water 15-60 clarification, Metal hydroxides, Brine ciarification Magnesium oxide, Lime I 70-130 sofGning, Brine softening Copper Tails, Iron Tails, 150-290 Coal refuse tank, Coal, Zinc or lead concentrates, Clay, Titanium oxide, and phosphate tails Uranium Counter Current 290+ Decantation (CCD), Iron Ore concentrate, Iron Pellet feed, Titanium Ilmenite

1335

K Factor (Ibs/ft) 1-4

I

5-9 20-40

40+

Lifts Rake lifts are used to protect the drive from high torque. These can be used with either bridge or center column designs. Traction type drives are generally not compatible with rake lifts, although some have been supplied using the cable-type design described below. Lifts are typically powered by a separate motor from the drive and must be designed for precise control. Lifts can prevent shutdown of the machine in plant-upset conditions when large amounts of coarse material are encountered. Their use in minerals applications is fairly ubiquitous, although there are many applications that do not need a lift, or where extra torque would make a lift unnecessary. Lifts have been used to aid in storing material, especially mineral concentrates, in a thickener. The rake is lifted allowing thickened concentrate to accumulate below the rake. The rake is then slowly driven into the dense concentrate to evacuate the thickener. This technique is not commonly used. However, it illustrates the need for the rakes to be designed to accommodate a downward force from the lift. Cable-type designs use an arm hinged at the center connection and are supported and towed by cables, so that the arms can pivot upwards when an obstruction or high torque is encountered. The advantages of this design include a streamlined arm design and low cost. The main disadvantages are the lack of positive control of the arm position and the inability of the rakes to lift significantly in the center, where the heaviest accumulations are usually found. Effluent Launders Most thickeners use a peripheral effluent launder to collect the clarified overflow and bring it to a single or double discharge point. The effluent should flow into the launders uniformly around the periphery of the tank, and should not be back-flooded into the thickener. V-notch weirs are often provided to assist in distributing the effluent around the periphery of the tank. The weirs can be built into the tank launder or can be a separate, adjustable element. Froth baffles can be located at the liquid level just inboard of the launder and are used to prevent floating material from getting to the launder. Flotation concentrate thickeners are almost always provided with froth baffles and often with some method of froth management such as water sprays. Other methods of handling effluent include radial launders, single point discharge, and bustle pipes. Radial launders are often used on solids contact clarifiers to help distribute the effluent removal evenly over the surface of the clarifier. Conversely, some applications can use a simple single point discharge nozzle and still have acceptable overflow clarity. Bustle pipes are often used where liquor storage is desired at the top of the tank or the tank liquor level is variable, using submerged pipes with orifices to distribute the effluent discharge. The size, slope, and number of discharge points of a launder are determined using hydraulic flow equations developed by the thickener manufacturers. Tank Design There are a number of possible tank styles that can be used, with attendant tradeoffs. Most mineral applications require good access to the underflow piping, and it is usually preferable to locate underflow pumps in close proximity to the thickener underflow outlet. This leads to the need for either elevated tanks or underflow access tunnels. Due to the complexity, cost and safety issues with underflow tunnels, elevated tanks are generally preferred for small to medium size thickeners, up to about 40 m diameter. Other benefits of elevated tanks are storage space underneath, unhindered pump access, and ease of leak detection and repair. On-ground tanks do not require the structural steel needed for elevated tanks, and so they are generally preferred for applications where the underflow pipe access is not critical as well as for large diameter thickeners and clarifiers, generally 50 m diameter and beyond. If the process allows the underflow pipe to be buried to the edge of the thickener, an on-ground tank is significantly less expensive. As a result, on-ground tanks are fairly common in clarification and water treatment applications. Within the realm of on-ground tank design, there are number of construction methods available using steel or concrete, Anchor channel construction refers to a steel channel embedded in a concrete footing at the wall, with the steel shell welded to the anchor channel. With this construction, the floor can be concrete, membrane or fill. The other method of construction for a steel wall tank is to use a steel floor on a compacted foundation. For concrete wall tanks, the floor

1336

can be concrete, membrane or fill. If fill material is used for a floor, it should be engineered, properly placed, and compacted as part of the overall tank construction. Various linings, covers and insulation can be applied to suit the process. SEDIMENTATION EQUIPMENT DESIGNS Clarifiers Standard clarifiers are often used for water and waste water applications. These units usually have relatively light mechanisms since the amount of raking, particle size and solids density are usually on the light end of the spectrum, and rake lifts are frequently not needed. In general, clarifiers are very similar in appearance to thickeners. The feedwell is generally larger in diameter and the tank depth higher. This is done to provide more time for feed slurry flocculation in the feedwell, slower velocities exiting the feedwell, and a longer liquor detention time in the clarifier. These features are necessary to achieve optimum overflow liquor clarity Classic clarifier hydraulic loadings start at about 1 m3/h/mz (0.4 gpm/ft2) and goes up as high as 6 m3/h/m2(2.5 gpm/ft2) for some solids contact clarifiers with optimal solids. Nominal sizing at 2.4 m3/h/in2 (1 gpm/ft2) is a good starting point for many applications. These numbers consider using coagulants or flocculants, which greatly aid particle settling rate. Chemical addition is almost always used in clarification applications due to its effectiveness at increasing the effluent clarity . Solids Contact Clarifiers This equipment is designed to internally or externally recirculate solids and to promote particle contact and flocculation in the reaction well or to enhance solids precipitation and hardness reduction. This can be done externally by pumping a portion of the underflow back to the feedwell. The same thing can be accomplished internally using a draft tube and a turbine. The design philosophy for solids contact clarifiers is to maximize the size and concentration of particles in the reaction zone by having the pumping capability to suspend a high concentration of particles. By recirculating centrally and directly above the tank floor, the heavier particles necessary for improved settling velocities are mixed with the incoming feed to allow particle contact and growth: Flocculation and recirculation are accomplished symmetrically within the reaction well for the most efficient use of reactor volume and turbine energy. A large feedwell provides the required detention time, and allows precipitation to take place prior to the slurry entering the clarification zone. Chemical addition is introduced into the recirculation drum in the presence of previously formed precipitated solids prior to passing through the turbine for optimum mixing and flocculation. Typical hydraulic loadings for solids contact clarifiers are 2 . 4 4 . 8 m3/h/m2 (1-2 gpm/ft2), upwards to 12-24 m3/h/m2(5-10 gpm/ftz) in some steel mill wastewater clarification applications. Inclined Plate Clarifiers Inclined plate clarifiers use plates to increase the effective settling area in a small tank. The plates are set at a 45'-60 O angle to allow settled solids to slough off. The plates are typically stacked with 50 mm spacing, although this can be varied for the application. The effective area is the sum of the horizontal projections of the plate areas. They can be supplied either with rakes or with steep cone bottoms, and are often available as packaged units. These units are very effective putting a lot of area in a small tank, but are susceptible to solids buildup and plugging. Conventional Thickeners There are many applications for which polymer is either not needed or may adversely affect downstream processing, and hence is not used. For these situations, a conventional thickener is all that is required. However, there are many conventional thickeners that use flocculant, to improve overflow clarity, handle a higher tonnage, or aid in achieving the desired underflow density. These units are often fairly simple, although relatively sophisticated mechanisms are used for particular applications. Due to the relatively large size, they are somewhat forgiving in operation

1337

and can have the storage capacity to absorb some plant upsets without affecting downstream operations. Typical features include a drive and rakes, a relatively shallow feedwell, and a bridge to support the feed pipe or launder and allow center access. The drive size is dependant on the application. Conventional thickeners can be either bridge, center column or traction design. Early thickeners mostly fall into this category. A table of conventional thickener sizing in included in the Appendix. High Rate Thickeners With the advent of synthetic flocculant, the terms High-Rate and High-Capacity emerged as a type of thickener, as the throughput rates for the now flocculated feed slurries were considerably higher than for unflocculated slurries. These terms now generally refer to designs optimized for use with flocculants, although further improvements on earlier optimized designs are frequently possible, as the technology has continued to evolve and improve. Thickener size or throughput is directly dependant on flocculant dose and feed slurry concentration, again as presented in Figure 4. Because of this, most high rate thickeners use feed dilution systems. The optimum size of these thickeners is governed by capital and the primary operating cost of flocculant. A smaller thickener may be less expensive to install, but more costly in the long run due to a higher overall flocculant cost, and vice versa. High Rate thickeners are generally small to medium sized bridge type thickeners, although large center column thickeners processing very high tonnage can also fall into this category. Flocculation is required and feed slurry dilution systems are often needed for optimal performance. These are the most common type of thickener in the minerals industry today. Grind, tails, leach, preleach, CCD, and concentrate thickeners often fall into this category. Common features include a deep self-diluting feedwell, heavy duty drive, streamlined rake arms, and large effluent launders and underflow outlets. High Rate Rakeless Thickeners This new class of sedimentation equipment utilizes a deep tank and steep bottom cone to maximize the underflow density while eliminating the rake and rake drive. The design is based on maximizing throughput rate in a small diameter while achieving good underflow density and overflow clarity. Flocculant is always used and feed dilution is usually built into the design. Since these units have a very low residence time, startup and shutdown are quick, typically requiring only about 30 minutes to reach steady state operation. They are operated as continuous process equipment, and cannot be used for storage. The lack of a rake mechanism makes these units very simple to operate. Current examples are the EIMCO E-CATTMClarifierIThickener and the Bateman Ultrasep Ultra High Rate Thickener. A diagram of the internal flow pattern of an ECATTMClarifiedThickener is shown in Figure 6 . High Density Thickeners This technology is an extension of high rate thickening, utilizing a deeper mud bed to augment the thickening capacity. Also known as high compression thickeners, these machines usually add depth to a high rate design to aid in increasing the underflow density. Deeper mud beds increase the mud compressive force, reducing the time required for thickening and increasing the underflow density. These machines may require significantly more torque (2-5x) than high rate machines due to the increased mud viscosity.

1338

Figure 6 E-CatTMClarifierlThickener showing the internal flow patterns. Alcan Deep Thickeners Alcan Deep Thickeners originated in work done by the British Coal Board in the 1960’s, utilizing steep bottom tanks without rakes to produce underflows with very high solids contents. The goal of the work was to produce a material that could be put on a conveyor, and while this was not achieved consistently, very thick material could be produced. More recent development in the alumina industry eventually led to commercialization of this technology outside alumina with applications ranging from high efficiency CCD washers to underground paste disposal and wet stacking of surface tailings. The EIMCO Deep ConeTMPaste Thickener and EIMCO Hi-Tonnage Paste Thickener are examples of this design. It is possible to produce material at the limits of pumpabilty with these units. In some applications, underflow with the consistency of paste can be achieved by high rate, high rate rakeless, or high density machines. However, deep thickeners are currently the best technology for achieving maximum underflow densities utilizing sedimentation equipment alone. These units typically utilize very deep mud beds in order to take maximum advantage of inud compressive forces for dewatering and provide sufficient time for the inud to dewater to a paste consistency. The tank height to diameter ratio is frequently 1:1 or higher. Due to the high underflow viscosities, mechanism torques can be 5-10 times higher than high rate machines on similar materials. Applications include surface tailings disposal by wet stacking, underground paste backfill, and countercurrent decantation.

1339

Major Factors Influencing Thickener Design a a

a a

m

a

a a

a

a

a

0

Process requirements for the overflow liquor quality and underflow slurry density. These determine the mechanism design. The quality of solids to be handled. Usually expressed as area per unit weight of dry solids per day. High rates usually require a combination of a stronger thickener mechanism and a lifting device. The amount of material larger than 250 micron (+60 mesh) in the feed. This affects tank bottom slope, drive and strength of mechanism. It may also require a rake lifting device. Specific gravity of the solids. The greater the specific gravity the more likely a stronger drive and mechanism will be required. Feed, overflow, and underflow systems capable of handling additional material when other thickeners are out of service. Feed and underflow material settling characteristics that may require special rake construction such as blades located a distance below the rake arms on posts or spikes on the blades to cut into packed solids. Scale build up tendency of feed slurry may require special arms and drive. An operating requirement to accumulate solids for defined periods of time will require a special mechanism design, as it is not a normal operating procedure. Froth control or removal may require sprays, froth baffles or skimmers. Slurry temperature, vapors, gases, etc. may require covered and/or insulated tanks with attendant seals. Soil conditions and ground water elevation affect foundation design and may determine tank and mechanism type. Climatic conditions may require special considerations, such as enclosures around the drive and instrumentation.

CONCLUSIONS Different sedimentation equipment designs are available for various applications and process objectives. Utilizing the proper design for an application can make the difference between a smooth operating process step and continual problems that prevent a plant from realizing its potential. Taking the time in the project planning stage to make sure the correct design is being used can make for a successful project. REFERENCES H. Cross, “A New Approach To The Design And Operation Of Thickeners”, Journal Of The South African Institute Of Mining And Metallurw, (February 1963) F. M. Tiller, D. Tarng, “Try Deep Thickeners and Clarifiers”, Chemical Engineering Progress, pp. 75-80, (March 1995). R. Klepper, T. Laros and F. Schoenbrunn, “Deep Paste Thickening Systems”, Proceedings of Minefill ‘98, Brisbane, Australia, (1998). R.C. Emmett, T.J. Laros, K.A. Paulson, ”Recent Developments in SolidLiquid Separation Technology in the Alumina Industry”, Light Metals, 1992, edited by E.R. Cutshall, pp 87-90. E.S. Hsia, F W . Reinmiller, “How to Design and Construct Earth Bottom Thickeners”, Mining Engineering, August 1977.

1340

APPENDIX Conventional Thickener Sizing ~~

APPLICATION

Yo SOLIDS

FEED

UNDERFLOW

UNIT AREA FT~ITONID AY

OVERFLOW RATE GPMIFT~

K* FACTOR FTLBS~FT~

(1) Torque based on scale load

*K = TI D2 also, Torque = KD2

1341

Design Features and Types of Filtration Equipment C.Cox' and F. Truczyp

ABSTRACT The chapter covers the various types of vacuum and pressure filtration equipment used in minerals processing, in addition to a few special designs for very fine particle applications. The basic design features of each type are presented, and the typical applications where the various types are used are discussed. Features within the overall equipment type designs which add to functionality within a specific application are also reviewed. INTRODUCTION The mechanical separation of solids from liquids is often the final major step in a mineral production process. Typically, all of a plants valuable product is handled by this stage; thus selection of the correct equipment is critical. Consideration must be given to feed characteristics, final product specification; i.e. cake percent moisture andor filtrate clarity, and production and availability requirements. This section will discuss the major types and respective design features of today's filtration equipment. Category of equipment to include vacuum and pressure, with discussion of filter media. The typical range of filter application with respect to particle size and final product moisture envelopes will also be presented. MECHANICAL DEWATERING BY FILTRATION Liquid-solid separation by filtration requires a differential pressure, AP , across a cake of solids. The AP required can be defined in a general way via Kelvin's Law, which describes capillary forces within the interstitial pores of a cake of solids and solution I:

AP = 4oT.Cos 8

D Where T = surface tension 8 = contact L D = pore diameter Thus the smaller the pore size the greater the force or AP required to overcome the capillary forces and displace the interstitial solution and achieve a desired final cake moisture. Pore size has a direct relationship to the p80 and p10 size of the material to be dewatered. For mineral dewatering applications requiring less than 1 Bar of AP, vacuum filtration methods are generally employed. At application requirements greater than 1 Bar AP, pressure filtration in its various forms is then utilized. The relationship of feed particle size to achievable, 1 Metso Minerals lnc., Colorado Springs, Colorado. 2 ElMCO Process Equipment Company, Salt Lake City, Utah.

1342

residual cake moisture by the typical range of vacuum and pressure filtration equipment is presented in Figure 1. In Figure 1, the region 1 represents the range of typical vacuum filtration equipment where as regions 2 and 3 indicate where pressure type filters are applied. It is important to note in Figure 1, that in today’s mineral processing world, due to the fact the mineral concentrate and tailings products generally have a p80 of 40p or less, pressure filtration is playing an ever-increasing role. As a general rule however, the greater the pressure drop required the greater will be the corresponding capital and operating cost per ton of product. In the subsequent sections, the various types and features of vacuum filters and pressure filters will be discussed. Specialty filtration devices, which also find application in plant design, such as the tube press and belt press, will also be reviewed.

Selection of Filters 0 2

ze

18

0

(D

3

20

e

22

K

24

z

26 28 1

10

100

1000

10000

Particle Size (micron)

Figure 1 Selection of filters

VACUUM FILTRATION Vacuum filtration is a well-established technique used in industrial dewatering. All vacuum filters operate on similar principle. Within a slurry tank a pressure differential between the filter medium surface and the inside of the drum,disk or belt is applied by means of vacuum. This pressure differential causes transport of liquid through the filtration surface while the filter medium arrests solid particles thus forming a cake. As the unit rotates, the cake rises above the slurry level and air is drawn through the cake, forcing out liquid. The liquid (filtrate) exits the filter through the internal piping and the vacuum head. A typical vacuum filtration installation is illustrated in Figure 2. The vacuum filter is fitted with a filter cloth or screen where the solids are deposited. A grid or drainage section supports the cloth where the vacuum is applied. Vacuum receivers are used to collect the filtrate. Filtrate pumps can either be mounted separately or attached directly to the side of the vacuum receivers. Multiple receivers are also used for horizontal belt filters and where individual filtrates may be separated or washing is used. Moisture traps prevent liquids from being pulled over into the vacuum pump. This is extremely important in the case of corrosive liquids or dry vacuum pumps. This can sometimes be eliminated when the solids or liquids contained in the filtrate would not be harmful and a liquid seal pump, such as a Nash, is used.

1343

The vacuum applied to the filters is the atmospheric pressure less approximately 0.2 bars for line losses. In cases where filter cake cracking occurs, the vacuum pressure can be reduced to prevent the air from short-circuiting. Note special mediums such as ceramic can eliminate air bypass by utilizing small enough openings to take advantage of capillary forces, which are greater than that which can be generated by a vacuum.

MOSSTURE TRAP FILTER VAR1-

SPEED DRIVE

VACUUM PUMP

TO CONVEYOR

REC~~VER

ISC CHARGE

O R HOPPEF

AIR SUPPLY

FILTRATE PUMP SEAL TANK

Figure 2 Typical vacuum filtration installation Disk Filters Vacuum disk filters are particularly suitable for relatively simple dewatering applications where high capacity is the principal requirement. The disk design permits a greater filtration area per unit of floor space compared to the drum or belt type designs. The disk filter consists of individual sectors placed together to form a circular disk. Vacuum filtration occurs on both sides of the disk and the filtrate is collected internally and fed through a collection pipe or center barrel. Vacuum is applied through a rotary distribution valve fitted to one or both ends of the unit. Individual sectors are covered with sector bags of cloth or screen. The life of the various mediums can vary from days to months depending on the type of material being filtered. The sectors are rotated in a tank that is mechanically agitated to prevent sanding and provide a consistent mixture of solids throughout the slurry. A variable speed drive is used to rotate the sectors at a normal speed of 1 to 10 minutes per revolution. Disk submergence can also be used along with cycle speed to adjust pick up and dewater times. Vacuum is applied in the submerged area or pickup zone on the disk for forming the cake. As the disk rotates out of the slurry cake dewatering continues until the cake discharge portion of the cycle is reached. For filters with conventional cloth media, the discharge area is separated by ineans of bridge blocks in the distribution valve and an airblow is applied to inflate the sector bag to discharge the cake. High pressure blows, or snap blow, is often used for difficult to discharge material. The advantages of the disk filter are the large filtration area, low initial investment cost, and minimum floor space requirements. The disadvantages are that the washing of the filter medium

1344

for cloth type systems is difficult and effective cake washing is not possible. Recent advancements have been made in hydraulic engineering to increase filtration capacity. Ceramac Disk Filter. A variation of the vacuum disk filter is the Outokumpu Ceramac Disk, Figure 3 (courtesy of Outokumpu Technologies Inc). In the case of the Outokumpu filter, the sectors are a complete one-piece ceramic design (see filter medium section). The primary driving force of the fluid is done by capillary action. The filter requires less power as the vacuum pump acts primarily as a filtrate pump. The capillary action can produce very low moistures that are generally less than conventional vacuum filtration and approach pressure filtration. The hydraulic capacity is somewhat limited, so high feed solids concentrations can produce the best results. The main applications are mineral concentrates. The Outokumpu Ceramic Disk provides the ability to maintain the sectors via ultrasonic cleaning and internal acid washing.

Figure 3 Outokumpu ceramic disk filter Drum Filters Vacuum drum filters have an extended range of application and are considered to have lower maintenance costs compared to disk units. Drum filters are preferred in applications that require lower moisture and or where effective cake washing is required. (Figure 4 courtesy of WesTech) Like a disk filter, the drum or shell is rotated in an agitated slurry tank. Drainage grids are mounted on the surface of the drum, which are then covered by filter media. The drainage grids are made in sections and each section contains a pipe or pipes that apply vacuum to that section. The pipes are collected in one or both ends of the drum to a rotary distribution valve. By having separate individual sections on the surface of the drum, the distribution valve can be used to change the proportion of cake form, wash, dewater and discharge zones. Also similar to the disk filter the submergence of the drum can also be varied. A displacement wash can be effectively applied to the drum by dripping or spraying liquid across the surface of the cake. When high wash efficiencies are required, two or more drum filters can be used in series. Repulp stages are used between filters to increase washing efficiency. More that three drum filters in a series are seldom used if flocculation is required as the cake usually becomes progressively harder to handle and filter.

1345

Figure 4 Drum Belt filters used for barite processing Drum filters offer the flexibility to handle a wide range of material types and pS0 sizes by modifying feeding point and by selection of the type of discharge method. For applications with coarse particle slurries a top feed drum filter should be considered. The top feed principle promotes segregation of the coarser particles in the feed box, which are then deposited first on the media forming a layer with large porosity, which increases filtration rate. Selection of the discharge method will depend upon material characteristic such as, tendency to stick to or blind filter media, fibrous content, and filtrate clarity requirements. Each type of discharge type has its own characteristics and is discussed separately. The most common discharge types are shown in Figure 5 (courtesy of WesTech).

Figure 5 Common types of drum filter discharge mechanisms Scraper Discharge. Scraper discharges consist of a scraper blade near or on the surface of the drum. During the discharge portion of the cycle, low-pressure air blow is used to expand the filter media and aid in cake release. Scraper discharge can be used with solids which do not blind the filter media, and the cake is readily released. For applications requiring high discharge blow pressures, the surface of the drum can be wound with wire to prevent damage to the cloth. Belt Discharge. In applications where cloth blinding is a problem, the Belt Discharge method is used. With a belt type drum, the cloth is removed from the drum after passing through the dewatering zone. The cake is discharged and the filter belt is the washed over its full width

1346

typically by high-pressure sprays, one above and one below the cloth. Wash water is confined and collected for separate discharge through a launder. The wash cloth is then returned to the drum for the next cycle. The belt discharge method is more involved as a cloth tracking system is required. Roll Discharge. Roll discharges are commonly used in the clay industry where material is sticky and difficult to release. The cake from the drum adheres to a roll which is in contact with the drum and which is driven at a slightly higher rotational speed. A scraper blade is set to remove or peel the cake leaving a heel in place on the roll. This discharge method is well suited for sticky materials which has a tendency to adhere to itself rather that the filter cloth. String Discharge. String discharge consists of a rotary drum similar to the scraper discharge filters. Strings are used across the face of the drum to remove cake or pulp from the surface and release it over a set of rolls. This is an effective discharge method when fibrous pulps or sufficiently strong cakes are produced. This method of discharge gives the advantage of a clean drum surface that can be washed intermittently from the outside. A draw back is the obvious maintenance of the string media. Precoat. Precoat filters are used either for clarification or for filtering of solids that are very slimy and do not produce a thick enough cake for discharge on other types of discharge mechanisms. The filter is first coated with a precoat material such as diatomaceous earth after which the slurry is introduced to the filter. Vacuum is maintained throughout the complete rotation cycle to ensure adhesion of the precoat. A doctor blade or knife is used to continuously cut a thin surface of the precoat to discharge the solids that have been deposited. This also presents a cleaned surface to the slurry for the next cycle. Precoat filters can produce a filtrate with a very low solids content thus they are often used for polishing applications. Horizontal Belt Filters Horizontal vacuum belt filters are generally used for handling coarse solids and, or where high washing efficiency is required. They also can achieve lower cake moistures compared to vacuum disk or drum types. Belt filter equipment size ranges from 1 square meter up to as large as 154 square meters. A typical belt filter is shown in Figure 6. The units consist of a rubber drainage belt supported by pulleys on both ends and travels over a vacuum box. Vacuum is pulled through holes in the center of the drainage belt, which is grooved to allow the liquid to flow to the center. The stationary vacuum box is joined to the moving drainage belt either by a series of traveling wear belts or lubricated wear strips. The vacuum box can normally be automatically lowered for easy access and maintenance.

Figure 6 Eimco Extractor horizontal belt filter A separate filter media travels along the surface of the drainage belt in the vacuum area and is removed in the discharge and washed before returning to the filter. The drainage belt is supported on a deck with air or water lubrication. Filter fabric media has been developed specifically for belt

1347

filters to provide single width heavy fabrics to eliminate tracking problems. The vacuum pan can be divided into multiple sections, and each section can have its own vacuum receiver. This type of filter is very sell suited for co-current or counter-current washing applications such as uranium or gold. When co-current or counter-current washing applications are used, the wash rations can be at any level desired since the equipment is not constricted geometrically. In addition, the pulp is only flocculated once and does not become more difficult to handle as in a series of drum filters. Thus compared to disk or drum, belt filters offer the most efficient washing capabilities. Major recent innovations include improved maintenance around the vacuum pan, support deck, and wear parts. Improved drainage belts with high curbing and 4.2 meter wide belts are becoming standard. Steam hoods can be used to decrease moisture content. Figure 7 shows a steam hood on an Eimco Extractor that utilizes a proprietary control system to prevent overheating of the drainage belt. The horizontal belt filters are used on a wide variety of minerals including concentrates, coal, industrial minerals, tailings, and washing applications.

Figure 7 Eimco steam hood PRESSURE FILTRATION The mineralogy of today’s ore bodies is such that economic liberation size of valuable components is becoming finer and finer for mineral concentrates. Final product p80 size of 30 microns or less is now common in Cu, Pb, and Zn processing as well as auriferous pyrites. Also shipping moistures and smelter schedules are requiring these fine size concentrates to contain moisture contents in the range of 8 to 10 weight percent. As a result, a greater pressure drop across the cake, than can be achieved by vacuum, is required. Thus the trend in new plant design and plant modernization is pressure filtration. Pressure filtration is an old concept. The challenge in the past has been to accomplish this with a cost effective, reliable method. The equipment industry has responded by developing ever larger machines and by taking advantage of state of the art manufacturing methods and incorporating up to date instrumentation and computer controls along with corresponding advancements in filter cloth technology. Today pressure filters handle the entire output of a large concentrator in one or two machines with online availability of greater than 95%. Operator supervision is minimal with supervision in some cases being managed by remote communications. Today’s filters have evolved from two basic roots: actuation either horizontally or vertically. Such filters have been adapted to address the specific difficult requirements of the minerals industry and are discussed in specific below.

1348

Horizontal Pressure Filters A picture of a typical Horizontal Pressure Filter is presented in Figure 8 (courtesy of Metso Minerals Inc.).

Figure 8 Horizontal pressure filter %

"

In this configuration filter plates, generally constructed from lightweight polymer are suspended upon a steel frame. The plates are linked together and are opened and closed by hydraulic cylinders. The plates have recessed chambers and between plates is hung individual filter cloths. The two cloths per chamber define the filter volume see Figure 9.

Arrangement of phke pack

Figure 9 Filter plate pack

Feed slurry is pumped to the filter and is distributed to completely fill each chamber. During the feed cycle, the pump output pressure is ramped up to typically 6 bar. Dewatering commences as soon as feed is introduced with filtrate exiting through both sides of the chamber. When the

1349

chambers are full of materials, select filters have membranes in one side of the filter chamber which are pressurized, typically by air, to hold the cake in place to prevent cracking and air by pass during the air dewatering cycle. A common misconception is that pressure filtration is accomplished by a mechanical squeeze of the cake. Since most concentrates are non-compressible the primary dewatering method is by pressurized air being forced through the cake. Such “air through blow” is performed at 5 to 8 bar effectively displacing liquid as it passes through the cake. The relative effect on dewatering by the feed, membrane squeeze, and air-drying is depicted in Figure 10 (dewatering curve). The curve was generated by a machine, which incorporates load cells to accurately determine the net cake weight throughout the entire filtration cycle. The use of a load cell ensures that the filter is properly filled with solids and that all solids have discharged at the end of the cycle. Also a load cell machine can be run based on timed cycles or weighed cycles. The curves show that the majority of the dewatering takes place during the air blow cycle. A wash cycle can be incorporated if required. During wash, a liquid takes the place of the air blow and is passed through the cake. After wash, air blow is restarted to achieve the final moisture.

Figure 10 Dewatering curve showing cake weight through a typical cycle time At completion of the air blow cycle the chambers are opened and the cake falls by gravity through a chute onto a load out conveyor. Chambers are open in sequence thus evenly distributing the discharge weight. Following discharge, the filter cloth is cleaned by a combination of vibration and spray wash. The wash down material is directed back to the filter feed tank for recovery. The chambers are then closed and the cycle repeats. The operation is batch, however with short cycle times of 8 to 10 minutes, and with the use of a filter feed tank with adequate storage, the filtration appears continuous to the overall operation. To match the demands of the industry, horizontal filters now include units with 2 meter by two-meter plates with up to 60 chambers, Figure 11. Such units can now economically handle applications such as dewatering mineral concentrates, tailings and even fine coal.

1350

Vertical Pressure Filters Vertical pressure filters differ from the horizontal units in that the individual chambers are stacked one on top of the other. The chambers are linked and are open and closed by vertically operating hydraulic cylinders as depicted in Figure 12 (courtesy of Larox Oy).

Figure 11 Filter frame for two x two meter plates

Figure 12 Vertical Pressure Filter Plate and chamber design incorporates membranes or diaphragms to provide high-pressure squeeze on the cakes. Unlike the horizontal units with the individual filter cloths hung between each chamber, the vertical units operate with a continuous cloth. The filtration cycle is similar to that described above for the horizontal type design. At the beginning of each cycle, all plates are closed simultaneously and slurry is fed to all chambers

1351

simultaneously. Filtration begins immediately as the cakes are being formed. When the chambers are full, high-pressure water is pumped behind the diaphragms to mechanically force additional liquid from the cake (air is used for Larox filters with filtration areas > 30~’). If washing is required, the diaphragms are drained of high-pressure water and wash water is pumped in on top of the cake. The diaphragms are again pressurized and the wash water is forced through the cake. For final dewatering, the diaphragms are again depressurized and high-pressure air is blown through each cake. Upon completion of air blow, all filter plates are opened and the cloth is advanced through the unit chambers achieving cake discharge followed by cloth washing. Recent improvements include automatic monitoring of the filter cloth condition, seam position and tracking. This eases maintenance and improves cloth life. Other developments include use of sensors on the filter to monitor and control cake thickness, cake pressing (squeezing) time, and air blow flow rate and time. The objective is to provide full process control to achieve consistent results during time of process upsets or mineralogy changes. Tube Press Select dewatering requirements of ultra fine (4 0 micron) materials required specialty equipment. The extremely powerful capillary forces demand greater pressure drops than can be achieved with the horizontal or vertical type units described above. For such ultra fines, pressure requirement to achieve target moistures exceed 100 bar. Thus filtration takes place in a pipe or tube to handle the mechanical forces of the high pressure. Tube presses were initially developed to dewater fine Kaolin. It has since been applied to a variety of difficult filtration operations. A cross section of the Tube press unit is shown in Figure 13 (courtesy of Metso Minerals Inc.). In summary the assembly consists of an outer casing which has a flexible membrane (bladder) fastened at each end. An inner “candle” provides the filtration service through surrounding layers of backing mesh, backing felt and cloth media. The candle is drilled with holes to provide drainage for filtrate.

Figure 13 Tube Press

1352

The filtration operations follow essentially the same sequence described previously for pressure units. Feed is introduced in the tube press under pressure. When a cake is formed, dewatering pressure is applied via the membrane and air blow. If a wash is required the membrane is relaxed and the wash liquor is supplied and then forced through the cake by resqueezing the membrane. A subsequent air blow can then be applied. When the filtration, washing and dewatering are complete, a hydraulic vacuum is applied to retract the membrane, and the candle is lowered to discharge the cake. Air can be blown in behind the filter media to aid in discharge. The candle is then reinserted into the casing for the next cycle. Plate and Frame The earliest types of filter presses date back to the Dark Ages when monasteries used filter presses for production of grape juice. The early modem day pressure filters were Plate and Frame filters. The filters consist of alternate plates and frames that are supported overhead or on side bars (Figure 14). The plates and frames are pressed together to form chambers that are filled under pressure with pumps. The plates are covered with filter media or can also use paper where slimes are present that would blind media. Plate and Frame filters are limited to about 7-bar pressure since the media extends outside the plates. Plate and Frame filters are low cost but still are labor intensive. They are widely used for low volume or long cycle time applications or where paper media may be required.

Figure 14 Plate and frame - side bar design Recessed Plate Pressure Filters Recessed Plate filters consist of single plates that are recessed on both sides . When the plates are pressed together, a chamber is formed. The plates can also be gasketed for applications where spillage of the filtrate is important. Recessed plate filters can operate at a higher pressure than Plate and Frame filters with normal operating pressures of 2-15 bars. Recessed Plate filters can also be fitted with membranes to help in cake washing and provide lower moistures. Figure 15 shows an overhtYad design recessed plate automatic filter (courtesty of Eimco Process Equipment CO.).

1353

Figure 15 Overhead design recessed plate filter The recent advancements in the Plate and Frame and Recessed filters include automatic plate shifting and high-pressure plate washing that reduces cycle time. Recessed filters have higher volume plates to reduce filter size requirements. Filter plate material of construction has also greatly improved with higher operating pressures and temperatures. Polypropylene plates can operate at temperatures of 95 degrees centigrade. Belt Filter Press Selected dewatering applications involve producing a material that can be handled, transported and stored as a solid. Reaching ultimate low moisture is secondary. Examples include aggregate or milling operations, which have minimal land available for tailing storage, and environmental considerations restrict the use of storage ponds. Such tailings typically contain a high percentage of minus 200 mesh material and as such are difficult to economically dewater to a state that can transported by conveyor or truck. Belt filter presses have been field proven to address the above challenge by dewatering slurries containing 40 to 50 percent solids to a solid phase material at 70 percent solids and higher. A diagram of a belt filter press cross section is provided in Figure 16 (courtesy of Phoenix Process Equipment Co.). The feed to the belt filter (1) press is dosed with flocculent, passes through an in-line-mixing device and is distributed evenly into the gravity drainage zone (2) . The feed is contained within the gravity drainage zone by a frame-mounted feed containment box. The gravity drainage section (3) is inclined to facilitate the drainage of free water through the lower filter belt. A 6ame mounted, wear resistant dewatering grid supports the lower filter belt in the gravity drainage zone. Filtrate is collected in a gravity drainage collection pan. The drained solids leaving the gravity drainage zone are feed into the adjustable compression zone where the upper filter belt converges (4) to gently apply compression. Liquid that has been pressed through the screen is collected in the wedge zone collection pan. Pressure and dewatering efficiency are increased by entry into the large roll compression/shear zone. As the belt and cake progress through the roll train (5) the diameter is progressively decreased to form the high-pressure high-shear zone. Discharge ( 6 ) of the dewatered cake from the press is accomplished by the use of plastic doctor blades, which peel the low moisture solids away from the filter belts. On the return to the feed section of the press, each filter belt passes through the continuous support functions of belt washing, belt alignment and belt tensioning. Belt filter presses are available in widths up to 3.0 meters and can handle up to 30 st/hr of dry solids. The main operating costs include belt replacement and flocculent consumption, which can range from 50 glt to 500 glt.

1354

Figure 16 Belt filter press diagram

FILTER MEDIA Vacuum and pressure filters all employ filter cloth or paper as the filtration medium. Experience has shown that specification of cloth type is critical to the success of a filter application, both in terms of operating cost and filtrate clarity. Media or cloth consumption is a major component of a filter’s operating cost per ton. Initial filter economics will include an assumption of the number of cycles or life expectancy of the filter cloth. Such assumptions will be based on experience with similar or like materials. However, often, the material encountered during operation can have different characteristics with regard to the degree of fines, clay content, or abrasiveness. As such the cloth pre-selected can blind or wear out prematurely. Assumed life cycles of for example 6000 can degrade to less than 1000 with an attendant major impact on cloth consumption cost, maintenance man hours, and filter availability. Therefore it is important to define as well as possible the cloth specifications for a given application. Following start up, cloth performance will be reviewed and monitored. Working with the filter supplier and their cloth suppliers is recommended to arrive at the best balance of cloth permeability, resistance to blinding and abrasion, chemical suitability, mechanical requirements, and cost options. Cloth technology is constantly evolving thus a good partnership with suppliers will ensure that new innovations can be incorporated and performance continually upgraded. The following filter cloth basics and a summary of filter cloth terms presented in Table 1 are provided by Crosible Filtration, Inc. There are two types of cloth, woven and non-woven. For the most part it has been determined that woven cloth offer the best combination of filtration efficiency and low blinding tendencies. Woven cloths have three basic yarn compositions: spun, multi-filament and monofilaments. A cloth can be made fiom 100 percent of one of these yarns or 50 percent of two of them; that is a different warp and filling. An all-spun fabric will have the highest particle capture rate for a given porosity; followed by multi-filament and then monofilament which has the lowest rating. As a trade off however the reverse order is the case with regard to the cloth’s susceptibility to blinding. Thus to meet a particular filter application requirements today’s cloths tend to be blend of multi/spun or mono/multi or even mono/spun to take advantage of each yam’s strengths. Yarns can be produced from several materials. The most popular is polypropylene as it is the least expensive, most chemically resistant and is easiest with which to work. Nylon and polyester are also considered when temperature and or chemical incompatibility issues preclude the use of

1355

polypropylene. Specialty fabrics like Nomex, Ryton and PTFE, which are very expensive and are employed in the harshest of chemical and temperature environments. The type of weave is also important in the performance of cloth. A plain weave is the tightest also most apt to blind. The twill weave, usually a 2 x 2, was a standard as a way to offset blinding and give a thicker cloth for filtration. Current design favors the sateen weave where a yarn will travel over several perpendicular yams then go under one then repeat the sequence. This praduces a very smooth surface, which resists blinding and easily cleaned. It is a matter of determining the proper yarn or yarns and type of weave for optimum filtration. In addition to conventional cloth technology, Outokumpu Mintec Oy has developed and commercialized ceramic media termed CERAMEC for vacuum disk application. A cross section of the CERAMEC disk is present in Figure 17.

Figure 17 CERAMEC disk cross section CERAMEC filter discs are patented sintered alumina membranes with uniform micropores to create capillary action. This microporous filter media allows only liquid to flow through. Even at a perfect vacuum there is insufficient pressure differential to force air through the media openings. The Capillary phenomenon is based on the Young-Laplace law, which states that the pores of a certain diameter cause a capillary effect due to surface tension and the contact angle of the liquid. As the discs are immersed into the slurry basin, a pressure difference maintained with a small vacuum pump causes cake formation on the surface of the discs and dewatering takes place as long as free liquid is present.

CONCLUSIONS Filtration equipment is available to meet dewatering requirements for a wide range of material characteristics taking into account feed particle size distribution, feed percent solids, desired product moisture and product handling specifications. Equipment capacities have expanded to address current large tonnage requirements while incorporating the latest in computer control and sensor design to allow automatic independent operation with self-diagnostic capabilities. In support of filtration equipment, media technology has progressed to offer a broad scope of materials and construction to provide the necessary cycle life and performance to ensure economic operation. REFERENCES Mapes, Christopher, Phelps Dodge Cop., “Development and Application of Ceramic Media Disk Filter Technology at Phelps Dodge Morenci, Inc. Petrey, Pete, Phoenix Process Equipment, “ Design Considerations in the Elimination of Slurry Ponds”Aggregates Manager June 1998

1356

Table 1

FILTER CLOTH T E m s Blinding Plugging up of a filter fabric resulting in reduced flow rates and filtration efficiency.

PolypropyleneA synthetic fiber manufactured from a petroleum industry by-product. Excellent resistance to full pH range and has max. Operating temp. of 190F. Is affected by oxidizing agents.

Cake Release Ability of a filter cloth to completely discharge the cake from the cloth.

Porosity (Permeability)The rate of flow of air under differential pressure through a cloth. Generally measured in e l m i n (at %,, water pressure).

MonofilamentSingle, long continuous strand of a synthetic fiber extruded in fairly coarse diameter. MultifilamentSmooth yarn consisting of two or more monofilaments twisted tightly together.

Satin (or Sateen) WeaveWeave in which warp yarns are carried uninterruptedly over many weft yarns to produce a smooth-faced fabric. Offers superior cake release and excellent resistance to binding.

Needled FeltA fabric (felt) made by the mechanical interlocking of individual fibers in a random orientation.

Spun (Staple)Yarns made from filaments that have been cut into short lengths, then twisted together. Staple fibers offer good particle retention; however, cake release may suffer due to the hairiness of the yarns.

NylonA synthetic fiber manufactured from, water and air. Has good resistance to alkalis, but is affected.by strong acids and solutions in pH range 1-6. Max. Operating temperature is 2253. Oxford WeaveModified plain weave where both warp and weft yams are multiple yams but not of equal number.

3 X 1Double WeaveSpecial weave in which both sides of the cloth show a smooth 3 X 1 broken twill surface. Offer excellent stability, retention and cake release properties.

Plain WeaveThe simplest and most common weave produced by passing the weft thread over and under each successive warp thread. Offers low permeability and excellent particle retention, however is susceptible to blinding.

Twill WeaveWeave in which the weft threads pass over one and under two or more warp threads to give the look of diagonal lines. Offers medium retention and blinding properties with high abrasion resistance and good flow rates.

PolyesterA synthetic fiber manufactured from Terephthalic acid and Ethylene Glycol, the resulting polymer has excellent resistance to most mineral acids and solutions in the H range 1-8. Max. Operating temperature = 285F.

WarpThe yam that runs lengthwise in cloth as it is woven on a loom. WeftThe yarns that run widthwise in cloth as it woven on a loom. Also known as filling yarns.

1357

Plant Design, Layout and Economic Considerations Mark Erickson. & Mike Blois*

ABSTRACT In the context of this handbook, solidfliquids separation predominantly refers to in-plant and tailings thickening (and tailings dam management to some extent), concentrate thickening, concentrate filtration and clarification. The incorporation of solidsfliquids separation steps within a process flowsheet affects the plant design and layout in two broad aspects: MACRO LEVEL - Impact of major plant sections on the overall block plan of the processing facility MICRO LEVEL - Impact of individual systems on the general arrangement / plant layout within one of the major plant sections Because of the large equipment sizes that are typically used, decisions regarding sedimentation equipment, e.g. thickener and clarifiers, generally have a greater impact on the macro level of the overall block plan. Environmental factors, such as weather and geotechnical aspects, are also influential and are presented in this chapter. The selection of filtration equipment usually impacts the plant design on the micro scale, such as the arrangement of filters within a building. At this level, an understanding of the system that includes the solidsfliquids separation stage is important. This chapter also presents issues such as design and layout implications for both upstream and downstream equipment.

INTRODUCTION The decision-making process for the choice of a plant siting and the development of a plant layout has been described as an “art rather than an exact science because the factual demands of process design must be combined with experience., e.g. to anticipate unsolved mechanical design problems, and to provide for the human element in operation and maintenance”.’ The literature goes on to say “The key to economical construction and efficient operation is a carefully planned, functional arrangement of equipment, piping and buildings. Furthermore, an accessible and aesthetically pleasing plot plan can make major contributions to safety, employee satisfaction and sound community relations. In fact, aside from the process design aspects, no single factor in a process plant project is as important as the physical layout of the equipment itself. A modern process unit erected today is likely to remain in use for 20 or more years. Any errors in the beginning will be costly to rectify later”. Although these opening thoughts were written for the chemical engineering industry, they are just as applicable to the minerals processing and metallurgical industries. Furthermore, these words, originally written over 30 years ago, remain just as valid in the “New Millennium”. To identify all the engineering details that are necessary to solve plant layout problems related to solidsfliquids separation is a task far beyond the scope of this chapter. This chapter illustrates, by means of examples, the range of factors that need to be reviewed when considering the plant design and layout of solidsfliquid separation facilities. The factors in this chapter can be used as a checklist to ensure that none are overlooked in the study and design stages.

* Bechtel Corporation, Denver, Colorado

1358

The concept of designing a plant layout is divided into two aspects of influence: macro and micro: 0 MACRO: The block plan (arrangement, and inter-relationship, of buildings or plant sections that make up the whole plant) 0 MICRO: The general arrangement (the layout within a building or plant section).2 This chapter emphasizes the importance of understanding the inter-relationship between individual systems within a solidsfliquids separation stage.

MACRO FACTORS INFLUENCING SOLIDS/LIQUIDS SEPARATION FACILITIES Overview of Macro Factors The macro factors that influence the layout of a solidsfliquids separation facility can be divided as shown in Table 1.

SITE CONDITIONS

Topographical Available Space Geotechnical Water Availability Environmental Issues Climatic Conditions

PROJECT CRITERIA

Expected Life of Operation

PLANT EXPANSION

I Expansion Possibilities

PROCESS SYSTEMS

Continuous Operation Batch Operation

Site Conditions Topographical. “The topography of a proposed plant site can be a factor of considerable importance. The ideal site is a level or gently sloping terrain, permitting optional arrangements of the mill and ancillary facilities (and optimal use of gravitational fluid flows). In rugged mountainous terrain, the selection of a plant site can be difficult. It is sometimes necessary to consider a longer ore haulage distance to a more acceptable site. Construction of plants in rugged terrain generally requires a large volume of excavation, extensive retaining walls, deep foundations or piling in filled areas and crowding of buildings because of the space limitation. The choice of plant layouts is restricted. It is usually necessary to locate the plant and auxiliary buildings at different elevations to obtain sufficient area for the complete in~tallation”~ The above paragraph was written with the whole of the metallurgical plant in mind but it is particularly true of the thickening stage in most solidsfliquids separation facilities. Available Space. For a green fields project, abundant space may be available; however plants that are excessively spread out can be problematical from a plant operator’s perspective and contribute to operating inefficiencies. Conversely, sites with little available space present other challenges to the layout engineer. An example of one of the most constrained sites for minerals processing plants would be the diamond recovery vessels operated by De Beers Marine Namibia and others. Vessels, such as the “Debmar

1359

Atlantic”, have the diamond separation plants located on board where space is obviously at a premium. Although soliddliquids separation is not used in the traditional sense, the heavy medium recovery circuits employ gravitational separators. As mentioned above, the nature of the topography is likely to be a major determinant of the available space. Thickening systems are more likely to be affected simply because of the space requirements of large diameter thickener tanks.

Geotechnical. The large areas covered by many modern thickeners make the use and understanding of applicable geotechnical information a critical issue. This is especially true of the thickeners installed at the large copper concentrators located in earthquake prone regions, such as the Andean region. The results from the geotechnical analyses could influence the economics of the method of construction of the thickener tanks, i.e. steel construction compared with concrete wall construction or in-situ configurations using membrane liners and earthen berm walls. As an example of the geotechnical influence on the design of a solid/liquid separation facility, the La Coipa operation in Chile utilizes tailings filtration to produce a filter cake that could be stacked as a method of tailings management. The high seismic potential of the area resulted in the following considerations: 0 The instability of conventional tailings deposition was considered to be too risky in the immediate vicinity of the mine. 0 Pumping of the tailings to a more stable deposition site was too expensive. The use of tailings filtration minimized the percolation of cyanide bearing solutions into the ground water. The use of tailings filtration also provided capabilities for increased recovery of both water, in an arid area of high evaporation rates and cyanide reagent for reuse and recirculation. Plant operating ex erience at La Coipa indicates that the “recovery of cyanide pays for the filter plant operation”! The La Coipa operation was originally designed to handle 15 000 t/d and the current throughput is approximately 18 000 t/d. The tailings filtration is carried out using twelve 100 m’ Delkor horizontal belt filters. Water Availability. The supply of water is a fundamental requirement for most mineral processing operations. In areas where water is in short supply, the focus of a solidsniquids separation stage may be for water recovery rather than separation of the valuable constituents. Water recovery process often takes the form of tailings thickening, the choice of which may also be impacted upon by the method of tailings management. The lO,OOOt/d Mantos Blancos operation, in Chile, is an example where the cost and availability of water are such that tailings filtration for water reclamation is economical. “Tailings from flotation are cycloned, with the coarse plus a portion of thickened fines being filtered. Originally, disc filters were used, but were subsequently replaced with three loom2 Delkor horizontal belt filters. In addition to the economic benefits of the water reclamation, the tailings deposition area had a high subsurface salt content. Deposition of the wet or unfiltered tailings could have resulted in serious salt dissolution problems resulting in soil instability problems under the dam site.”6 Environmental Issues. Although many large capacity concentrators now discharge flotation tailings directly to the tailings management facility, the use of tailings thickeners is also common. The diameter of these large thickeners can exceed 120 metres and therefore can present a significant visual impact. The environmental issues associated with the storage and management of concentrator tailings are beyond the scope of this chapter; however prevention of breaching and containment of seepage are primary considerations. The use of vacuum as a method of filtration requires the use of vacuum pumps. These units tend to be noisy and specific enclosures, e.g. brick buildings and sound proofing, are usually required to ensure that the vacuum pumps operate within the noise regulations.

1360

Climatic Conditions. The drive to higher tonnage operations and single line facilities has resulted in an increase in the size of conventional and high rate thickeners. Most of these large thickeners are installed outside of buildings. However mining operations in cold climates such as Northern Canada require that thickeners need to be installed inside the concentrator building, both for operator protection and to prevent freezing of the process streams. An example of a cold weather installation is the EKATITMdiamond mine in Canada. The EKATIT” mine, which commenced operations in October 1998, is located in the Northwest Territories approximately 300 kilometres NNE of Yellowknife in Canada. Fine tailings, consisting of minus 0.65 millimetre material, is thickened in deep-bed compression ultra-high rate thickeners, supplied by Wren Technologies (Pty.) Ltd.+ “The advantage of this unit lies in its small footprint, high capacity per unit area and the fact that there is no raking mechanism and therefore no moving parts.” “Below is a list of the advantages and disadvantages of the Wren unit (E-Cat thickener) that were considered, by BHP Diamonds Inc., as most important for this application, as compared with those of other thickening and de-watering devices.” Potential advantages of the Wren unit (E-CAT) include: “high throughput rates low capital cost, mainly due to the simplicity of construction small floor area required Potential disadvantages of the Wren unit (E-Cat) include: ‘konsiderable pumping capacity required to overcome the large static head, because of the large height to diameter ratio limited solids surge capacity; therefore, any problem at the thickeners or downstream requires a rapid shutdown of the solids feed to the thickener A Wren pilot scale CAT was used at BHP’s Koala bulk sampling plant, prior to the design and construction of the main E K A T P process plant. Extensive testing of this pilot unit proved its suitability for the application. In this particular instance, the deep bed compression type thickener was the only feasible option due to the limited ~ p a c e . ” ~Figure 1 below shows an indoor installation.

Figure 1 E-Catm Clarifier/Thickenersinstalled indoors for cold climate operation Wren Technologies (Pty.) Ltd. was acquired by Eimco Process Equipment Company. These deep-bed compression ultra-high rate thickeners are now called the “Eimco E-Cat thickeners.”

1361

Project Criteria The overall project criteria will define the expected life of a potential mining project. The life of a mining operation / mineral processing plant typically depends on the size of the deposit and the rate of mining, and the design life will affect the selection of solifliquid separation equipment and facility arrangement.

Plant Expansion ‘The history of the mining and metallurgical industry abounds with expansions of operations beyond the original capacity. This usually results from changes to the original design criteria such as “the mining plan” and “the final pit limits.”’ The space requirements of thickeners, particularly tailings thickeners, are often substantial. A clear understanding of the plant expansion possibilities is needed and this understanding should be incorporated into the Plant or Process Design Criteria so that potential future misunderstanding can be minimized. Unless it has been defined as a prerequisite, the plant designer of the original plant cannot be expected to anticipate future expansion needs or provisions. However he should investigate the operator’s vision, and strive for acceptable practicable means to allow for future flexibility. “Generally speaking, the expansion of a metallurgical plant embraces two categories of equipment: 0 that which can be increased in capacity by simply adding more units within existing space limitations 0 that which is one of a kind and simply has to be enlarged or replaced by a larger machine with attendant waste of investment capital and loss of produ~tion’’~ Process Systems Another of the more important aspects of designing a plant layout is an understanding of the type of process system and the grouping of these process units. There are three types of process streams that are commonly found in the mineral processing industry Batch Systems. Batch systems in the mineral processing industry tend to be confined to relatively small capacity, high value products. An example would be the use of plate and frame type filter presses for copper and other concentrates; these units can operate in a “fill, pressure squeeze, air blow and cake discharge” cycle. During this cycle, the filling of the filter may only occupy three minutes out of a twenty-minute cycle, i.e. filling only takes place for 15 percent of the time. As a result, sufficient storage of both the filter feed slurry and the product filter-cake must be provided if the filter is part of a continuous system. Batch systems tend to require more frequent and a greater degree of sophisticated attention by the plant operator. Further these systems tend to have a greater degree of instrumentation associated with them.

Continuous Process Stream System. In this system, the layout follows the process flowsheet. Many mineral processing plants are designed around this concept. The pipe routes tend to be minimized with this layout. Functional Grouping System. In this system, similar process functions are grouped together. This system is common in the chemical engineering industry. An example of this continuous functional grouping system would be the grouping of filters with different duties within a single filtration building. The piping routes tend to be longer in this system but the ancillary services such as the reticulation for vacuum and compressed air systems would be more localized.

1362

MICRO FACTORS AFFECTING SOLIDSLIQUIDS SEPARATION SYSTEMS Overview of Micro Factors The micro factors that influence the design of a soliddiquids separation facility can be divided as shown in Table 2. Table 2 Micro Factors FEED

Feed Streams

DISCHARGE

Solids Discharge Liquid Discharge

SAFETY ISSUES

Safety in Operation Safety in Construction

OTHER ISSUES

Process Considerations Plant Height Operability Maintainability ConstructionConsiderations

MICRO FACTORS AFFECTING THICKENING SYSTEMS Introduction The two key aspects of the micro factors that affect soliddliquids separation facilities are: 0 Input or feed streams Output streams covering both the solids and the liquids discharge These three streams, for the thickening stage of solids separation are usually in the form of slurries. Feed Streams The feed stream to a thickening stage is either by gravity flow or from the discharge of a pumping stage. Thickeners are often fed by gravity when the feed stream is the overflow from a cycloning stage in a milling circuit as the cyclones are usually positioned at a height sufficient to permit gravity flow. Similarly, the tailings stream from a flotation plant often utilizes gravity flow to the tailings thickeners. As an operating precaution, the gravity flow from the flotation cells allows the flotation cells to be partially drained in the event of a loss of power or control. The cell discharge valves would operate in the “fail-open” mode. This arrangement allows some of the slurry volume, held within the cells, to report to the tailings thickener and thus preventing the cells from sanding up. In the case of the very large copper concentrators, a modified control system could be adopted to reduce the loss of un-floated valuable constituent to tailings and resulting loss of revenue. In the case of the ultra high rate thickeners, where the height to diameter aspect ratio is high, the thickeners usually need to be pump fed. This additional pumping stage would add to both the capital and the operating costs. Whatever the method of feeding, the incoming slurry to a thickener should enter the feed well in a manner which minimizes turbulence and air entrainment, The Fitch type of feed well with its opposing “race-ways is an example of a split feed launder design that has been used for conventional thickeners. If the slurry is being pumped, the problem of air entrainment can be greatly reduced by pumping from a level-controlled sump so that the pump does not aspirate air when the slurry level in the sump drops to the level of the pump suction.

1363

Discharge Streams “Underflow pumping has four basic arrangements: (1) The underflow pump at the tank perimeter with buried piping from the discharge cone, (2) the underflow pump at the tank perimeter or directly under the discharge cone in a tunnel or elevated tank, (3) the underflow pump at the tank perimeter with a peripheral discharge from the tank sidewall and (4) the underflow pump in the center of the tank or at the perimeter with center column piping”5 “A high percentage, if not the majority, of thickener operating problems can be accounted for by deficiencies in the underflow removal system. As a general rule, the closer the pump or throttling system, such as an orifice or control valve, is to the bottom outlet of the thickener or clarifier, the better. Exceptions can be made for underflow slurries that are relatively fluid and have little tendency for sand separation such as magnesium hydroxide and waste treatment suspensions. Also some fine-size mineral applications in which very low flocculant dosages are utilized; these dosages would be insufficient to affect the slurry vis~osity.”~ For the majority of mineral applications, speed regulated centrifugal pumps provide adequate control for the underflow pumping system. The high suction head on the underflow pump means that gland seal water is usually needed on a centrifugal pump. To prevent excessive dilution of concentrate slurries that are pumped to filtration stages, the use of variable speed peristaltic pumps, such as the Bredelm pump, have found increasing acceptance; further they are able to handle dense slurries. The use of pulsation dampers on the discharge piping is required with this type of pump and should be regarded as a critical design issue. The use of gravity discharge for thickener underflows is not uncommon in the mineral processing industry. Those systems, which involve large tonnages, generally work best since larger orifices can be used and plugging from tramp oversize is less likely. This gravity discharge approach can be used where the disposal of the slurry, usually tailings, is downhill from the thickener. Southern Peru Copper Company’s (SPCC) Toquepala concentrator has three peripheralltraction driven 100 metre diameter tailings thickeners and the underflow from these gravitates to the disposal area. Safety Issues There is always the possibility of someone falling into a thickener and drowning (or poisoning in the case of cyanide solutions). Life buoys should be located along the length of the thickener bridge. If there is operator access to the thickener overflow weir, life buoys should also be placed here. Although not strictly a personnel safety issue but rather more of an equipment interruption issue, hard hats worn by the operators have a tendency of falling into the thickener and eventually causing plugging problems in the thickener underflow system. In order to reduce the potential of the hard hats falling into thickeners, operators and maintenance staff working on or near thickeners are typically required to remove their hard hats and place them in a box located near the edge of the thickener; as an alternative hard hats can be equipped with chin straps. As part of a HAZOP analysis of the thickener installation, the possibility of a failure in the underflow piping must to be considered. The risk of failure in the underflow pumping system is greater if the pumping system is located under the center of the tank compared with the pumps located at the periphery. However, the pump suction piping may be more prone to plugging if the pumps are located at the periphery. With an elevated thickener, the spillage from the underflow system is likely to be collected within a bunded area. However, if the thickener is installed at ground level and the underflow system is installed beneath the thickener, the possibility exists of flooding the underflow pumping area and the access tunnel. The tunnel could be sloped towards the center so that personnel move to shallower ground on leaving the tunnel; obviously there would have to be a significant spillage handling system located at the center. Alternatively, the tunnel could be sloped away from the center, in which case the spillage system would be at the periphery. In either case, and especially for large thickeners a personnel emergency escape tunnel should be considered.

1364

Other Issues Process Issues. The use of Counter Current Decantation (CCD) circuits is not confined to the recovery of dissolved valuable components. Newmont’s Batu Hijau operation in Indonesia uses seawater as the flotation medium. “The final product, copper-gold concentrate containing 30 percent to 33 percent copper, is washed in a counter-current decantation system to remove chloride before being pumped as a slurry to the filter plantvy8 The chemical composition of the liquid phase must be fully understood. In acid base metal recovery process, the liquid phase contains varying concentrations of sulfuric acid. In the potash industry, the liquid phase may contain varying concentrations of salts and especially the chloride ion. Obviously the chemical composition of the liquid phase will impact the selection of materials of construction with which the fluid comes in direct contact. However, care must be taken to understand the impacts on the materials with which there is no direct and intentional contact. For example, the acid mist generated in the filtration sections of some base metal refineries requires that the building structural steel and cladding be suitably protected. In addition to structural steel protection, the chloride environment requires that consideration be given to the protection of the outside of piping, electric motors and concrete. Construction Considerations. A number of recent thickeners have been built using the compacted bed approach with either a concrete thickener floor or a plastic membrane covered with sand being placed upon the compacted bed. There have been a number of reported failures of the compacted beds; these include the Hartley Platinum Project in Zimbabwe, and the Murrin Murrin nickel laterite project in Australia. Whilst the specific reasons for each failure are not known, these failures would suggest insufficient quality control in either or both the design and construction stages. MICRO FACTORS AFFECTING FILTRATION SYSTEMS Introduction The two key aspects of the micro factors that affect solidsfliquids separation facilities are: Input or feed streams Output streams covering both the solids and the liquids discharge The soIids discharge stream for the filtration stage of solidsfliquids separation is usual!y in the form of a filter cake containing 5 percent to 20 percent moisture depending on the characteristics of the material being filtered. “Although modern automatic pressure filters have short cycle times, they are nevertheless batch units. Filter feed and cake discharge is intermittent. This must be taken into account when placing pressure filters into a flowsheet for a continuous process such as a flotation plant. Adequate upstream and downstream surge capacity must be provided. It is critically important to understand the stages of the pressure filtration cycle, and the relationship between average and instantaneous Feed Streams A filter feed surge tank is needed to provide buffer capacity because the thickener underflow is a continuous stream but the feed to the filter is intermittent. The size of the buffer storage depends on the type of filter selected and the number of operating filters; however good practice is to have a minimum of one hour’s storage at maximum production rates. Manufacturers of filtration equipment are recommending that trash screens be installed prior to the filter feed surge tanks. “All process streams contain trash, which can block pipelines and filter plate feed ports. The source of trash can be a mystery, and the type of trash varies from stones and hardened slurry to rubber gloves and cable ties. Trash removal is essential for good operation and to prevent blocked pipelines or filter feed ports that could damage equipment. A self-cleaning screen discharging trash to a container outside the bund is more reliable than a manually emptied

1365

screen basket.” Delkor linear screens are becoming increasingly accepted as trash screens ahead of both thickener and filter facilities. The design of the filter feed system is critical to the successful operation of a filter. Some filter suppliers prefer to be responsible for the filter feed system and therefore they can include a smaller intermediate feed tank and the feed pumps within their supply. “For most mining and metallurgical applications, the filter feed pump runs only during filter feeding, and is started and stopped under control from the filter’s PLC. In special cases, the feed pump runs continuously, with a recycle (at lower pump speed) to the feed tank during other phases of the filtration cycle. A pressure filter is a batch unit and the filling or filtration stage usually takes only 20percent of the total cycle time. As a consequence, the instantaneous flow rates during filtration are quite high. Centrifugal slurry pumps with gland service water are the standard choice. Gland seal arrangements are best determined by the pump manufacturer, but they need to account for the stardstop operation of the pump and the wide range of operating conditions. Where full flow water flushing of the gland is provided it can cause considerable slurry dilution between filling cycles, and solenoid control of gland water, or reduced flow sealing systems should be considered. “The pump runs against increasing head as the filter cake builds (Figure 2), and an understanding of the system curve is important for correct pump sizing. Once the chamber is about 50 percent full, the pressure will rise as the cake resistance increases, causing the operating point to move back up the pump curve. At the end of the filtration stage, the pump will be operating at close to full pressure and the flow will have decreased to 10 to 40 percent of the initial flow rate. Where slurries are known to have a highly abrasive nature, the maximum flow rate may have to be lower. Where pump selection shows that flows greater than the recommended maximum are possible, a variable speed drive pump should be used. Pump speed may then be increased as system head increases. To calculate the flow, the frictional losses for the pipeline and static head need to be known. The resultant curve should then be drawn onto the pump curve for the proposed pump. The intersection of the system curve with the pump curve at the proposed speed will then determine the maximum feed’flow rate. Where large feed tanks are used, the flow rate should be checked with full and empty tank levels. With flat system curves there can be a significant change in flow with changes in tank level. Similarly, where there are significant changes in feed density, the flow with minimum and maximum densities should be checked. Once the maximum flow rate has been determined the Net Positive Suction Head (NPSH) required should be checked against the NPSH available with the proposed suction pipe and tank d e ~ i g n . ” ~ FEED PUMP SYSTEM GRAPH for Pressure Filters

-

30 25

I

20

15 10

5 0 0

0.2

0.4

0.8

0.8

1

1.2

Flow (U. .m?

Figure 2 Filter feed pump and system curves (modified from Townsend)

1366

1.4

Discharge Streams

Solids Discharge Streams. An important criterion for the design of a filter plant is the arrangement for handling of the solids product (filter cake) from a filtration stage. The height of installation of the filters is often determined by the handlinghansfer arrangement of the filter cake. Filters can be located directly above stockpiles and thus the discharge would be by gravity directly onto the stockpile. Although this would minimize the initial handling of the filter cake, there would be two distinct disadvantages. Firstly, the elevation of the filters would likely be higher than if the product was transferred away from the filter; as a result both the capital and operating costs of the installation would probably be higher. Secondly the control of spillage and hose-down water would be difficult; it is not good practice to allow the spillage to fall down onto the stockpile. In general, the filter cake should be transported from the filter and the surrounding wet area into a dry area. The method of discharge of the filter must be understood prior to the design of the transport system; the instantaneous discharge rate can be significantly higher than the average hourly rate. A horizontal belt filter discharges the cake continuously and therefore at a rate very close to the average hourly rate. A plate and frame filter generally drops the filter cakes one at a time throughout the discharge period; therefore the instantaneous discharge rate is not as great as for a Larox type filter whose discharge period is usually between 10 and 30 seconds per cycle. Filtrate Streams. As with the solids product, the type of filter determines the approach to the collection of the filtrate and especially whether the filtrate flow is continuous or batch. The filtrate is usually at atmospheric conditions for pressure filters and can be gravitated; however for rotary drum and horizontal belt filters, the filtrate needs to be separated from the vacuum system. Vacuum systems require the use of filtrate receivers, barometric legs and moisture traps. These units should be installed in a manner that minimizes the pressure loss between the filter and the vacuum pumps. Excessive pressure losses result in energy wasted in overcoming these losses. The minimization of pressure losses, in the form of flow restrictions, between the filter and the receivers, will reduce wear caused by filtrate containing abrasive solids. “The line to the vacuum pump should also be designed with the same objective and if a velocity of approximately 70 feet per second is used in selecting the pipe size, the pressure drop will be minimized in the line. With dry vacuum pump installations it is essential to include a moisture trap between the receiver and the pump; this is possibly followed by a scrubber if a corrosive solution is being filtered. The height of the moisture trap is not important as long as the discharge line, which will act as the barometric leg, is of a length greater than needed to sustain a column of water under full vacuum, typically a height close to 35 feet. The barometric leg should be sized to handle the total filtrate volume at a line velocity not greater than 5 feet per second. Water sealed vacuum pumps generally will not require moisture traps unless the solution is corrosive or valuable, or contains free lime. Filtrate pumps must be selected with NPSH characteristics that match the expected operating condition^."^ The Krogh pump is a good example of a filtrate pump that mounts directly onto the filtrate receiver, thus doing away with suction piping and balance legs. The batch nature of pressure filters means that, like the solids, the filtrate flow will vary from a maximum at the start of slurry feeding to almost zero at the end of the feed cycle. If an air blow step is included, there is likely to be very little filtrate entrained in the exhaust air stream. Thus the filtrate collection system needs to be sized for a range of conditions and needs to recover the fluid from what can be a very high airflow. At the start of the feed cycle, the filtrate can contain a significant quantity of suspended solids; as a result the filtrate is usually returned to a thickening step for the recovery of the solids. If the thickening stage is omitted, the implications and quantity of the suspended solids must be understood prior to selecting the discharge point of the filtrate. Safety Issues “As a result of the new labor reducing designs and new types of filter cloth developed by the cloth manufacturers, the pressure recessed plate filters are being used in applications that have

1367

traditionally been processed on vacuum disc or drum filters.”” The improvements in the design of both the filter cloth and especially the recessed plates has lead to increases in both the pressure used for closing the filters and the slurry inlet feed pressures. The filter manufacturers have placed an increased emphasis on the safety hspects of the filter operation. The use of light curtains, which stop the closing of the presses, have become common. Filter cake removal systems and cloth washing, in both the vertical type of filter and the horizontal suspended recessed plate filter, have ensured that good cloth seals are obtained upon filter closing. Thus the chances of pressurized slurry streams squirting from between the plates have been reduced. Other Issues Process Considerations. The selection of filtration for a solidsfliquids separation stage can also be influenced by process considerations. Twenty-six 83 m2 Delkor horizontal belt filters were installed at the Nchanga Copper Tailings Leach Plant, in Zambia. “The major process consideration was whether to recover the copper solutions produced in the acid leaching process with either: conventional counter current decantation (CCD) thickeners, or 0 large horizontal belt vacuum filters recently applied to other high-tonnage mining applications. The principal advantage of the filters was that they recover valuable solutions from unwanted solids in one step at greater efficienc and at a lower capital cost than a train of say five thickeners, each 76 metres in diameter.

,,A

Plant Height. The discharge arrangement for the solid product, as mentioned above, is often the determining factor for the height at which filters can be installed. Both capital and operating costs are increased with increasing height of installation. Operability. “ Pressure recessed plate type filters have undergone extensive redesign to minimize operator attention. Almost all of these pressure filters have automatic opening and closing mechanisms to either open all of the plates at the same time, or to index and open 1 to 6 plates at the same time.”” The increased use of reliable automation has improved the operability of filtration equipment, especially pressure filters. The increased the use of automation has enabled filter systems to operate unattended. However, it is good practice to locate them in an accessible and frequently visited part of the plant to ensure that regular checks and routine maintenance are carried out. Maintainability. In order to provide an efficient maintenance environment, there should be adequate clear space around the filter and the building should allow for the safe lifting of components from the ground floor to the filter. In most installations, an overhead crane is required above the filter. The crane’s access to the filter should not be impeded by filter feed pipes or other process connections.

LAYOUT METHODOLOGY “There is no single technique leading to the best arrangement in any layout problem; several stages may be required with different techniques appropriate to each. The development of a plant layout is an interactive process between all factors both macro and micro. The fact that plant layout problems are 3-dimensional in nature is initially the main reason for the interactive process. During the detailed design phase, additional information now available may require re-evaluation of some earlier constraints. As a result, re-examination of earlier alternatives may be required.”’ “There are three basic principles of layout planning: 0 Plan the whole then the detail: Individual aspects must be subservient to the whole and sub-optimization avoided.

1368

0

Plan the ideal and from it the Dractical: The ideal is free from restrictions and gives a datum, the cost of departing from which can be set against the advantages to be gained. Plan more than one layout: It is seldom that a single layout is “best” for each criterion. Planning more than one permits comparisons and leads to greater confidence in making the final

CONCLUSIONS This chapter has, by means of a variety of examples, illustrated the range of factors that need to be reviewed when considering the plant design and layout of solidsfliquid separation facilities. The relative importance of each of these factors is very much situation dependent. The diversity of solidsfliquid separation situations faced by mineral processing and design engineers is large and therefore this chapter has deliberately not focused on one particular aspect. The factors highlighted in this chapter can therefore be used as a checklist to ensure that none are overlooked in the study and subsequent design stages.

REFERENCES 1 House, F.F. “An Engineer’s Guide to Process-Plant Layout” Chemical Engineering 120 (July 28, 1969) 2 Blois, M.D.S. and Talocchino, L. “Plant Siting and Layout - A Metallurgist’s Perspective”. South African Institute of Mining and Metallurgy, School of Metallurgical Process Design in the 90’s. 1993 3 Weiss, N.L. “SME Mineral Processing Handbook”. SME of the American Institute of Mining, Metallurgical and Petroleum Engineers Inc. New York (1985) 4 Mohns, C.A. and Paradis, T.G. “Deep Bed Thickener Operation at the EKATITM Diamond Mine”, Paste Technology for Thickened Tailings Symposium November 1999. School of Mining and Petroleum Engineering, University of Alberta (1999). 5 King, D.L. and Baczek, F.A. “Characteristicsof Sedimentation-BasedEquipment” Design and Installation of Concentration and Dewatering Circuits SME of the American Institute of Mining, Metallurgical and Petroleum Engineers Inc. New York (1986) 6 Minson, D.N. and Williams, C.E. “Filtering Systems for Dry Tailings Deposition” Canadian Institute of Metallurgy, 38th Annual Conference of Metallurgists. Quebec (1999) 7 Emmett, R. “Private Communication” September 7th, 2001 8 DeMull, T.J., Spenceley, J. and Hickey, P. “Planning and Teamwork Lead to Successful Start-up at Batu Hijau” SME Annual Meeting 2001. SME of the American Institute of Mining, Metallurgical and Petroleum Engineers Inc New York (2001) 9 Townsend, I. “Plant Design Considerations for Automatic Pressure Filtration of Flotation Concentrates” XXV Convencion de Ingenieros de Minas del Peru. Arequipa, Peru (2001) 10 Moos, S.M. and Klepper, R.P. “Selection and Sizing of Non-Sedimentation Equipment” Design and Installation of Concentration and Dewatering Circuits SME of the American Institute of Mining, Metallurgical and Petroleum Engineers Inc. New York (1986) 11 Hampsheir, P.R. “Design and Construction of the Nchanga Copper Tailings Leach Plant Stage 3”, The Institution of Mechanical Engineers Volume 200 No. 131 (1986) 12 Mecklenburgh, J.C. Plant Layout “A Guide to the Layout of Process Plant and Sites.” The Institution of Chemical Engineers, London (1973)

1369

Selection and Sizing of Slurry Pumps Michael J. Bootle’

ABSTRACT The mineral processing industry utilizes centrifugal slurry pumps in a wide range of key applications, including hydrotransport, grinding circuits, flotation circuits, thickening, and tailings disposal. Optimum system operation is dependent upon a large number of critical factors. Foremost, a detailed knowledge of the solid and slurry properties is required for sump and piping design, in order to prevent settling of the solids and to minimize air entrainment, as well as, to determine the flow, friction losses, head and pressure requirements of the pump. This same knowledge of slurry properties is also used to properly select a pump with the appropriate combination of geometry and materials to yield the best balance of the often conflicting pump priorities of stable operation, maximum wear life, and minimal energy consumption. These factors and choices will be discussed along with methods of applying the appropriate slurry corrections to the final pump selection.

INTRODUCTION/OBJECTIVE The objective of this chapter is to provide the reader with a basic introduction to the critical factors, which need ,to be considered to create a successful solids handling pumping system. It is hoped that upon completion of reading this chapter, the reader will have a better understanding of the following: The basic construction features of a centrifugal slurry pump. The advantages of various impeller and casing designs. The benefits and limitations of elastomers and metals for wear components. A basic understanding of pump performance curves, system curves and their interdependentrelationship with respect to point of operation, i.e. flow and head. The difference between settling and non-settling slurries, viscous Newtonian and viscous non-Newtonian slurries. The effect of each of the above slurries on pump performance and how to apply the appropriate corrections to account for these effects. The importance of determining the settling velocity of the solids in the sluny and a method of calculating settling velocity for a heterogeneous slurry. Which of the above factors will most likely be involved in the various applications within the mineral processing industry and what pump type and appropriate correction factors apply. In an attempt to accomplish all of the above in one short chapter, it is obvious that for most of the above topics only the basic details will be discussed. It is the writer’s deliberate intent to give preference to an overview of the system design process and very basic, but important potential pitfalls to be avoided at the expense of excessive detail. When appropriate, references will be given to allow for additional research, if and when required. ‘Weir Slurry Group, Inc., Madison, Wisconsin. Copyright 0 Weir Slurry Group, Inc. 2002

1373

PUMP CONSTRUCTION

General The design of a horizontal centrihgal slurry pump is a balance of design considerations to best meet the requirements of a particular slurry duty. These requirements may include one or more of the following: The ability to pump high density abrasive slurries with adequate wear life. The ability to pass large diameter solids. The ability to handle air entrained andor viscous fluids with reliability and minimal performance corrections. When compared with clear liquid pumps, the above requirements often result in the slurry pump being larger than its clear liquid counterpart and sacrificing maximum efficiency in exchange for the ability to achieve the above goals. Within a slurry pump it is expected that the components which come into contact with the abrasive slurry will wear. It will be shown that minimizing wear is done through appropriate pump design, proper material selection and proper pump application. Figure l a illustrates a typical unlined horizontal centrifugal slurry pump and Figure Ib illustrates a klly lined version of the same pump. The corresponding wear components for each have been appropriately marked. To maximize wear life, thick casting sections are provided on the impeller, the casing of the unlined pump, and the wear liners of the fully lined pump. Often, the unlined casing casting thickness is greater than fully lined version casing liners. This additional thickness requirement is due to a greater need for the unsupported, unlined casing to safely handle the internal pressure of the pump with an adequate factor of safety to account for wear. The lined pump also allows for the use of a wider variety of materials, such as elastomer liners, which often outperform metal in fine particle and corrosive applications. Therefore, while the unlined pump may offer the lower initial capital cost, the lined version allows for a greater number of material choices, which may have longer wear life and lower replacement spares cost. The lined design is also lnherently safer from a pressure containment standpoint. Large clearances are provided within the impeller and casing to allow for the passage of large diameter solids, while also reducing internal velocities and corresponding wear.

SUCTION SIDE LINER IMPELLER

IMPELLER BACK LINER.

BACK LINER

SHAFT SLEEVE

SHAFT SLEEVE

VOLUTE CASING LINER

CASING

Figure la: Wear components on an Unlined Figure lb: Wear Components on a Fully Lined Centrifugal Slurry Pump. Centrifugal Slurry Pump.

1374

Slurry pump impellers tend to be larger than their clear liquid counterparts. To minimize speed and maximize wear life of both the impeller and suction side liner, rarely are slurry pump impellers trimmed in diameter to meet the duty point. On low horsepower applications (below 300 kW, 400 hp), belt drives are the most popular means of achieving the required speed for the duty point. Belt drives are inherently quiet and also allow for relatively easy speed changes if required. For higher horsepower applications, gearboxes are commonly used to meet the desired speed and duty point. On applications where variable flow is required, variable frequency drives are used to provide the necessary continual speed changes. Most often these VFDs operate with other speed reduction devices, such as belt drives and gearboxes, to allow for the use of higher speed, lower cost motors. Bearing Assembly The bearing assembly of a heavy-duty slurry pump should incorporate larger diameter shafting and bearings than those found on a clear liquid pump. Often, roller bearings are used to provide further additional load carrying capacity to handle the added forces associated with the greater specific gravity liquid, the inherent imbalance due to wear, the shock loading fkom large particles and the hydraulic imbalance from air entrained liquids. On belt drive applications, the drive end bearing handles the majority of the side load due to belt pull, which can be substantial. Most properly sized and properly tensioned v-drives exert approximately 9 KN of belt pull for every 100 kW of motor power (1500 lbf for every 100 Bhp). Figure la illustrates a bearing assembly with identical low angle, angular contact bearings on both the pump end and drive end. The low angle bearing is well suited to the radial loads from the impeller or belt pull from the drive (if so equipped). Figure l b illustrates a higher capacity bearing assembly, which fits within the same packaging dimensions. This assembly features a duplex angular contact roller bearing on the pump end and a cylindrical roller bearing on the drive end. The duplex angular contact bearing has a higher bearing angle than the bearings shown in Figure la, making it more suitable to handle higher axial loads, such as those seen with smooth backed impellers, open faced impellers and series pumping applications. In this configuration, the cylindrical roller bearing on the drive end handles none of the axial load, but has tremendous radial load capability, making it well suited to belt drive applications. Grease or oil are both suitable means of providing lubrication to the bearings. Grease has the advantage of providing greater contamination protection. Oil has the advantage of higher speed capability and is easier to change if it becomes contaminated. Oil lubrication also provides the opportunity to install a bearing cooling system should it be judged necessary. With remote installations, care must be taken with oil lubrication to insure level mounting. When the shaft seal design allows it, shorter shafts with reduced impeller overhangs result in reduced bearing loads, shaft stress and deflection through the seal area. When packing is used as a shaft seal, a hardened andor ceramic coated shaft sleeve is recommended to prevent shaft wear. Impeller Design As mentioned above, slurry pumps typically have impellers that are larger than their clear liquid counterparts. This is to lower the impeller speed required to achieve a given head and to provide more material for wear purposes. For high wear applications closed impellers are preferred. In coarse particle applications, expelling vanes are recommended on the face of the front shroud. The purpose of these expelling vanes is to prevent large particles from becoming trapped between the impeller and suction side liner and to minimize recirculation. The benefit is a reduction in gouging and recirculation wear, at the expense of two to three percentage points of efficiency. Expelling vanes are also often used on the back shroud of the impeller in coarse particle applications to prevent the trapping of large particles between the impeller and back liner. In this location they also serve to reduce the forward axial load (improving bearing life) by lowering the pressure acting on the back shroud and beneficially reducing the pressure at the hub and packing. This reduces the pressure differential at the shaft seal and reduces the tendency for slurry leakage from the pump. As with expelling vanes on the front shroud, “backvanes” usually absorb two to three percentage points of efficiency.

1375

To combat wear and allow for the passage of large diameter solids, heavy-duty slurry pump impellers feature thicker main LARGER CLEARANCE -14) SHORT THICK MAIN PUMPING VANES pumping vanes and fewer of them. Both of these factors further contribute to a reduction in efficiency when compared with a clear liquid counterpart. While a clear HEAVY DUTY IMPELLER liquid impeller usually has five to nine vanes, most slurry pump impellers have two to five, with four and five vane designs DECREASED 16) THIN LONG-WRAP being the most common. Two and three WIDTH MAIN PUMPING VANES vane designs are usually reserved for very large particle passing, as required in dredging applications. Figure 2 illustrates the difference HIGH EFFICIENCY IMPELLER between a four vane heavy duty slurry design with front and back expelling vanes and a six vane high efficiency slurry design with smooth front and back shrouds. Note the short “blocky” vanes on the heavy-duty design and the thin long length, long wrap Figure 2: Differences between heavy duty and vanes on the high efficiency design. high efficiency style impellers. Work has been done which shows that on large particle applications, the large solids follow a different path from the fluid due to the inertial effect of the heavy solids (reference 1). This is illustrated in Figure 3. This results in gouging wear at any location where the fluid is required to make an abrupt change in direction. For this reason, on large particle applications, heavy duty designs with blunt leading edges, wide between shroud spacing, and thick back shrouds are recommended to combat the impact of the large particles. Short main pumping vane lengths with minimal vane overlap allow the large particles to pass unimpeded. This heavy-duty geometry results in a head versus capacity curve that is flatter than the typical clear liquid pump. Figure 4 shows a typical performance curve of a four vane heavy-duty impeller for a 10 inch suction, 8 inch discharge horizontal slurry pump. The total combination of fewer, thick, short main pumping vanes, combined with expelling vanes on the front and back shroud can result in slurry pump efficiencies, which are as much as ten percentage points lower than a comparable clear liquid impeller. These differences are minimized on larger pumps. On fine particle service (dss < 100 microns) it has been shown that the particles follow the fluid path (Figure 3). In these instances, high efficiency designs, which minimize the presence of vortices, have been shown to not only improve efficiency, but also improve impeller life. In these fine solid particle applications, the lack of expelling vanes on the front and back shroud has also been shown to improve side liner wear life. Figure 5 illustrates the performance for a four vane high efficiency design, similar to the six vane design illustrated in Figure 2. This impeller is an alternative COARSE PARTICLEI impeller for the 10/8 slurry pump with heavy-duty 4 vane performance illustrated in Figure 4. Note the steeper head versus capacity curve for the high efficiency and lower Figure 3: Fine and coarse particle trajectories in design with longer wrap an impeller. impeller vane exit angle. Also note higher efficiency (8 1% versus 75%). FINE PARTICLES

1376

WEIR SLURRY GROUP

WEIR SLURRY GROUP WPA 10BADJM

WPA lOBAD7hi

ISSUED SEP 1990

JULY 1995

Figure 4: Performance curve for 10/8 slurry pump with heavy-duty 4 vane impeller.

ISSUED

Figure 5: Performance curve for 10/8 slurry pump with high efficiency 4 vane impeller.

Specific Speed Specific speed is a dimensionless number which defines the relative proportions of the impeller, and pump in general, in terms of its hydraulic performance at the best efficiency point (BEP) for any given speed. The equation for specific speed, N,, is:

N = pump speed in rpm, Q = capacity in m3/sec at BEP, H = total head per stage in meters at BEP.

where:

For the heavy duty impeller with performance illustrated on curve WPA108A03M in figure 4, the N, is calculated as:

N,

=

1000rpm~(1600m3 / hr)/(3600sec/hr)

(74my4

= 26.4

Note: 1000 rpm was arbitrarily chosen as the reference speed, but similar values would have been obtained using BEP performance at other speeds. Figure 6 (reference 2) illustrates the relationship between specific speed and impeller geometry. Slurry pumps typically have specific speeds in the range of 15 to 40 for N, derived from terms of m3/sec, m head and rpm. As of the time of this writing, in North America, specific speed is still mainly derived from units of USGPM flow, feet of head and rpm, resulting in slurry pumps with specific speeds in the range of 750 to 2000 N,, using these terms. The conversion from N, in terms of m3lsec and m head to N, in terms of gpm and ft of head is a multiplier of 5 1.67. As the equation for N, incorporates BEP flow and head conditions (in the numerator and denominator, respectively), it is clear that high specific speed pumps are better suited to high flow, low head applications, while low specific speed pumps are better suited to low flow, high head applications. The impeller geometry illustrated in Figure 6 supports this idea, as flow is proportional to inlet area and head is proportional to impeller diameter (squared) for any given speed. Thus, as illustrated, the low specific speed pumps have larger outlet diameter to inlet diameter ratios. Values of Specific Speed, Ns, using m3lsec and meters: 10

20

Radial-Vane Area

D2 ->2 D1

40

30

50

60 70 80 90 100

Francis-Vane Area

_ D2 - 1.5TO2 D1

Mixed-Flow Area

200

300

Axial-FlowArea

D2 -<1.5

D2 _ -1

Di

D1

Figure 6: Impeller geometry and its effect on specific speed (courtesy of Hydraulic Institute).

1378

Maximum efficiency occurs at N, of 42. Nevertheless, from a wear standpoint it may not be desirable select a pump with this high a specific speed, as higher specific speed pumps (with smaller diameter impellers) are required to operate at higher speeds to achieve any given head. Similarly, low specific speed pumps are not always desirable due to their lower efficiency and poor ability to pass large diameter solids. Pumps with N, of 20 to 25 are recommended for severe duty applications (reference 3). For medium to heavy-duty applications an N, of 25 to 33 is recommended. Experience has shown pumps with Ns of 42 can be safely used on medium duty applications with heads of 30 meters or less.

Casing Design The three basic casing design types used in a centrifugal slurry pump are the true volute, near or semi-volute, and the circular volute. These three volute types are illustrated in Figure 7a, 7b, and 7c, respectively (reference 4): The true volute, Figure 7a, has the highest efficiency due to its tight impeller to cutwater clearance. The cutwater is the "V" shaped diverterfflow splitter between volute and discharge. This tight clearance increases efficiency by minimizing recirculating flow at the best efficiency point. At flows below the best efficiency point, however, the increased recirculating flow has a high velocity through the small impeller to cutwater area. This results in high wear at low flows for this design in the area at and just beyond the cutwater. This design also has highest hydraulic radial loading at flows other than the BEP. The near volute, Figure 7b, is similar to the true volute, only the impeller to cutwater clearance has been increased to reduce velocity at the cutwater area at lower flows. This design results in significantly reduced wear at this area at low flows. This design also has significantly lower hydraulic radial force at conditions away from the BEP. For most slurry applications, this design offers a good balance of wear life and efficiency. The circular volute, Figure 7c, has uniform area between the impeller and casing all around the volute. This volute design has the lowest hydraulic radial force at conditions off the BEP. This makes it ideal for high head, low flow applications.

(a)

(b)

Discharge Casing Geometry H R

H R

--

H

f ... I I

R

I I I

I

'BEP

Head (H)and Hydraulic Radial Load (R) characteristics Figure 7: Slurry Pump Casing Types.

1379

'

Normal recommended operating conditions for the three volute types are listed below: True volute: 80 to 120 percent of BEP flow Near volute: 60 to 110 percent of BEP flow Circular volute: 40 to 100 percent of BEP flow For severe duty slurries, one would llke to select a pump with an operating point in the middle of the above ranges to minimize volute wear. For relatively light slurries, the above ranges can be widened.

MATERIALS General The primary materials used for wear components in centrifugal slurry pumps are hard metals, elastomers, and to a lesser extent ceramics. In very simple layman’s terms, hard metals and ceramics combat erosion due to their high hardness values. Elastomers combat erosion by their ability to absorb the energy of the impacting particle(s) due to their resilience and tear resistance. Elastomers generally outperform hard metals, in terms of erosion resistance, in those applications where particle size is smaller than 250 microns, impeller tip speed is within the limits of the elastomer, and there is no risk of large particle “tramp” damage. Metals The three basic types of metals used to combat erosion in centrifugal pumps fall under ASTM A532 (class I, 11, and 111). These are the Martensitic White Irons (class I), the ChromiumMolybdenum White Irons (class II), and the High Chrome Irons (class III). These materials consist of hard carbides within a supporting ferrous matrix. The two types of carbides found within these materials and their approximate Vickers hardness range are listed below: Iron Carbide 850 to 1000 H V Chromium Carbide (Eutectic (Fe,Cr)7C3) 1200 to 1500 HV The matrix types and corresponding hardness ranges are listed below: Ferrite Austenite Martensite

150 to 250 H V 300 to 500 HV 500 to 1000 HV

The bulk hardness of the material is dependent not only upon the carbide and matrix type, but also upon the volume of the carbides within the matrix. For medium to large particle applications the bulk (combined) material hardness is of primary importance. For small particle applications, a fme microstructure, with smaller intercarbide spacing, is more important to minimize erosion of the softer matrix. For very large particle applications (greater than 4 inch diameter) fracture toughness of the matrix is most important. Ni-Hard 1 and Ni-Hard 4 are martensitic white irons, which fall under ASTM A532, class I. They consist of iron carbides in a martensite matrix with some retained austenite. In the as cast form, the material typically has a hardness in the range of 500 to 550 Brine11 (540 to 600 HV). Heat-treatment is used to reduce the retained austenite and increase the matrix hardness and, therefore, bulk hardness, however the erosion resistance is not as good as most of the following metals. The low chromium content (approximately 3%) also provides little corrosion resistance. For these reasons, the martensitic white irons have been largely superceded by the chromiummolybdenum white irons and high chrome irons in heavy duty slurry applications.

1380

The chromium-molybdenumwhite irons (ASTM A532, class 11) and the high chrome irons (ASTM A532, class 111) consist of extremely hard chromium carbides (approximately 20 to 30% by volume) in a martensitic matrix with retained austenite. Heat-treatment of these materials not only increases hardness by reducing the retained austenite through conversion to martensite, but also initiates the precipitatiodformation of secondary (fine) chromium carbides in the matrix. As the chromium carbides are refractory materials, they are generally inert to corrosion from most slurries and the corrosion resistance of these material is generally dependent upon the chromium content and corrosion resistance of the matrix. 15-3 Alloy is a 15% chrorr1ium-3% molybdenum white iron that falls under class II,which can be fully hardened to 750 Brinell. It suitable for use in slurry pumps where high erosion resistance is required and where impact loading conditions and corrosion rates are minor to moderate. 27% high chrome iron falls under class III and can be fully hardened to 650 Brinell. It is suitable for use in slurry pumps where erosion resistance, corrosion resistance and fracture toughness requirements are moderate to high. Many propriety variations of the high chrome iron are available: Lowering the carbon, and raising the chromium results in less of the chromium being removed from the matrix to produce chromium carbides. This lowers the percent by volume of carbides and correspondingly the hardness, but retains additional chromium in the matrix for added corrosion resistance. Increasing the chromium content provides further gains in corrosion resistance. Currently, there are proprietary alloys with 450 Brinell hardness and a duplex (austenite-ferrite) matrix with the corrosion resistance of CD4MCu. Advanced high chrome irons are available with 50 to 75% by volume carbides and hardness above 700 Brinell. Through proprietary techniques, these materials are able to be produced with a relatively fine microstructure, which when combined with the increased volume of carbide, reduces the intercarbide spacing and have shown the ability to achieve two to three times the life of standard high chrome iron. Tougher high chrome irons with austenitic matrix and high impact resistance are available for extremely large particle dredging applications.

Elastomers Typically are used in applications with particle diameters not greater than 10 rnm. Elastomers can be broadly broken into two categories: natural rubber and synthetic elastomers. Based solely on erosion, natural rubber is the clear winner due to its sigmfkantly greater resilience and tear resistance. Resilience is a measure of how high a ball of the material will bounce measured as a percentage of the initial drop height. Typically, this value ranges from 65 percent to 90 percent dependent on the rubber blend. Tear initiation resistance for natural rubber is typically in the range of 30 to 110 N/mm, dependent upon blend. For natural rubber, tear resistance tends to improve with increasing hardness, while resilience tends to decrease with increasing hardness. For fine particles applications (less than 100 microns) resilience has been shown to be more important in combating wear. For larger particles (greater than 500 micron) tear strength (resistance) is more important. As mineral processing slurries have a mixture of particle size, the best performing natural rubber will be one with the optimum combination of resilience for fine particle wear resistance and tear resistance to prevent larger particle damage. Synthetic elastomers are used in small particle applications where natural rubber would be subject to chemical attack, causing swelling, hardening or reversion of the natural rubber. The most commonly used synthetic elastomers for wear materials and some of their typical applications are listed below: Nitrile: Generally used in fats, oils and waxes. Moderate erosion resistance. Limited resistance to acids and alkali environments.

1381

Butyl: Hydrochloric acid, phosphoric acid and sodium hydroxide. Sulphuric acid causes degradation and chlorinated hydrocarbons cause swelling. Hypalon: Primary use in acid conditions with some resistance to vegetable and mineral oils. Not recommended for use in ketones or chlorinated solvents. Neoprene: Moderate resistance to oils, fats, grease and some hydrocarbons. Can also be used in some mild oxidizing acids. Synthetic elastomers also have higher temperature limits than natural rubber. While natural rubber is limited to 75 to 85 degrees C, dependent upon blend, the above synthetic elastomers have temperature limits ranging from 95 degrees C for Nitrile to 110 degrees C for Hypalon. Further comparisons can be made with respect to mechanical properties. The tear resistance for the above synthetic elastomers range from 30 N/mm for Nitrile to 50 N/mm for Neoprene (compared to 30 to 110 N/mm for natural rubber). Neoprene has the highest resilience of the synthetic elastomers at 58 percent (compared to 65 to 90 percent for natural rubber). The superior mechanical properties of natural rubber indicate it should be used in preference to synthetic elastomers, unless temperature andor chemical resistance are overriding factors. Polyurethane is an elastomer that is used in applications where there is a good chance of large particle “tramp” damage, which would otherwise cut natural rubber and synthetic elastomers. It excels at its cut resistance due to its high tear strength (50 to 100 N/mm). Polyurethane is generally limited in temperature to 70 degrees C, due to swelling from hydrolysis attack, although polyurethanes are currently being produced, which reportedly can operate at 110 degrees C with no degradation. Generally speaking, in most small particle applications, where “tramp damage” is not a problem, natural rubber will outperform polyurethane. Tip speed limit Another important factor to consider in the selection of materials is the impeller tip speed limit. For elastomers, the concern is vibration or fibrillation of the material due to the relative motion of the impeller with respect to the side liners. This vibration can lead to heat generation within the elastomer and a thermal breakdown of the material. For natural rubber this often results in the elastomer reverting back to its natural “gummy” state. Tip speed problems on elastomers are easy to diagnose, as the damage is always most severe at areas with highest relative speed. For elastomer impellers this would be the periphery of the impeller and for elastomer side liners, this would be the area adjacent to the periphery of the impeller. Generally speaking, the tip speed limit is a function of the hardness of the elastomer and its ability to dissipate heat. Typically, natural rubber has a tip speed limit of approximately 27.5 d s e c . Highly wear resistant soft natural rubber can have a tip speed limit as low as 25 mfsec, while proprietary blends with improved thermal conductivity can operate at 30 d s e c . For a centrifugal slurry pump this tip speed limit is important, as the head generated (meters) is approximately equal to the quantity (0.5 times the square of tip speed in meters per second) divided by the acceleration due to gravity (9.8 dsec’). Approximate speed limits and corresponding approximate BEP head limits for various materials are listed below: Highly wear resistant soft natural rubber Typical natural rubber Anti-thermal breakdown rubber

25.0 d s e c 27.5 d s e c 30.0 d s e c

32 meters head 39 meters head 46 meters head

Nitrile Butyl Hypalon Neoprene

27.0 d s e c 30.0 d s e c 30.0 d s e c 27.5 d s e c

37meters head 46 meters head 46 meters head 39 meters head

Polyurethane

30.0 d s e c

46 meters head

Hard metal (impellers)

38.0 d s e c

74 meters head

1382

For the hard metal impellers, the 38 d s e c (7500 ft/min) tip speed limit is based on the limited ductility of the material and not thermal breakdown.

PUMPWEAR In general, wear in a centrifugal pump will consist of various modes of abrasion and erosion. Abrasion, which is the forcing of hard particles against a (wear) surface, only takes place within a slurry pump on the shaft sleeve and between the tight tolerance wear ring section of the impeller and the suction side liner. Erosion is more commonly used to describe the progressive wear loss from the interaction or impingement of the fluid and particles against the wear components. The three primary modes of erosion and their wear locations within a centrifugal slurry pump are described below: Deformation wear:

Direct impact to the leading edge of the impeller vanes, the back shroud of the impeller and the "protruding" cutwater within the volute.

Random impingement:

Random impacts to the impeller shroud and trailing edge of the main pumping vanes.

Low angle impact:

Wear from the tangential or near tangential movement of particles against the volute casing or vane surface.

Of these modes of erosion, deformation wear (direct impact) is the most severe and low angle impact is the least severe. The degree of wear is dependent upon the following: The kinetic energy of the particle: particle mass (specific gravity) and velocity. The particle shape: sharp particles have small contact area and high local stress, so wear is more severe than with rounded particles. The slurry concentration: higher percentages of solids result in more impacts for a given flow. Based on the above, slurries can be classified according to the following criteria (reference): Heavy duty: Cw>35%,dp5>400pm, SGh2.0, sharp particles. Medium duty: 20%1.4, angular particles. Light duty: Cw<20%,dp5<150pm, SG,>1.4, rounded particles. In Addition to the general specific speed limits and the material tip speed limits discussed previously, to maximize wear life, the following general impeller tip speed limits can be applied based on the severity of the duty: Heavy duty: Medium duty: Light duty:

25 d s e c max. 32 meters head @ BEP 32 d s e c max. 52 meters head @ BEP 38 d s e c max. 74 meters head @ BEP

Further recommendations can be made with respect to impeller type and flow range: Heavy duty: Medium duty: Light duty:

heavy duty impeller @ 0.60 to .80 BEP heavy duty impeller @ .70 to .90 BEP high efficiency impeller @ .80 to 1.1 BEP

1383

CAVITATION: Cavitation is another cause of wear, which should be briefly mentioned. It is both system and pump design dependent. It is due to the local velocities at the impeller inlet reducing the suction pressure, which is system dependent, to a level below the vapor pressure of the liquid at the temperature being pumped and causing the fluid to boil. The increase in pressure as the fluid flows through the impeller results in a collapse and implosion of the vapor bubble(s) resulting in localized wear.

A SUCTION SUPPLY OPEN TO ATMOSPHERE

B

WITH SUCTION LIFT

SUCTION SUPPLY OPEN TO ATMOSPHERE WITH SUCTION HEAD

HS

NPSHA = pa+ H,-(V

I

; CLOSED SUCTION SUPPLY WITH SUCTION LIFT

I

I

P

+ hf)

D CLOSED SUCTION SUPPLY WITH SUCTION HEAD

.-

-

NPSHA = PT+H,-(V

P

+hf)

I t

1.4pa = Barometric pressure in meters absolute. Vp = Vapor pressure of the liquid at the pumping temperature in meters absolute.

PT = Pressure on the liquid surface inside a closed tank in meters absolute.

HI,= Static suction lift and head in meters absolute.

Hs hf = Friction loss in suction pipe in meters. Figure 8: Calculation of NPSHA for various suction conditions.

1384

DeDending on the severitv of the cavitation. high noise and vibration mav be present and a Whether or not cavitation occurs is dependent upon the suction characteristics of the pump and the system in which it is placed. The suction characteristics of a pump are expressed using the term net positive suction head required, or NPSHR. The suction characteristics of the system are expressed using the term net positive suction head available, or NPSHA. Even though cavitation is a pressure phenomenon, as friction in the pipeline and head output from the pump are expressed in terms of meters of head, so are NPSHR and NPSHA. For cavitation not to occur the NPSHA should exceed the NPSHR by 15 to 30 percent. Values for NPSHR for a particular pump are a function of the pump design and are read off the manufacturer's performance curve at the duty point. Generally, N P S H R increases with flow and pump speed, although NPSHR can often rise at extremely low flows near pump shut off. Values of NPSHA for the system are not simply the height of liquid above the centerline of the pump. Instead, NPSHA is a function of the local atmospheric pressure, the height of the liquid relative to the pump centerline, the friction losses in the suction pipe and the vapor pressure of the liquid at the pumping temperature. Means of calculating NPSHA for various suction conditions are illustrated in Figure 8 on the previous page (reference 4). Under suction lift conditions as shown in Figure 8a, it is recommended the atmospheric pressure (pa) be reduced by dividing by the specific gravity of the liquid. This is to account for the weight of the liquid in the column and the negative effect this has on the suction pressure at the pump. It is not recommended that any positive correction be applied under flooded conditions. The reader is encouraged to review references (2), (4)(5) and ( 6 ) for more mformation on NF'SHR and NPSHA. Cavitation can easily be prevented during the system design process by insuring there is adequate suction height, minimal suction friction and proper pump selection. After system construction, these factors are far more difficult to change.

HYDRAULICS: Because the success or failure of a centrifugal slurry pump is highly dependent upon its operating point, it is important to have a clear understanding of the relationship between the pump performance curve and the system curve. As seen in Figure 4 and Figure 5 , centrifugal slurry pumps typically have slowly drooping head versus capacity curves, with the highest head for a given speed being produced at the lowest flow. Changes in speed result in the generation of a new head versus capacity curve, which follows a set of rules called the affinity laws, These rules indicate that the relative location of points of equal efficiency will occur at a capacity equal to the original capacity multiplied by the ratio of speed change and at a head equal to the original head multiplied by the ratio of the speed change squared. Within reasonable speed changes, this useful tool allows one to plot the characteristic pump performance curve at a multitude of speeds given a curve for one reference speed. Nowadays, most Centrifugal slurry pump manufacturers plot performance at a wide range of speeds (as shown in Figure 4 and figure5), so it is very easy to estimate the characteristic pump performance curve at a given speed without the use of these rules. It is important to note that the point of operation (flow and head), is system dependent and does not necessarily follow the points of equal efficiency predicted by the affinity laws, as will be shown. The system curve is a graphical representation of the piping systems resistance to flow. It consists of a fixed component called the static head and a variable component called the friction head. In systems that are open to atmosphere, the static head is the elevation of the discharge pipe (or discharge liquid level for submerged discharge), minus the elevation of the liquid in the suction tank. In systems where the suction or discharge are pressurized or under vacuum these conditions must be accounted for after being converted to appropriate head values. The friction head is the resistance to flow through the piping and fittings, which varies with the flow rate. The reader is encouraged to refer to (2), (4), (5) and ( 6 ) for more information on system curves.

1385

The point of operation for a centrifugal pump is the intersection of its head versus capacity curve with the system curve. Figure 9 illustrates a typical system curve with 10 m of static head and friction that increases exponentially with flow. Also shown are pump head versus capacity curves for the heavy duty impeller illustrated in Figure 4 at 600 and 800 rpm. For each speed, the operating point OCCUTS at a flow where the head output fiom the pump matches the total resistance in the system at the same flow. As stated above, this is the intersection of the head versus capacity curve and the system curve. For this pump in this hypothetical system, the expected operating points are 1000 m 3 h @ 26 m head at 600 rpm and 1400 m3hr @ 44.5 m head at 800 rpm.

Effect of Pump Speed Variations 70.0

I

60.0 50.0

+800

rpm pump performance

-600

rpm pump performance

A

E 40.0 a 8 30.0

Y

=

20.0

i

10.0

+System curve with 10 m static head

0.0 500

0

1000

1500

2000

Flow (m3/hr)

Figure 9: Effect of speed on point of operation. Figure 10 illustrates the effect of a change in static head for the same system. In this instance we have increased the static head fiom 10 m to 25 m. Note the system curve shifts directly up by the increase in static head (15 m). The result is the intersection of the system curve with the 600 rpm and 800 rpm performance curves occurs at lower flows, reducing pump output. This effect is observed when there is a drop in suction level or an increase in elevation, such as an increasingly tall tailings dam.

Effect of Changes in Static Head 70.0

60.0 50.0

T

E 40.0 m8 30.0 v

=

20.0 10.0 0.0

-800

rpm pump performance

+600

rpm pump performance

7 .

+System curve with 10 m static head

I 1 1 -System curve with 25 m static head

0

500

1000

1500

2000

Flow (m3/hr)

Figure 10: Effect of changes in static head on point of operation.

1386

Next Page

Figure 11 illustrates the effect of a change in friction. In this case we have increased the friction, which results in a steeper system curve and correspondingly lower output from the pump at both 600 and 800 rpm. Examples of this effect are the partial closing of discharge valve (not recommended in slurries), a change to smaller diameter pipe, a lengthening of the discharge pipe or increase in friction due to an increase in concentration of the slurry. The opposite effect of that shown in figure 11 is a decrease in friction, which results in a shallower system curve and increased pump output. This condition usually occurs when an engineer, with good intentions, but poor judgement, has added a factor of safety to his friction calculations.

Effect of Changes in Pipe Friction 70.0 1 -

-m- 800 rpm pump

60.0

50.0 E 40.0 TI 2 30.0 20.0

performance +600 rpm pump performance +System curve with I 0 m static head

=

10.0 0.0

, +System curve with increased friction

0

500

1000

1500

2000

Flow (m3/hr)

Figure 11: Effect of changes in pipe friction on point of operation. Figure 12 illustrates the effect of having two identical pumps at the same speed (in this case 600 rpm) operating in series. The net effect is a summation of the individual heads of each pump at a given flow. This results in a steeper combined pump head versus capacity curve and increased head at any given flow. This, of course, results in an intersection with the system curve at a higher flow and hgher flow output. Typically, the use of two or more pumps in series is not to increase flow output per se, but instead to overcome a large amount of friction, such as in a long tailings line or to improve wear life by lowering pump speed.

Effect of 2 Pumps in Series 70 , +600

rpm two pumps in series

+600

rpm one pump performance

+System curve with 10 m static head

0

500

1000

1500

2000

Flow (m3/hr)

Figure 12: Effect of 2 pumps in series on point of operation.

1387

Selection and Sizing of Slurry Lines, Pumpboxes and Launders Baha Abulnaga, Maza'ak International Inc., S u m s , Washington, USA Ken Major and Peter Wells, HATCH Associates Ltd., Vancouver, BC, Canada

ABSTRACT A significant component in the design of a mineral processing plant to ensure an efficient transition from concept (engineering) to reality (plant operation) is materials handling, the moving of ore through the different unit operations. The design of slurry systems using pipelines and launders needs to consider a number of different variables including slurry rheology, density, viscosity and particle size. In this paper the different regimes of slurry flows are reviewed outlining methodology for sizing full flow pipes, launders and upcomers. With the advent of modern personal computers, it is possible to size up slurry lines more precisely using specialty programs. INTRODUCTION Mineral processing is the combination of the many unit operations that are required to produce a marketable product from an ore. Although the coarse size reduction steps (crushing) are dry, in most project flowsheets the recovery steps are in slurry form. In mineral processing, slurries are a mixture of water and/or chemical solution with ground rock. A simplified process flowsheet will have four or five unit operations. The design of the slurry handling system is very important to the efficient operation of the process because it provides the links between these unit operations. A problem with surging flow in the first unit process, typically grinding, has the potential to impact the behaviour of the downstream processes. SIZING OF PIPING SYSTEMS A mineral processing plant typically involves a large number of unit operations. Each of these different operations may have specific slurry system requirements depending on the slurry characteristics. The piping system must be designed to ensure that the opportunity for sanding and plugging of the pipeline is minimized while at the same time ensuring that the line velocities to prevent this from happening do not lead to excessive wear conditions of the pipeline. To maintain suspension of a particle in the slurry the velocity of the slurry in the pipeline must be greater than the critical velocity. The main design factors to consider when selecting the pipe size are:

0

Particle size Slurry density Viscosity Flowrate Friction losses

1403

Particle size in a mineral processing plant will vary based on the unit operation and process criteria. For example, grinding circuits are typically characterized with large particles (+19 mm SAG grinding, +2mm rod mill, ball mill) and high densities (45% solids to 60% solids). A flotation concentrate stream after multiple regrind stages may have a Pgoof 10 microns. As the slurry density increases andlor the particles become finer the hindered settling velocity in the slurry will increase making it easier to keep the larger particles suspended. The finer particle size will also increase the apparent viscosity of the slurry. This is observed in many of the Nevada gold mines with high clay contents in the ore. For all the accuracy that can be applied to pipeline sizing through calculations to determine settling rates, slurry velocity and pipe diameters the process becomes flawed when operation needs are defined. Many of the ore deposits have multiple ore types with varying processing requirements. In low grade copper and gold mills economics are predominated by the need to maximize throughput. With a change in ore from hard to soft, relative to SAG grinding conditions, throughputs can change radically. Doubling the feed grade in a flotation plant will result in doubling the concentrate production.

Figure 1: Computer representation of a SAG mill circuit showing both pumped and open channel flows. A typical plant layout in a SAG mill circuit, Figure 1, may involve pumped flow such as from the discharge of the SAG mill to the hydrocyclones or gravity flow in launders from the underflow of the cyclones to ball mills. The location of hydrocyclones high in a plant, may also involve gravity flow to downstream processes. It is important to appreciate these two different kinds of flows.

SLURRY PIPE FLOWS Slurry flows are classified as: heterogeneous (settling flows). homogeneous (non-settling) flows.

1404

Heterogeneous flows with typical particle size, d5&4 microns constitute the majority of circuits encountered in plants. The flow is characteristically Newtonian. There are however instances where clays, very fine grinding of the ore leads to non-Newtonian flows particularly in certain bauxite (red mud, or kaolin), and even gold-copper associated with clays at weight concentration in excess of 40% There are essentially four main regimes of Newtonian flow in a horizontal pipe: Flow with a stationary bed. Flow with I a moving bed and saltation (with or without suspension). Heterogeneous mixture with all solids in suspension. Pseudo-homogeneous or homogeneous mixtures with all solids in suspension.

Mean velocity

c

Figure 2: Regimes of Newtonian flows for slurry mixtures in a horizontal pipe"' Transitional Velocities These four regimes of flow can be represented by a plot of the pressure gradient versus the average speed of the mixture as illustrated in Figure 2. The transitional velocities are defined as:

(')

v1

=

v2

=

V3 orVD

=

v4

=

the velocity at or above which the bed in the lower half of the pipe is stationary. In the upper half of the pipe some solids may move by saltation or suspension. velocity at or above which the mixture flows as an asymmetric mixture with the coarser particles forming a moving bed. velocity at or above which all particles move as an asymmetric suspension and below which the solids start to settle and form a moving bed. velocity at or above which all solids move as a symmetric suspension.

Abulnaga B.E. 2002 - Sluny Systems Handbook - McGraw-Hill

1405

of flow

Figure 3: Regimes of velocity for Newtonian flowsof settling slurries in horizontal pipes.”’ The deposition velocity VDor V3 is usually established by the Durand equation VD

FL

= v3 = =

v3

=

Di

= = = =

g Ps

PL

FL

(2

* g * Di [(ps - PLYPLJ 1”

(EQ 1)

is the Durand factor based on grain size and volume concentration the critical transition velocity between flow with a stationary bed and a heterogeneous flow. pipe inner diameter (m) acceleration due to gravity (9.81 d s ) density of solids in a mixture (kg/m3) density of liquid carrier (kg/m3)

The Durand factor FDis typically represented in a graph for single or narrow graded particles after the work of Durand (1953). However, since most slurries are a mixtures of particles of different sizes, this plot is considered to be too conservative. The Durand velocity factor has been refined by a number of authors. Schiller (1991) proposed the following equation for the Durand velocity factor based on the d5o of the particles FL = {( 1.3 x Cy0.125 )( 1 - exp ( -6.9 dso))}

(EQ 2)

Equation 2 can be solved using a simple computer program, The Durand factor F L is related to the Froude Number by the following equation: Fr = FL* 4 2.

1406

2.

Based on Schiller

h

,o

ZR .G 0

c4

1.

2

9

z

-0

R

0 0

1.o

2.0

3.0

Particle diameter (mm)

Figure 4: Comparison of the conventional Durand factor for single graded slurries and the factor using Schiller equation.’’’ The Schiller’s equation is valid for viscosities of the order of 1 cP. For settling mixtures with higher viscosities. To estimate the deposition velocity V3, Gilles et al (1999) developed an equation for the Froude Number based on the Archimedean number Fr=aArb

(EQ 3)

p~ = viscosity in Pa-s d,

= particle diameter in meters

p~ = density of liquid in kglm3 ps = density of solids in kg/m3 g = acceleration due to gravity in mls2 Fr is non-dimensional For An540, a= 1.78, b= -0.019 For 160cAr-540, a= 1.19, b= 0.045 For 80
This correlation is useful in the range of:

1407

The Wislon-Judge method requires a computer program due to the various ranges of Archimedean numbers. To determine the drag coefficient the actual density of the liquid should be used while the viscosity should be corrected for the presence of fines and for volumetric concentration (Figure 5). For volume correction to the viscosity, Thomas (1965) proposed the following equation with an exponential function:

clrn =1+K1Cv+K2C,2+Aexp{BCv) PL

K1 is the Einstein constant of 2.5 10.05 KZ A 0.00273 B 16.6 CV = volumetric concentration The magnitude of K2,A and I),may however changer for very fine particles:

100

90

E a

aa 3.

a -

80

\E

70

50

0

i

2

30

c

.z

8

20

.->

0 O c

.-

2

v)

>

lo

.

0 0

.

I

10

1

-

. . . . . . 20 30 40 50

.

. . . . . . . . 60 70 80 90

.

CV Volumetric concentrationof solids

Figure 5: effect of volumetric concentration on viscosity of a slurry mixture (after Thomas 1965) From the calculated deposition velocity, the pipe size can be determined based on the flow rate. With the short piping runs generally found in a mill operation there are some simplified guidelines that can be adopted for selecting pipe diameters: Coarse particle applications Medium particle sizing Fine particle sizing

SAG mill discharge Leach I Flotation Feed Concentrates

3.5 to 5 m/s (12 to 16 ft/sec) 2.5 to 3.5 m l s (8 to 12 ft/sec) 1.5 to 2.5 m / s (4 to 8 ft/sec)

The selection of the pipeline diameter should also consider the slurry density, with higher velocities used at the lower densities, and slurry viscosity. The selection of a pipe diameter that results in a line velocity higher than required will result in an increased wear rate in the pipe and an increase in operating costs. The smaller pipe diameter will also result in a higher friction loss per unit length that will increase the power requirements for the pump and the unit operating costs.

1408

The selection of a pipe that is too large for the projected flow regime will result in settling in the line and the creation of a dead zone. The dead zone will continue to form a deeper bed in a horizontal line until the apparent line velocity is sufficient to keep the particles suspended in the slurry or moving through the pipe by saltation. The formation of the dead bed will create an irregular shaped pipe with a very rough surface factor increasing the unit friction loss and power requirements for the pump. In addition the formation of a bed will result in two contact lines along the wall of the pipe that will be exposed to an increased wear rate from the sliding bed. An oversized pipe on a vertical pipe will create a different set of issues. Coarse particles that cannot be suspended in the nominal slurry conditions will be in transition between settling and suspension until the apparent density is sufficient that the solids are carried as a high density “slug” to the next unit operation or until enough mass builds up in the horizontal to vertical transition to plug the line. A typical example would be a SAG mill discharge application. The size of the coarsest particle in the SAG mill discharge pumping system (combined SAG mill / ball mill cyclone feed) will be dependent on the mill grates and the discharge screen openings. Normal maximum particle size is 19 mm (3/4”) with coarser particles experienced with grate and/or screen failures. The slurry stream will also contain some loading of steel chips from the grinding media. These ball chips will have a higher specific gravity and irregular shapes and have a greater tendency to sink rather than be carried by the slurry. A low velocity line that plugs requires a significant effort and mill downtime to clean out. A low velocity line that tends to “slug” can result in plugging problems at the cyclones and/or create surging in downstream operations.

COMPOUND MIXTURES It is now an accepted fact that mineral slurry typically consists of a mixture of coarse and fine solids. The coarse flow at the lower bottom of a horizontal pipe while the finer particles flow above the bed. Wasp et a1 (1977) proposed therefore that the pressure loss for each layer be computed. The concentration of particles of a certain size are established in reference to a layer “a” which is typically 8% of the pipe diameter. The Wasp method was derived from extensive research by various authors on coal slurries. The Wasp method requires repetitive iterations. A suitable computer program for such a method is presented by Abulnaga (2002). The Wasp method is limited to slurries with d5d44 microns. For finer slurries, non-Newtonian models should be used. Various models are however based on a bi-modal distribution, meaning fine and coarse. A review of these models was presented by A.R Khan and J.F. Richardson 1996. Equations are then developed for what is essentially two layers, a bottom layer of coarse particles and an upper layer of fine particles. The two-layer model has gained acceptance in the oil sand industry and computer models are available from appropriate research labs such as the Saskatchewan Research Institute in Canada. The two-layer method is limited to slurries with d50>74 microns. Certain grades of oil sands have been found to yield d5674 microns. The plant design engineer should be aware of the limitations of both methods. In many instances the routing of the pipe is short and involves essentially static rise or drop, with minimal friction losses when the pipe is properly sized. Engineers have been able to use simple formulae based on a Hazen-Williams factor or a Zandi factor. For long in-plant piping or pipelines more correct methods are required. Models for non-Newtonian flows have been developed by various authors such as Darby, Heywood, Torrance, Wilson and Thomas, Slatter. LAUNDER SIZING In process plants where the lay out is conducive to utilizing gravity transportation of slurries open launders and gravity pipes are often used. The design of open launders and gravity pipes has traditionally been based on empirical formulas, The well-known Manning formula is usually used for designing launder systems however it determines the slope and configuration as if it were transporting water. Recently the Graf-Acaroglu relationship to size open launders as a function of density, particle size, hydraulic radius and volumetric concentrate has been used.

1409

Open launders are usually utilized where access to the slurry is require ie for visual inspection where reagents are added. The slope of the launder is critical as too much slope increases the equipment elevation unnecessarily and too little slope results in spillage. Launders and pipes should be sized to run half full to alleviate the problems associated with surges in plant throughput. By definition an open channelllaunder is not full. The hydraulic diameter is the defined as the equivalent diameter of flow for an open channel. The hydraulic radius is defined as the ratio of the area of the flow by the wetted perimeter. It is also called in certain European books the hydraulic mean depth. RH=

A P RH

= = =

A P

(EQ 7)

area in m2 perimeter in meter hydraulic radius in meter

The Manning number is correlated to the hydraulic radius RH,and to the Fanning friction factor for flow in a launder by the following equation. n = RH116

14

n = manning roughness number g = acceleration due to gravity or 9.8 m/s2 fN = fanning friction factor if the Darcy friction fD factor is used instead of the fanning friction factor, the Manning roughness number is expressed as

For a fully developed and uniform flow, the slope or energy gradient of an open launder is established in terms of the head loss per unit of length (Henderson 1990).

The slope S is expressed in decimals.

1410

Figure (6) shows typical values of the hydraulic radius.

8 -

Figure 6 Hydraulic Radius of open launders Table 1:Typical values for the Manning’s Number “n” for water flows (do not use for slurries) Channel Surface Manning factor “n” Manning factor “n” in ,-1n 5-1 in ft“” S -1 Glass, plastic, machined metal surface Smooth steel surface Sawn timber, joints uneven Corrugated metal Smooth concrete Cement plaster Concrete culvert (with connection) Glazed brick Concrete, tifiber forms, unfinished Untreated gunite Brickwork or dressed masonry Rubble set in cement Earth excavation, clean, no weeds Earth, some stones and weeds Natural stream bed, clean and straight Smooth rock cuts Channels not maintained Winding natural channels with pools and shoals Very weedy, winding and overgrown natural rivers Clean alluvial channels with sediments

0.01 1 0.008 0.014 0.016 0.0074 0.01 1

0.009 0.009 0.014 0.015 - 0.017 0.014 0.017 0.020 0.025 0.020 0.024 0.034 0.067 0.033 0.040 0.075 0.150 0.031 (d75)’16

-

using d75 size in feet After Manning R( 1895) and Henderson (1990)

1411

0.016 0.012 0.02 1 0.024 0.01 1 0.016 0.013 0.013 0.0208 0.022 - 0.0252 0.0208 0.0252 0.022 0.037 0.030 0.035 0.050 - 0.1 0.049 - 0.059 0.111 -0.223 0.0561 (d75)’16

using d75 size in m

Designing Launders for Slurry The presence of solids accentuate the slope of the launder as determined by the size of the particle. Tournier and Judd (1945) reported that the specific gravity of the ore is an important factor to consider. Heavier ores require more slope to be transported in an open channel; as shown in Figure 7. h

E al r

a e

3

* 0 al

a 0

v1

0

20 40 60 80 Weight concentration (%)

Figure 7: Launder Slope as a function of specific gravity and solids concentration.'" Tournier and Judd (1 945) reported that the size of the particles play an important role, and larger particles require more slope as shown in Figure 8.

* 0

particle size (mm)

Figure 8: Launder slope as a function of weight The first step when designing a launder for slurry is to determine the deposition velocity for solids. Dominguez et al. (1996) published an equation based on experimental data measured at Codelco and at the Chilean Research Center of Mining and Metallurgy. For cases where the viscosity effects are negligible:

VD= 1.833 W ~ R (Ps H - P m Y P m 1" ( ~ s ~ / R H ) ~ " ~ ~ VD g ps pm ds5 RH

= deposition velocity in m/s = acceleration due to gravity or 9.8 mls2 = density of solids (kg/m3) = density of mixture (kg/m3) = 85% passage diameter of solids in m = hydraulic radius in m

1412

(EQ 10)

However, in cases where the viscosity of the carrier liquid is instrumental, such as with alkaline water, Dominguez et al. (1996) derived the following equation: 1.2 (3'100n) VD= 1.833 [8& (ps - p,)/p, 3" (dg5/R~)~.~" J

= RH(gRH)"hn

VD g ps pm dgs RH p,,,

= deposition velocity in m/s = acceleration due to gravity or 9.8 m/s2 = density of solids (kg/m3) = density of mixture (kg/m3) = 85% passage diameter of solids in m = hydraulic radius in m = the absolute viscosity of the mixture in Pa-s

(EQ 11)

These two equations clearly indicate that the deposition velocity is a function of the hydraulic radius, the density as well as the particle diameter. It is a far cry from the Manning based equations. The characteristics of the slurry flow in open launders depends on the Froude Number defined as: Fr=V At low Froude number (Fr
Uf = friction velocity 2,

= shear stress at the wall in Pa

p S g

= density of liquid in kg/m3 = slope of the launder in decimals = acceleration due to gravity in m / s 2

By assuming that the absolute roughness of the bed is equal to the particle diameter, Acaroglu and Graf (1968) proceeded to define the shear intensity parameter as:

1413

The power consumed with friction or head losses in the open channel is expressed in terms of the energy slope (head loss per unit length) and a non-dimensional transport parameter is derived as:

By examining data from various authors and by regression analysis, Graf and Acarogla extrapolated the following relationshi . (PA=

10.39

(

YA

)

-2%

(EQ 16)

or

Cv = volumetric concentration of solids in decimals Uav = average velocity in the launder in m/s RH =hydraulic radius in m S = slope in decimals d, = average particle diameter in m g = acceleration due to gravity or 9.8 m/s2 ps = density of solids (kg/m3) p~ = density of liquid (kg/m3) This equation was obtained for finely graded sand with a particle diameter between 0.091 mm and 2.70 111111. (0.0036 - 0.1063 in) and was studied in rivers and open channel flumes. This equation applied to both closed conduits and open channels as Graf (1971) explained. It is particularly well suited for saltation flow in closed channels.

1o2

10

1

10''

Figure 9: The Graf-Acaroglu relationship to size open launders as a function of density, particle sue, hydraulic radius, volumetric concentration - adapted from Graf and Acaroglu (1968)

1414

Non-Newtonian gravity flows are encountered with concentrate pipelines and with thickener underflow. Abulnaga (2002) discussed the methodology for design of such launders.

GRAVITY FLOW PIPELINES During development of the plant equipment layouts it is important for designers to look for opportunities to use gravity to transport slurry between unit operations or between stages. The ability to use gravity eliminates unnecessary pump installations and reduces power requirements. Slurry transportation using gravity will typically use launders (flotation concentrates) or pipes (tank interconnectionsfor leach and CIL tank trains). Sizing of gravity flow pipelines for slurry follows a similar approach to that for pressurized lines. Determining and maintaining a slurry velocity is important to prevent plugging of the pipeline and some design allowance must be provided for varying flowrates. With the design of gravity flow pipelines it is important to remember that for a given flow and pipeline slope the line velocity in a ?hfull pipeline is the same as for a full pipeline as a result of friction losses based on the slurry volume flow rate and wetted perimeter. For a given flow maximum velocity is through a % full pipeline. In selecting a pipeline the diameter can be selected to allow capacity to increase by about 100%. From the Manning equation for open channel flows of water: Qf = 0.463*d8"*S'n/n Qf d S n

(EQ 18)

= full flow volume, cubic feet per second

= pipe diameter in feet = energy loss, ft per ft of conduit length, approximately the slope of the conduit invert = pipe roughness (Manning number)

Vf = 0.590*d**S'n/n

(EQ 19)

Vf = full flow volume, feet per second

From the process design criteria; nominal plant throughput, slurry density and the full volume flow rate is known. Utilizing the target pipeline velocities discussed previously it is possible to solve EQ (18) and EQ (19) to provide a preliminary sizing for pipe diameter, d and the slope, S.

UPCOMER DESIGN The development of leaching to recover gold and carbon-in-pulp (CIP) and carbon-in-leach circuits (CIL) the design of gravity flow systems between tanks became very important. The movement of slurry in and out of the tanks to ensure good mixing conditions and to minimize short circuiting was also identified as a key area of piping design. Upcomers are provided to allow the slurry to be collected near the bottom of the tank, drawn to the top for feeding to the top of the next successive tank, typical of a leach tank operation. For CIP and CIL circuits the use of an upcomer or downcomer is dependent on the type of in-tank carbon screen and the plant layout. The screens are mounted at the surface of the tanks and it is desirable to direct the slurry discharging through the screen near the agitator blades in the subsequent tank to effect the best mixing. Downcomer design must ensure that there is no restriction to the flow, as this can create a back up in the tanks preventing maximum throughput opportunities. Oversizing a downcomer is not critical. The downcomer shouldn't extend below the bottom agitator blades. Critical to downcomer operation is maintaining a consistent slurry density. A lower density slurry discharging to the top of an agitated tank will be distributed in to the tank slurry quickly. A lower density slurry delivered to a tank through the downcomer must have sufficient head to overcome the difference in density or a back up will occur in the tank train with the possibility of overflowing tanks. Upcomer design considerations are more specific. It is desirable not to oversize the upcomer because the upflow velocity will be too low to carry the larger particles. Underdesigning the

1415

upcomer could lead to flow restrictions at higher throughputs. Recommendations from agitator suppliers suggested an upcomer flow velocity of 6 times the hindered settling velocity. The hindered settling velocity, as before, is a function of particle size and density. Upcomer design should be evaluated for nominal conditions. A check can be easily made for upset conditions that will be a result of low density, coarser grind and reduced throughput. Determining the terminal settling velocity'2': V=

(2gr2)(dl-d2)

(EQ 20)

91-1g = acceleration of gravity cm/sec2 r = radius of particle cm d, = density of particle g/cm3 d2 = density of medium g/cm3 p = viscosity of medium dyne sec/cm2

Hindered settling velocity:

u, = v (1 - CV)"

(EQ 21)

Cv = Volume percentage of solids n =4.65 Upcomer velocity: V, = U, x 6

PUMPBOX DESIGN The pumpbox is an integral component of the pumping system and its design is critical for the successful pumping of slurries. In determining a pumpbox size it is normal practice to keep the pumpbox height to a minimum whilst allowing sufficient volume for fluctuations in flow and sufficient retention time. The retention time is normally set at one minute to allow sufficient time for entrained air to escape. Plant layouts sometimes make it difficult to achieve this especially when sloped pumpboxes are used. Plant capacity will also limit retention time. It may not be practical to install a pumpbox with sufficient capacity to achieve the desired retention time. Pumpbox level sensors controlling pump speed and/or water addition will be important for minimizing process flow surges and reducing spills.

Figure 10: Typical Rectangular Pumpbox with a sloped bottoms ~~~

(2) Chemical Engineers'

Handbook, Peny Chilton

1416

A typical rectangular pumpbox with a sloped bottom is shown in Figure 10. The sloping of the pumpbox forces the solids into the pump before they can settle out. If the solids are allowed to settle periodic sliding of the settled material into the pump suction can occur. When this type of sliding of coarse material into the pump suction occurs the pump will operate erratically and can choke. The design of pumpboxes for froth applications requires particular attention due to the blinding effect of froth on the pump and subsequent loss of pumping action. Standpipes are commonly used where the plant layout is suitable this allows a crushing effect on the froth and prevents it from being drawn into the pump suction. Froth factors are usually plant specific and normally range from: 1.1 to 1.5 - Brittle froth 1.5 to 2 .O - Tenacious froth

In certain plants froth factors of up to five have been used. A common error in pumpbox design is to have the pipe delivering the slurry discharge to close to the suction of the operating pump. This will result in air being drawn into the pump suction adversely affecting the pump performance. When a pumpbox has a standby pump consideration must be given to prevent or minimize a build up of solids in front of the standby pump when it is not in operation. A quick release dump valve on the pump suction and/or pumpbox should also be installed. Where varying flows are expected the pumps should be fitted with VFD’s unless the pumpbox has sufficient volume to cope with flow variations. Where abrasive materials are encountered pumpboxes should be rubberlined. It is difficult to effect repairs to a rubberlined pumpbox. The main wear area will be around the discharge nozzle. This should be designed using sacrificial inserts.

PLANT PIPING LAYOUT The routing of slurry piping will have a direct influence on the successful operation of a slurry transportation system. Pipe routing should be kept as straight as possible as every bend, elbow and ‘T’ piece results in a potential area of wear and subsequent failure. Methods of reducing pipe wear is to rotate straight lengths of pipe, utilize long radius bends and hoses. Short radius bends, tees and elbows should be avoided where possible. In high tonnage (high flowrate) plants the application of 3D or greater bends is often impractical because of support locations or space limitations. The use of “T’s” or fabricated wear boxes provide effective solutions. The routing and layout of large diameter (+150 mm) process plant piping should be taken into consideration in the early stages of design. Priority routing should be provided to the main process slurry lines because of the potential production impacts resulting from sanded lines or line failures. Wherever possible pipe runs should be designed in the horizontal and vertical planes. This facilitates design of the pipe supports. In addition, pipelines installed at an angle will wear out faster as a result of the sliding bed that will form on the bottom. Horizontal pipelines should be installed with a 2” to 3” slope to promote drainage from the line during shutdown. This slope is not expected to drain the solids. Some of the solids will settle to the bottom of the line but will be picked back up once the system is charged and pumping slurry again. The pipelines should be sloped back to the pump and/or to the discharge point, if opened to the atmosphere. Drains, or break connections, should be installed near potential plug points, for example, the horizontal to vertical transition on a cyclone feed line. Plant piping design should minimize valves in slurry applications. Also pump suction lengths should be kept as short as practical. Long suction lines or obstructions (screen, valves) on the line can have significant negative impacts on the operation of the pumping and piping system. The use of 3D models assists greatly in preventing clearance, maintenance and pipe support problems. For gold plants incorporating carbon absorption technology, carbon transportation will be either carbodwater or carbodprocess slurry. Piping systems must be designed to minimize

1417

carbon degradation as this will impact gold recovery. Where practical, operations have achieved success field running HDPE pipe to the natural bending radius.

PIPING MATERIAL SELECTION For slurry operations, steel pipe is rubber lined with natural rubber with a shore hardness of 40. Polyurethane is an acceptable substitute to rubber particularly for grooved pipes connected by Victaulic and alternative grooved couplings. The maximum length of rubber-lined pipes is usually 12 m or 18 m depending on the capabilities of the fabricator. High Density Polyethylene (HDPE) has been accepted as a substitute to rubber for fine particles of mild abrasivity up to 3.3 m l s but has failed on taconites and certain nickel and laterite ores: However HDPE can be used as a lining for pipes up to a length of 1 km, as was done on the Collahusi copper concentrate pipeline. The maximum rating of HDPE piping is typically 1.4 Mpa in Northern climates. HDPE is particularly suited for granular carbon pumping systems, sulfuric acid processes, and cyanide leaching. For slurry with coarse particles, larger than 6.4 mm (%-in), hardened steel pipe will be an alternative to rubber-lined steel. Some fittings are cast in ASTM A532 grades (Ni-hard, 16% chrome to 28% chrome white irons). Steel pipes can be manufactured for very high pressures up to 17 Mpa at 232°C. Mineral processing applications this would apply to include autoclave feed, paste-filling, long distance concentrate and tailings pipelines. To minimize layout issues with piping material conducting hose has been used successfully for slurry applications. It is not common to find material conducting hose used for the full length of the line because the line needs to be completely supported along the horizontal. The main applications for hose have been to substitute for elbows (providing long radius bends for cyclone feed line applications, etc.) or for providing alignment and connection between sections of a pipeline. Phosphate and phosphoric-acid based slurries, such as those used to pump phosphate matrix attack steel pipe, this creating a mechanism of erosion-corrosion. Some manufacturers offer special alloys, which are forms of stainless-steel reinforced with carbides. The selection between a plastic (HDPE) and a steel pipe should be based on more than just the cost of the pipe. The plastic pipe will require more supports and on the rate of expansion. HDPE in warm South American or Middle Eastern and Australian environment the plastic pipe will tend to expand at higher rates than steel. VALVE SELECTIONS FOR SLURRY APPLICATIONS A number of specific types of valves are used in a typical mineral beneficiation plant and these each have specific uses. The vales normally utilized in process plants slurry applications are: 1. 2. 3. 4.

Knife Gate Valves Pinch Valves Diaphragm Valves Tech Taylor 5 . Ceramic ball valves

Knife Gate Valves 0

0

Knife gate valves have in recent years been extensively used in slurry pumping applications. Knife gate valves provide relatively trouble free odoff control for slurry applications. The knife gate utilizes a thick elastomer sleeve which has replaced troublesome hard seats, guides and packing. Various elastomer and knife materials are available to suit the required application. The knife gate valve can be operated manually or with actuators. One of the major advantages of the wafer style knife gate valve is that it takes very little room to install.

1418

The knife gate valve should only be used in odoff conditions and should not be used to control flow.

Pinch Valve

0

Pinch valves offer a cost effective method of odoff and flow control for slurry applications. This type of valve has the advantage that the slurry only comes in contact with the rubber sleeve and the mechanism air or mechanical operates outside of the slurry. A disadvantage of this type of valve is that its bulky in construction especially where higher pressures are required. The overall length of the valve can also create problems with the piping layout. When using the valve as a control valve it offers durability and relatively accurate control. Pinch valves come in a variety of materials such as rubber, neoprene and urethane and this allows for the selection of the material to suit the application.

Diaphragm Valve Diaphragm valves are similar to a pinch valve however the diaphragm closes against a valve seat which is in contact with the slurry. The diaphragm valve is usually used for flow control. As the slurry passes between a valve seat and the diaphragm high maintenance costs result due to high failure rates.

Tech Taylor Valves

0

0

The Tech Taylor type of valve is used where a standby pump delivers into a common line with the operating pump. The pressure from the slurry in the operating line keeps the valve ball in place over the stand-by pump discharge line. When the stand-by pump is engaged and the operating pump shutdown the ball is relocated by the slurry flow. The advantage of the Tech Taylor valve is that it operates automatically and allows either or both pumps to operate without any external operator. The Tech Taylor valve also has the advantage that it reduces the costly high maintenance that is required for this type of pump layout. Plug, gate and butterfly valves are normally not suitable for use in an abrasive sluny application.

Ceramic Ball Valves Ceramic ball valves are available for tailings and concentrate lines up to a diameter of 150 mm (6").Because they are very expensive their use is limited to very high pressures, as those associated with positive displacement pumps.

INSTRUMENTATION AND MONITORING Instrumentation and monitoring of the pipelines will be related to the process requirements and are discussed in detail elsewhere in the text. The main instruments related to piping for plant design will typically include:

0

0

magnetic flow meter radiation gauges ultrasonic flow meters diaphragm-sealed pressure gauges float switches and level indicators ultrasonic level sensors

1419

0 0

bubble tubes differential pressure (dp) sensors

An important aspect of the instrumentation is to control the speed of flow in pipes and to maintain it above deposition speed. A magnetic flow meter, a nuclear radiation gauge or ultrasonic meter is non-intrusive. Various pieces of equipment from a pump, to a tumble mill must operate at the correct density and volume of slurry and need to be adjusted through instrumentation. A level switch, float switch on a pump box is often used to adjust the flow of makeup water or the speed of a pump. A float switch on a floor sump is used to activate and shutdown a vertical slurry pump. CONCLUSION

The last thirty years have seen considerable advancement in the science of slurry pumping. Process engineers have gradually increased the concentration of slurry, and the technology of grinding permits to produce much finer slurries for better recovery of ores. It is now accepted that the slurry in a plant is far from single-graded but consists of mixtures of fines and coarse or particles of different sizes. The new models for the deposition velocity take in account the dm particle size. Friction calculations are also based on the characteristics of the layers of fines and coarse whether a multi-layer model or a two layer model is used. For fine mixture with dmc44 microns, it is recommended to use non-Newtonian models. A better understanding of open-channel flow is now achievable. The conventional Manning equations are now limited to very dilute slurry, as they do not account for particle size, volumetric concentration of solids and bedforms. Recent work from Dominguez et al (1996) based on extensive tests in Chile has yielded an acceptable equation for deposition speed in launders. The Graf-Acaroglu equation establishes a correlation between the volumetric concentration, the hydraulic radius, the dgs size, and the specific gravity of the ore. More research is needed to understand nowNewtonian gravity flows as they are being observed more frequently in tailings disposal, thickener underflow and concentrate circuits.

REFEWNCES 1. Abulnaga B.E. 2002 - Slurry Systems Handbook - McGraw-Hill 2. Dominguez, B., R. Souyris, and A. Nazer. 1996. Deposit velocity of slurry flow in open channels. Paper read at the symposium, Slurry Handling and Pipeline Transport. Thirteenth annual International Conference of the British Hydromechanic Research Association, Johannesburg, South Africa. 3. Durand, R., and E. Condolios. 1952. Experimental investigation of the transport of solids in pipes. Paper presented at Deuxieme Journee de l’hydraulique, Societe Hydrotechnique de France. 4. Gillies, R. G. J. Schaan, R. J. Sumner, M. J. McKibben, and C. A. Shook. 1999. Deposition Velocities for Newtonian Slurries in Turbulent Flows. Paper presented at the Eng. Foundation Conference, Oahu. 5 . Graf, W. H. 1971. Hydraulics of sediment transport. New York McGraw-Hill. 6. Graf, W. H., and E. R. Acaroglu. 1968. Sediment transport in conveyance systems. Part I. Bulletin. Intern. Association of Sci. Hydr., XIIIe Annee, no. 2 7. Green, H. R., D. H. Lamb, and A. D. Tylor. 1978. A new launder design procedure. Paper read at the Annual Meeting of the Society of Mining Engineers, March, in Denver, Colorado. 8. Hanney K.E.N. 1982. Selection and sizing of slurry lines, pump boxes and launders. Design and Installation of Comminution Circuits. SME. 9. Henderson, F. M. 1990. Open channelflow. New York: Macmillan Publishing Co., Inc. 10. Manning R.1895. On the flow of Open Channels and Pipes. Transactions, Institution of Civil Engineers of Ireland, Vol 10, pp 161-207;14 11. Khan, A. R., and J. F. Richardson. 1996. Comparison of coarse slurry pipeline models. Paper presented at Hydrotransport 13 pp 259 -281 published by the BHR Group, Cranfield, UK 12. Schiller, R. E., and P. E. Herbich. 1991. Sediment transport in pipes. Published in Handbook of Dredging. Edited by P. E. Herbich. New York McGraw-Hill, Inc 1420

13. John C. Loretto and ET Laker. Process Piping & Slurry Transportation. 1978. Mineral

Processing Plant Design. 14. Thomas D. G. 1965. Transient characteristics of suspensions: part VIII. A note on the viscosity of Newtonian suspensions of uniform spherical particles. J. Colloid Science 20:267. 15. Toumier E.J. and E.K.Judd.1945. Storage and Mill Transport - Section 18 - Handbook of Mineral Dressing - John Wiley & Sons. 16. Wasp, E. J., J. P. Kenny. and R. L. Gandhi. 1977. Solid-Liquid Flow - Slurry Pipeline

Transportation. Aedermannsdorf, Switzerland:Trans-Tech Publications 17. Yalin, M. S . 1977. Mechanics of sediment transport. 2"dEdition. Toronto: Pergarnon Press.

1421

Slurry Pipeline Transportation Brad L. Ricks’

ABSTRACT

Before the 1950’s slurry pipelines were used mainly for waste (tailings) disposal where the volumetric flow was high and the distances were relatively short. Solids transport was achieved via gravity systems with high line velocities. However, over the last few decades, slurry technology has made sufficient advances to become a very reliable long distance transportation2 method for many useful products such as steam coal, mineral concentrates (iron and copper in particular), phosphate, and other mineral^.^ Also, in order to address environmental and public safety issues, slurry technology advances have been utilized for long term and long-distance tailings disposal systems. This sub-section (D-G-4) deals with the current technical status of longdistance slurry pipeline systems. INTRODUCTION

The idea of hydraulic transport of solids is not new and in fact demonstration models and patents were issued prior to the turn of the century. However, the development of the modem slurry pipeline industry really began in the early 1950’s with the systematic experimentation of coal slurries by Edward J. Wasp, then of Consolidation Coal. Wasp synergistically combined the works of several scientists and engineers to develop his model that relates the homogeneous (uniform solid’s distribution throughout the pipe or conduit’s cross section) and the heterogeneous (non-uniform solid’s distribution) flow characteristics of slurry for the design of long distance slurry pipelines. (Thompson 1981.) During the course of the development effort, Wasp realized that the key to designing reliable slurry pipeline systems was an understanding and control of the slurry properties rather than in the development and selection of exotic materials or special equipment. (Wasp 1977.) This philosophy was instrumental in making the industry viable. Long distance slurry pipelines were then developed using conventional equipment and construction practices developed by the oil and gas industry in prior decades. The hndamental problems were and remain: 1. 2.

Stable flow conditions throughout the pipeline and Control or elimination of the pipe’s internal corrosion and/or erosion to the extent that design lives up to 40 years are possible.

Based upon these concepts the industry has built several pipelines. In fact, known to this author, over 4,000 km of long distance slurry pipelines have been constructed with over 125 million tones per year combined capacity. Nevertheless, in recent years, special equipment and materials have been developed that are utilized to enhance the economics and reliability of long distance slurry pipelines. Also, special construction techniques and procedures have been utilized.

’ Brad Ricks Advanced Slurry Systems {BRASS},San Ramon, California {www.brassengineering.com}

The definition of ‘long-distance” systems is somewhat arbitrary. Some systems lengths are only a few kilometers while other are hundreds of kilometers. Other minerals include, zinc concentrates, lead concentrates, mineral ores, limestone, Gilsonite, and kaolin.

1422

And lastly, slurry preparation techniques have been developed for not only hydraulic transport of the solids, but also to better meet the needs of the end user. The following article will mention many of these advances. TRENDS

Not only are slurry pipelines economic but, with concerns for the environment, long distance slurry pipelines have become more attractive as a means of bulk solid’s transportation, mainly because they are: I! buried and out of sight, and 2/ the route is quickly reclaimed by the native environment. In the case of tailings disposal, impoundment sites are selected more for environmental reasons than for convenient locations, thereby necessitating longer pipelines. By necessity, modern materials, equipment, slurry preparation and designs are employed. Whether, the slurry pipeline transports mineral products or tailings, in most cases the slurry pipeline is the most pragmatic choice. This is because they must transverse rough terrain were other haulage methods can not. Shown in Figure 1 is a bridge that spans a gorge in a mountainous sector.

Figure 1: Alumbrera Copper Concentrate slurry pipeline’s bridge spanning the Arroyo Cangrejillo Gorge. Photo was provided courtesy of Minera Alumbrera Ltda., Catamarca, Argentina (Ricks, Connelly, and Moreiko, 1998).

In many cases beneficiation processes produce mineral concentrates that are suitable for long distance sluny pipeline transport without any further processing. Common examples are iron, copper, zinc, phosphate, limestone and kaolin. In other cases, such as steam coal, the slurry is prepared to meet specified characteristics. Nevertheless, in order to optimize the development of a

1423

mineral deposit, one must consider the ore dressing process, the slurry transportation requirements, and utilization requirements of the overall system. Slurries can be classified into one of three categories: Homogeneous (due to the force of gravity, it is recognized that truly homogeneous slurries are nonexistent but rather are pseudo-homogeneous) 2. Heterogeneous where the concentration of the slurry particles varies greatly from top to bottom of the pipe’s cross section 3. Complex (not completely homogeneous nor heterogeneous) 1.

Heterogeneous slurries are typically composed of coarse particles which are not uniformly suspended in the pipeline and tend erode the pipe bottom. Associated with heterogeneous slurries are high line velocities and high-energy consumption. Because of these drawbacks, heterogeneous long distance pipelines are generally not feasible. Most slurries are either homogeneous or complex and behave in a non-Newtonian manner. Typically, with the exception of coal, the solid’s volume fraction is less than 0.40. These slurries are considered to be “conventional”. In general, for long distance pipelines the slurry properties are such that during transport, the slurry is uniform from top to bottom in the pipe. That is to say, the slurry should be a homogeneous fluid having the solids uniformly suspended through the pipeline’s cross section. I[n reality complete homogeneity is approached but not achieved (pseudo-homogeneous). Usually the slurry specifications require a somewhat “fine” particle size distribution (PSD) and sufficient concentration to produce a “suspending” viscosity. The line velocities are such that friction losses are reasonably low in order to be economic and erosion of the pipe wall does not occur. Yet at the same time, the line velocities must be sufficient to generate enough turbulence to uniformly suspend the slurry particles. Normally, the long distance pipeline transports conventional slurry For short distance pipelines, the degree of homogeneity is not as critical. Therefore, the design approach can be quite different than for a long distance system. For example, the slurry PSD may be relatively coarser for the short distance pipeline resulting in a fluid with non-uniform characteristics across the pipe’s cross section (heterogeneous flow). Usually the line velocities are higher, relative to long distance systems, and the system design allows for some erosion of the pipe’s wall. These pipelines are designed to be periodically rotated, or the pipe material may be erosion resistant, or it could be steel pipe with erosion resistant liners. The most common erosion resistant materials are rubber and polyurethane using a pipe joining method that employs a suitable flange or coupling. Alternately, “non-conventional” slurry transport systems are emerging. The non-conventional slurries have high solids concentrations and are highly viscous. Associated with these systems are high friction losses and relatively low line velocities. The main advantage is that carrier fluid (usually water) quantities are reduced and any end “de-watering” process is eliminated.

RELIABILITY

Slurry pipelines are now regarded as a very reliable transportation mode For example, the Black Mesa Pipeline in the USA has historically been over 99.5 percent reliable (Montfort 1981) in its 32 years of operation. The Samarco Pipeline in Brazil has a similar reliability (Ricks 2000). In one case, a grassroots slurry pipeline system was constructed and put in operation in spite of the fact that it paralleled an existing railroad for much of its route. In that case the pipeline was selected not only for its economic benefits but also because of high reliability. Shown in Figure 2

1424

is “Availability versus Probability” plot of a long distance slurry pipeline system derived from a detailed Monte Carlo Simulation. The figure considers natural events, corrosion, third party incidents, major component failure, and system component maintenance and repairs.

Long Distance Slurry Pipeline Availability

I

1.00 \

I

0.99

.-E E 2 .-rs

4

~

-

0.98

0.97 0.96

--

_.

~

__Typically, long distance slurry pipelines are very reliable The plot shows that 0 50 -- probability corresponds to 0 9925 availability

__ -

~

_

_

.

_

_

_

-

_

_-

.

_

~~

__

~

~~~~~~

-

-.

~~

0.95 0.94 0.00

I

0.20

0.40

0.60

0.80

1.oo

Probability -

_ _ _ _ _ _ . ~ __ . _ . -_ ~_ _ _ ~ ~ ~ ~

~~

-

~

Figure 2: Long distance slurry pipeline availability versus probability derived from a detailed Monte Carlo Simulation. Note that the average result (0.50 probability) is an availability of 0.9925. SLURRY PIPELINE DESIGN In every case, the decision to develop a project is political. We use technical and economic projections to persuade the involved parties to proceed. Historically, slurry pipelines have been technically and economically persuasive. They enjoy the same type of economy of scale as gas and petroleum pipelines. In general this is true for relatively high volume and moderate distances. However, depending upon the slurry and terrain, it would be remiss not study the feasibility of slurry pipelines for almost every bulk material transportation project. For example for coal slurry pipelines, distances over 80 km and mass rates of 4 or 5 million tones per year are viable. For copper concentrates, distances over 15 kilometers and volumes over 0.75 million tones per year are viable. For a typical slurry pipeline, the major contributions to unit transportation costs are capital related fixed costs that are not subject to inflation. The major elements of a slurry pipeline system are: the slurry preparation facilities such as a mineral concentrator; carrier liquid supply (e.g. water) slurry surge storage at the preparation facility; the mainline with pump stations, valve stations, and in some cases energy dissipation stations communications and control system intermediate slurry storage facilities (if required) Terminal slurry storage facilities.

1425

Technical Design Bases Although cost factors are normally considered an important element in system design, they are generally very specific to each project and are not discussed in this article. So the purpose of this discussion is held to technical issues. The technical design bases for slurry pipelines fall under three general categories: 1. 2. 3.

Slurry Properties Capacity Terrain

Slurry Properties. Slurry characteristics are determined by the carrier fluid (usually water), the solids particles, and interaction between the carrier fluid and solids in the mixture.

Liquid. Several carrier fluids have been proposed and studied, such fuels, oil, alcohol, and liquid carbon dioxide (Ricks 1988). However in nearly all the operating systems the carrier fluid has been water at or near ambient temperatures. The relevant physical properties for the carrier fluid are 1/ specific gravity, 21 bulk modulus, 3/ viscosity, and 41 vapor pressure. These properties are well understood as a function of temperature. Notwithstanding, for most slurries with volume concentrations greater than 30 percent, the temperature of the carrier fluid has a diminished effect upon the overall slurry properties. Solids. The overall slurry properties are affected by the following solid’s properties: 1/ specific gravity (SG), 2/ particle shape, 3/ particle size distribution (PSD), and 41 any surfactants acquired during the beneficiation process. Particle Top Sized vs. Solid’s SG 2.5

-----

0.50

--I

0.45

E

2.0

0.40

0.35

5

.y 1.5

0.30

E

P 0 I-

0.25

E

6

v)

s0

.-

E

E 0.20 2

1.0

5

0.15

a

0.5

0.10

0.05 0.0

I

1.o

2.0

3.0

4.0

0.00

5.0

Solid’s SG

Figure 3: The relationship between solids specific gravity and “conventional” slurry top particle size and solid’s volume fraction. The volume fraction relationship assumes that the carrier fluid is water at ambient temperatures. The volume fraction’represents the higher values. Many tailings systems operate with lower volume fraction values and are not shown in the Figure.

1426

Although it is possible to pump virtually any combination of solids and carrier fluid, strict restriction on solids PSD and concentration are required to ensure a slurry: 1/ will maintain predictable and stable flow conditions, 21 that will it not wear the pipe bottom, 3/ and it can be shutdown and restarted. Inter-related is the solids SG with both concentration and PSD. Shown in Figure 3 is a rough relationship between a conventional slurry particle‘s “top size”, volume concentration and the solid’s SG. Mixture. The solid’s concentration, particle shape, and acquired surfactants effect the particle’s settling characteristics and the slurry rheology. In general, most slurries are characterized as nonNewtonian. A comparison of selected rheological types is shown in Figure 4 showing rheogram curves for each type,

INewtonian Pseudoplastic Dilatant Bingham Plastic Yield-Power Law

T T T T T

I

= MY = Ky”

n
= KY”

1. e n ..

=

To

=

T,+

+

qy rly“

Figure 4: Rheogram comparison of selected constitutive equations for rheological models associated with slurries. The most common rheological models used for conventional slurries a r e the Newtonian model and Bingham-Plastic. Technically the nomenclature for the above is: “p” is defined as the Newtonian viscosity; “K” is defined as the power-law consistency index; “n” is. the power-law flow behavior index; and “17” is defined as the Bingham-plastic coefficient of rigidity.

1427

Establishing the Slurry Concentration Range. Holding the PSD, solid’s SG, and process reagents (such as pH and process surfactants) constant for a given slurry, the rheological properties are a strong function of concentration. For most conventional slurries, a solid’s volumetric concentration up to 40 percent is readily handled per Figure 3. For these slurries, the solid’s surface area to volume ratio is relatively low. However for some slurries such as sludge and minerals with fine clays, the solid’s surface to volume ratio is quite high. In these cases, the solid’s volumetric concentrations of 15 percent or less is often the practical limit. For these slurries, knowledge of the rheological characteristics is important. As a rule of thumb, usually a comfortable concentration is 10 to 15 volume percent less than the final settled concentration (Aude and Thompson 1985.) The settled concentration can be determined by allowing the slurry to settle and then calculating its settled concentration by comparing the column height of the fluid to the column height of the liquid-slurry interface. Shown in Figure 5 is a typical plot of slurry viscosity4 versus the slurry concentration. The concentration range is from 59 to 64 percent. This range produces viscosities that provide sufficient suspending characteristics yet the slurry is not extremely sensitive to concentration changes. Viscosity Plot 60

50

40 n

2;

’3 30

s

Typi--. concc. ... ation range. Viscosity is -sufficient for suspension, yet not extremely sensitk to concentration changes.

A

5

-I -

20

10

0 0%

10%

20%

30%

40%

50%

60%

70%

Concentration, wt %

Figure 5: Typical slurry viscosity versus solid’s concentration plot for conventional slurry. Note the pumping range is from 59 to 64 percent

4

In this case the reference to viscosity is loosely applied. It could refer to viscosity for a Newtonian fluid or the coefficient of rigidity for a Bingham plastic fluid.

1428

Slurry Hydraulic Gradient Plot

Deposition Velocity

Transition Velocity Line Velocity, mls Vehicle Properties

Data Set #3

Figure 6: Hydraulic gradient (friction losses) versus line velocity for a selected slurry test. A Bingham plastic rheological model is used. A comparison is shown for the carrier fluid (water a Newtonian fluid) and calculated losses based upon slurry homogeneous properties (homogeneous fluid) and vehicle properties (complex fluid). Note as the line velocity increases, the homogeneous and vehicle calculations become indistinguishable. For buried long distance systems, such as concentrate pipelines, the “safe” operating range is well above the deposition velocity. For short systems such as tailings disposal, the normal operating range need not be as conservative. Predicting Friction Losses. The mainstay of the slurry design calculation is the Wasp model. At given flow conditions in a pipe, the model determines the degree of heterogeneity (or conversely, the degree of homogeneity) of the solid’s particles. It then determines the friction losses contributions of the vehicle (the supporting pseudo-homogeneous slurry) which suspends the heterogeneous solids (the bed). The total friction loss is calculated by summing losses due each. However, the model assumes Newtonian behavior of the vehicle. Most conventional slurries fall within the experience represented by the model, particularly in the turbulent flow regime. In this case the degree of L‘non-Newtonianess’’is not significantly large to affect its predictive accuracy. Refer to Figure 6 showing a plot of friction losses in terms of a hydraulic gradient versus line velocity for selected slurry tests. Note the plots compares the calculated friction losses for the carrier liquid to that of the slurry based upon “homogeneous” properties and “vehicle” properties. Note also, that at higher velocities, where the fluid turbulence is sufficient, the vehicle and homogeneous friction losses correspond. This is because, at these velocities, the turbulence is

1429

sufficient to produce a pseudo-homogeneous suspension throughout the pipe or conduit’s crosssection. In this case the vehicle and homogeneous conditions are indistinguishable. Referring to Figure 6. If the line velocity is too low, say below the deposition velocity, then a bed of slurry forms and it is dragged along the bottom of the pipe. Pipe bottom wear results. If the velocity is too high, then high friction losses are encountered. If the line velocities become too high, then abrasive conditions develop. The preferred operating conditions are in the sector denoted by the “safe operating range.” The velocities are sufficient to suspend the solids yet sufficiently low to keep friction losses within economic levels. At the same time the velocities are low enough that abrasion does not occur. Refer to Figure 7 showing the calculated solids distribution across the pipe cross-section. It shows the concentration ratio of a particular particle size at the top of the pipe compared to the bulk concentration.

Figure 7: Particle concentration distribution plot showing the ratio of the concentration of a particular particle size at the top of the pipe compared to the bulk concentration. As the line velocity increases, turbulence becomes sufficient to suspend the particles more uniformly. Note the larger particles are not as readily suspended as the fine particles. Also, a t low velocities the slurry is mainly a heterogeneous fluid, but as the particles for each size are suspended, the slurry becomes a complex mixture, and finally for the fine particles, the slurry becomes nearly homogeneous. Conventional Mineral Slurries. To date mineral slurries have been conventional slurries and they are usually prepared as a natural consequence of their beneficiation process. These process usually involve crushing and grinding, and separation of refuse material through washings, flotation, or other separating techniques. The mineral concentrates are usually thickened and held in agitated slurry storage tanks prior to being committed to the pipeline. In these cases, the slurry pipeline designer does not have much control over the PSD but has some latitude over the concentration and line velocities. The majority of mineral concentrates are iron, copper, phosphate, and limestone.

1430

Conventional Tailings Slurries. Due to environmental reasons, the length of tailings disposal pipelines have increased. Gone are the days of dumping the tailings at the most convenient downhill slope. Typical of many tailings systems, the beginning is at high elevations and the impoundment is at lower elevations. Conventional tailings slurries lack the high degree of concentration and particle size control normally associated with concentrate slurry pipelines. Therefore, the tailings pipeline designer prepares for many scenarios using heterogeneous and homogeneous models in order to account for all possibilities. Sometimes systems must dissipate excess energy in order to control the pipeline velocities. Choking is evolving as preferred methods of energy dissipation though cascades and drop boxes are used. Non-conventional Slurries. Non-conventional slurries are highly concentrated and are nonNewtonian, with most being significantly non-Newtonian. Consequently, other rheological models have been used to describe their behavior. Refer to Figure 3 for Bingham plastic, pseudoplastics, dilatant, and yield-power law fluids. These slurries have a high degree of homogeneity and are treated accordingly. It is argued that if a slurry is modeled for laminar flow (rheogram), then the same momentum transfer will also be carried into the turbulent regime. For example, if a slurry is described by a Bingham plastic model, then terms such as the yield stress must be accounted for in the turbulent flow calculation. These models have been worked out and are now being used (Hanks 1978; Darby 200 1; Hanks and Ricks 1975; Hanks and Dadia 1971.) Thickened Tailings Disposal (Non-conventional). In recent years, the concept of thickened tailings disposal (TTD) is gaining recognition (Robinsky 1999). It has the advantage of disposing of highly concentrated tailings and therefore reduces the storage volume required. Also, it eliminates the need to install water reclaim systems. To put the difference in perspective, conventional tailings slurries usually are transported between 15 and 35 percent concentration by weight while TTD concentrations are around 65 to 70 percent. Because the tailings are waste material, a wide range of solid’s particles and tramp material is introduced into the disposal line. A TTD slurry is viscous and homogeneous throughout the flow range. For conventional slurries the velocities are usually from 2.5 to 3 meters per second in order to maintain the slurry solids in suspension. In contrast, TTD transports with line velocities below one meter per second. Non-conventional Steam Coal Slurries. Non-conventional steam coal slurries have been proposed that avoid the de-watering process prior to being fed to the boiler. Slurries such as coal-oilmixtures (COM), and coal-water-mixtures (CWM), slurries using liquid carbon dioxide, and other fuel mixtures are being developed. Of the above, the most advanced to date is CWM. CWM consists of a highly concentrated slurry that can be directly fired in a boiler. The slurry is homogeneous and the line velocities are low (laminar flow). Pilot systems have been built and tested. A commercial system has been built in Russia, The results are not well known, it but appears that there is more development required.

Determining Minimum Velocities. Referring to Figure 6 , note that two velocities are specifically annotated, namely “transition velocity” and “deposition velocity.” The transition velocity corresponds to the critical velocity where laminar flow transitions to turbulent flow. For a Newtonian fluid this would correspond to the laminar-turbulent transition for the classic Reynold’s number of 2100. The deposition velocity corresponds to when the slurry particles begin to saltate. Usually the large particles require higher velocities for suspension than do the finer particles. Transition Velocity. The transition velocity is determined for homogeneous fluids. For a Newtonian fluid, that velocity would correspond to the well know critical Reynold’s number value of 2 100. For non-Newtonian fluids, the “critical Reynold’s number” value varies depending upon the degree of non-Newtonian properties of the fluid. Since a Bingham plastic model is used for most conventional slurries, it is used to illustrate the transition velocity determination. The degree of non-Newtonian behavior can be described by a dimensionless group of factors known as the

1431

Hedstrom number (NHe). The larger the Hedstrom number for a given slurry the more nonNewtonian its behavior. It has been shown that there is a relationship between the critical Reynold's number (NReC)and the Hedstrom (Hanks and Dadia 1971):

Where

16,800 (1-50C)

Similar relationships have been developed for pseudo-plastics and yield-power law models (Hanks and Ricks 1974, 1975 .)

Deposition Velocity. The deposition velocity is related to complex slurries and heterogeneous flow. Durand did the classic work where he proposed that the deposition velocity was proportional to the square root of the pipe's diameter multiplied by a density factor (Durand 1952):

Deposition Velocity

S

-

\i

D * ( S - I)

Solids S G I Liquid S G

Wasp added to the above an additional factor to account for the solid's particle size (Wasp 1977): ( d I D)"'

However, according to experience the slurry concentration has a significant role in the deposition phenomena. It has been observed that for conventional mineral concentrates the degree of heterogeneity or complexity of slurry decreases as the concentration increases. Therefore, the deposition velocity decreases with concentration up to the point where the viscosity (concentration) is sufficient to produce a homogeneous slurry. Thereafter, the deposition velocity and the transition velocity become virtually identical. To illustrate, refer to Figure 8 showing the relationship between deposition and transition velocities for a selected concentrate slurry. A much more sophisticated calculation is used to determine this relationship.

1432

Figure 8: Limiting velocity plot of a selected conventional slurry showing transition velocity and deposition velocity relationship. Capacity and Diameter Selection. Determining the pipeline diameter is an iterative process. Friction losses, limiting velocities, concentration ranges, power consumption, energy dissipation, material costs, and installation costs are all balanced in order to design the most cost-effective system. Furthermore, a overall approach to the entire project may be conducted in order to consider the mine, ore beneficiation, transportation, and utilization of the mineral in order to arrive at an overall system optimization. Usually the concentration range is established and then to start, for the concentration range, candidate diameters are determined based upon the capacity requirements of the system. Normally, the velocities will fall in the range of 1.25 to 2.1 meters per second, with the most likely bulk velocities falling inside 1.65 to 1.85 meters per second. In the end, the selected pipeline diameter must result in bulk velocities above the minimum operating velocity yet still low enough to keep friction losses within reason in order to provide economic transportation. Turndowns are generally controlled via flow and concentration control and are of the order of 0.75. to 0.80. Terrain. In many cases slurry pipelines are selected because of the terrain. Usually the route is in difficult and remote locations. The slurry pipeline lends itself to such situations for the following reasons: Pipelines are easier to construct than railroads and truck haul roads because the pipeline grade (slope) requirements are less restrictive Construction rates are faster using pipeline construction techniques developed over many years The land is naturally reclaimed in a relatively short time period The pipeline is underground and less susceptible to third party incidents The system is remotely controlled Major maintenance is usually at pre-determined locations such as pump stations, valve stations and terminal facilities. Maintenance at intermediate points along the route is infrequent and usually does not affect production while being done.

Usually candidate route corridors are identified and studied regarding technical and economic issues using topographic maps or preliminary surveys. Once the corridor is determined, the route is “staked” based upon length, slope criteria, construction criteria and property issues. Right-ofway and easements are finalized and detailed surveys are performed. Finalizing a route can be iterative. As the route is finalized, the system design is refined and finalized. Computer Simulations. Due to the development of computer technology over the last couple of decades, dramatic changes in design methods and analysis have transpired (Ricks and Aude 1992.)

1433

Over the last ten years, computer simulations have been employed from a project’s beginning through to commercial operation and control: namely conceptual engineering, basic engineering, detailed engineering, commissioning and startup, and into operation and control. Shown in Figure 9 is a graphic display of a simulation program. The ability to view results of steady state and dynamic graphic displays has given the engineer immediate feedback on cases under analysis. Also, from a practical standpoint, the input data and design assumptions are checked and verified. The benefits of computer simulations have resulted in: Reduced engineering labor costs More cost effective design Better designer understanding of operations and better development of operating procedures Better operator understanding of system operations And finally designs that are rigorously proven to be within safety codes.

Figure 9: Graphic display of a computer simulation program used for system design, operating procedure development and system monitoring. Provided courtesy of Minera Alumbrera Ltda.

Codes Since 1986, slurry pipelines used the American National Standard (ANSI), American Society of Mechanical Engineers (ASME) B31.11 as the design code for slurry systems. Its origin stems from the initial ASME code for pressure piping, B31, and in particular it is a modification of ASME B3 1.4, “Liquid Petroleum Transportation Piping Systems.” Prior to 1986, B3 1.4 was used for cross-country slurry pipeline design. B3 1.1 1 is basically the same as B3 1.4 with the following exceptions:

1434

For non-hazardous water based slurries, provisions in the code for flammable and explosive petroleum products are dropped or not as stringent Internal corrosion and abrasion aspects are treated more stringently The design factor has been increased from 0.72 to 0.80 An update is issued approximately every three years. System Components Support components such water supply, slurry preparation, and dewatering facilities are not discussed herein. Also, not discussed are communications and controls components. Those component covered here are:

0

slurry surge storage tanks high pressure mainline pumps slurry flow control valves mainline pipe materials energy dissipation

Slurry Storage. It is common practice to install surge storage at facility interface locations. The most common surge storage location is between the slurry pipeline or its first pump station (if required) and its preparation facility. For most long distance mineral slurry pipelines, agitated storage tanks are used. For many systems, the surge storage is a sump. The surge capacity may range fiom a few minutes of pipeline flow to several hours, or weeks, or months. The storage can be classified as active or inactive. Active storage refers to slurry that is kept homogenized (solids in suspension) and is ready for transportation. Inactive storage refers to long term storage where the solids are allowed to settle. The purpose and location of slurry storage considers the following:

at pipeline feed locations, such as concentrators or preparation plants, storage is used to separate different types of slurry such as copper concentrate and zinc concentrate, or caps and tails for coal systems at intermediate pump stations to facilitate for re-homogenization of slurries restarted after shutdowns, or slurries with density waves, or to facilitate batch operations at pipeline terminals as surge storage between the pipeline and terminal processing facilities at pipeline feeder junctions or distribution junctions emergency dump storage (inactive storage) Each storage facility is selected based upon logical criteria appropriate to any specific project. Agitated Slurry Storage Tanks. Agitated slurry storage tanks are designed using API Standard 12D, “Specification for Field Welded Tanks for Storage of Production Liquids”, or most commonly, API Standard 650 “Welded Steel Tanks for Oil Storage” for guidance. Tank tops are not required and the tank design must be structurally modified to mount an agitator on a bridge and account for the forces generated by the agitator. The tanks are cylindrical, baffled and having a “flat” bottom’ with a wear plate and bump ring installed directly below the agitator’s impeller/s. Sizing criteria are usually based upon the number hours pipeline capacity considered appropriate for the project under consideration. These values can range from a few minutes to several hours, but are normally from 8 to 24 hours of pipeline capacity. Since vessel materials are minimized for a fixed volume at about 1.05 height-to-diameter ratio (Z/D) it has been common practice to select 5

The tank bottom has a slight slope from the center to rim.

1435

Z/D ratios between one and 1.05. However, one study has shown that this ratio is not necessarily the most economic. Depending upon the circumstances, the optimized Z/D ratio can vary from 0.63 to 1.05 (Von Essen and Ricks 1999). Two types of axial flow impellers are used: 1/ hydrofoil, and 2/ pitched-blade turbine. The hydrofoil type has the advantage of high pumping efficiency, meaning low power consumption. It has the disadvantage of a low live volume to total volume ratio because the impeller must be mounted about to 25 to 30 percent of the tank height above the floor. Recent installations have employed the both impeller types, using hydrofoil main blade and a smaller pitched blade turbine near the tank floor. This is done to utilize as much of the total volume as possible. In Figure 10 is a sketch showing typical dimensions for an agitated mineral-concentrate slurry storage tank.

Figure 10: Typical Agitated Slurry Tank. Data for the sketch were provided courtesy of Philadelphia Mixers, Palmyra, Pennsylvania. Mainline Pumps. Depending upon the pipeline’s route and pipeline size, pump stations may or may not be required. If pump stations are required then is becomes a matter of whether to use centrifugal pumps or positive displacement pumps (PD pumps).

Centrifugal Pumps. Centrifugal pumps are used for relatively low discharge pressures and large flows. Centrifugal pumps are not as sensitive as PD pumps to solid particle size or tramp material in the slurry. For PD pumps the solid’s particle size is usually limited to 6 or 7 millimeters in order to protect their valves. However, centrifugal pumps operate and lower efficiency than the PD pump. Usual efficiency values are around 0.60 to 0.65. For slurries with coarse particles, the head ratio (ratio of “tdh” of slurry to that of water) is usually less than one (1.00). For most concentrate slurries, the head ratio value can be considered as unity. For low head systems, the centrifugal pump trim is usually rubber-lined or lined with other polymers. For higher head systems, hard metal trim is used. The initial cost of centrifugal pumps is much less than PD pumps and since they are commonly used in most process facilities, maintenance and operating personal need not be specifically trained. For PD pumps personnel require special training. If impeller tip speeds can be keep low (low head requirement system), then pump durability is good.

1436

On the other hand, if head requirements are high requiring high pump speed, then wear parts are consumed at a high rate. Some “grinding” of the slurry particles occurs as the slurry passes through a centrifugal pump. This is due to the high shear conditions inside the pumps. Consequently, the slurry properties are affected, in particular the PSD and rheological properties. These changes can not only affect the pipeline transportation process, but also, processes downstream of the pipeline system. Therefore, for centrifugal systems, inordinate numbers of stages or numbers of pumps (and pump stations) in series are avoided. Furthermore for multiple pumps, dilution from seal water can adversely affect the delivered slurry.

PD Pumps. Normally for high-pressure systems positive displacement pumps are used. The genesis of these pumps was in the “oil patch.” Early pipelines were driven by oil field pumps that were modified for heavy continuous duty with a design life of 25 years or more. The largest pumps in commercial operation have a frame size of the order of 1600 or 1700 horsepower.6 For high volume pipelines, such as would be required for coal slurry pipelines, larger pumps have been designed and tested up to a frame size of 3600 horsepower. They have been designed in modular form from the power-end to the fluid-end so that singleplex, duplex, or triplex machines can be fabricated from the modular components. None of these are in commercial ~ p e r a t i o n . ~ PD pumps can generate high discharge pressures (currently over 21,000 kPa), but their flow capacity is limited. The following table shows the PD pump types that have been used:

I

PD PUMP TYPE Piston Plunger Piston-diaDhraEm

I SLURRY TYPE

I Non abrasive

I Abrasive I

Abrasive

I

Abrusivity. To select the PD pump type, an “abrasivity” test is performed. Refer to ASTM G7589 (Miller and Miller 1989). A Miller number is an index of how abrasive slurry may be in contact with moving pump parts. For slurries having a Miller Number less than 60, a piston pump could be used. For slurries with Miller Numbers greater than 100 a plunger or piston-diaphragm pump should be used. As in all cases, these values are not considered rigid criteria for pump selection. The designer’s experience is very important. For a given project, the designer may wish to determine a slurry’s SAR number. SAR means Slurry Abrasion Resistance. Rather than use standard test specimens, “non-standard” specimens are used. The non-standard specimen material is usually specified in order to be pertinent to a specific each project. Piston Pumps. Of the above listed PD pumps, the piston pump is probably the simplest and least expensive. The origin of slurry pipeline pumps evolve from mud pumps used for oil well drilling that were “beefed-up” for use with slurry pipelines. One of the early uses for piston pumps was the Black Mesa pipeline. This pipeline has been operation for more than 30 years. With a piston pump the slurry comes in direct contact with the pump’s seals, piston, liner, and valves. Refer to Figure 1 1. The wear parts are: piston seals ~~~

6

The frame size is not indicative of the motor horsepower. The manufacture usually determines the frame size based upon the anticipated push-rod and bearing loads. 7 Actually there is a large pump with a frame size of 2000 horsepower in operation. It is installed on a large copper concentrate pipeline. Its flow capacity is over 300 cubic meters per hour. Another large pump that is approximately the same frame size should be commercially operating by the time this article is published. It is installed on a magnetite concentrate pipeline. Its flow capacity is approximately 200 cubic meters per hour.

1437

0

0

piston rubbers piston liners . Valve sets.

Figure 11: Piston pump fluid end view provided courtesy of National-Oilwell, Houston, Texas. Plunger Pumps. Plunger pumps are used for “abrasive” slurries. The pumping action occurs as the plunger displaces slurry in the fluid end. The abrasive action of the slurry is mitigated by “flush water” that is injected around the plunger in order to avoid direct slurry contact between the plunger and stuffing box components. The stuffing box components include lantern rings for even distribution of flush water around the plunger, and packing sets that are held between bushings for sealing. Refer to Figure 12. The wear parts are: 0

0 0

plunger barrel throat bushings and lantern rings packing sets And valve sets.

The purpose of the plunger flush system is to increase the wear parts utility life; however, it introduces more complexity for maintenance and operations. If the slurry system requires multiple pump stations, the design must account for slurry dilution due to flush water injection. The usual flush water volume is 2.5 to 4 percent of the total pump flow.

i438

Figure 12: Plunger pump fluid end view provided courtesy of National-Oilwell, Houston, Texas.’

Piston-Diaphragm Pumps. Piston-diaphragm pumps incorporate a design that avoids contact of the stuffing box components with abrasive slurry. A fluid such a lubricated-water or oil is used in the stuffing box to drive a flexing diaphragm. The diaphragm action draws and displaces the pumped volume through the valves. Refer to Figure 13 showing a typical pump power end, propelling fluid chambers, and fluid end where the slurry is pumped. Since most of the wear parts are in contact with the propelling fluid, their utility lives are relatively long. The wear parts in direct contact with the slurry are limited to the: 0 0

Diaphragm And valve sets

8

In Figures 11 and 12 the suction-discharge valve arrangement is known as an “overhnder” fluid end block because the suction valve is directly under the discharge valve. In many installations, the fluid end valve arrangement is modified into a “L” shaped fluid end block in order to facilitate suction valve maintenance.

1439

Pulsation dampener Discharge valve Pressure relief valve Diaphragm housing Pump diaphragm Piston Suction valve

Liquid end

Power end

Figure 13: Piston-diaphragm pump sketch provided courtesy of EnviroTech Pumpsystems (Geho), a Weir Company, AE Venlo, The Netherlands. Note that the slurry is on the left-hand side of the diaphragm and propelling fluid is on the righthand side. These pumps are more sophisticated and somewhat more expensive than piston pump and plunger pumps but they reduce and simplify routine maintenance. Slurry Flow Control Valves In a number of recently designed systems, the routine operating procedures allow for pipeline shutdowns while full of slurry. With this operating plan in mind, the importance of critical valves has increased significantly and could be considered equal to mainline pumps. The value duty may, in some cases, be considered even greater than that of the pumps, since there is usually an installed spare pump. Installation of installed spare flow paths and valves is not common practice. The valves are used in an “openhhut” application and not used for throttling flow. The severity of the slurry valve duty can be put into perspective by the following example. It is commonly believed that abrasive wear for given slurry increases by the velocity to a power of somewhere between 2 and 3.5 (Wiedenroth 1984; Rao and Buckly 1984). The maximum velocity through a valve during the opening/closing cycle can be estimated using its characteristics and differential pressure. If a moderate differential pressure across a valve of 10,000 kPa is assumed, as the port opens or closes, velocities up to 90 meters per second can occur. Therefore, if the slurry is mildly abrasive at a velocity of three meters per second, then it will be at least 900 times more abrasive at 90 meters per second (Ricks and Aude 1989). Low and Moderate Pressure Slurry Valves. There are variety of low pressure slurry valves, such knife gate, pinch, ball and plug. Low pressure meaning ANSI Class 150 and below and moderate pressure meaning up to ANSI Class 300 rating. The most commonly used low-pressure valves are knife gates. Typically the maximum pressure across these valves are less than 500 kPa. Shown in Figure 14 is a typical knife gate.

1440

Figure 14: Slurry knife gate valve. Cut-away view of AK Knife Gate Slurry Valve provided courtesy of the Weir Slurry Group, Madison, WI. For moderate pressures, such as branch line isolation or choke line isolation, plug valves or hard metal ball valves have been used. In recent years, valves whose trim is lined with abrasion resistance rubbers or polymers are being developed. High Pressure Slurry Valves. Most high-pressure applications are found at pipeline terminals and intermediate valve stations. These valves are used to shut-off or start slurry flow beginning with high differential pressures. Because of the specific gravity of typical concentrate slurries and the attendant mountainous terrain, these valves must be rated up to ANSI Class 1500 (26,000 kPa), the most common rating being ANSI Class 900 (16,000 Wa.) However, due to technological advances, higher pressures are becoming the norm rather than the exception. The most common valves for this duty have been lubricated plug valves and hard metal ball valves. Lubricated plug valves are typically used in “openlshut” applications where a full round opening is not required. The plug and body cavity must have hard, wear-resistant surfaces. Conscientious application of lubricantlsealant is very important to tight shut-off and long life. The sealant lubricates the plug and at the same time purges slurry solids from between the plug and valve body. Shown Figure 15 is a typical lubricated plug valve. Hard metal ball valves are used as station isolation valves where full port openings are required in order to pass a pipeline “pig”. They also used for opedshut applications and as isolation valves for high-pressure mainline pumps. As with lubricated plug valves, preventing accumulation of solids between the moving surfaces in the body cavity is essential. Also; the slurry must be prevented from restricting movement in the stem’s seal rings. Show in Figure 16 is a typical hard metal slurry ball valve.

1441

Figure 15: Lubricated slurry plug valve. Cut-away view of a lubricated plug valve provided courtesy of the Nordstrom Valve Incorporated, Sulfur Springs, Texas.

Figure 16: View of a hard-metal slurry ball valve courtesy of Mogas Industries, Inc., Houston, Texas. Mainline Pipe Materials . Due to high pressure usually required in slurry pipelines, high test carbon steel continues to be the main material used. For short distance lines or low pressure lines other material may be considered. CorrosiodErosion Protection. For many long distance systems, the design life is from 20 to 40 years. The pipe wall thickness must be sufficient to contain the internal pressures after allowing for losses due to internal corrosion and/or abrasion. Therefore control of internal corrosion and erosion is important. More often than not, the slurry particles increase the corrosion rate because they tend to scour protective corrosion products from the pipe wall leaving a clean new surface ready for further attack.

To control corrosion in pipelines using ferrous pipe material (usually steel), the slurry pH is adjusted, if necessary, to values between 9 and 11 in order to chemically inhibit the steel from corrosive attack. If necessary, the dissolved oxygen content is reduced by using oxygen scavenger reagents.

1442

Inserting corrosion resistant liners into the pipe can eliminate corrosion. For new pipelines, the costs of the liner and its installation are weighed against the cost of extra steel (corrosion allowance), corrosion inhibitor systems, and inhibitor reagents and consumable supplies. Since the liner is a passive protection form, it is usually preferred over corrosion inhibitor systems that require constant maintenance and process attention. Steel. Carbon steel continues to be the principal pipe material used. It is relatively inexpensive and easily obtained. The pipeline design can choose from many grades of steel. Usually for low and moderate pressures Grade B material is used. For high-pressure applications, high test line pipe material is used. These grades can vary from API 5L X42 to X70. The most common grades are X60 and X65. The grade and wall thickness is interrelated depending upon the corrosion allowance, availability, and construction criteria under consideration. Long distance slurry pipelines are designed and operated under flow conditions such that erosion of the pipe wall does not occur. However, corrosion control is required either by chemical inhibitors or corrosion resistant liners. For lines where wear can occur (usually bottom wear in tailings systems or similar), steel lines are designed so that they can be periodically rotated in order to maintain uniform wear around the pipe or abrasion resistant liners such as rubber, polyurethane, or concrete are used.

Rubber. Many abrasion resistant materials have been developed such as basalt lined pipe, ceramic lined pipe, hardened metals, and many polymers. However, rubber is still the most commonly used in pipes, pumps, and other process equipment. It has provided good service for slurries that are uniformly abrasive. High Density Polyethylene (HDPE). HDPE has seen substantial increase in service over the last ten years. It is relatively inexpensive and its applications are quite flexible. Improvised pipelines have been constructed in very short time periods by using a field joint fusing process that is quick and effective. HDPE appears to be effective for slurries that are uniformly abrasive. For slurries with large and/or jagged particles that may cut the HDPE, it is not as effective against wear. HDPE is a good corrosion resistant material. It can be used on its own as the primary pipe when internal pressures are less than 1500 kPa. For high pressure applications it is used as an effective liner inside steel pipes. In this manner the corrosion resistance properties, and somewhat abrasion resistance properties, are combined with the strength of steel. Pipelines with pressures well over 2 1,000 kPa are using HDPE liners. For long distance pipelines, HDPE liner is pulled through existing steel pipelines at about one kilometer intervals. This technique is used to rehabilitate old pipelines or depending on project economics, it is installed in new pipelines.

Polyurethane. Polyurethane has gained a reputation for toughness against abrasion. Even for abrasive conditions such as for cutting and gouging, polyurethane performs well. Generally it is more expensive than HDPE or rubber but it can be shown to be cost effective for many applications. It finds use as pipeline liners and also it is used for coatings and liners for variety equipment, valves, and components. Normally a joint is made by using a flange or coupling. For Long distance pipelines these methods can be expensive in terms of materials and installation labor. In order to reduce costs techniques are being developed that employ joint welding similar to steel pipeline construction. Energy Dissipation Stations. In many cases, slurry pipelines transport material from high mountainous locations to lowland facilities. In order to avoid abrasive line velocities and maintain control of the flow energy dissipation is required. Several mean are available such as drop boxes, cascades, and choke stations. For most long distance mineral concentrates, choking is used to

1443

control the flowrate. The choke material is composed of abrasion resistant ceramics or composite materials.

Figure 17: View of slurry pipeline choke bean and holder assembly. SUMMARY AND CONCLUSIONS Over the last several years technical advances in terms of slurry fluid mechanics, equipment and materials have made slurry pipelines (minerals, concentrates, and tailings) a very reliable and dependable transportation method. It is expected that advances will continue.

REFERENCES Aude, T. C., Thompson, T. L., (Edited by Kulwiec, R. A.), Materials Handling Handbook. 2”d Edition, John Wiley & Sons, Inc., 1985. Darby, R., “Take the Mystery out of Non-Newtonian Fluids,” Chemical Engineerinrr, March 2001. Durand, R., “The hydraulic transportation of coal and other materials in pipes,” Colloq. Of National Coal Board, London, U. K., November 1952. Hanks, R. W., “Low Reynolds Number Turbulent Pipeline Flow of Pseudohomogeneous Slurries”, Fifth International Conference of the Hydraulic Transport of Solids in Pipes, May 1978, Hanks, R. W., Ricks, B. L., “Laminar-turbulent transition in flow of psuedoplastic fluids with yield stresses”, Journal of Hydronautics, October 1974. Hanks, R. W., Ricks, B. L., “Transitional and Turbulent Pipe flow of Pseudoplastic Fluids”, Journal of Hydronautics, January 1975. Hanks, R. W., Dadia, B. H., “Theoretical Analysis of the turbulent flow of non-Newtonian slurries in pipes”, AIChe Journal, May 1971. Miller, J. E., Miller, J. D., “Slurry Abrasion Testing - The Miller Number and The SAR Number”, White Rock Engineering, Inc., Dallas, Texas Montfort, J. C., “Operating experience is described for Black Mesa coal- slurry pipeline,” Oil and Gas Journal, 27 July 1981. Rao, P. V., Buckley, D. H., “Solid Impingement Erosion Mechanisms and Characterization of Erosion Resistance of Ductile Metals,” Journal of Pipelines, Elsevier Science Publishers B. V., Amsterdam, 1984. Ricks, B. L., Yearbook of Science and Technology, “Slurry Pipeline Transportation”, McGrawHill (Parker, S. P., Publisher), 1989. Ricks, B. L., private communication, June 2000. Ricks, B. L., Aude, T. C., “Slurry Pipeline Valves”, Proceedings of International Freight Pipeline Conference, May 1989. Ricks, B. L., Aude, T. C., “90’s Pipeline Design Methodology - Dynamic Modeling”, 17” International Conference on Coal and Slurry Technology, CSTA, Washington, D.C., April 1992. Ricks, B. L., Connelly, M., Moreiko, F., “The Alumbrera Pipeline, The World’s Longest Copper Ore Concentrate Slurry Pipeline,” 1998 SME Annual Meeting, Orlando, Florida, March 1998.

1444

Robinsky, E. I., Thickened Tailings Disposal in the Mining Industry. E. I. Robinsky Associates Limited, Toronto, Canada, November 1999. Thompson, T. L., “From Reynolds to Wasp - Development of an Industry,” presented at the Sixth International Technical Conference of the Slurry Transport Association, Las Vegas, Nevada, March 198 1 . Von Essen, J. A., Ricks, B. L., “Optimum Design of Pipeline Slurry Storage Tanks and Agitators,” Chemical Engineering Progress, November 1999. Also presented at Hydro Transport in The Netherlands, September 1999. Wasp, E. J., Kenny, J. P., Gandhi, R. L., Solid-Liquid Flow S l u m Pipeline Transportation, TransTech Publications, Clausthal, Germany, 1977. Wiedenroth, W., “An Experimental Study of the Wear of Centrifugal Pumps and Pipeline Components,” Journal of Pipelines, Elsevier Science Publishers B. V., Amsterdam, 1984.

1445

The Selection and Sizing of Conveyors, Stackers and Reclaimers Greg Barfoot', Dave Bennett' and Martin COP

ABSTRACT The selection of the conveying, stacking and reclaiming equipment requires not only determination of the width, speed and size requirements, but also the optimization of the layout and equipment interactions. This paper outlines the traditional methods of selection and sizing of conveying, stacking and reclaiming equipment as well as reviewing computer simulation techniques that are valuable tools in the detailed design of these systems. Selection and sizing equipment as a system produces a plant that minimizes capital and operating cost and maximizes availability. INTRODUCTION In recent years the tools available to the material handling engineer for the design of conveying, stacking and reclaim systems have improved dramatically. This improvement has been in two main areas. The capacity and availability of computer design tools, and the ability to measure and document the performance of existing systems. This paper will focus on the selection process. Determining what type of system is appropriate and developing the information that can be used for more detailed calculations. It will also assist the reader to interpret some of the results generated by advanced design tools. BELT CONVEYOR SYSTEM TYPES AND SELECTION CRITERIA This section will briefly outline the basic types of conveyor systems available and the key features for each system that influence the decision making process. Conventional Troughed Conveyor The most common type of conveyor system. Very versatile catering to a wide range of tonnages and applications. Components for this type of system are freely available from multiple sources all over the world. Capable of throughputs up to 10,000 mtph, this type of system can operate at speeds up to 10 d s . Conventional belts can typically negotiate inclines and declines up to 18 degrees and have been used in horizontally curved applications with radii of a few thousand meters depending on the belt strength and tensions. Key Features: 1.

2. 3. 4. 5.

High tonnages, large lump size Readily available components High speeds Up to 18 degree inclines Horizontal curves.

' Fluor Daniel, Denver, Colorado

* Fluor Daniel, Vancouver, Canada RAHCO International, Spokane, Washington

1446

Pipe Conveyor Pipe conveyors are similar to conventional troughed conveyors in terms of the components used but rather than forming a trough the belt is formed into a pipe. This type of system has the advantage of enclosing the transported material. This has benefits where the material needs to be protected or where dust from the material may be a problem. The disadvantages are lower tonnages and smaller lump size. Key Features: 1. 2.

3. 4.

Lower tonnages than conventional troughed systems Smaller lump size, may require extra crushing Readily available components (uses most of the same components as conventional belts Capable of higher inclines and tighter horizontal curves than conventional belts.

Cable Belt Unlike conventional and pipe conveyors, cable belt systems utilize two steel cables to provide the driving force of the system. The belt itself rests on top of these cables. These systems have some particular advantages for long distance conveying systems. Because the driving force is delivered through the cables intermediate drives can be used without the need for transferring material. The system can negotiate tighter horizontal curves than conventional systems. Key features: 1. Lower tonnage and smaller lump size than conventional systems 2. Specialized components 3. Capable of tighter horizontal curves than conventional systems 4. Some technical advantages for long distance systems.

Steep Angle Conveyor Systems For applications where inclines exceed 18 degrees, even to vertical applications, there are a number of systems, which are capable of conveying material. Conveyors where the material is kept in place by buckets and may also include walls as part of the design. Sandwich type systems where the material is kept in place by a second belt pushing down on top of the material. Key features: 1. Capable of high inclines, up to vertical 2. Specialized components 3. Lower tonnages and smaller lump size than conventional systems.

BELT CONVEYOR SYSTEM COMPONENTS Regardless of the type of conveyor used the basic components of a conveyor system utilize similar names. Although the following is based on a conventional troughed conveyor the basic components can be applied to most types of conveyor system. Figure I shows the basic components that make up a belt conveyor system. Feed Chute Loads the material onto the conveyor system. It is designed to minimize impact on the belt and provide some acceleration of the material in the direction of travel. This area can employ rock boxes or specially designed curved chutes to minimize wear and are typically lined with wear resistant material.

1447

FEED CHUTE

DISCHARGE CHUTE TROUGHED CARRYING IDLERS CONVEYOR BELT

TAKE UP SYSTEM

Figure 1 Basic conveyor system components

Loading Skirts This area prevents spillage of the material until the load has stabilized after the loading area. It also ensures that the material is centrally loaded on the conveyor. Usually employs a sealing mechanism to prevent spillage and to aid in dust collection. Troughing Carrying Idlers The carrying idlers form the belt into a trough and carry the material load. Typically forming the trough into three sections, the outside sections are typically at 20, 35 or 45 degrees to the center. Higher capacity systems will utilize the higher toughing angles. Troughing Conveyor Belt Usually manufactured from rubber around a reinforcing carcass of fabric layers (lower strength applications) or steel cables (high strength applications). Other materials such as PVC are utilized for special applications such as high temperature or flameproof applications. Discharge Chute Takes the material stream from the conveyor and directs it to the next component in the system. Head Pulley The belt flattens out as it passes over the head pulley to be returned to the loading point along the return strand of the system. This is usually the high-tension area of the system and as such, the head pulley is often the drive pulley (the place where the motor drives the system). The power from the motor is transmitted to the belt by the friction between the pulley and the belt surface. Take-Up System This system maintains the tension in the system. The tension must be high enough to prevent too much sag between idlers, particularly on the carrying side, and to prevent slip at the drive pulley due to the power being imparted into the belt. Tension is adjusted by physically adding or removing belt from the system using the pulleys. The movement can be created by gravity type systems, which use a large weight, or by a winch arrangement.

1448

Return Idlers These rolls carry the empty belt back to the loading point. Because there is no material loading these a spaced at longer intervals than the carrying idlers and can be flat or on longer systems form a small ‘v’ trough to keep the belt running in the center of the structure. Impact Idlers At the loading point the vertical forces produced by the material falling onto the belt must be absorbed by the impact idlers. To accommodate this, these idlers are spaced very close together and are usually covered in impact absorbing material such as rubber. Tail Pulley Re-directs the belt back to the loading point from the return strand of the system. BASIC BELT CONVEYOR SYSTEM CALCULATIONS As with the previous section, these calculations are based on the conventional troughed conveyor system although they can frequently be applied to other types of systems as well. Conveyor System Power and Belt Strength There are a number of methods and standards for calculating the power and belt tension (required belt strength) for a conveyor system. They are all, however, based on the same principle. The conveyor system is modeled as a mass being dragged over a surface as shown in Figure 2. The friction between the mass and the surface is used to represent the losses in the idlers, indentation of the belting itself and the flexing of the material as it passes over the carrying idlers. The forces required to lift or lower the material itself is a matter of simple physics. The calculation outlined below is a combination of a number of methods and forms a very simple calculation for tension and power. This calculation is designed to give basic results suitable for conceptual layouts, cross checking of other results and for basic technical viability decisions. It is strongly recommended that more detailed analysis be undertaken for the detailed design of a conveyor system.

-

I -

Figure 2 Conveyor System Model Material Loading In order to proceed with a basic conveyor calculation we need to establish some design criteria. The first being material throughput. Based on the required tonnes per day, and an estimate for conveyor availability, a design throughput can be established. For example: If a system is required to deliver 36,000 tonnes per day with an expected availability of 18 hours per day, the required design throughput for the conveyor is 2000 tonnes per hour. The next criteria to select are belt speed. Standard conveyor speeds vary from 0.5 m/s up to 6.5 d s . The lower speeds are suitable for low capacity systems, a few hundred tonnes per hour, or where the properties of the material may require low velocity. Such as very light material easily moved by air currents. Typical speeds for most applications will fall between 1.5 m / s and 3.5 d s .

1449

For high capacity systems or long conveying lengths, greater than 2 km, speeds up to 6.5 m / s can be used. In order to determine the material loading required in following calculations we use Equation ( 1). m=- 1000 T 3600 v

Where m = mass of material [in kg/m] T = design throughput [in tonnes/hour] v = belt speed [in m/s] Belt Width Selection The bulk density of the material is the next criteria required. Standard values for the bulk density of most common materials can be found in many design standards. This value combined with the result of Equation ( I ) allows us to calculate the cross sectional area of the material stream using Equation (2). A=-

m

P Where A = cross sectional area of material [in m2] m = mass of material [in kg/m] p = material bulk density [ in kg/m3] Once the required cross sectional area of the material stream is known, the profile of the conveyor belt and material can be estimated. This can be done graphically or by using one of the methods described in conveyor design standards. The result of this exercise will be the width of the belt that is actually in contact with the material. A fully loaded conveyor belt requires a certain amount of clearance at each edge. This clearance prevents spillage of material. As the maximum size of material lump increases the required edge clearance to prevent spillage will also increase. The calculation of usable belt width (or the amount of belt actually in contact with the material), for a standard three idler configuration where material lump size is not a factor, is given in Equation (3). b = 0.9 B - 0.05 ($or B < 2m) b = B - 0.25 ($or B > 2m)

(3)

where b = usable belt width B = actual belt width

Conveyor system calculations are an iterative process. At this point a mass for the conveyor belting should be selected and used for the following calculations. Once we have our first result, the conveyor belt rating is known and the selected value for belt mass can be checked. It may be necessary to repeat the process several times to obtain the final result. Idler Loading An estimate of idler mass is required to proceed with the calculations. This can be determined from manufacturers data based on the idler vertical loading that is calculated using Equation (4). Li = !E!L

(4)

a

1450

Where Li = idler load [in kg] m = material loading [in kg/m] m/, = belt mass [in kg/m] a = idler spacing [in m]

Slope Resistance The first part of the driving force for the conveyor, the force to lift or lower the material, is called the Slope Resistance and is described by Equation (5). F, =m.g.h (5) Where F, = force to lift (or lower) material [in N] m = mass of material [in kg/m] g = acceleration due to gravity h = change in elevation from loading to discharge [in m] Main Resistance The Main Resistance of the conveyor is the force required to move the belt over the idlers overcoming rolling (idler bearings and seals), indentation (of the rubber belt) and flexure (of the material) resistance. The Main Resistance is calculated using the formula in Equation (6). F,= f Lg[mR+rnc+(2m,+m,)cosd]

Where F, =Force to move the belt [in N] f =Friction coefficient L=Length of conveyor[in m] g =Acceleration due to gravity mR=Mass of return idler rotating parts [in kg/m] m, =Mass of carry idler rotating parts [in kg/m] m, =Mass of belt [in kg/m] mM =Mass of material [in kg/m] 8=Angle of Incline[in degrees]

The mass component of the basic friction force equation shown in Figure 2 is replaced with the mass of the conveyor system calculated from the length of the conveyor and the weight of the components in mass per unit length. This illustrates the basic principle of how a conveyor system is modeled and the forces required to drive the system calculated. The selection of the friction factor f is crucial to the calculation outlined in Equation (6). A typical value for this friction factor is 0.02, but conditions can exist which vary the value between 0.013 and 0.03. Conditions that require higher values of friction factor include short conveyor systems, poor alignment or difficult conditions, low tensions and high sag values and large spans between idler sets. Conversely, some conditions that require lower values of friction factor are long conveyor systems, well aligned low resistance systems, and low energy loss designs for belting and components. Secondary Resistance The secondary resistance is a component in the calculation which accounts for the forces created by belt cleaning devices, losses in pulley systems, drag forces from skirting and other miscellaneous items not included in the main calculation. Generally, these items represent a small fraction of the total forces but for short length conveyors, less than one kilometer, they can be significant. A reasonable approximation of secondary resistance can be made by increasing the main resistance value by 5% for conveyors between one kilometer and five kilometers in length.

1451

Longer than five kilometers the value of secondary resistance is negligible for this type of calculation. For conveyors less than one kilometer in length the secondary resistances are significant and worthy of individual calculation. Formula for calculating these are available in most of the recognized conveyor design standards. Once these basic calculations are complete there is sufficient information to make an estimate of the size of the conveyor components required for the system. Some iterative calculation will be necessary to make an accurate selection of belting and idlers for the application. In most cases, further calculations will be necessary to develop the detailed design of the system. These further calculations are beyond the scope of this paper and would generally be done by a knowledgeable material handling specialist.

ADVANCED BELT CONVEYOR SYSTEM DESIGN TOOLS The calculation outlined in the preceding section is designed to allow a rough calculation of conveyor forces for the purpose of estimates, feasibility studies or as a cross check of information supplied by others. There are of course, many detailed calculations related to conveyor system design that are outside of the scope of this paper. However, the reader should be aware of these and what purpose they serve in the development of a conveyor system design. It is recommended that once the project has progressed passed the basic conceptual stage a material handling engineer be employed to complete the detailed calculations and design of the system. Drive Calculations Once the tensions and driving force for the conveyor system has been calculated, the layout of the drive system can begin. The detailed calculation will include the required wrap angle for the drive. The wrap angle is simply the amount of belt in contact with the drive pulley surface; the more contact area, the more force can be transmitted to the belt by that drive pulley. In a high capacity conveyor, i t is usual to have multiple drive pulleys to effectively increase the contact area between the belt and pulleys and allow transference of sufficient power to drive the system. This calculation will include variables such as angle of wrap, friction coefficient (between the pulley and belt) and belt tension before and after the drive. Take-Up Calculations The size, type and location of the take-up system will vary from system to system. Take-up location is a function of the drive location and the belt tensions. For example, in a high lift conveyor system the drive will most likely be located at the top, or head end, of the conveyor. This is the high tension area of the system and locating the drive at this location means there are higher forces between the belt and the pulley surface making the chance of the drive pulley slipping very low. As one of the functions of the take-up is to maintain tension at the drive during start up of the system it is often beneficial to locate the take-up immediately after the drive, however, locating a take-up in a high tension area means the take-up itself needs to be very large. It may make more sense to locate the take-up at the tail end of the system thus reducing the size of the take-up required. Vertical Curve Calculations This involves selecting the radius for vertical curves in the system (where the incline of the conveyor changes). The selected radius needs to ensure that the belting will not lift out of the idlers, for a concave curve, or exceed the rating of the belting and/or idlers, for convex curves. The calculation takes into account the mass of the belting, how the belt tensions vary with material loading, particularly for partially loaded conditions, and how the running forces of the system may vary with temperature and operating conditions. Horizontal Curve Calculations Horizontally curved conveyors are becoming more common. The ability to have a conveyor system negotiate curves can eliminate transfer points and optimize the design of the system. Calculating the minimum radius for a horizontal curve is a matter of balancing the gravitational

1452

forces, the weight of the belt and material, with the tension forces trying to pull the belt into the curve. T o balance these forces the idlers for the system are angled opposite to the direction of the curve, i.e. opposite to a camber used on a highway. As with vertical curves this calculation needs to account for varying load conditions and the effect of temperature and operating conditions.

Dynamic Calculations Some of the more advanced computer models for conveyor systems are capable of simulating the changing tensions and velocities that occur during starting, stopping and load variations of the system. These calculations are essential in designing long, complicated belt conveyors and can identify major problems with operation at the design stage. Typically, however, such levels of calculation are not required for standard, straight belt conveyors. For a given conveyor system these programs can give good estimates for drive torque requirements, test starting control options, identify peak tensions for vertical and horizontal curve calculations, evaluate braking and stopping functions, determine take-up requirements and identify potential tension problems arising during starting and stopping. Figure 3 and Figure 4 show some examples of' the results generated by dynamic calculations.

Figure 3 Dynamic Calculation of Belt Tension

Figure 4 Dynamic Calculation of Belt Velocity

1453

STOCKPILE SYSTEM TYPES AND SELECTION CRITERIA Stockpiles in bulk solids handling system generally provide a buffer between two processes or modes of transportation. At a transshipment facility, stockpiles absorb the difference in the interarrival time of incoming and outgoing product. At a mine or quarry stockpiles are commonly used to provide surge capacity up stream of a mill or crushing and screening circuit to ensure a uniform feed is available to the process. In addition to providing surge storage, stockpiles at iron and steel plants and other process facilities are often used to blend different materials before being reclaimed and delivered to the downstream process. Coal fired power plants use stockpiles to ensure both short term and long-term coal stocks are available. One additional and unique “stockpile” is the stacking of and sometimes reclaiming of heap leach material (most commonly found in the gold and copper industries). The type, size and purpose of the stockpile have a direct impact on the methods of stacking and reclaiming to be used. Before a specific method of stacking and reclaiming can be considered it is important to established some basic criteria, the following are some of the main factor that must be considered:

1. Does the pile have to be covered? 2. Does material have to be stacked and reclaimed simultaneously? 3. Is a first in, first out (FIFO) stacking and reclaim system required? 4. How many products have to be stockpiled? 5 . Is cross contamination of products a concern? 6. What are the material characteristics? Bulk density; Moisture content; Size distribution; Friability; Hardness, Susceptibility to spontaneous combustion or oxidation, Stickiness, Abrasiveness, etc. 7. What is the method of product delivery to the stockpile facility and how uniform is it? 8. What is the method of product delivery from the stockpile facility and how uniform is it? 9. How much storage is required? 10. What proportion of live to dead storage is required? 11. For a blending bed or stockpile, the proportions and availability of the constituent elements have to be defined. 12. In heap leaching applications, is it a multiple lift (permanent pad) or an on/off (reusable pad)? This paper cannot cover all of the potential applications and materials that require stockpile facilities; consequently, it will deal with the typical requirements at coal and hard rock mines and transshipment facilities. Once the basic criteria have been established the next step in the design of a stockpileheclaim system is to select the basic layout. This section reviews the basic conceptual layouts and some of the advantages and disadvantages of the commonly used systems: Stacking systems 1 . Fixed stacker - conical pile 2. Overhead tripper or shuttle - longitudinal pile 3. Radial stacker - circular or kidney pile 4. Cascading conveyor system with radial stacker - rectangular or odd shaped pile 5 . Traveling stacker or stacker heclaimer - longitudinal pile 6. Mobile conveyor with traveling tripper - longitudinal and/or circular piles 7. Silo storage - fed by a conveyor.

Reclaim systems

’

Front end loaders to a loading hopper 2. Gravity to a feeder system below the pile or silo 3. Bucket wheel reclaimer 4. Scraper reclaimer 5. Drum reclaimers. 1.

1454

STACKING SYSTEMS Fixed Stacker A fixed stacking system typically consists of a cantilevered conveyor, which creates a conical stockpile generally up to 50m in height, although larger systems are possible. The reclaim from such a pile can be by front-end loader for intermittent, low through put facilities. For facilities requiring an automated uniform reclaim rate, vibrating, apron or belt feeders are normally located in a tunnel beneath the pile. The feeders draw material from the pile and discharge it onto a belt conveyor for transportation to the down stream process. The number, type and size of the feeders selected are normally based on the material flow characteristics, feed rate required and the live to dead ratio required in the pile. For most situations, this is lowest cost stacking/reclaim system. However, it is not efficient in terms of the live to dead ratio. With a single line of feeders the live capacity often less than 20% of the total even with relatively free flowing material and with only a single feeder with less free flowing material the live capacity may be less than 10%. However, this is frequently a good option where space is readily available and the dead portion of the pile can be utilized to provide feed to the plant during infrequent interruptions in the stacking system. When stacking and gravity reclaiming from the live portions of the pile a FIFO (first in first out) regime exists and size segregation is not normally a problem. However, during initial pile building, natural size segregation occurs which results in coarser material collecting at the edges of the pile with finer products at the center. To reduce potential dust problems, a telescoping chute can be fitted to the head of the stacking conveyor but these units can be a major wear and maintenance item in situation where abrasive coarse ore is being handled. If necessary, the system can easily be covered to improve dust control and protect the product from the elements. Overhead Tripper or Shuttle A fixed conveyor with a traveling tripper or a reversible shuttle conveyor is supported on a structural frame over the stockpile. An example of this type of system is shown in Figure 5. Material is discharged from the tripper or shuttle directly into the stockpile or storage bin system below. This stacking system allows materials to be stacked in different piles or to be blended in a single pile. The other main features associated with the fixed stacker system are also applicable to the overhead tripperhhuttle, with the exception that a telescoping discharge is not practical on a tripper. The most significant differences between the overhead tripperkhuttle and the fixed stacker are the size of stockpile and the possibility to use automated reclaimers, as an alternative to gravity reclaim. The overhead tripperhhuttle system is also considerably more expensive than a simple fixed stacker. Reclaiming from a longitudinal pile formed by an overhead tripperkhuttle can be by gravity to feeders beneath the stockpile, as described above for fixed stackers, or by automated mechanical reclaimers such as bridge mounted bucket wheels, drum reclaimers, portal or full face scraper reclaimers. These automated mechanical reclaimers generally have limitations on the lump size they can handle and depending on the product, they can be a source of high wear and maintenance. Additional aspects of these reclaimers are discussed below in the section on traveling stackers. When under pile reclaim feeders are used they can be arranged along the length of the pile, as is commonly used at gravel plants, or across the width of the pile to feed different process lines as often used at base metal concentrators.

1455

Figure 5 LAXT Coal Terminal (Courtesy KRUPP) Radial Stacker There are several configurations for this type of stacker. The most common are those that comprise an inclined belt conveyor supported on a pivot at the loading point and a structural leg approximately two thirds along the conveyor. The structural leg can be mounted on pneumatic tires or steel wheels and a curved rail. The stacker slews about the pivot point to form a kidney shaped pile. The gravity reclaim systems used with this type of stacker are similar to those described above for both fixed stacker and overhead tripperkhuttle type systems. The arc described by the inclined radial stacker can be 270 degrees or more. However the most efficient live to dead ratio that can be achieved using a single reclaim tunnel occurs when the center line of the reclaim tunnel forms the cord of an arc described by a 60 degree slew of the stacker. Automated mechanical reclaimers are not normally used on kidney shaped stockpiles and it is difficult to efficiently cover this type of stockpile. This type of stockpile is commonly used in quarries and gravel plants. Another type of radial stacker that is used in automated systems for blending, stacking and reclaiming comprises a pedestal mounted radial conveyor boom. Product is discharged from a feed conveyor to the radial stacking boom at the central pedestal. The stacking boom can luff and slew as it slowly rotates around the central pedestal, thus allowing product to be discharged in chevron layers or as cone shells to form a continually rotating “kidney shaped” stockpile. The open part of the “kidney” is where the reclaimer operates, it follows in the same direction as stacker boom, reclaiming material from the pile and discharging it at the central pedestal, onto an under ground transfer conveyor. Reclaimers in this type of system are normally of the bridge mounted bucket wheel or scraper reclaimer type. They are supported at one end by the central pedestal and the other on a circular rail around the perimeter of the pile. This type of system can be fully automated and is ideal for stacking and blending relatively fine material (minus 25 mm). It is not suitable for a coarse ore storage system. The maximum economically practical Reclaimer Bridge defines the limit on size for this system. Although 40m span reclaimers have been built, 35 m spans may prove to be a more economic length, this would result in a system diameter of over 75 m and if required could be fully enclosed. A key feature of these mechanical reclaim systems is that the entire stockpile is live and they are better at handling less free flowing materials that may be very difficult to reclaim with a gravity based system. Cascading Conveyor System The cascading or commonly called “grasshopper” system is often used to build piles of ore for heap leaching. This system incorporates a series of portable conveyors, which feed another in a “cascading” fashion. The last portable conveyor in the line discharges material onto a “bridge” conveyor, which is, positioned 90” to the portable system. The bridge conveyor, a horizontal conveyor mounted on either rubber tires or crawler tracks or a combination of the two, is generally 20- 1 10% longer than the portable cascading conveyors to allow flexibility when removing one (or

1456

two) portable conveyor(s). In turn, the bridge conveyor feeds a radial stacker, which can be either rubber tire or crawler track mounted. The heaps are constructed in a retreat fashion. An example of this type of system is shown in Figure 6.

Figure 6 Cascading or “Grasshopper” System (Courtesy RAHCO) Traveling Stacker or StackerIReclaimer These machines are commonly used on longitudinal stockpile systems at transshipment terminals and process plants. Although traveling stackers can be used with gravity reclaim systems they are usually combined with mechanical reclaimers and are often built as combined stackerh-eclaimers. The stackers run on rail tracks along the side of the stockyard and are equipped with a tripper that raises the feed conveyor belt to transfer the product onto the stacker’s boom conveyor, which in turn discharges it into the stockpile. There are several different configurations of stacker boom such as: fixed or luffing, single or double booms and single boom machines that luff and slew. The selection of stacker type must take into consideration the reclaim system to be used as well as the stockyard layout to be adopted. The length of the stacker boom is directly related to the width of stockpile required and may be influenced by the method of stacking to be adopted. Although not commonly used today windrow stacking was used as a method of reducing the effects of size segregation that occurs with chevron stacking. The cost of the longer boom required to build windrows is difficult to justify in most applications so that IuffingMewing stackers that reach the center of the stockpile and build chevron or cone shell piles have become more common. Boom mounted bucket wheel reclaimers also run on rail tracks along side of the stockpile and varies in boom length and capacity. Boom lengths of up to 60 m are common at coal and iron ore terminals with reclaim rates in the order of 6,000 t/h to 10,000 t/h. Many terminals prefer to use multiple combined stacker/reclaimers that provide additional versatility when managing a large number of grade or types of material in a stockyard. Traveling stackers are also used to stockpile material in blending beds and in stockpiles where portal scraper reclaimers are used. To build an efficient blending bed the entire constituent parts of the blend must be laid down in the pile so that they will occur consistently in a predetermined mix when the pile is reclaimed. The best blending bed reclaimers are therefore machines that take a full slice from the exposed face of the stockpile. Typical machines that fall into this category are drum reclaimers, bridge mounted bucket wheels and scraper reclaimers. The maximum width of this type of stockpile is set by the economic span of the reclaimer that is in the order of 40 m. Where precise blending is not required larger wider stockpiles can be reclaimed with a portal type scraper reclaimer that can span pile widths of 60 m.

1457

The maximum economic capacity of reclaimers for stockpile use varies depending on the type of machine and the materials to be handled. Bucket wheel and drum reclaimers with capacities of 6,000 t/h are relatively common in coal terminals and machines with capacities up to 10,000 t/h are not uncommon at iron ore terminals. The main advantages of the rail mounted stacker and reclaimer systems are their flexibility to handle multiple products with varying flow characteristics with little or no dead storage. While these systems can be designed to handle lump sizes of 150 mm, they are better suited to handle material of less than 75 mm. For most facilities using this equipment, it is not practical to cover the stockpile system and alternative methods of dust control have to be adopted. Figure 7 shows an aerial view of a traveling stacker system. Figure 8 shows a closer view of this type of stacker and reclaimer.

Figure 7 Traveling Stacker and Reclaim Systems (Courtesy KRUPP)

Figure 8 Blending Yard Stacker and Portal Style Reclaimer (Courtesy FAM)

1458

Heap Leach Stacking and Reclaiming System The reusable or on/off leach pad has become very common particularly in the copper industry. One of the main differences between the standard blending yard or port facility and the reusable leach pad is the strict “first-in, first-out’’ material flow. Typically, a pair of rectangular leach pads are placed side-by-side with feed and discharge conveyors in the corridor between the pads. A mobile tripper (usually rail-mounted) transfers material from the feed conveyor to the Mobile Stacking Conveyor (MSC) as shown in Figure 9. The MSC is mounted on crawler tracks and has a rail-mounted tripper, which travels the length of the MSC as it discharges material to form the heap for leaching. Once leached, the material is reclaimed using a Bucketwheel Reclaimer (BWR) and fed into a rail-mounted hopper, which travels the length of a second crawler-track mounted mobile conveyor, the Mobile Reclaim Conveyor (MRC) as shown in Figure 10. Once the BWR has completed one pass along the face of the heap (parallel to the MRC), the BWR and MRC both take a step toward the pad and repeat the process Approximately 1.O- 1.5 m of material is removed with each pass. The reclaimed spent ore is then transferred to a mobile hopper, placed on the spent ore conveyor and stacked on a permanent spent ore pile. A shiftable conveyor and a large crawler mounted spreader or a third mobile conveyor typically stack the spent ore.

Figure 9 Typical Reusable Heap Leach Pad (Courtesy RAHCO)

1459

Figure 10 Mobile Stacking Conveyor (Right) and Mobile Reclaim Conveyor with Bucketwheel Reclaimer (Courtesy RAHCO) Silo storage This type of storage system is used in special circumstances where relatively small quantities of product are to be stored, or some process or product requirements favor such an enclosed environment. The main use of these silos in a mining related environment is at high-speed rail load out facilities for coal. A typical 10,000 t capacity silo is filled by a conveyor directly from the coal preparation plant and discharges directly into the moving train as it passes through the silo. The silos are normally slip or jump formed concrete structures with multiple loading gates at the bottom to control the flow of coal into the railcars, refer to Figure 1 1. Silos are also used for fine ore storage at some hard rock mines as surge storage ahead of agglomerators at heap leach operations where typical capacities would be in the order of 6,000 t.

Figure 11 Silo Storage (Courtesy FAM)

1460

Table 1 Common Repose Angles, CEMA Belt Conveyors for Bulk Material, Fifth Edition, Pages 29-43 Material Ash, coal Ash, fly Bauxite, mine run te, -75 mm Lime Cement. Portland Clay, dry, fines Coal Coke Copper Ore Diatomaceous earth Earth, dry as excavated Earth, wet containing clay Gravel, pebbles Gypsum -12 mm Gypsum, 30-60 mm Ilmenite ore Iron ore Iron ore pellets Kaolin Clay -75 mm Lignite, air dried Limestone, crushed Phosphate rock, broken, dry Phosphate rock, pulverized Potash (muriate), dry Potash (muriate). ,, mine run Potassium nitrate Potassium, sulfate Quartz Salt, common dry, fine Salt cake, dry, coarse Sand, bank, damp Sand, bank, dry Sandstone, broken Soda ash, briquettes Soda ash, heavy Soda ash, light Sodium phophate Taconite, pellets Zinc ore

Angle of Repose 45" 42" 31" 30-44" 30-44" 30-44" 35" 27-40" 30-44" 30-44" 30-44" 35" 45" 30" 40" 30" 30-44" 35" 30-44"

35" 30-44" 38" 25-29" 40" 20-29" 30-44" 20-29" 45" 20-29" 25" 38" 45" 35" 30-44" 22 35" 37" 37" 30-44" 38"

1461

BASIC STOCKPILE SYSTEM CALCULATIONS Calculating Live Storage Calculating the live storage of a stockpile is pure geometry. Using the estimated reposes and draw down angles of the material, refer to Table 1, a geometrical shape of the full and empty (fully reclaimed) stockpile is generated. Many three dimensional CAD packages are extremely useful for creating these shapes. The live storage of the stockpile is the volume of the completely reclaimed stockpile minus the volume of the full stockpile.

CONCLUSION The handling of bulk material via conveyors is quite common and has become much more flexible in the last decade. A wide variety of materials are conveyed, stacked and reclaimed using a broad range of equipment in many configurations. This paper has attempted to introduce the selection and sizing of this equipment and will provide a base from which a conveying, stacking and reclaiming system could be developed. As many a user of such equipment has said, “the only limit is our own imagination.” ACKNOWLEDGMENTS The authors would like to thank the equipment manufacturers noted in this paper for the use of information and photographs of material handling systems. REFERENCES Belt Conveyors for Bulk Material, Fifth Edition, Conveyor Equipment Manufacturers Association. I S 0 5048 - Continuous Mechanical Handling Equipment, Belt Conveyors with Carrying Idlers, Calculation of Operating Power and Tensile Forces. 1989

1462

Selection and Sizing of Concentrate Drying, Handling and Storage Equipment Michael E. Prokesch, P.E., FFE Minerals Inc., Grant Graber, P.Eng., AMEC

ABSTRACT The selection of appropriate drying equipment is a function of a material’s physical properties and drying rate, and desired product characteristics. Proper material characterization and analysis must be followed by equipment selection based on established criteria. This paper covers material characterization, reviews available drying systems with a focus on rotary, fluid bed and flash systems, and outlines basic selection criteria given the advantages and disadvantages of each. Concentrate handling and storage equipment and systems are also reviewed, taking numerous factors into account including moisture content, bulk material properties, climatic conditions, transportation logistics, and market conditions. DRYING Drying is the use of thermal energy to reduce the moisture content of a material through vaporization. Typically, free moisture is removed, as opposed to bound or crystalline water that . often requires higher temperatures and specific energy input. Since mechanical dewatering generally cannot reduce concentrate moisture content below 10-20%,drying is frequently required to attain moisture levels that maximize the thermal efficiency of a downstream process, provide the desired consistency for handling and blending, satisfy end-use specifications, etc. Available drying systems handle materials ranging from slurries to filter cakes to free-flowing crushed ores. Many of these systems also dry pyrophoric ores, using controlled gas composition and temperature to limit the potential for oxidation. Proper system selection must be based on a thorough understanding of material properties and product requirements. DRYER SELECTION CRITERIA Material Properties. The selection of a dryer is highly dependent on the material’s physical and specific drying characteristics. Drying is a three-step process (see Figure 1): first, sensible heat brings the material to the vaporization temperature. The surface then remains at a constant temperature until the critical moisture point, at which all its moisture has been removed. Final drying occurs at a decreasing rate, as surface temperature increases and internal moisture is driven to the surface and volatilized. During this stage, materials with high levels of internal moisture may be subject to high stresses, resulting in structural degradation. The drying rate curve characterizes a material’s basic drying properties. This is developed by drying a material at a constant gas temperature and determining the rate of weight loss. Figure 2 shows drying curves for oil shale, clay and garnierite ore. The curve does not provide an absolute timekemperature relationship for dryer design, but gives a relative comparison to other materials, so that required dryer operating conditions can be calculated from experience. The drying rate curve is used to select dryer operating parameters, rather than dryer type.

1463

FALLING RATE PHASE

1

CONSTANT RATE PHASE

CRITICAL MOISTURE POINT

K

w LL u)

SENSIBLE HEAT PHASE

z a K

/

I-

i

I LL

0 W

5K

MOISTURE CONCENTRATION

0

Figure 1 Drying cycle The physical properties of the dryer feed stream have the largest impact on dryer selection and design. These include moisture content, disposition of moisture (surface and internal), particle size distribution, density, particle integrity (i.e., friable), temperature sensitivity, abrasiveness and corrosivity. Of these, particle size distribution and consistency are used to define the category of dryers suitable to an application. Dryer Classification. Dryers may be classified as adiabatic (direct heat transfer) and nonadiabatic (indirect), or by continuous or batch operational mode. Some designs incorporate both direct and indirect heat exchange, but their application is limited. Most concentrate drying applications require large continuous throughputs and maximum thermal efficiency, so this paper will describe only continuous adiabatic systems. For the limited concentrate applications requiring non-adiabatic andlor batch operations, an expert should review the many equipment options available. These applications often include materials that cannot be exposed to combustion gases or an oxidizing atmosphere during the drying process. 25

20

z t .-a

15

10

5

n

0

3

6

9

12

15

Time-min

Figure 2 Drying rate curves

1464

18

21

24

27

30

Dryer Selection. Table 1 provides general guidelines for the initial selection of an adiabatic, continuous feed dryer system based on feed consistency and particle size. Other adiabatic, continuous feed designs are available, but these are typically modifications of the basic systems shown in Table 1, and have been excluded to simplify the selection process. Table 1 Dryer selection based on feed properties Slurries Pumpable Sludge Spray dryer Spray dryer Rotary tray dryer Filter Cakesmhick Sludges: Fine (under 6 mm) Flash dryer wlrecycle Rotary dryer Fluid bed dryer Rotary tray dryer whnert bed Band dryer Hammermill dryer wlperforming wlrecycle Free-Flowing: Fine (under 6 mm) Flash dryer Rotary dryer Fluid bed dryer Rotary tray dryer Hammermill dryer

Rotary dryer Hammermill dryer wlrecvcle

Coarse Rotary dryer Hammermill dryer

Rotary tray dryer Band dryer

Coarse Rotary dryer Hammermill dryer

Rotary tray dryer Band dryers

The term “recycle” refers to conditioning the wet feed stream with a portion of the dry product stream to obtain flow properties compatible with the dryer system. For example, a sticky sludge containing 50% moisture may be mixed with dry product to reduce its moisture content to 20-25% and produce small free-flowing agglomerates that may be properly entrained in a flash flyer’s gas stream. This mixing may be performed using a simple pug mill device. The inert bed for filter cakelsludge processing in a fluid bed usually comprises a coarse stone (i.e., 13 mm crushed limestone). This stone is maintained in a highly active state of fluidization, which enables it to break up sticky agglomerates and disperse the material through the bed for rapid drying. The inert material is replenished as it slowly degrades in the bed. Following selection of an appropriate system, additional feed properties should be considered to narrow the system options. Flash dryer, fluid bed dryer with inert bed and hammermill dryer systems are not recommended for friable materials for which fines generation must be minimized. This must also be considered when designing a rotary dryer with internals, due to the increased potential for particle degradation. Temperature-sensitive materials are not well-suited to gas suspension flash drying, because during the falling rate phase, the system’s short retention time (1-3 sec) requires a high temperature differential between the particle surface and core. Materials at risk for hightemperature degradation or reactions (i.e., sulfide ores) are better suited to rotary, fluid bed, rotary tray and band dryer systems, which have longer material retention times. Drying capacity, spatial requirements and capital cost may further narrow the options. Rotary tray and band dryers are high-cost and limited to several tons per hour, giving them a high cost per per unit of material processed. Rotary dryers require a larger footprint than other systems. Flash dryers have the greatest vertical clearance requirement (10-15 m). Design summary. A number of items must be considered in dryer system selection (although not are necessary for initial selection and sizing): Material drying properties - drying curve Free moisture level - loss at 105°C

1465

Particle size distribution Wet and dry bulk density Specific gravity Feed flow properties: slurry, sticky sludge, free-flowing, etc. Particle degradation issues Temperature sensitivity Corrosiveness Potential emissions requiring control Capacity in wet or dry mtph Mode of operation - 24 h/d or campaigned Product moisture specification Fuel type Installation area description / limitations Operating environment.

DRYING EQUIPMENT Rotary Dryers. Rotary dryers are the commonest system for mineral processing, since they offer the greatest flexibility in terms of capacity, retention time, operating temperatures, ease of operation, operational availability, the ability to process different feeds, handling variations in feed properties, and operating at reduced throughputs. A single unit can process several thousand tons per day of wet feed. A typical rotary dryer system (Figure 3) includes an unlined carbon steel shell with supports and a drive mechanism. Product purity requirements or corrosive applications may dictate more robust materials of construction. Hoods at both ends of the tube accommodate the transfer of solids and gas. Fuel is combusted in a refractory lined external combustion chamber, and the resulting hot gas enters the tube and flows parallel with the material stream. Material drops off the far end of the tube into a material handling circuit, and the dryer offgas is filtered, scrubbed if necessary and then emitted to atmosphere. An induced draft fan downstream of the offgas filter controls the static pressure (slightly negative in the discharge hood) inside of the dryer.

Fuel

Product

Figure 3 Rotary dryer system

1466

The earliest application of the rotary dryer used a straight tube. Since then, efficiency has been improved through the use of internals such as chains, lifters and dams. Figure 4 includes examples of several common lifter configurations. The purpose of these devices is as follows: Chains break apart agglomerates for increased surface area and absorb heat from the gas stream for transfer to the material load. Lifters shower the material through the gas stream. Lifter geometry is varied to adjust the angle at which the material is showered through the gas stream. Lower angles are used for fine or friable materials to minimize entrainment and degradation, or to accommodate sticky materials. Lifter sets are staggered to maximize effectiveness. Dams increase the material loading and retention time in the tube. A typical loading rate is 10-15%. Material retention time in a rotary dryer typically ranges from 15-30 minutes. Most dryers include a variable speed drive to permit changes in retention to maximize capacity. Retention time is a function of material angle of repose, shell diameter, shell length, shell slope, shell speed, gas velocity and the design of internals. Retention time changes linearly with shell speed changes.

Straight

45'

90

Figure 4 Rotary dryer lifter configurations Typical design ranges for rotary dryer systems are as follows: Diameter: 6 m maximum Length: equivalent to 5-10 times the diameter Retention time: 15-30 minutes Exit gas velocity: 3-5 mps Loading: 10% of total shell volume Shell slope: 2.4" Shell peripheral speed: 0.1-0.5 mps Inlet temperature: approximately 900°C maximum Outlet temperature: 125-150°C Evaporation load: 30- 1 10 kg/h/m3, depending on material properties. The rotary can operate in counter-flow or parallel-flow mode. Most operate in parallel mode, which permits operation with a high inlet gas temperature to maximize dryer capacity and efficiency without overheating the solids. The higher differential gas temperature in the parallel mode equates to lower specific energy consumption per unit of dry product than counter-flow. Parallel-flow may be used to dry pyrophoric ores. This requires limiting the oxygen content in the system to a maximum of 10% through gas injection or flue gas re-circulation. The rate of pyritic sulfur oxidation at this concentration is significantly retarded. In addition, the particle surface temperature must be limited to <3OO0C during the falling-rate drying period. The rotary dryer's disadvantages include high price, space requirements and maintenance costs. These should be weighed against their flexibility and ease of operation.

1467

Fluid Bed Dryers. Fluid bed dryers use fluidization or suspension of solids in a gas stream to transfer heat between gas and solids. Close contact is maintained between gas and solids through vigorous mixing, promoting higher product uniformity than a rotary dryer. These mixing and heat transfer properties, coupled with a relatively long retention time (several minutes), enable a fluid bed to easily meet low-moisture specifications. They can also functioning as classifiers to segregate coarse and fine fractions. A conventional fluid bed dryer system (Figure 5 ) includes a cylindrical vessel with an air distribution plate in its lower portion to form a plenum below the plate, and bedlfreeboard zones above the plate. Hot air from a heater passes through the plate and enters the bed. Wet feed is added through the side of the vessel at the top of the bed or through the top of the unit. Dry product either overflows from the top of the bed across from the feed inlet, or is discharged through an underflow at the level of the distribution plate. Gas containing entrained fines passes through the expanded freeboard zone, followed by a cyclone andlor baghouse for particulate removal. These fines may be kept separate from the coarse product or recombined. An ID fan downstream of the baghouse maintains a neutral static pressure in the fluid bed freeboard zone.

Wet Feed

Baghouse

h I I

I

I

I

I

Fuel Air Heater

Product

Figure 5 Fluid bed dryer system The air distribution plate generates a minimum pressure drop of -3.4 kPa, to ensure even gas distribution across the bed cross-section, essential for proper fluidization. Uneven distribution would allow the gas to channel through and over-fluidize portions of the bed; with sulfide ores, the remaining static portions would overheat and begin to oxidize. Figure 6 illustrates several modes of fluidization; active fluidization, or “bubbling bed” mode, is preferred. Static

Bed

Minimum Fluidization

Active Fluidization

Figure 6 Modes of fluidization

1468

Pneumatic Transport

A material's fluidization properties are determined in the laboratory by a cold fluidization evaluation. This enables the designer to observe the fluidization behavior of the material, determine the velocity required for minimum fluidization, determine the velocity range for active fluidization, and measure the quantity and particle sizing of the fines elutriated from the bed during active fluidization. Improper fluidization leads to poor drying efficiency, a reduction in capacity and overheating of the material bed. Round orifices, bubble caps or nozzles usually distribute the air. Plate design is influenced by material particle size distribution, abrasiveness, fluidization velocity, target pressure drop and operating temperature. Most fluid bed dryers operate with a maximum inlet temperature of 800"C, employing a stainless steel plate. Higher temperatures are possible with exotic alloys, ceramics or refractory materials. The key to the success of fluid bed technology is a feed material with acceptable flow characteristics and consistent size. The incoming feed must break up and disperse in agglomerate sizes comparable to or less than the top size of the fluidized bed of particles. Otherwise, the large agglomerates will sink to the air distribution plate, remain semi-static and become overheated. This accumulation eventually disturbs the fluidization of the entire bed. Sticky sludges or hard filter cakes are usually processed by simply mixing dry product into the wet feed to improve its flow properties. A second method uses an inert bed of coarse particles such as limestone, slag or other hard material that will not contaminate the dry product as it slowly degrades in the bed. The inert particles are fluidized at high velocities (10-15 fps), which serves to break apart and disperse the incoming feed throughout the bed. As the feed breaks apart and dries, it is swept out of the fluid bed to a cyclone for collection. No bed overflow or underflow is used. The pressure drop across the bed is monitored to track the inert material inventory. As the material degrades and the pressure drop decreases, fresh inert material is added to maintain the inventory. The inert fluid bed system has been successfully used to dry pyrophoric concentrates, which are broken down in the inert bed zone and quickly quench the incoming hot gas stream to well below the reaction temperature. The fine concentrate particles are entrained and then carried through the relatively low-temperature freeboard zone. To further minimize the potential for oxidation reactions, oxygen concentration in the process gas stream is reduced to below 10% through nitrogen injection or gas recirculation. The inventory of material in the fluid bed zone enables the unit to handle minor fluctuations in feed moisture content or flow rate. The unit is also very simple to operate once the fluidized bed has been established. However, it cannot accommodate wide variations in feed particle size or moisture content without jeopardizing system stability. Typical ranges for fluid bed dryer design and operation are as follows: Diameter: 5 m maximum Height: 2m maximum Retention time: 5+ minutes Bed gas velocity-conventional: 0.15-2 mps Bed gas velocity-inert bed: 3-5 mps Distributor plate pressure drop: 3.4 kpa minimum Plenum temperature: 800°C maximum Bed outlet/freeboard temperature: 125-150°C. Fluid bed drying systems have relatively low capital/maintenance costs, and a small footprint (except for fine materials that require very low velocities, which have a much smaller capacity per unit bed). Their specific power consumption costs are higher, due to the pressure drop across the air distribution plate and suspended bed. Flash dryers. Flash dryers are becoming more popular due to their low cost, small footprint, efficiency, low maintenance and ease of operation. Material is introduced to a hot gas stream at a velocity sufficient to entrain all feed particles and carry them vertically up the flash tube. The excellent gas/solids contact enables most materials to be completely dried in 1-3 seconds. The

1469

gas/solids mixture passes through a cyclone collector for solids removal, followed by a baghouse for fine particulate capture. Figure 7 illustrates a typical gas suspension flash drying system, with provisions for dry product recycle to handle feed materials that are not free-flowing. Adequate velocity must be maintained to ensure that all feed particles and agglomerates are entrained and carried through the unit. In addition, particularly with large diameter flash tubes, the material must be dispersed uniformly across the tube. This can be accomplished using multiple feed locations and by using splash plates at the feed inlets. A velocity of 3-4 mps is needed to convey particles up to 850 p with specific gravity <3. Units may be designed for velocities >18 mps to convey 5 mm dense particles (SG up to 4.5). However, as gas/material retention time is a function of flash tube length, cost becomes prohibitive at higher velocities due to the height of the tube and support structures. Tube wear also becomes an issue with coarse abrasive materials, even with the use of wear resistant refractory materials. Therefore, most flash dryers are designed for particles >2.5 mm with a maximum velocity of 10 mps and a retention time of 1-2 seconds. When a fine feed is unsuitable for a conventional flash dryer due to the presence of a small percentage of +2.5 mm particles, a hybrid system can be considered. This hybrid system combines the features of fluidized bed and flash drying. A shallow, low-pressure drop fluid bed zone is developed in the bottom of the tube using an air distribution plate or high-velocity throat. This zone operates at 4 mps to support active fluidization of a dilute bed of coarse particles. The freeboard zone above the bed is extended to provide 1-2 seconds of retention time at 4 mps, to serve as the flash drying zone for the fine fractions. The dried coarse fraction from the bed overflow and the dry fine fraction from the collection cyclone are then combined outside of the dryer. This allows the dryer system to accommodate much wider variations in particle size.

Figure 7 Flash dryer system Typical flash dryer design and operating parameters are as follows:

0

Diameter: 5 m maximum Height: 8-30 m Retention time: 2-3 seconds Gas velocity: 4-10 mps Feed top size: 2-2.5 mm Inlet temperature: up to 1000°C with refractory inlet Outlet temperature: 150-200°C.

1470

Air-Swept Hammermill Dryer. This dryer has the ability to mill and dry materials with a wide range of particle sizes and moisture levels. If size reduction is desired or is not an issue, then this system offers the best combination of performance, flexibility and ease of operation. The circuit (Figure 8) resembles a flash dryer with a hammermill between the air heater and flash tube. Material is introduced to the mill with the hot gas stream. As the material passes through the high-speed rotor, it is reduced in size, entrained in the gas flow and carried into the flash drying zone. This zone provides 1-2 seconds of retention time to complete the drying process. In most cases, a static or dynamic separator in the upper portion of the flash drying zone separates and returns coarse particles to the mill for further reduction. The gas and material passing through the separator are directed to a cyclone and baghouse for product collection. If the feed material is too sticky to flow through the mill housing, dry product recycle is used to condition the feed. The system can produce a fine, dry product from coarse feedstocks with moisture levels exceeding 50%. For limestone or software materials, top size can be reduced to 80-100% passing 150 p. While footprint and ease of operation are similar to a flash dryer, a hammermill dryer is costly, consumes much power and is not easy to maintain. A rotary dryer and crushing/rnilling circuit combination may be equally worthy of consideration.

IDFan

7

Uaghouse

1,

Fuel

Hammermill Dryer

Figure 8 Hammermill dryer Typical ranges for hammermill dryer design and operating parameters are as follows: 0

0

Capacity: >200 mtph dry product Rotor width: 4 m maximum Rotor diameter: 4 m maximum Heat input: 100 MM Btu/hr maximum Inlet temperature: 870°C maximum Outlet gas temperature: 80-95°C Gas retention time: 1-2 seconds.

Rotary Tray Dryer. The rotary tray dryer, also known as the turbo tray dryer, can handle feeds ranging from viscous slurries to free flowing solids, and can provide long residence times. It provides gentle handling and heating of friable and temperature-sensitive materials. Its capacity is lower, and its capital cost per unit output higher, than most adiabatic drying systems.

1471

The dryer (Figure 9) comprises a series of round trays stacked in a vertical plane. Feed is dropped onto the top rotating tray where it is leveled. Following approximately one revolution, the material is conveyed onto the tray below using a stationary wiper. Additional fixed devices mix and level the material on this tray. This mode of transport and mixing continues until the material is discharged from the bottom tray. Retention time is controlled by tray speed. Hot gas enters the bottom of the dryer and is circulated around the trays in a counter-flow orientation using central fans. The gases exit the top of the unit and are directed to a dust control circuit.

Figure 9 Rotary tray dryer Typical rotary tray dryer design and operating parameters are as follows: Diameter: 10 m maximum Height: 20 m maximum Capacity: 15 mtph product maximum Evaporative load: 1,100 kg/hr maximum Material travel on tray: 8 0 4 5 % of circumference Tray speed: 0.1-1.O rpm Material retention time: 30-60 minutes.

Band Dryer. The band dryer, or conveyor dryer, dries a permeable bed of coarse particles or fine particles that have been preformed into large agglomerates such as briquettes or extrusions. It is applied before end use or high-temperature processing. The feed lies on a perforated belt that travels in a horizontal plane similar to that of a conventional belt conveyor. The belt sits inside a chamber supplied with hot gas that circulates around and through the bed of static material. The drying chamber is usually divided into cells, with gas recirculation to control the drying temperature profile. The last cells are often used for ambient air cooling. The most common configuration includes a single conveyor dryer. Multiple conveyors (in a multi-stage dryer) may be used in a vertical configuration when the material is very heat- sensitive, to improve efficiency and/or capacity. A single-stage band dryer is shown in Figure 10.

1472

+

Product

Figure 10 Band dryer The dryer’s gentle handling virtually eliminates degradation of even the most friable materials. The most critical part of the operation is the delivery of material to the belt in a manner that establishes a uniform bed depth to promote even air distribution and drying. Breaking devices may be employed at one or more locations along the belt to provide some degree of material agitation and mixing. The typical ranges for band dryer design and operating parameters are as follows: Belt width: 0.5-3.0 m Belt length: no typical range Belt velocity: 0.001-0.02 mps Bed depth on belt: 20-150 mm (typically <40-60 mm to limit dP) Superficial velocity through bed: 0.5-2.0 mps Drying temperature: 100-200°C.

Spray Dryer. The spray dryer is the most effective for drying pumpable slurries. The feed is atomized to produce fine droplets of uniform size characterized by a high surface area (35,0003 10,000 m2/m3 slurry), resulting in almost instantaneous drying that achieves very low moisture levels without particle overheating due to the high rate of evaporation. For slurries with multiple components, of spray drying is effective at fixing particles that are homogeneous in composition. The primary components include an atomizer, drying chamber and product collection circuit. The most common configuration (Figure1 1) introduces the atomized slurry into the top of the vessel. Hot gas also enters the top of the vessel, flows down and sweeps the dried particles out the bottom to the product collection device. This figure illustrates concurrent flow, but flow may also be countercurrent or mixed.

1473

Hot Gas

Product

Figure 11 Spray dryer system Three methods of slurry atomization are utilized: single fluid nozzle, pneumatic nozzle and centrifugal discs. The single fluid nozzle is generally used to produce fine droplets (120-250 p), with the droplet size being a function of pressure drop across the nozzle. The pneumatic nozzle offers more flexibility in terms of droplet size, with size being a function of nozzle velocity and slurry properties. This design can also produce the smallest droplet size of the three nozzle options (10-20 p). The centrifugal disc is the most widely applied method of atomization due to its simplicity, ease of operation, ability to handle abrasive slurries, rare occurrence of blockages, and atomization is not as sensitive to slurry properties. Slurry is dropped onto a disc rotating with a linear speed of 90-210 mps, and droplet size (30-120 p) may be controlled by adjusting disc speed. A larger diameter drying vessel is required when the centrifugal disc is utilized in order to accommodate the wider spray pattern. Typical spray dryer design and operating parameters include:

0

Evaporative loading: up to 7,000 kg/h Inlet gas temperature: 93-760°C Fluid pressures: Single fluid nozzle: 2,000-27,000 kPa Pneumatic nozzle: 0-415 kPa Centrifugal disc: 0 Retention time: 3-30 seconds.

CONCENTRATE HANDLING Bulk mineral concentrates are handled by an array of equipment including trucks, railcars, ocean vessels, front-end loaders, and conveyors. (Slurry pipelines are also used to transport concentrates, but these are addressed separately.) Concentrates are typically stored in bins, buildings or open stockpiles at one or more points between production and delivery to the end user, which may be a smelter, transport contractor or other customer. Flow Properties. Flow properties of concentrates vary significantly not only with mineral type, but also within a particular concentrate due to variations in moisture content, grain size, climate conditions, and consolidation (i.e., due to storage time). The need to understand the flow properties under all conditions cannot be overemphasized when designing and selecting handling equipment. For example, chute and hopper valley angles of 45" may be sufficient for reliable flow for one particular concentrate, but not another. In fact, some concentrates require slopes of at least 75". Analysis of concentrate flow properties is commonly omitted to save money, but the resulting design often proves problematic, with operating and maintenance cost increases that more than

1474

offset the front-end savings. A simple flow test program provides useful design information such as minimum chute valley angles and hopper opening sizes, eliminating the risk of material “hangup” in chutes and hoppers, and bridging and ratholing in bins. Moister concentrate is usually harder to handle, primarily because it tends to adhere to conveyors as well as surfaces such as the corners of truck and railcar bodies. Most concentrates are produced with moisture content between 4 and 8%, though higher content is not uncommon. For every concentrate and concentrator operation, there is a point at which the incremental cost of reducing moisture is no longer offset by the expense of transporting water to the end user. Long-term storage (e.g., several weeks or months) can introduce storage and handling problems if consolidation effects are not addressed in the handling equipment design. For example, as moisture migrates out of the pile over time due to consolidation and/or vibration, the concentrate can become further compacted, which may introduce reclaim and handling problems. The exposed exterior of the pile will effectively dry out, which can lead to fugitive dust problems. Extended exposure to cold can form frozen lumps in stockpiles, which may not break up during reclaim operations. Handling Equipment. The most common concentrate handling equipment items are frontend loaders and belt conveyors, with the latter including belt feeders, overland conveyors, stackers, and tripper conveyors. Mechanical conveying devices such as screw, drag, and pneumatic conveyors are also used, but their application is usually limited to low flow rates of concentrates with moisture content below 4%. Belt conveyors provide economic, reliable handling, provided that transfer chutes are properly designed and belt cleaning devices are properly specified and maintained (especially for moist concentrates). Chute valley angles must be steep enough to promote reliable flow under all conditions. Chute liners, such as ultra-high-molecular-weight (UHMW) polyethylene and stainless steel, are a relatively low-cost means of preventing hang-up, and can be installed new or as retrofits. Properly specified and maintained belt-cleaning devices will minimize the carry-back of concentrates on the belt, reducing cleanup and maintenance costs. In the case of outdoor belt conveyors, wind and precipitation can result in material loss, fugitive dust generation, and product degradation. Conveyors can be provided with partial or full belt covers, or fully enclosed in a steel structure, to minimize or eliminate such problems.. Belt conveyors can usually be inclined 15’ or higher, depending on the concentrate characteristics. Finer grain size and/or higher moisture content usually permit steeper inclines due to greater internal cohesion. Concentrates usually have higher bulk density than coal or crushed ore. Over time, accumulation and subsequent compaction of accumulated dust and/or spillage can result in relatively high loads on conveyor support and access structures. It is therefore critical to consider such loads in the design of these structures. Some specialized belt conveyor systems are seeing increased application in concentrate handling. Sandwich-type high-angle conveyors, PocketliftTM and pipe conveyors offer several advantages over conventional belt conveyors. High-angle and PocketliftTM conveyors allow inclines of up to 90°, while pipe conveyors can provide effective dust containment without the need for enclosures. These conveyors are usually more sophisticated than conventional belt conveyors in terms of design and componentry, and equipment manufacturers should be consulted when considering their application. Relatively free-flowing, low-moisture concentrate may permit the use of mechanical conveying systems such as drag, screw, and pneumatic conveyors. However, the abrasiveness and particle size of the concentrate must be considered. Bottom-dump, end-dump and side-dump trucks are commonly used for concentrate transportation, along with railcars (usually bottom-dump type). Concentrate flow properties must be considered when selecting a truck body type. Bottom-dump hoppers must be designed for reliable discharge of consolidated concentrates, while the tipping angle of side- and end-dump trucks must be sufficient to’ promote sliding discharge without the need for manual intervention. UHMW or stainless-steel liners in truck bodies are often helpful.

1475

Weighing and Sampling. Concentrate weighing and sampling are usually performed at one or more points in the handling process. Conveyor belt weigh scales or weightometers are commonly used for process feedback control by providing instantaneous mass flow rate information, and for load measurement. For example, integration of the instantaneous flow rate data over a discrete duration can provide an estimate of concentrate discharged into a bin, railcar, or truck. When installed on a conveyor forming part of a truck, train or shiploader, higheraccuracy “certified for trade” scales can also be used for accounting purposes. Concentrates can also be weighed indirectly via truck and railcar scales, draft survey or in batch weighing systems under loadout bins. Sampling is normally performed at one or more points for analysis of moisture content, grade, bulk density, etc. This is normally done by manual or automated grab sampling, or with automated sweep or cross-cut type systems installed over a belt conveyor. The method is dictated by accuracy and precision requirements as well as concentrate flow properties. Dust Suppression, Safety, and Environmental. Concentrates inevitably generate fugitive dust when handled or exposed to wind. As most contain various metals and other hazardous elements, dust emissions can present significant safety and environmental hazards if not mitigated. The amount of dust generated depends primarily on the handling method. concentrate moisture content and particle size. Conveyor transfers, front-end loader operation, and loading of railcars, trucks and ocean vessels will generate dust due simply to the displacement of air; clearly, dust generation can be reduced by minimizing the number of times that concentrates are transferred from one point to another and the heights through which they are dropped. Design of storage facilities should include effective sealing of building panels and doors. Conveyor system transfer chutes should be designed with minimal opening sizes. For relatively free-flowing and not overly abrasive concentrates, chutes can be designed such that the material slides along the inner chute surfaces, rather than hitting them. Cartridge or baghouse dust collection systems are also effective in controlling fugitive dust, provided that they are properly designed for the concentrates’ flow rate, particle size, specific gravity, and losses from chute openings and system inefficiencies. Water spray dust suppression systems can be useful, although attention must be given to runoff collection and their effect on concentrate moisture content. In enclosed storage buildings where concentrates are reclaimed by front-end loaders, emissions of both concentrate dust and equipment exhaust can pose potential health and safety hazards. Adequate positive building ventilation is typically impractical for large storage buildings due to the prohibitive costs of fans and power. Operators should wear dust masks, and emissions control devices (such as scrubbers and diesel particulate traps) should be installed on the equipment. STORAGE Capacity. Concentrates are usually stored for some period of time upon discharge from the concentrator, for periods ranging from a few days to several months. Short-term storage facilities are typically sized to effectively decouple concentrate production and its subsequent handling operation; therefore, production rate and availabilities of the mill and downstream transportation are key criteria. A properly sized facility eliminates the need to reduce or stop production when, e.g., a bulk carrier is unavailable for loading or a railcar loading facility is down for maintenance. Conversely, as a fundamental part of planning mill shutdowns, the storage facility can be “topped up” to allow production to stop without interrupting concentrate loading and transportation. Sizing of long-term storage facilities is sometimes mandated by system availabilities or market conditions. Supply contracts are typically be linked with storage capacity. Mills that produce more than one type or grade of concentrate require storage facilities that allow different volumes of different products to be stored depending on market conditions. Many operations in Arctic regions require facilities that store up to one year’s production, since the ocean transport of product is limited to short ice-free shipping seasons.

1476

Storage Facility Design. Concentrates are typically delivered to a storage facility by truck, rail, front-end loader or conveyor, which may comprise a fixed conveyor feeding a stockpile or bin, radial stackers or overhead tripper conveyor. While concentrates can be stockpiled in the open, they are usually stored in bins, silos, or buildings for protection from the weather and to comply with environmental regulations. Without protection, wind can create dust problems and erode valuable product. Precipitation can increase concentrate moisture content, adversely affecting its flow properties and increasing water weight. In some short-term storage facilities or in dry areas, covered storage may not be economically justified; in such cases, water spray systems may be used to mitigate fugitive dust emissions. Stockpiles may or may not require underlying concrete slabs, depending on geotechnical conditions, product or ground contamination, and reclaim method. In many cases, a compacted fill floor is adequate and less expensive; alternatively, a compacted fill floor can be initially covered with a thin (e.g., 0.3 to 1.0 m) layer of compacted low-grade concentrate to prevent soil from being excavated with the concentrate during reclaim operations. Multiple types and/or grades of concentrate can be stored in the same area, provided piles are separated. Cast concrete or concrete block bulkheads are a useful means of separation. Bins are common for relatively short-term storage requirements, such as smelter feed systems or surge bins. In designing bins, it is essential to understand (and preferably test) the concentrate bulk material properties under all potential conditions, so that hopper geometry and reclaim equipment are properly designed to ensure reliable reclaim operations. The effects of ambient conditions, moisture content, and consolidation on properties such as internal and wall friction angles, and bulk density can have highly detrimental results on reliability. Reclaim. Reclaiming of concentrates from open or covered stockpiles is typically performed either by front-end loaders (which transfer concentrates directly to trucks, railcars or to a hopper feeding a belt conveyor), or by an automated reclaim system. Vibratory feeders draw concentrates from under a stockpile through one or more draw-down openings and onto a belt conveyor; a system with multiple draw-down points and vibratory feeders can permit blending of product from more than one stockpile, but vibratory feeders are better suited to relatively free-flowing concentrates. Bucket-wheel or scraper reclaimers are relatively expensive, but offer relatively low operating costs when handling high throughputs on a continuous basis. In specifying such equipment, it is important to consider the worst-case scenario of consolidated or frozen concentrate in the stockpile, since this can pose a hazard to downstream handling equipment. If lumps are relatively rare, they can usually be broken up using mobile equipment. Otherwise, a feeder-breaker unit can be installed. CONCLUSIONS AND RECOMMENDATIONS The wide array of available drying systems may make the selection process seem difficult, but each has distinct advantages and disadvantages depending on material feed properties, capacity and installation requirements, and other factors. Options can usually be narrowed down to one basic system type, after which plant personnel may contact suppliers to determine the system configuration that best suits their needs. The design and operation of storage and handling systems requires attention to many factors including economics, moisture content, flow properties, ambient climate conditions, production rate, storage duration, transport method, and customer requirements. Concentrate flow properties should be analyzed prior to design and selection of handling equipment to ensure reliable, cost-effective operation. REFERENCES Williams-Gardner, A. 197 1. Industrial Drying. Leonard Hill. London, England. Van? Land, C. M. 199 1. Industrial Drying Equipment-Selection and Application. Marcel Dekker Inc. New York, NY. Kolthammer, K.W. 1980. Concentrate Drying, Handling and Storage. Mineral Processing Plant Design, 2"dEdition, Society of Mining Engineers, USA.

1477

The Selection and Sizing of Bins, Hopper Outlets, and Feeders Dr. John Carson and Tracy Holmes, Jenike d Johanson, Inc.

ABSTRACT This paper provides practical guidelines for the selection of bins, feeders, hopper outlets and gates, outlines the principles in their selection, and touches on the basic need to know the bulk flow properties of the material being handled. INTRODUCTION To cover this very broad topic in a manner that is easy to follow, we have divided this paper into three individual sections: 0

0

Bin selection Hopper outlet sizing Feeder selection.

BIN SELECTION A bin (silo, bunker) generally consists of a vertical cylinder and a sloping converging hopper. Based on the flow pattern that develops, there are three types of bins: Mass Flow, Funnel Flow and Expanded Flow, all of which will be discussed in more detail below. Whatever type of bin is selected, it needs to have the desired capacity, be capable of discharging its contents reliably on demand, and be safely constructed. Here are some of the important things that need to be done in order to ensure that a bin will perform these functions adequately: Step 1. Define your storage requirements Identify the operating requirements and conditions. Some of the more important include: 0

0

0

0

0

0

Capacity. This will vary with your plant's operating philosophy, and where the bin is to be located (e.g., start of your process, at an intermediate process step, or at the end). Discharge rate. Consider average and instantaneous rates, minimum and maximum rates, and whether the rate is based on volume or mass. Discharge frequency. How long will your material remain in the bin without movement? Mixture and material uniformity. Is particle segregation a concern in terms of its effects on material discharge or, more importantly, downstream processes? Pressure and temperature. Consider differences between the bin and upstream and downstream equipment. Environmental. Are there explosion risks, human exposure concerns, etc.? Construction materials. Abrasion and corrosion concerns may limit the types of materials you can use to construct your bin.

Step 2. Calculate approximate size of your bin Initially ignore the hopper section. Use the following formula to estimate the approximate height of the cylinder section that is required to store the desired capacity:

1478

H = ( C I y A) avg where H = cylinder height, ft C = bin capacity, ft3 y = average bulk density, lb/ft3 avg A = cross-sectional area of cylinder section, ft2 The actual cylinder height will have to be adjusted to account for volume lost at the top due to the material's angle of repose as well as for the volume of material in the hopper section. In general, the height of a circular or square cylinder should be between about 1.5 and 4 times the cylinder's diameter or width. Values outside this range often result in designs which are uneconomical or have undesirable flow characteristics. It is important to recognize that a bin's storage volume and its active (live, useable) volume are not necessarily the same. With a funnel flow or expanded flow pattern (described below), significant dead (stagnant) volume may need to be taken onto account. Step 3. Determine your material's flow properties The flow characteristics of a bulk solid must be known in order to predict or control how it will behave in a bin or hopper. These characteristics can be measured in a solids flow testing laboratory under conditions that accurately simulate how the solid is handled in your plant. Tests should be conducted on-site if your solid's properties change rapidly with time or if special precautions must be taken. The most important bulk solids handling properties that are relevant to predicting flow behavior in bins and hoppers are listed in Table 1. Each of these parameters can vary with changes in the following: 0 0

0 0

Moisture Particle size, shape, hardness and elasticity Temperature Time of storage at rest Chemical additives Pressure Wall surface.

The appropriateness of these bin design parameters has been proven over the last 40 years in thousands of installations handling materials as diverse as fine chemical powders, cereal flakes, plastic granules and mined ores. Table 1 Important flow properties Parameter Measured by Cohesive strength Shear tester Frictional properties Shear tester Sliding at impact points Chute tester Compressibility Compressibility tester Permeability Permeability tester Segregation tendency Segregation tester Abrasiveness Friability

Abrasive wear tester Annular shear tester

Useful for calculations of Outlet sizes to prevent arching and ratholing Hopper angles for mass flow, internal friction Minimum angle of chute at impact points Pressure calculations, bin loads, feeder design Discharge rate calculations, settlement time To predict whether or not segregation will occur To predict the life of a liner Maximum bin size, effect of flow pattern on particle breakage

1479

Step 4. Understand the importance of flow patterns Although it is natural to assume that a bulk solid will flow through storage or conditioning vessels in a first-idfirst-out sequence, this is not necessarily the case. Most bins, hoppers, silos and conditioning vessels move solids in a funnel flow pattern. With funnel flow, some of the material moves while the rest remain stationary. This firstiflast-out sequence is acceptable if the bulk solid is relatively coarse, free flowing, nondegradable, and if segregation is not important. If the bulk material and application meet all four of these criteria, a funnel flow bin is the most economical storage device. Unfortunately, funnel flow can create serious problems with product quality or process reliability. Arches and ratholes may form, and flow may be erratic. Fluidized powders often have no chance to de-aerate. Therefore, they remain fluidized in the flow channel and flood when exiting the bin. Some materials cake, segregate or spoil. In extreme cases, unexpected structural loading can result in equipment failure. These problems can be prevented with storage and conditioning vessels specifically designed to move materials in a mass flow pattern. With mass flow, all material moves whenever any is withdrawn. Flow is uniform and reliable; feed density is independent of head of solids in the bin; there are no stagnant regions, so material will not cake or spoil and low-level indicators work reliably; sifting segregation of the discharge stream is minimized by a first-idfirst-out flow sequence; and residence time is uniform, so fine powders are able to de-aerate. Mass flow bins are suitable for cohesive materials, powders, materials that degrade with time, and whenever sifting segregation must be minimized. A third type of flow pattern is called expanded flow. In this, the lower part of a bin operates with flow along the hopper walls as in mass flow, while the upper part operates in funnel flow. An expanded flow bin combines the best aspects of mass and funnel flow. For example, a mass flow outlet usually requires a smaller feeder than would be the case for funnel flow. This flow pattern is suitable for storage of large quantities of non-degrading solids. It can also be used with multiple outlets to cause a combined flow channel larger than the critical rathole diameter. Step 5. Follow these detailed design procedures Step 5A. Mass flow. In order to achieve a mass flow pattern, it is essential that the converging hopper section be sufficiently steep and have low enough friction to cause flow of all the solids without stagnant regions, whenever any solids are withdrawn. In addition, the outlet must be large enough to prevent arching and to achieve the required discharge rate. Typical design charts showing the limits of mass flow for conical- and wedge-shaped hoppers are given in Figure 1. Hopper angle (measured from vertical) is on the abscissa, and wall friction angle is on the ordinate. For example, mass flow will occur in a conical hopper which has an angle of 20" and is constructed from or lined with a wall material which provides a wall friction angle of 23" or less with the stored bulk solid. Making the hopper walls less steep by 4" or more could result in funnel flow. Alternatively keeping the wall angle at 20" but increasing the wall friction angle to 28" or more would also result in funnel flow. Calculating the outlet size needed to overcome arching is more difficult. It involves measuring the cohesive strength and internal friction of the bulk solid, then following the design procedure outlined in Reference 1. Sizing the outlet for discharge rate is covered in this article. Step 5B. Funnel flow. The key requirements for designing a funnel flow bin are to size the hopper outlet large enough to overcome arching and ratholing, and to make the hopper slope steep enough to be self-cleaning. Minimum dimensions to overcome arching and ratholing require knowledge of your material's cohesive strength and internal friction. Design procedures for funnel flow are also given in Reference 1 The requirement for self-cleaning can usually be met by making the hopper slope 10" to 15" steeper than the wall friction angle.

1480

Step 5C. Expanded flow. Consideration must be given to both the mass flow and funnel flow sections. In the lower mass flow section, the procedure outlined above for a mass flow hopper should be followed. In addition, the flow channel must be expanded to a diagonal or diameter equal to or greater than the material's critical rathole diameter, which can be calculated using the procedure in Reference 1. Here too, the hopper slope in the funnel flow portion should be steep enough for self-cleaning. Step 6. Consider the bin's shape. At first glance, it might appear that a square or rectangular straight-sided section at the top of a bin is preferable to a circular cross-section. Such cylinders are easier to fabricate and have greater cross-sectional area per unit of height. However, these advantages are usually overcome by structural and flow considerations. A circular cylinder is able to resist internal pressure through hoop tension, whereas flat walls are subjected to bending. Thus, thinner walls and less external reinforcement are required with circular cross-sections. In addition, there are no corners in which material can build up. This is particularly important when interfacing with a hopper at the bottom.

40' ($1;

Wall friction angle

30' 20"

4':

30'

Wall friction angle

20'

MASS FLOW

10"

0"

0" 0 '

10'

8,

0'

20' XI" 40" 50"

10" 20'

8,

Conical hopper angle, from vertical

30' 40' 50" 60" : Planar hopper angle, from vertical

Figure 1 Typical chart determining mass flow wall angles Several factors to consider when choosing hopper geometry are listed below: Sharp versus rounded corners. Pyramidal hoppers usually cause a funnel flow pattern to develop because of their inward-flowing valleys that are less steep than adjacent sidewalls. Conical, transition and chisel shapes are more likely to provide mass flow because they have no corners. See Figure 2. Headroom. Typically, a wedge-shaped hopper (e.g., transition or chisel) can be 10" to 12" less steep than a conical hopper and still promote mass flow. This can provide significant savings in hopper height and cost, which is particularly important when retrofitting existing equipment in an area of limited headroom. In addition, a wedge-shaped hopper design is more forgiving than a cone in terms of limiting hopper angles and wall friction. Outlet sizes. In order to overcome a cohesive or interlocking arch, a conical hopper has to have an outlet diameter that is roughly twice the outlet width of a wedge-shaped hopper (provided the outlet length is at least three times its width). Thus, cones generally require larger feeders. Discharge rates. Because of the increased cross-sectional area of a slotted outlet, the maximum flow rate is much greater than that of a conical hopper. Capital cost. Each application must be looked at individually. While a wedge-shaped hopper requires less headroom or a less expensive liner than a cone, the feeder and gate (if necessary) may be more expensive. Discharge point. In many applications, it is important to discharge material along the centerline of the bin in order to interface with downstream equipment. In addition, having a single inlet point and single outlet, both located on the bin's centerline, minimizes flow and structural problems. Generally, conical hoppers are better for these situations, particularly if only a gate is used to stop and start flow.

1481

0

Mating with a standpipe. If material is being fed into a pressurized environment, a circular standpipe is often preferred to take the pressure drop.

It- Inlet diameter, D 4

It Inlet diameter, D 4

Outlet width,

Inlet diameter, D Inlet length, E

4

d

H-

Outlet lenath, L I

length, L

Figure 2 Hopper geometries (conical-topleft; chisel-top right; pyramid-bottomleft; transition-bottom right) Step 7. Consider other important factors Some additional considerations include:

0

0

Gate. A slide gate at the outlet of a bin must generally only be used for maintenance purposes, not to control or modulate the flow rate. Therefore, it should only be operated in a full-open or full-closed position. Feeder. The feeder's design is as important as that of the bin above it. The feeder must uniformly draw material through the entire cross-section of the bin's discharge outlet to be effective. The section below on Feeder Selection covers this in more detail. Mating flanges. The inside dimensions of the lower of two mating flanges must be oversized to prevent any protrusions into the flowing solid. The amount of oversize depends on the accuracy of the construction and erection. Usually 1 inch overall is sufficient. If gaskets or seals are used, care must be taken to ensure that these too, do not protrude into the flow channel. All flanges should be attached to the outside of the hopper with the hopper wall material being the surface in contact with the flowing solids. This ensures that the flange and gasket do not protrude into the flowing solids. Interior surface finish. Whenever possible, welding should be done on the outside of the hopper. If interior welding is necessary, all welds on sloping surfaces must be ground flush and power brushed to retain a smooth surface. After welding, all sloping surfaces must be clean and free of weld spatter. The surface finish is most critical in the region of the hopper outlet. Therefore, any blisters in this area from exterior welding must be brushed smooth. Horizontal or diagonal welded connections should preferably be lapped

1482

0

with the upper section on the inside so the resulting ledge does not impede flow. If horizontal butt welds are used, care must be taken to avoid any protrusion into the flowing solid. Liner attachment. Inside liners, such as stainless sheet or ultra-high molecular weight (UHMW) polyethylene, must be placed on sloping surfaces with horizontal or diagonal seams lapped with the upper liner on top in shingle fashion. Vertical seams may be either lapped or butted. Abrasive wear considerations. In mass flow, a bulk solid flows against the hopper and cylinder walls. Handling an abrasive bulk solid may result in significant abrasive wear of the wall material including coatings and liners. Therefore, when designing a mass flow hopper, it is important to assess the potential for abrasive wear. Generally, a hopper surface becomes smoother with wear. However, occasionally a wall becomes rougher, which may upset mass flow. The life of a given wall material can be estimated by conducting wear tests. Access doors and poke holes. In general, poke holes are not recommended in mass flow bin designs as they have a tendency to prevent flow along the walls, thus creating a problem that mass flow bins are intended to solve. Access doors are also a frequent cause of problems. If they are essential, it is better to locate them in the cylinder, rather than in the hopper section. Structural design issues. It is important that the bin be designed to resist the loads applied to it by both the bulk solid and external forces. This is particularly important when designing, or converting, an existing bin to mass flow because unusually high localized loads may develop at the transition between the vertical section and the mass flow hopper. Bulk materials of inferior flowability (e.g., more cohesive with larger critical arching and ratholing dimensions than the material upon which the design was based, or more frictional requiring steeper wall angles) should not be placed in the bin because flow obstructions are then likely to occur. Such obstructions may lead to the development of voids within the bin and impose dynamic loads when material collapses into the voids. Bin failures have occurred under such conditions. Prefabrication drawing review. Before fabrication of the bin and feeder, an engineer trained in solids flow technology should review all detailed design drawings. This review is necessary to ensure that the design follows the recommendations and that any design details or changes are consistent with reliable bulk solids flow.

HOPPER OUTLET SIZING In most applications, a feeder is used to control discharge from a bin or hopper (see section below on Feeder Selection). For such applications, the maximum achievable flow rate through the hopper outlet must exceed the maximum expected operating rate of the feeder. This ensures that the feeder will not become starved. This is particularly important when handling fine powders, since their maximum rate of flow through an opening is significantly less than that of coarser particle bulk solids whenever a mass flow pattern is used. In addition, any gates must not interfere with material discharge. The following step-by-step procedure will assist in properly sizing a hopper outlet.

Step 1. Calculate the ratio of outlet width or diameter to particle size Flow stoppages due to particle interlocking are likely if the diameter of an outlet is less than about six times the particle size. With an elongated outlet, problems are likely if the ratio of outlet width to particle size is less than about 3: 1.

1483

Step 2. Determine into which of the following categories the material fits: Coarse, Easy Flowing, and/or Fine Powder For purposes of flow rate calculations, a bulk solid is often considered coarse if no more than 20% will pass through a 1/4 in. screen. Whether or not a material can be considered easy flowing depends upon the cohesiveness of the bulk solid, the dimensions of the container in which it is stored, and whether or not any excess pressures are applied to the material (e.g., the container is vibrated after being filled). If the combination of these factors results in no flow stoppages at the vessel's outlet (due, for example, to arching or ratholing), the material can be considered easy flowing in that application. A fine powder is a bulk solid whose flow behavior is affected by interstitial gas. Common household examples of fine powders include flour and confectioners (icing) sugar. Step 3. Determine maximum achievable flow rates Step 3A. The bulk material is coarse but not easy flowing. Either the outlet size must be increased or the material's cohesive strength must be decreased to allow the material to flow. Step 3B. The bulk material is coarse and easy flowing. If the ratio of outlet size to particle size is sufficiently large to prevent particle interlocking, then the maximum achievable rate through an orifice of a coarse, easy flowing bulk solid such as plastic pellets is given by the following equation: 1/2

Q = 3600 yA [Bg/(2(1 + m) tan e)] Where Q = flow rate, lb/hr y = bulk density, lb/ft3 A = outlet area, ft2 B = diameter of circular outlet or width of a slotted outlet, ft 2 g = gravitational constant, 32 ft/sec m = 0 for long slotted outlet, 1 for conical hopper 8 = flow channel angle (measured from vertical), deg

(2)

A modification can be made to Equation (2) to consider particle size. This modification is only important if the particle size is a significant fraction of the outlet size. Note that, with a mass flow hopper (see section on Bin Selection), the flow channel coincides with the hopper wall; hence 8 is the hopper angle. On the other hand, with a funnel flow hopper, the flow channel forms within stagnant material and, while it is steeper than in mass flow, it is variable. Thus, the maximum flow rate from a funnel flow hopper is generally higher but less predictable than the flow rate from a mass flow hopper having the same outlet size. Step 3C. The bulk material is a fine powder. Fine powders are often mishandled in funnel flow bins. As noted in the section on Bin Selection, fine powders have little or no chance to deaerate in such bins. Instead, they often remain fluidized in the flow channel and flood uncontrollably when exiting the bin. A funnel flow pattern should therefore be avoided when handling fine materials. Mass flow bins, on the other hand, can provide predictable and controlled rates of discharge of fine powders as well as other bulk solids. Unfortunately, a fine powder's maximum rate of discharge through a mass flow hopper outlet is often several orders of magnitude lower than the limiting rate of a coarse particle material having the same bulk density. This severe flow rate limitation is the result of the upward flow of air through the hopper outlet caused by the slight vacuum condition, which naturally forms in the lower portion of a mass flow hopper as material flows through it. The limiting rate of material flow through a mass flow hopper outlet can be calculated once the powder's permeability and compressibility have been measured.

1484

If the limiting discharge rate is too low, there are several possible ways to increase it: Increase the outlet size. Since the limiting rate is approximately proportional to the crosssectional area of the outlet, doubling the diameter of a circular outlet increases the maximum discharge rate by roughly a factor of four. Decrease the level of material in the bin. Fine powders do not behave like fluids. Thus, lower heads result in higher discharge rates, although the effect is generally not too pronounced. Provide an air permeation system as shown in Figure 3. This has the effect of partially satisfying the vacuum condition that naturally develops. As a result, there is less need for air to be pulled up through the outlet counter to the flow of particles. With such a system, the maximum flow rate can often be increased by a factor of between 2 and 5. If none of these will allow the desired discharge rate, fluidization, as discussed in the next step, can be considered.

Figure 3 Air permeation systems Step 3D. The bulk material is a fine powder and the required discharge rate is high. If the limiting flow rate from a mass flow hopper is still too low, consideration should be given to fluidizing the fine powder and handling it like a liquid. For this technique to be successful, it is generally necessary that the powder have low cohesion and low permeability. Low cohesion allows the material to fluidize uniformly, so the air does not channel around large lumps. With a low permeability material, significant pressure gradients can be established, and the material takes a long time to de-aerate. The Geldart chart, shown in Figure 4, provides a rough indication of whether or not a particular material is a good candidate for fluidization. Powders falling within classifications A and B are generally considered good candidates, while category C materials are difficult to fluidize. Category D materials are acceptable for fluidization, but the bed settles quickly and high gas rates are required. If the bin is small, it may be practical to fluidize the entire contents. With larger bins, this is neither practical nor necessary. However, if only localized regions are fluidized, consideration must be given to the potential for arching and ratholing in non-fluidized regions. In some cases, it is necessary to only fluidize intermittently, while in other cases, continuous fluidization is required during discharge. Whether batch or continuous discharge is required will influence this, but there are other factors to consider as well.

1485

When considering the fluidization option, several operational requirements must be evaluated: 0

0

0

0

0 0

The bulk density of the discharging material will be lower than if the material were not fluidized. Therefore, a given mass will occupy more volume, which could result in downstream equipment (such as a bulk bag, hopper or railcar) receiving less than the desired mass even though it is full. The material’s bulk density will also vary with time depending on the degree of fluidization of the discharging material. This can present process problems downstream. Some materials are hygroscopic, while others are explosive. In such cases, dry air or inert gas may be required for fluidization. Higher energy and gas consumption rates must be taken into account as an additional operating cost. The feeder controlling the discharge must be capable of metering fluid-like materials. This technique should be avoided if particle segregation is a concern. 10,000

Particle Density (kg/cu. m) 1

,ooo

10

100

1,OOo

10.000

Mean Particle Diameter ( pm)

Figure 4 Geldart’s fluidization classification FEEDER SELECTION A feeder is used whenever there is a need to control the solid’s flow rate from a bin or hopper. Conveyors are incapable of performing this function, because they only transport material and do not modulate the rate of flow. Dischargers are sometimes used to encourage material to flow from a bin, but by themselves cannot control the rate at which material flows. Thus, they are not a feeder, either. Consequently, when a discharger is used, a feeder is also required to control the flow rate from a storage system. Criteria For Feeder Selection Regardless of what type of feeder is used, it should provide the following: 0

0 0

0

Reliable and uninterrupted flow of material from some upstream device (typically a bin or hopper). The desired degree of discharge flow rate control over the necessary range, Uniform withdrawal of material from the outlet of the upstream device. This is particularly important if a mass flow pattern is desired, perhaps in order to control segregation, provide uniform residence time, or to minimize caking or spoilage in dead regions. Minimal loads acting on the feeder from the upstream device. This minimizes the power required to operate the feeder, as well as minimizes particle attrition and abrasive wear of the feeder components.

1486

Plant personnel often prefer a certain type of feeder because of past experience, availability of spare parts, or to maintain uniformity to make maintenance easier throughout the plant. Such personal preferences can usually be accommodated because, in general, several types of feeders can be used in most applications, if they are designed properly.

Volumetric or Gravimetric Feeders can be divided into two basic types - volumetric and gravimetric. A volumetric feeder modulates and controls the volumetric rate of discharge from a bin (e.g., cu ft/hr). Many types of volumetric feeders are available on the market. The four most common types of such feeders are screw, belt, rotary valve and vibrating pan. Each has inherent benefits and limitations, many of which are spelled out in this guide. Special designs can often overcome many of the weaknesses stated. In contrast to a volumetric feeder, a gravimetric feeder modulates the mass flow rate. This can be done either on a continuous basis, in which the feeder modulates the mass of material discharged per unit time; or on a batch basis, in which a certain mass of material is discharged and then the feeder shuts off. Follow this step-by-step procedure for selecting a helical screw, belt, rotary valve (sometimes referred to as a rotary air lock), or vibrating pan for your application. Step 1. Determine maximum particle size of your material If it's less than about 1/2 in., almost any type of feeder can be used. If over about 6 in., the choices are limited. In most industrial plants, this generally means either a belt or vibrating pan feeder. Step 2. Establish whether particle attrition is a concern If particle attrition were a concern, a vibrating pan would be a good choice. Feeders that have pinch points (screws, rotary valves) should be avoided. Step 3. Evaluate likelihood and frequency that material will drop directly onto the feeder, such as when the bin is empty A vibrating pan is a good choice for this application, since it is more rugged than a belt yet it has a smooth surface, which limits buildup such as can occur with screws or rotary valves. Step 4. Identify outlet configuration of hopper to which feeder will be attached Square and round outlets present no restrictions in choice of feeder. Elongated outlets, on the other hand, generally require either a screw or belt. An elongated rotary valve, called a star feeder, can be used to feed across the narrow dimension of a slotted outlet. A vibrating feeder can also be oriented to feed across this dimension. This kind of feeder may require several drives to accommodate extreme width, although the drives will be small because of the feeder's short length. Step 5. Decide whether volumetric or gravimetric control is required Screws, rotary valves, and vibrating pans can only control flow on a volumetric basis. Belts, on the other hand, can be used for either application. Step 6. Determine maximum operating temperature A belt is generally limited to about 450"F, unless special materials of construction are used. Screws, rotary valves, and vibrating pans can be used with temperatures in excess of 1000°F. Step 7. Determine design throughput Step 7A. Maximum. The highest throughput can generally be achieved with a belt feeder, followed by a screw, and then a rotary valve or vibrating pan. For example, with material having a bulk density of 100 pcf, the maximum capacity of a typical belt feeder is about 3,000 ton/hr. The maximum capacity of a rotary valve or vibrating pan is about one-sixth this value. As discussed

i487

above, as well as the mechanical limits of the feeder, consideration must also be given to the maximum achievable discharge rate through the hopper outlet to which it is attached. Step 7B. Minimum. The minimum rate of throughput is required in order to determine the turndown required for the feeder. Most feeders can easily achieve a 10:1 turndown. If significantly higher turndown were required, a good choice would be a vibrating pan.

Step 8. Look at other specific operational requirements If the bulk material is to be fed into a pressurized environment (e.g., positive pressure pneumatic conveying line) a rotary valve is an excellent choice. A screw feeder can be used if it is designed with a moving plug at the discharge end. If return spillage is a concern, a belt feeder should be avoided. Any of the other types of feeders being considered would not have this problem. Step 9. Determine material characteristics that might affect feeder choice With fine dry material, flooding and dust generation are likely to be a concern. Therefore, having a feeder that seals the outlet and/or is totally enclosed is important. A rotary valve is an excellent choice as can be a screw, if it designed properly. If the only concern is dust generation and not flooding, then either of these two types of feeders or an enclosed vibrating pan feeder is a good choice. If the bulk material is pressure sensitive, avoiding pinch points and minimizing excessive compaction are important considerations. A belt feeder is often the best choice for these types of materials. When handling materials that degrade easily, belts and vibrating pans are good choices. They are easy to clean and by their nature do not have stagnant zones within the feeder itself. Screws and rotary valves are not as good in this application as they do have stagnant zones; however, they can be designed for quick disassembly for cleaning. If the bulk solid is expected to contain tramp material, belts and pan feeders are good choices whereas a screw is only a fair choice and a rotary valve is a poor choice. Screw Feeders The key to proper screw feeder design is to provide an increase in capacity in the feed direction. This is particularly important when the screw is used under a hopper with an elongated outlet. One common way to accomplish this is by using a design as shown in Figure 5.

Figure 5 Mass flow screw feeders draw uniformly from the entire outlet Uniform discharge across the entire hopper outlet opening is accomplished through a combination of increasing pitch and decreasing diameter of the conical shaft. Unfortunately, normal tolerances of fabrication are such that extending the length under the hopper outlet to greater than about six to eight times the screw diameter often results in a poorly performing screw. This length can be extended with special design and fabrication techniques. Through special design techniques, a moving plug can be formed at the discharge end of the screw, allowing material to be fed into a higher pressure environment while minimizing leakage back into the feed bin, that can potentially create arching and ratholing problems.

1488

Belt Feeders As with screw feeders, the key to proper belt feeder design is to provide increasing capacity (draw) along the length of the bin outlet. Without this, material will channel at one end of the hopper and disrupt mass flow, potentially creating arching and ratholing problems. An effective way to increase capacity is to cut a converging wedge-shaped hopper in such a way that it is closer to the feeder at the back of the outlet than at the front. This provides expansion in both plan and elevation as is shown in Figure 6 .

Fccd direciiiw

Ruhhcr skin

Belt l d c r

Figure 6 Typical mass flow belt feeder interface It is important that the bed depth of material at the front of the outlet be at least 1.5 to 2 times the largest particle size to prevent blockage.

Vibratory Feeders Vibratory feeders are excellent in providing a nearly continuous curtain of material discharge. Electromagnetic vibratory feeders are extremely rugged and simple in construction; thus, they are well suited to being used in hostile and dirty environments. Like screw feeders and rotary valves, they can be enclosed to eliminate dusting and product contamination. They are, however, limited for the most part to feeding from round, square, or slightly elongated openings. Rotary Valves Rotary valves are generally limited to being used with hoppers having circular or square outlets. Thus, they are not as useful when handling cohesive bulk solids as, for example, a screw or belt feeder. A rotary valve can also be used as an air lock when feeding into a higher or lower pressure environment, such as a pneumatic conveying line. Gates To make it easier to perform maintenance on a feeder, various types of gates, such as clamshell or slide gates, are used to isolate the feeder from an upstream bin. If the bin is designed for mass flow, it is important that there be no protrusions into the flow channel when the gate is open. Thus, the inside dimensions of the gate must exceed those of the bin outlet. In addition, gates must generally be operated only in a full-open or full-closed position, and not to modulate the rate of solids flow. This is the job of the feeder. A partially opened gate will allow stagnant regions to form above it, resulting in a funnel flow pattern. The height of the gate should be minimized to reduce the additional head pressure on the feeder.

1489

REFERENCES Jenike, A.W. Storage and Flow of Solids, University of Utah Engineering Experiment Station, Bulletin No. 123, Nov. 1964. Carson, J.W. and Marinelli, J. Characterize Bulk Solids to Ensure Smooth Flow, Chemical Engineering, Vol. 101, No. 4, April 1994, pp. 78-90.

1490

Design of Barrick Goldstrike’s Two-Stage Roaster David Warnica’, Andy Cole:, and Stanley Bunk3

ABSTRACT Two fluid-bed roasters were constructed in 1999 to pretreat carbonaceous refractory ores at Barrick Goldstrike Mines. These are the largest gold-ore roasters in the world, which use almost pure oxygen to remove contained organic carbon and sulfide sulfur prior to a conventional carbonin-leach process for gold extraction. This article describes the overall roasting process and design features that enable these roasters to exceed their design objectives. INTRODUCTION Indications in the early 1990’s that ore reserves were becoming increasingly carbonaceous at Barrick Goldstrike Mines led to reviews of alternate pretreatment technologies. These carbonaceous ores are called double refractory, because the gold is locked within sulfide minerals and the carbonaceous matter adsorbs gold during conventional cyanide leaching (Chryssoulis and Cabri, 1990; Afenya, 1991). The existing pretreatment using pressure oxidation (autoclave leaching) would not overcome the preg-robbing characteristic of carbonaceous material, which would significantly hurt gold recoveries. The carbonaceous ores were to be mined starting in late 1999, and comprise a reserve of approximately 80 million tons with an average grade of 0.177 troy ounces per ton. An alternate pretreatment had to be determined in the meantime to maintain annual gold production at about two million ounces. Two metallurgical processes for pretreatment were considered: Acid or alkaline autoclaving followed by ammonium thiosulfate (ATS) leaching Whole ore roasting Both pretreatment options were the subject of bench-scale metallurgical testing and pre-feasibility engineering studies. These studies concluded that whole ore roasting was the most proven technology at the time. Other pretreatment options were discounted if they were inherently unsuitable to the Goldstrike ores or had not received significant commercial applications. Different roasting options were therefore considered, including Lurgi’s circulating fluid bed (Folland and Peinemann, 1989) and Freeport-McMoRan’s oxygen-enriched roasting (Smith, McCord, and O’Neil, 1990). Based on extensive testwork, oxygen-enriched roasting was selected for pretreatment due to its ability to achieve high gold recoveries from the Goldstrike ores. This technology was also well proven in commercial operation (Brittan, 1995), with two plants operating since 1989 at Jerritt Canyon and Big Springs (now decommissioned). By mid-1997, a preliminary process flowsheet was defined with three roaster circuits. Subtle changes in the mine plan and further engineering studies enabled the use of only two roasters, with

’ Hatch Associates Ltd., Mississauga, ON.

Barrick Goldstrike Mines Inc., Elko, NV. Technip-Coflexip, San Dimas, CA.

1493

corresponding refinements in the oxygen plant and the gas cleaning system. One of these roasters is shown in Figure 1, with parts of the ore feed and calcine quench systems.

-BUCKET ELEVATOR AIR SLIDE

F L U I D I Z E D FEEDER-

ROASTER S I L O

AIR SLIDE ROASTER F E E D D I S T R I B U T I O N BOX AIR SLIDE

WEIGH B E L T

OUENCH LAUNDER QUENCH TANK

Figure 1: Barrick Goldstrike Roaster (showing parts of the ore feed and calcine quench systems)

1494

Detail engineering was based on the following unit operations for pretreatment of carbonaceous ore reserves at an average rate of 12,000 short tons per day (STPD) (Cole et al., 1999 and 2001):

0

Primary and secondary crushing run-of-mine ore to obtain 80% passing 1.9 cm (0.75 in) Dry grinding of crushed ore using two double-rotator mills to obtain 80% passing 74 microns Whole ore roasting of ground ore using two 2-stage, oxygen roasters to remove contained organic carbon and sulfide sulfur by oxidation Two gas quench and dust scrubbing systems, one for each roaster, including off-gas condensers and mist eliminators One common gas cleaning train, including electrostatic precipitation and removal of mercury, SOz, CO, and NO, Conventional carbon-in-leach (CIL) process for gold extraction from the roasted ore.

The existing autoclave plant was to continue processing non-carbonaceous ores, and gold desorption and recovery were to be carried out using existing facilities at the Goldstrike plant.

PROCESS DESCRIPTION Ore Mineralogy The majority of the carbonaceous ore reserves at the Barrick Goldstrike property are in distinct zones w i t h the footprint of the Betze-Post pit. The remainder of this ore comes from underground, primarily the Rodeo and Griffen deposits and to a lesser extent Barrick’s Meikle mine. The mineralization is generally within a Devonian Popovich formation. The host rock is typically decalcified muddy limestone and silicified sedimentary breccias. The gold is mainly present as colloidal gold occluded in the arsenian pyrite and marcasite, so the gold concentration generally increases in the fine-grained sulfide minerals. There are also trace amounts of orpiment, realgar, stibnite, arsenopyrite, and cinnabar present in the deposit. Sulfide sulfur concentrations vary from 0.5%-3.5% throughout the reserves. An average sulfide grade of 1.9% is expected initially but is predicted to drop slightly in the later years of processing. There is significant variation in the carbonate content throughout the reserves. For the first five years of processing, the carbonate concentration is expected to be about 5%, whch is near the historical Betze-Post level. A significant increase is projected as mining progresses to the west of the Betze-Post pit, where carbonate values are anticipated in the range from 15%-20%. Organic carbon content in the ore ranges from 0.5%-4%. Tests conducted on the carbonaceous material indicate the ore is strongly preg-robbing as defined by the standard pregrob and Barrick Goldstrike Mines (BGMI) bleach leach procedures. Further mineralogical evaluations find that the strongly carbonaceous matter is hghly amorphous and generally has a small crystalline structure. All the carbonaceous materials analyzed from Goldstrike ores are comparable to anthracite or higher-grade coal. Roaster Chemistry Various chemical reactions occur in the roaster, including: Combustion of organic carbon Combustion of carbon monoxide Combustion of sulfide sulfur Fixation of sulfur dioxide with ore, lime, and hematite Reaction of lime with carbon dioxide Dehydration of ore

1495

0

Vaporization of mercury Oxidation of nitrogen.

These reactions are described in more detail below. Based on chemical analyses of the reactants and products in pilot and commercial-scale tests, this process chemistry provides an adequate model for mass and energy balances of the roasting process. Combustion of Organic Carbon. Carbon from all sources (ore, coal, and oil) oxidizes within the roaster fluid beds to carbon monoxide and carbon dioxide with a resulting split between CO and C02 estimated at 4% and 96% (percent carbon by weight). Carbon monoxide may oxide further in the roaster freeboard to reduce CO levels in the off-gas. The carbon reactions occur as follows:

The overall extent of organic carbon oxidation in the ore is estimated to range from 81% to 89%, depending on the ore mineralogy. The extents of coal and diesel-oil oxidation are approximately 99.5% and 100% overall, if applicable. Carbon monoxide remaining in the roaster off-gas is removed by a CO incinerator. Combustion of Sulfide Sulfur. Oxidation of orpiment (As2S3),realgar (AsZS2),and arsenopyrite (FeAsS) proceed simultaneously until they are fully reacted. Due to the highly oxidizing environment, essentially all of the arsenic is converted to solid ferric arsenate (FeAs04) or arsenic pentoxide ( A s ~ O ~which ) , report to the calcine:

Decomposition and partial oxidation of pyrite (FeS2) proceeds to pyrrhotite (Fe7Ss), most of which is hrther oxidized to hematite (FezO3):

The overall extent of sulfide combustion is estimated at 99%, with 97.5% reacting in the firststage bed and 1.5% in the second-stage bed. Fixation of Sulfur Dioxide. Sulfur dioxide is fixed by reactions with minerals in the ore (carbonates and hematite) and with the lime added to the ore prior to dry grinding. Based on pilot test results, sulfur dioxide fixation ranges from 54.5%-89.5% by reaction with the ore minerals. The lime addition rate is controlled at 50% of the stoichiometric requirement for the remaining sulfur dioxide, with a lime utilization of 60%. Therefore, the lime fixes 30% of the sulfur dioxide remaining after reaction with ore minerals. The total sulfur dioxide fixation in the roaster is estimated to range from 68.1%-92.7%, depending on the carbonate content of the ore. Sulfur dioxide reacts with carbonate and hematite in the ore:

Additional SO2fixation is by reaction with lime added to the ore:

1496

Sulfur dioxide remaining in the roaster off gas is removed by the gas-cleaning system and SO2 scrubber, which has an overall removal efficiency of 99.95%. Reaction of Lime with Carbon Dioxide. Ten percent of the lime is estimated to react with carbon dioxide: CaO(s)+ C02(g)+ CaCO,(s)

Dehydration of Ore. The dehydration of clays or other hydrated minerals occurs predominantly in the first-stage bed as evaporation of water to superheated vapor. Moisture in the roaster feed ore and water vapor in air additions are also superheated. Vaporization of Mercury. All of the mercury in the feed ore vaporizes to elemental mercury in the roaster off-gas, which is removed by the gas cleaning system.. Oxidation of Nitrogen. Nitrogen is thought to be oxidized in the roaster with a 9 to 1 volumetric ratio of NO to NO2:

Although there is usually little conversion of atmospheric nitrogen to NO, at temperatures below 980°C (1,80O"F), the presence of NO, in the off-gas may be due to feed nitrogen or the high partial pressure of oxygen.

Roaster Operation The roasting process is accomplished in two parallel circuits to heat carbonaceous refractory gold ores and oxidize the contained organic carbon and sulfide sulfur. As shown by Figure 2, each roaster is comprised of two bubbling, fluid-bed reactors in a single vessel with an average design capacity of 6,000 STPD. This equipment includes a fluidized feed system, first and second-stage cyclone systems, and ancillary systems. The following description refers to one of the roaster circuits. Key process parameters are established from extensive testwork on Goldstrike ores. As shown by the process flow diagram in Figure 3, the fluidized feeder distributes ore continuously from its hopper to the first-stage (upper) bed of the roaster. The feeder is fluidized with air and overflows into the roaster through standpipes extending into the first-stage bed. The roaster feed ore is 80% passing 74 microns. Up to 30% of the ground ore is smaller than 10 microns, depending on the extent of particle agglomeration. The upper bed is maintained at constant temperature in the range from 524"-593"C (375"1,100"F). The heat source for the first stage is provided by the ore's net heat of reaction. Coal may be added to the ground ore to provide additional heating for low fuel-value ores, or fresh water may be injected to cool the first-stage bed for high fuel-value ores. Water may also be sprayed into the first-stage freeboard to cool the roaster off-gas, if excessive freeboard combustion occurs. The ore is oxidized predominantly in the first stage. Solids discharge continuously from the first stage through the inter-stage solids transfer system to the second-stage (lower) bed, which is maintained at a constant temperature in the range from 524"-621°C (975"-1,15O"F). If heating is required, diesel oil is injected through oil guns around the bed circumference. Water may be added with the diesel oil to reduce flame temperatures and sintering, in amounts up to 75 percent relative volume. Oxidation is essentially completed in the second stage. The overall oxidation is typically 99 percent for sulfide sulfur and 88.5 percent for organic carbon. Solid product discharges by gravity from the second-stage bed to the calcine quench system. Low-pressure, high-purity oxygen (99.5% by volume) is introduced as the fluidizing medium through the cold windbox to the second-stage bed and the solids transfer boxes. Hot exhaust gases from the second-stage reactor are conveyed through the inter-stage gas transfer system and the hot windbox to fluidize the first-stage bed. The hot, oxygen-rich gas from the second stage promotes

1497

FLUIDIZED FEEDER FIRST-STAGE CYCLONES F I R S T STAGE FREEBOARD WATER GUN

(

TYP

)-

FIRST-STAGE F L U I D BED F I RST -STAGE W I NDBOX

(

HOT

)-

-ASPIRATOR ( T Y P )

SECOND-STAGE CYCLONES

INTERSTAGE SOLIDS TRANSFER SYSTEM SECOND-STAGE FREEBOARD

OIL GUN ( T Y P )

INTERSTAGE CONE VALVES SECOND-STAGE F L U I D BED

-

DISCHARGE CONE VALVES

F L U I D I ZED TRANSFER BOX ( T Y P )

Figure 2: Barrick Goldstrike Roaster (showing equipment details) rapid oxidation of the organic carbon, sulfur, and fuel in the first-stage bed. The first-stage exhaust gases are de-dusted in the gas discharge system before passing to the gas quench and cleaning train. The exhaust gas from each roaster stage is de-dusted in a set of high-efficiency, primary and secondary cyclones. In each stage of the roaster, a coarse fraction of elutriated solids is recovered by two primary cyclones and returned to the bed. A fine fraction is recovered by two secondary cyclones and fed forward through the reactors, to reduce accumulation of tines withm the roaster. Fines from the first stage are therefore fed forward to the second stage, and fines from the second stage are fed forward to the calcine product. Dust is carried over to the roaster off-gas cleaning circuit from the first-stage cyclone separators. The off-gas system is designed to process dust in amounts up to 4.5% of the dry ore feed, but the actual amount is much less in practice and depends on the off-gas rate. Dust collected within the roaster area is recovered in the roaster dust collection baghouse and returned to the roaster feed distribution box. Cleaned air is emitted to the atmosphere through the roaster baghouse stack.

1498

.to Gas

Handling Secondary Cyclone Primary Cyclone

Freeboard

(First Stage)

Fluid Bed

Freeboard (Second Stage)

Fluid Bed

QumchHead

Figure 3: Process Flow Diagram Typical operating curves are shown in Figure 4, which describe many aspects of the Goldstrike roasters. These curves reveal relationships between different equipment capacities and summarize roaster performance over a range of ore composition. At first glance, these curves appear complicated, but they provide an invaluable tool for understanding the roaster operation. Figure 4 shows three graphs with respect to the ore fuel value along a common abscissa. The ore fuel value is correlated with the contained organic carbon (%TCM) and sulfide sulfur (%S) as follows: Ore Fuel Value [BTUAb] = 120 x %TCM + 72.6 x %S where units of composition are percent of dry weight.

1499

The bottom graph shows how the maximum instantaneous ore rate can vary with the fuel value. The total oxygen plant supply is limited to 1,100 STPD for two roasters, so the maximum ore rate decreases with increasing fuel value and hence oxidant requirement. This theoretical ore rate tends to be low during early years of the mine plan, due to the high carbon and sulfur contents, which decrease during later years when more ore can be processed with the same oxygen supply. The total ore rate for two roasters is an average rate assuming 90% operating availability. The grinding limit of 12,000 STPD for two mills therefore corresponds to an instantaneous rate of 278 short tons per hour (STPH) per roaster. The retention limit (36 minutes) corresponds to a bed inventory of 208 tons in one roaster, which is the combined inventory of both stages. This minimum retention time is established by testwork to yield the desired gold recovery. The coal rate shows that first-stage heating is only required in a few years with low ore fuel values. Water is injected for bed cooling in the other years with higher fuel-value ores. The autogenous limit (225 BTU/lb) is the vertical line that separates these operating regimes. The bottom graph of Figure 4 therefore shows that the roasters can process a wide range of ore compositions at the desired average rate (12,000 STPD), corresponding to the dry-grinding capacity. The oxygen plant has some excess capacity for most of the ore feeds, except very high fuel-value ores that may require additional bulk oxygen (LOX) to supplement the plant oxygen supply. The top two graphs represent the operating characteristics of each stage, showing gas volumes and the corresponding superficial gas velocities. In each of these graphs, a single curve is able to represent the gas volume, and bed-top and freeboard velocities, due to inherent relationships between these parameters. A different curve represents the bed bottom velocity. A common ordinate is used to read both bed velocities (top and bottom). Gas velocities differ between the bed top and bottom, due to differences in vessel diameter and the possible presence of first-stage water injection. These graphs are prepared for nominal design temperatures where the first-stage (upper) reactor is maintained at 552°C (1025°F) by either coal addition or water injection. The water rate increases with the ore fuel value so there are corresponding increases in gas volumes and velocities. Operating conditions of the second-stage (lower) reactor are relatively uniform, regardless of fuel value. The temperature of lower reactor is nominally increased to 566°C (1050°F) by oil injection, but may be allowed to float. The actual operating temperatures of both stages depend on characteristics of the feed ore and its amenability to cyanide leaching.

1500

20

i

I

,

,

,

,

,

,

,

,

YPPFR RFACTOR

loo

gJ10

,

,

,

-

-

-

-

_ _ _ _

/

~

,

*

,

I

1

-

I

1

I

I

-

+

I

/

,

I

/

- 1.5

/

GasVolume; Bed Top 8 Freeboard Velocities

-

H

1.0 Q)

v)

E .-20.

- - - - _ _ Bed Bottom Velocity

-

-30

>

0

>

Q) >

u)

U

-

8= $ 5

0.5

0

I!

0 20

I

I

I

I

I

I

I

I

,

I

,

I

,

,

I

,

0

Y

1 1

0.0

II

E n 8 2!

d

0.0

L RFACTOR

0.5

2.0

2.5

2.0 W

Oil Rate

0

1.0

= 0 n W

rn

0.0

0.5

E-

0

/

I

I

I

I

I

I

I

I

I

I

I

I

I

I

I

1

1

I

1

I

1

I

~

I

I1

Il

I1 I

1

I

~

I

I

1

I1

1

0.01

2.5 350 - I

: E

-

2.0

-

325

I B

I

E 2

- 2 - k

1.5

L

3

- g 1 300

0

-

2

- w ;1.0 - g! 0

0

u)

-

0.5 -

2

275

I C

- c 0.0 -

- 250 175

200

225 250 Ore Fuel Value [BTUllb]

275

Figure 4: Barrick Roaster Operation

1501

300

1

0.0

EQUIPMENT SELECTION AND SIZING Project Design Criteria In general, the project philosophy was to use proven technology and equipment where possible to avoid custom engineering and unpleasant surprises during start up. It was also a project standard to install spare mechanical equipment and avoid unnecessary roaster shutdowns for routine maintenance. These criteria had implications for equipment selection and flowsheet development. For example, unsupported refractory domes were limited to the typical maximum diameter that has demonstrated operational experience and long-term campaigns (6.7 1 m [22 ft]). Also, installed spare equipment was supplied for most blowers, and all pumps and solids transfer systems. The only exceptions were the start-up and preheat air blowers, which are used infrequently and can be tested off line. The roasters themselves are custom designed by virtue of the specific process objectives and their size. Key features of the process are: Fluid bed roasting using almost pure oxygen as the process gas Two roasting stages in a single vessel to ensure a high level of both organic carbon and sulfide sulfur oxidation. The Goldstrlke roasters are the largest of their lund, with freestanding vessels about 33.5-m (1 10-ft) tall. Their shear size is evidence of the level of custom engineering that was required for process, mechanical, and structural engineering. Process Design Basis The process design basis incorporates many aspects, but only a few are presented to reveal some of the main considerations for equipment selection and sizing. The base-case ore composition was assumed to have 0

1.9% sulfides as S 1.2% organic carbon as C 5.5% carbonates as CO,.

with a fuel value of 282 BTU/lb. This composition represented an average feed for the first five years of operation. The corresponding reactions were modelled using the roaster chemistry to yield a mass and energy balance of the system, which was generally used for sizing the vessel and other equipment. The oxygen plant supply was to be 1,100 STPD for two roasters, with a purity of 99.5% by volume. The process was to be controlled so the concentration of oxygen in the off gas is not less than 10% (wet basis). A special design case was also used to represent the maximum velocities and temperatures expected in the first-stage freeboard, for sizing this part of the vessel and the off-gas system. The maximum velocities were assumed to occur for a theoretical ore composition with the maximum concentrations of organic carbon and sulfide sulfur, not necessarily for the same year of the mine plan. The corresponding ore fuel-value was 302 BTU/lb. The maximum temperature was selected at 760°C (1,400"F) to account for possible combustion of carbon monoxide in the frst-stage freeboard. Number of Roasters Three roasters were initially thought to be required, but changes in the mine plan and consideration of equipment capacities made it possible to use only two roasters. The two-roaster configuration was able to process the same amount of ore as three roasters (12,000 STPD), but used less oxygen (1,100 STPD instead of 1,350 STPD). This equipment reduction was made

1502

possible by a revised mine plan with lower ore fuel values than what were initially projected. The lower oxygen consumption resulted in lower gas volumes and therefore reduced the required cross-sectional areas for gas flows.

Internal Dimension

Vessel Location First-stage Diameters - Freeboard

10.82 m

(35.5 ft)

- Bed Top

8.38 m

(27.5 ft)

- Bed Bottom

6.71 m

(22 ft)

6.71 m

(22 ft)

- Bed Top

6.71 m

(22 ft)

- Bed Bottom

5.79 m

(19 ft)

6.40 m

(2 1 ft)

- Bed

3.96m

(13 ft)

- Freeboard

6.17 m

(20.25 ft)

- Bed

5.18m

(17ft)

Second-stage Diameters - Freeboard

First-stage Heights - Freeboard

Second-stage Heights

Bed Temperature Control Due to inherent trends of the mine plan, the roaster design had to be sufficiently flexible to accommodate wide variations in the feed ore composition. Depending on the organic carbon and sulfide sulfur content, which affect the ore fuel value, the overall system energy requirement

1503

ranges from slightly endothermic to highly exothermic in the frst-stage reactor. Laboratory tests reveal that subtle temperatures of the first and second-stage reactors would affect gold recovery. A critical relationship exists between roasting temperature, residence times, and the subsequent amenability of the ore to cyanide leaching. Precise control of these parameters is essential, so independent temperature control is required for each stage. If the organic carbon and sulfide sulfur content of the feed ore is low, coal can be added to the feed streams upstream of each roaster. If the feed ore contains a large amount of organic carbon or sulfide sulfur, an excess amount of heat can be generated in the first stage, which must be removed. Two options were considered for cooling the first-stage bed: direct quenching by water injection and indirect cooling by bayonet heat exchangers. Water injection was selected instead of indirect cooling, due to its lower capital cost, maintenance requirements, and engineering complexity. Using this system, the first-stage bed temperature can be controlled precisely over the desired range from 524"-593°C (975"-1,100"F). The feed ore is oxidized predominantly in the first stage, and the remaining fuel value is only sufficient to offset some heat losses from the second-stage and the sensible heating of the fluidizing oxygen. Since gold recovery may be enhanced by heating the second stage to a hgher temperature than the first, instead of letting it float, two options were considered to heat the second-stage bed: preheating the fluidizing oxygen and direct fuel injection. Oxygen preheating was dismissed to avoid the increased capital cost and engineering complexity of adding heat exchangers and replacing steel gas distributors with refractory domes, because oxygen temperatures greater than 315°C (600°F) would have been required. Direct fuel injection was therefore selected, but special considerations were given to reduce flame temperatures and avoid sintering the bed, which is thought to occur at bed temperatures above 650°C (1,200"F). Provisions were made to inject an emulsion of diesel oil and water, which can control secondstage bed temperatures over the desired range from 524"-621°C (975"-1,15OoF) and reduce the tendency for sintering. Roaster Off-Gas Cleaning The roaster off-gas system quenches the hot off-gas and then processes it for removal of particulate solids, mercury, SOz, CO, and NO,. Independent off-gas systems were considered initially for each of three roasters, but this system was reviewed when the number of roasters was reduced. The outcome of this review was an independent gas quench and dust scrubbing system for each of two roasters, and a common gas cleaning train for removal of mercury, SO2, CO, and NO,. Each gas quench system included an off-gas condenser and mist eliminator, and the common gas cleaning train included electrostatic precipitation. Oxygen Plant The basis for design of the oxygen plant is a cold box capacity of 1,100 STPD with a total production capability of 1,200 STPD by periodic receipt of up to 100 STPD of external LOX. This significant reduction in the required capacity is due to changes in the mine plan. Oxygen is supplied to the roasters at two pressures. Low-pressure oxygen up to 100 kPa (15 psig) is used predominantly for fluidization. Aspirators that purge solids from the hot windbox use medium pressure oxygen at 550 kPa (80 psig). All oxygen content is 99.5% by volume. The use of high-purity oxygen at elevated temperatures in the roasters required careful attention to compatibility with materials of construction. Mild steel is used generally where gas velocities are low and there is no impingement. Stainless steels or nickel alloys are required in other critical locations such as orifice plates and aspirators. Thread sealants, gaskets, and packing materials also required careful selection.

1504

DESIGN FEATURES Solids Feed and Distribution System Based on other plant experience, multiple feed chutes are used to improve operations and product quality. Splitting a high solids feed rate into several equal streams presents a challenge. The fluoseal feed distributor aerates the solids, and overflow weirs yield uniform distribution to each of six feed chutes. This device has been used successfully for many calcining and roasting applications. The feed chutes extend into the first-stage (upper) fluid bed, to maintain a pressure seal and avoid venting hot roaster gases to the dust collection system of the feed distributor. Feed chutes are vented to the roaster freeboard to avoid solids slugging and maintain constant flow. The internal support arrangement of the feed legs allows thermal expansion in a dust-laden environment, using lateral braces with guides on the vessel shell. Solids Transfer Systems In general, the solids flow easily when aerated, but not after being stagnant during shutdown periods. An almost vertical angle of repose is evidence of this undesirable behaviour of stagnant solids. All solids transfer chutes are therefore vertical, except very short runs from the fluoseal feeder to the feed legs. These angled runs do not pose transfer problems, however, because solids entering these pipes are generally well aerated and the pipes are never filled with stagnant solids. Purge air is introduced at critical locations such as the cone valves which are used to control bed levels. Fluidized transfer boxes are used exclusively to convey solids between chutes and fluid beds. Transfer boxes on the Goldstrike roasters are the largest ever built, due to the large difference between freeboard and bed diameters. All transfer boxes are hung from spring hangers for additional support that accommodates thermal expansion of the vessel. Thorough fluidization of the transfer boxes is an important aspect to aerate the solids and maintain flow, especially in regions below the incoming chutes. First-stage Gas Distribution and Hot Windbox Hot exhaust gases from the second-stage freeboard are conveyed to the hot windbox that separates the first and second-stage reactors. The hot windbox is formed by the space between the two selfsupporting refractory domes. Gas enters the side of this windbox and flows upwards through 701 tuyeres to fluidize the first-stage bed. Although the gas is cleaned to some extent by primary and secondary cyclones, fine particulate solids are unavoidable, which pose long-term maintenance requirements of either solids build-up (scaling) or erosion of the tuyere holes. Tuyeres are designed to minimize these possibilities, by using relatively large vertical holes to reduce gas velocity and pressure drop. Based on test experience, the hole size is selected to allow solids bridging and avoid draining the bed during shutdowns. During normal operation, the tuyere holes gradually plug due to scaling, which is removed during periodic maintenance shutdowns. Tuyere plugging is detected by increased pressure drop. Aspirators Particulate solids that enter the hot windbox with the gas stream and by sifting through the firststage tuyeres are removed continuously during normal operation by aspirators. Twenty-four holes are distributed over the bottom refractory dome of the hot windbox, between the hot windbox and the second-stage reactor. Using principles of fluid induction, medium-pressure oxygen is injected through each hole by an aspirator to induce gas flow and purge solids from the windbox. The design of these aspirators is based on experience with phosphate calciners. The removal of solids from the hot windbox significantly reduces the maintenance requirements due to tuyere scaling and erosion.

1505

First-stage Water Injection Water guns are located around the bed circumference at two elevations. Twelve guns at the higher elevation are generally used during normal operation with full bed levels. Six guns at the lower elevation are used only during start up when the bed level is low. Water is injected typically into the splash zone near the top of a fluid bed, where there is intimate contact with solids circulating between the bed and freeboard but water vapour doesn’t increase bed velocities. Second-stage Oil Injection Depending on the desired temperature, the second-stage bed may be heated by injecting an emulsion of diesel oil and water through oil guns located around the circumference. Diesel oil is supplied evenly to each of 16 oil guns by positive displacement pumps with multiple heads. Water is then mixed with the oil by an in-line mixer located just upstream of each oil gun. Water can be added in amounts up to 75% by volume to reduce flame temperatures as required, depending 011 operational experience and gold recovery. Air from the purge air blower conveys the oil and water emulsion into the bed, and keeps oil guns clear of solids when not being used. Air is also used to cool the oil guns. Freeboard Water Sprays Six atomisation nozzles are installed in the roof of the first-stage freeboard, to spray water and protect the roaster and off-gas system from high temperatures due to possible combustion of carbon monoxide. Fresh water is delivered by a high-pressure turbine pump, as required at rates up to 1.4 m3/hr (6 USgpm). Different numbers of nozzles admit water, depending on the freeboard temperature. This type of system has been used for numerous fluid-bed reactors, primarily for incineration applications, Start-up Air System The roasters are normally operated using high-purity oxygen, but started up using air. One multistage centrifugal blower is connected to both roasters, and isolated as required to start one roaster at a time. There is a smooth transition from start-up air to oxygen as the desired bed temperatures and levels are achieved. Preheat Burner System Two preheat burners are mounted on the second-stage freeboard to heat a shallow first-stage bed above its autoignition temperature. Preheat equipment is used only for cold starts or after extended shutdowns. Ore is introduced and bed levels are increased after the desired temperature is achieved. Each burner is rated at 2,930 kW (10 million BTU/hr). One multistage centrifugal blower is connected to both roasters, and isolated as required to start one roaster at a time. The preheat burner blower supplies 100% excess air to two burners. Purge Air Blower System Air is injected into the roasters at various locations to 0 0

Fluidize the feed system Provide the transport medium for diesel oil and water injection Cool oil guns Purge solids from instruments and other equipment.

Three positive-displacement, purge air blowers are used for this air supply. One blower provides the air requirements of each roaster, and an additional blower is connected as an installed spare to provide continuous operation during blower maintenance. Each blower has a capacity of 1,200 Nm3/hr (700 ft3/min),which is small in comparison with the oxygen supply.

1506

OPERATING EXPERIENCE The commissioning and start-up of the roasters proceeded smoothing. Start-up of the roasters was separated by one month for both construction and ease of commissioning reasons. T h s proved to be an advantage for the start-up of the second train in that design deficiencies were corrected prior to going hot. Design capacity for the first train was reached in about two weeks after start-up while the second train reached capacity within hours of start-up. There were very few modification required during the start-up phase and those problems needing correction were minor. The most significant issue was associated with solids feed and distribution system. The feed system turned out to be over-sized, causing the feed to slug flow and break seal, which ended up producing fluctuation in the roaster’s freeboard pressure control. A very simple fix of blinding two out of the six feed legs was all that was required to stabilized the feed rate. Dust leaks were probably the most troublesome struggle during the first few months of operation. Frequent cycling of the roasters due to upstream and downstream problems caused the standard stainless bolts installed during construction to stretch and lose strength. Those bolts were eventually replaced with bolts made of alloy steel having a greater tensile strength and increased ductility. Another problem was erosion of vent lines providing equalization between the cyclones and fluoseal and the roaster. This would not have been as serious of a problem if isolation valves had been provided on both ends of the vent line. Eventually valves were retrofitted on every connection coming to and from the roaster. Probably the most significant problem has been with the second stage of the roaster. Upon defluidization pressure is retained in the bed causing severe back-sifting of calcine into the cold windbox. The result has been accelerated erosion of the club head tuyeres in the second stage. The cause of the problems is associated with the deeper second stage and the design of the discharge from the underflow of the secondary cyclones. Previous designs had the cyclone discharging to a quench head rather than a common transfer box. It is thought that with a quench head the pressure is more easily relieved. The solution has been a two-fold approach. First, a pressure relief system has been designed to relieve pressure from the freeboard of the second stage to the freeboard of the first stage. When defluidization occurs, valves open to release the pressure. Ceramic tuyeres are also being tested in hopes of extending tuyere life. The back sifting problem has necessitated the replacement of all second stage tuyeres with less than four years of service. On the positive side, one of the most significant and pleasurable surprises has been the success of the aspiration system. Time between cold shut downs for most roasters is dictated by the accumulation of solids and the ultimate build of pressure across the hot windbox. The staggering of the aspirators and the use of pure oxygen has made the system reliable and effective. Based on the first two years of experience solidpressure build-up across the hot windbox will not be an issue with these roasters. The performance of the roasters during the first two years of operation has exceeded expectations with respect to both throughput and operational availability. As shown in Figure 5 and Figure 6, the performance and reliability have been exceptional. The increase in production is the result of a trade-off between fuel value (BTU/lb ore) and processing rate. The roasters were designed to treat a base case heat value (282 BTU/lb ore at 278 STPH) for a set amount of oxygen. Instantaneous processing rates have increased by leveraging the roasters’ capacity to treat a given BTU per unit of time.

1507

16,000

7

I

12,000 0,

10,000

p?

2

8,000

0

6,000

Q

4,000

2 Q

2,000 0 3Q 2000

4Q 2000

1Q 2001

2Q 2001

3Q 2001

4Q 2001

1Q 2002

Quarter

Figure 5: Roaster Throughput by Quarter

.E .-

100.0

m m

95.0

Q

2rn .-C m b

90.0

+I

85.0

0" c

E Q

80.0

2 Q

n

75.0 3Q 4Q 2000 2000

1Q 2Q 3Q 2001 2001 2001

4Q 2001

1Q 2002

Quarter

Figure 6: Roaster Availability by Quarter ACKNOWLEDGEMENTS The authors gratefully acknowledge contributions of Barrick Gold, Barrick Goldstrike Mines, Hatch, SNC-Lavalin, Crescent Technology, Technip USA, Air Products, and Freeport-McMoRan, and express gratitude to the dedicated people who contributed to the success of this project. REFERENCES Afenya, P. M. 1991. Treatment of carbonaceous refractory gold ores. Mining Engineering. 4:711~1043-1055. Brittan, M. 1995. Oxygen roasting of refractory gold ores. Mining Engineering. 47:2: 145-148. Chryssoulis, S. L. and L. J. Cabri. 1990. Significance of gold mineralogical balances in mineral processing. Transactions of the Institution of Mining & Metallurgy, Section C. 99: 1-10, Cole, A,, S. Bunk, S. Dunn, and T. McCord. 1999. Refractory gold ore treatment by fluidized bed roasting for Barrick Goldstrike. Proceedings Randol Gold & Silver Forum '99. 79-84.

1508

Cole, A., J. McMullen, K. Thomas, and S. DUM. 2001. Barrick Goldstrike roaster facility: Roasting and gas handling. Proceedings 33"' Canadian Mineral Processors. Folland, G. and B. Peinemann. 1989. Lurgi's circulating fluid bed applied to gold roasting. Engineering & Mining Journal. October: 28-30. Smith, J. C., T. H. McCord, and G. R. O'Neil. Treating refractory gold ores via oxygen-enriched roasting. U S . Patent No. 4,919,715 (24 April 1990).

1509

Selection of Materials and Mechanical Design of Pressure Leaching Equipment Ken Lamb, AMEC Mining & Metals, James Gulyas, SNC-Lavalin

ABSTRACT Numerous pressure oxidation and pressure leaching autoclave plants have been installed throughout the world over the past 50 years. The technology is reaching maturity, as evidenced by the successful design and construction of plants to treat a wide range of feedstocks. Important aspects of the design of such plants include selection of materials, slurry pumping, slurry heating and heat recovery, pressure control and let-down, vessel design, agitator design, safety, and ancillary systems. INTRODUCTION The technology involved in pressurized hydrometallurgical processes was first developed in the early 19OOs, when the Bayer process was introduced for the production of alumina. The development of pressure hydrometallurgy for the base metals industry started in the 1950s, with pressure acid leaching of nickel laterites and ammonia leaching of nickel sulfides, and has continued through the subsequent decades. Autoclave circuits have been used or proposed to process a variety of metals, including copper, gold, molybdenum, nickel, titanium dioxide, uranium, and zinc. The circuits may require different autoclave operating conditions and reagent additions, but can generally be grouped as either pressure oxidation, requiring oxygen addition in some form, or pressure leach, operating in alkaline or acid conditions. This paper is intended to provide the reader with an overview of the important factors for mechanical design and operating practice related to autoclave circuits. OVERVIEW OF CIRCUIT DESIGN Typical unit operations in a pressure leach or oxidation flowsheet are indicated in Figure 1. The operations include: Slurry preheat Autoclave processing Flashing Vent scrubbing. The mechanical design of these circuits must address: Materials of construction Vessel design Piping and valve design Preheat circuit design Pump design Pressure relief Pressure let-down system design Agitation Seals and seal water system.

1510

t t T’

Oxygen

Figure 1 Typical unit operations in a pressure leach leach or oxidation flowsheet Materials of Construction In common with most industries, the selection of the most economic materials for specific applications is one of the first tasks facing the design engineer. Pressure oxidation and leach circuits deal with a variety of fluids, many of which are highly corrosive andfor abrasive, and operate at high temperatures. The basic input data required for material selection include the design flows, pressures, temperatures, and fluid compositions that are usually given in the process heat and mass balances. However, the mechanical designer must be aware of potential excursion limits in determining the final design parameters. Factors that must be considered in the selection of metallic alloys are corrosion rate, material cost, fabrication costs, availability of the alloy in all component forms (plate, bar, pipe), and anticipated maintenance costs over the life of the project. The use of high-purity oxygen in pressure oxidation circuits further complicates material selection owing to safety issues related to the potential for spontaneous ignition. Consideration must also be given to the non-metallic materials that are commonly used for membranes and brick linings in autoclaves and flash vessels, and to other materials suitable for the relatively low temperature sections of the circuit. Material Selection. Equipment and material suppliers can usually assist in material selection for equipment that handles the more common fluids encountered in the industry, such as reagents, including sulfuric acid. For autoclave applications, however, the range of suitable materials is relatively limited; typically, high-grade alloys or refractory metals are required. Non-metallic materials have been used as autoclave membranes and lining materials, and polymers have been developed for valve seat materials to withstand 260°C. Some vendors of specialty alloys and industry organizations such as the Nickel Development Institute and NACE (National Association of Corrosion Engineers) can provide technical assistance in materials selection through their substantial databases. Unfortunately, the databases cannot cover the almost infinite variations in slurry composition, impurities, and temperatures encountered in autoclave circuits, and designers have to rely on predictions of the anticipated alloy performance.

1511

Key parameters in the selection of materials are operating temperature, acid concentration (typically sulfuric) or pH, chloride content (particularly important for stainless steels and titanium), and concentration of metallic ions such as nickel (Ni2+), copper (Cu2+), and iron (Fe”). In addition, fluoride ion can have a detrimental impact on titanium and, to some extent, on stainless steels and brick linings. In ascending order of corrosion resistance, and usually in cost, materials used in autoclave circuits include the 300 stainless steels, super austenitic and super duplex steels, nickel alloys, and titanium. It should be noted, however, that there are some anomalies in this ranking; for example, Inconel Alloy 625 has been used in the construction of autoclave internals in gold circuits and has experienced very high corrosion rates. For this application it has therefore been replaced with duplex stainless steel alloys such as Ferralium Alloy 255, which has proved to be much more corrosion resistant. However, Inconel Alloy 625 weld overlay, applied to the carbon steel autoclave nozzles and protected from the direct process environment by brick and mortar, has been relatively successful. Titanium is the only material that has been found to withstand the autoclave environment without significant corrosion, but there is reluctance to use it in oxygen dip pipes because of the potential for spontaneous ignition in the presence of high-purity oxygen. The development of an titanium-niobium alloy has alleviated this concern to some extent, although i t could still ignite (see Figure 2 ). 400 350 300

E 200 0

f

f

150

100 50 0 0

100

200

300

400

500

600

700

Oxygen Pressure (psi)

A Nb-55Ti No Ignition 0 Ti No Ignition

+

Nb-55Ti Fire

Nb-55Ti Ignition

0 Ti Ignition

+I+ Nb-55Ti (Approx. Curve from Teledyne Wah Chang Data)

X Ti Fire

+Ti

Ignition Curve

Figure 2 Effect of temperature on spontaneous ignition of ruptured unalloyed titanium & niobium-titanium in oxygen In North America, materials are identified by UNS (Unified Numbering System) numbers, which specify an allowable chemical composition range for the alloy. They are not material supply specifications, however, and as such must be used with caution. Some vendors may supply material at the lower end of the chemical analysis which would reduce costs, but would also result in poorer corrosion resistance and mechanical properties. A more-restrictive chemical analysis, and corresponding new UNS number, has been applied to some alloys to overcome these problems. For example, the more general chemical composition wrought Duplex Alloy 2205, UNS S3 1803, has been supplemented with a tighter chemical specification UNS 32205.

1512

Nickel alloys, stainless steels, and titanium are supplied to an ASME (American Society of Mechanical Engineers) or ASTM (American Society for Testing Materials) specification that includes requirements for ordering, manufacturing, testing (including tensile tests and minimum tensile requirements), inspection and marking, and allowable dimension variations, as well as for chemical composition. At present, titanium alloys are the only ones not referred to by their UNS numbers in these specifications; however, this is expected to change in future editions. Current metallurgical trends are towards more highly alloyed austenitic and duplex stainless steels, which may be used for equipment and piping outside the high-temperature autoclave environment. Corrosion resistance generally increases with increasing chromium, molybdenum, and nitrogen content in stainless steel alloys. More super austenitic and super duplex alloys offering higher pitting resistance equivalent numbers (PRE, or PREN when nitrogen is included) are becoming available, which offer improved localized pitting and crevice corrosion resistance in aqueous solutions containing halides (chlorides and fluorides) at low pH. Halides increase acid corrosion by making it difficult to maintain the passive surface layer, and high chloride concentrations and low pH, as well as high temperatures and stagnant conditions, increase the probability of pitting and crevice corrosion. The PREN numbers establish a measure of ranking for pitting and crevice corrosion resistance by calculating the sum of the most important alloying elements in a weighted form. The higher the PREN number, the better the resistance to crevice corrosion. Table 1 shows typical PREN numbers.

Table 1 Typical PREN numbers UNS

Material

Type

Form

%Ni

%Cr

%Mo %W

%N

Other

2.5 0.13 12.5 17 SS Wrt S31603 316L 11 19 2.5 J92800 CF3M ss cast Wrt 35 20 2.5 SS NO8020 Allov2ocb3 3.0 Cu 5.5 255 2 Dup Cast J93370 Cd4MCu 3 0.15 5.5 22 Dup w r t S31803 2205 2 0.2 3.0Cu 5.5 255 DUD Cast N/A Cd4MCuN 1.5Cu NO8904 904L ss w/c 25.5 21 4.5 3 0.18 2.0Cu 5.5 25.5 W/C Dup S32550 Ferraliurn255 1.oCu 32 27 ss w r t 3.5 NO8028 Sanicro28 4 0.25 7 2 5 Dup w r t S32750 2507 3.7 0.7 0.25 0.7 Cu 6.5 25 Dup Wrt S32760 Zeron100 0.2 0.75 Cu 18 20 6.25 S31254 254SMo ss w r t 24.5 21 6.5 0.2 SS Wrt NO8367 Al6XN Base 22.5 7 1 3.4 c o Wrt NO6985 HastG3 Ni Wrt Ni NO6625 Alloy625 Base 21.5 9 7.3 0.5 0.5 Cu ss w r t 22 24 S32654 654SMo 3.0FeMx NO6455 AlloyC4 Ni Wrt Base 16 15.5 13.5 3 3.0 Fe NO6022 AlloyC22 Ni Wrt Base 22 16 3.75 5.5 Fe N10276 Alloy276 Ni Wrt Base 15.5 PREN numbers calculated from: %Cr + 3.3 x (%Mo + OS%W) + 16 x (%N) W/C - wrought or cast Wrt - wrought

1513

PREN

27.3 27.3 27.3 32.1 34.3 35.3 35.9 38.3 38.6 42.2 42.4 43.8 45.7 47.3 51.2 56.1 67.2 71.5 74.5

Ignition of titanium. The application of titanium in oxygen environments has received considerable attention in pressure oxidation applications because a freshly exposed surface of titanium alloy may spontaneously ignite in oxygen concentrations greater than 35% by volume. The oxygen content in the vapor space of typical pressure oxidation autoclaves is usually less than 35% during normal operation, with only the oxygen dip pipes subject to high oxygen partial pressures. Designs and materials have been developed to mitigate the damage from oxygen usage. One example is the use of heavy-weld neck flanges on the oxygen dip pipe inside the vessel to act as a heat sink and limit flame propagation if it should ignite. Also, titanium should not be used in areas where a high-velocity, oxygen-enriched gas jet may occur, such as at a seal between autoclave pressure and the atmosphere. Alloy Ti-45Nb has a wider operating regime than the more-common Grade 2 titanium in oxygen environments with respect to reduced potential for spontaneous ignition, and has been used where gas enriched with oxygen can be anticipated. Figure 2 shows the ignition limit for titanium and titanium-niobium, A number of titanium fires have occurred at autoclave installations over the years, causing considerable equipment damage and in some cases personal injury. The safety aspects of material selection cannot be taken lightly. A full HAZOPS analysis, together with careful design by experienced personnel, will reduce the likelihood of such incidents. Vessel Design General. The most common codes used in the design of pressure vessels for pressure hydrometallurgy are ASME Div l(North American), AS-1210 (Australian), and BS-5500 (British). While the codes are similar, they have some variances in allowable stresses and design requirements (e.g., for masonry lining) that will result in small differences in shell thickness. This section outlines design considerations based on the ASME code. In North America unfired pressure vessels are designed to ASME Section VIII, Division 1 or Division 2. Division 1 uses approximate calculation methods that are adequate for most services. Division 2 provides an alternative to the minimum construction requirements of Division 1. Division 2 rules are based on more-precise calculation methods and are more restrictive. They do not permit the use of some materials allowed by Division 1, they prohibit some common design details, and they specify which fabrication procedures may be used. Engineering and manufacturing costs for materials meeting Division 2 criteria are therefore higher than those meeting Division 1 criteria. Where the stress intensity is controlled by ultimate or yield strength, Division 2 permits the use of higher design allowable stress values in the range of temperatures covered. Hence, for autoclave applications, Division 2 may be considered where savings in materials and labor justify the costs of the necessary engineering analysis and more rigorous construction requirements. Where expensive materials or large vessels are being used, this may be the most cost-effective approach. In some cases, material savings accompanying the Division 2 criteria for carbon steel are not substantial until the shell thickness approaches 100 mm (4 inches). ASME Section VIII pressure vessel code changes. Various criteria are used in establishing maximum allowable stress values for pressure vessel codes. It is important to note that the Division 1 safety factor was recently changed from 4: I (in use since 1944) to 3.5:l and brings the stresses closer to those of most European pressure vessel codes. This change was issued on July 1, 1999. The temperature at which allowable stress values are affected has also been changed. Previously, a design temperature of 650°F or less did not affect the allowable stress value of SA516-70, a carbon steel plate material in the pressure vessel code commonly used in vessel fabrication. That temperature value has been changed to 500°F. However, the allowable stress value for SA-106-B, a carbon steel seamless piping material also used in vessel fabrication, does not change until 650°F. Each item needs to be reviewed to determine the allowable stress at the vessel design temperature. It should also be noted that the design temperature always affects external pressure calculations and flange ratings.

1514

Another code change has been made in the area of required minimum hydro-test pressure; this has changed from 1.5 times MAWP (maximum allowable working pressure) to 1.3 times MAWP. Vessels designed to the new Division 1 criteria will be thinner, lighter, and cost less than vessels installed under the earlier code criteria. The material cost savings between Division 1 and 2 vessels will not be as significant under the new code rules. The increase in allowable stresses could lead to existing vessels being re-rated for higher design pressures, which could benefit some processes. A National Inspection Board Code interpretation indicates that re-rating due to code changes in allowable stresses is acceptable as long as certain criteria are fulfilled; however, some jurisdictions will still not allow it. Autoclaves. Currently accepted construction options for autoclave vessels are titanium clad steel or a carbon steel shell protected by membranes of lead or lead with fiberglass, and acidresistant brick. Brick-lined vessels. Historically, the most common design for pressure oxidation applications has been horizontal vessels with a carbon steel shell, lead and/or vinyl ester membrane, and two layers of acid-resistant brick lining. In the case of brick-lined vessels, the steel shells, internals (compartment walls, nozzle inserts, dip pipes and baffles, typically made from non-ferrous materials), and brick lining have usually been specified and purchased separately. Where a complete unit was supplied, the lining manufacturer took the coordinating role and subcontracted the other work. This purchasing concept - where the vessel design is done in coordination with the lining design - reduces the work done by the owner’s agent, i.e., the engineering consultant, but does restrict the number of companies that are willing to bid. Other than accounting for the lining weight, the ASME pressure vessel codes d o not specifically consider brick linings. However, the codes do state that the additional loads placed on the shell from the lining must be determined. The British code, BC-5500 contains design guidelines for lined vessels. A primary objective of lining design is to balance the stresses between the steel shell and the lining. Since the tensile strengths of the acid brick and mortars are relatively very low, they must be maintained under compression or, at worst, with a small amount of tension. On the other hand, the compression must always be below the compressive strength of the brick to prevent spalling or crushing and not impose severe additional tension on the steel vessel wall. A number of variables contribute to lining stresses, such as: Selection of materials for acid resistance and erosion resistance (includes brick mortar and membrane). The fact that bricks irreversibly swell in wet acidic conditions must be considered in selecting materials for the bricks. Thermal properties of the brick and mortar (used to calculate thermal gradient across lining system and hence the temperature of each component and its corresponding expansion) Number of brick layers and thickness Operating pressure, temperature, vessel diameter and material, and corresponding shell thickness. In lined pressure vessels, stresses are purposely introduced into the lining system before operation. A newly lined vessel is cured in an acid solution at a temperature around the boiling point. During this curing time, a chemical reaction occurs in the brick that causes it to swell irreversibly, thereby increasing the stresses and “tightness” of the lining system. Bricks have been known to swell during an initial period of a few weeks, stop, and continue months later. The lining system will be subject to various stresses during start-up, normal operation, and shutdown and must be designed accordingly. In addition, procedures must be developed to minimize stresses that could damage the lining during heat-up and cool-down. The procedures include rates for pressurization and depressurization, temperature increase and decrease, and time for soaking.

1515

Additional loads can be imposed on the lining system by the vessel itself. For example, if the vessel is out of roundness, or the vessel weight and support locations introduce unacceptable deflections at the vessel centreline. Tolerance criteria for out of roundness and deflection have been developed to minimize the transfer of these additional loads to the brickwork. Three major companies design and install pressure vessel linings in North America: Stebbins, Didier, and Koch. Each company has recommendations about materials for membranes, bricks, and mortars, as well as installation details and techniques. For example, Koch prefers to supply a sulfur-enhanced elastomer (Pyroflex) membrane, which, owing to its flexibility, will take up some of the brick expansion. Stebbins use its vinyl ester (AR 500) membrane and Hydromet (lead based) mortar for difficult vapor zone applications in pressure oxidation applications. Didier has developed a potassium silicate (Stellakit A) mortar for the vapor zones. The companies have bricks that differ in swell properties and resistance to vapor zone conditions of steam condensation, where the bricks and mortar tend to soften. Lead has been the traditional membrane lining for pressure oxidation service. Given the increasing awareness of the health hazards of lead, the development of materials to replace it has become a priority for vendors. The installation of a lead liner requires substantial safety precautions, is expensive and time consuming, and may still incur corrosion in areas around the vapor space nozzles where operating temperatures are highest. One of the brick lining companies is testing other types of polymer membranes to replace lead. An experienced pressure vessel lining company, with a history of successful design, should be selected. Such companies are most familiar with the performance and limitations of their products, such as the thermal expansion and swell properties, and are therefore in the best position to determine vessel design loads, roundness, and deflection tolerances and to pass this information along to the vessel fabricator. Titanium-clad vessels. These vessels are fabricated from carbon steel with an explosively bonded titanium cladding (see Figure 3). Titanium and iron are not metallurgically compatible at high temperatures, and under the conditions normally used for weld overlay or hot roll bonding, they instantly react to form brittle compounds. Consequently, explosion cladding is the preferred process for the manufacture for titanium-clad steel. This is a solid-state metal-joining process that uses explosive force to create an electron sharing metallurgical bond between two metal components. The titanium cladding is used as a corrosion barrier only; its strength is not taken into account when designing the shell wall thickness. Optimum bond mechanical properties and plate sizes are produced when the yield strength of the titanium and base metal is below 345 MPa. Therefore, the optimum bond strength and toughness of titanium cladding results from a combination of titanium-clad steel to a moderate-strength, pressure vessel steel such as SA 5 16 Gr 70. Titanium grades 1, 11, and 17 exhibit similar yield strength and bond performance and have been used to clad autoclaves used for nickel processing. Special considerations must be taken in design, fabrication, welding, and testing to ensure a reliable product. Typically, a batten strap technique is used for cladding fabrication, as shown in Figure 4. The cladding is applied to a flat sheet of carbon steel thick enough for the pressure vessel design plus an additional minor allowance for compression of the steel during cladding. The plate is formed into heads and rolled into sections. Joining two sections together usually requires the removal of approximately 12 mm (0.5 inches) of titanium from each side of the weld preparation edge. The steel is welded conventionally and the vessel is then cleaned and prepared for titanium welding. In the batten strap technique, a filler (metal strip of copper or titanium) is used to fill the space where the titanium has been removed, a titanium batten strip is applied over the plate, and the edges are fillet-welded to the clad titanium. The wider the joint, the higher the stresses in the fillet-weld during operation. Large-diameter nozzles are frequently fabricated from clad plate using the same procedures. Small nozzles are typically lined with titanium sleeves and seal-welded in the shell.

1516

Flxed Support

0

Figure 3 Explosively bonded titanium cladding

0.5 ins

v

4

3 ins (max)

m-

0.5 ins

f--

;r;5&

Batten Strip Filler Strip(Copper/Titanium) Fillet Weld Titanium Cladding ( Explosion Bonded)

Figure 4 Batten strap technique More than one of the nickel autoclave plants currently in operation have had problems related to titanium material selection, principally due to chloride stress cracking. The batten strips on the vessels have cracked due to thermal expansion. Batten strip design and test procedures, such as hot cycle testing, have been developed to minimize failures during start-up and operation. Vessel heads. Some options are available for the heads of vessels in a pressure hydrometallurgical circuit. The selection is generally based on cost and delivery. For higherpressure systems, the head selection is between hemispherical and ASME 2: 1 elliptical. While a 2:l head may be preferable because the agitator is placed in the “volumetric” center of the compartment, the difference that a hemispherical head brings to bear on this criterion does not effect operations. Both types of head can be brick lined or explosion clad. Flash and preheat vessels have been supplied variously with 2:l elliptical heads, conical bottom heads, and, for lower-pressure systems, ASME Flanged and Dished heads. The conical heads were used ostensibly to aid in slurry flow, but they are more expensive, and the other head profiles have been just as successful. Nozzles. Historically, the most troublesome issue in the maintenance of brick-lined autoclaves concerns the vessel nozzles - particularly the nozzles located in the vapor space. An individual vessel can have up to 30 nozzles. To minimize penetration in the brick linings, nozzles are grouped together (usually two to four nozzles) into a “multiple nozzle,” with one lining penetration. In the initial autoclave designs, the lead membrane in the vessel shell extended through the nozzle and over the face of the flange. One of the principal problems was keeping the lead at a low enough temperature (<85”C) to prevent creep and corrosion. The annular space provided in the nozzle was insufficient to allow for the temperature drop from the operating temperature to an acceptable membrane temperature. Insulating bricks, polyethylene blocks, and insulating rope have been tried with limited success. Designs with fiberglass over the lead, or replacing the lead with a sulfur-enhanced elastomer, have also been attempted. The most successful design has used an Inconel 625 overlay through the length of the nozzle and into a “bull’s eye” in the vessel shell, with the overlay protected by brick and/or mortar and an insert. The temperature is not critical in this design, and the Inconel has withstood the environment that exists behind the brick lining. The insert is used to protect the bricks in the nozzle from mechanical damage and limits the exposure to condensate that would otherwise form and flow down the bricks and mortar, causing premature failure.

1518

The terminations of the membrane(s), the nozzle insert, and the sealing surface are all at the face of the nozzle flange. This requires the fabrication of a custom-designed Inconel or other highalloy seal ring to be fitted onto the face of the flange. The rings are designed to prevent the autoclave environment from reaching the steel shell by sealing the ends of the membrane(s) and across the nozzle insert, and include machined sections for the gasket(s). Spiral-wound gaskets and non-metallic gaskets have both been used successfully. Because spiral-wound gaskets have very high seating stress requirements, the gasket width should be carefully selected so that the bolting and flange thickness do not have to be custom designed. A typical sealing face arrangement for brick-lined and titanium-clad vessels is shown in Figure 5.

Flange /Cover Flange Gasket

Vessel Nozzle

Titanium CladdinglLining Titaniu Nozzle Lining

(4 Carbon Steel Shell

Titanium Cladding

Figure 5 Typical sealing face arrangements for brick-lined and titanium-clad vessels Nozzles on pressure vessels may be supplied as integrally reinforced components or may be fabricated from pipe and flanges with reinforcing pads. The Pressure Vessel Code requires that the area of the shell removed by the opening must be replaced with reinforcement, within dimensional limits set by the code. Some installations have extremely large agitator nozzles that allow the agitator to be withdrawn in its entirety to reduce maintenance downtime. These nozzles require significant reinforcement and usually custom design of the flange and bolting. The design also results in additional stresses within the lining system. When a large section of the lining is removed, additional loads are transferred to the remaining sections, further increasing the compressive loads on the bricks. Other design considerations are the need to reinforce the nozzle insert to handle the increased lining stresses and possible compromise of the self-supporting nature of the lining system when such a large section of the arch is removed. This design is costly, not only because of the large nozzle and the associated blind flange, but because of the need to increase crane capacity and building height and to provide a structure to house the agitator assemblies. The maintenance savings must therefore be judged in light of the capital costs and lining issues.

1519

Piping and Valve Design Piping design and material selection are critical to the safe and prolonged operation of autoclave circuits. The design must accommodate the hot acidic slurries and high operating temperatures and pressures, and the piping layout must allow for ease of operation, maintenance access, and egress in the event of an emergency. Some design considerations are:

0

The use of lower alloy or carbon steel back-up flanges for high alloy and refractory pipe to reduce cost Double check valves on oxygen, steam, and high-pressure water The use of “blowback traps” for steam and oxygen as backup to check valves Low velocities in autoclave vent systems Correct design and specifications for oxygen piping systems, i.e., materialhelocity and cleanliness Proper selection of valves, mainly for seat materials for the temperatures and fluids handled Double block and bleed for oxygen, steam, and water into the autoclave, along with proper purge and flush connections Consideration of drainage or pumping out of the autoclave Proper pressure relief design Provision of control valve bypass systems Adequate pipe stress analysis during design Adequate inspection during piping fabrication and installation Consideration of dual feed lines.

Piping materials will vary for the different circuits and materials handled. A selection of piping materials for a typical refractory gold installation are outlined in Table 2 . Table 2 Piping materials for a refractory gold pressure oxidation circuit Fluid Material Demineralized water, seal water Stainless steel TP 304L or T P 3 16L Stainless steel TP 3 16L Oxygen, air, steam, quench water (autoclave to trap) Oxygen, air, quench water (before trap) Carbon steel A-53 Steam Carbon steel A-106 Process slurry Ferralium 255 Titanium Ti Gr 2 Process slurry (high temperature) Titanium - Niobium Ti-Nb 45 Autoclave vent Duplex stainless steel 2507 or Zeron LOO Flashed slurry The high-temperature autoclave piping is usually designed to ASTM B3 1.3. Flange Ratings. In designing the piping systems, care must be taken in understanding and applying the flange designations used in various places in the world. The class designation system is used in the United States, Canada and some other countries, whereas the nominal pressure (NP) designation system is used in Europe and most other parts of the world. In both cases, the numerical designation offers a convenient round number for reference purposes; however, the PW designation is nominally the cold working pressure in bar. A flange rating for a specific material and class will give the allowable working pressure at various temperatures. Ratings are calculated using the material allowable stress and yield values taken from the Pressure Vessel Code. As indicated earlier, stress values have recently been increased for some materials. It could be expected that allowable working pressures for flanges would also increase, but this is not the case. The original allowable pressures, which were developed over 70 years ago, were generally based on asbestos and elastomer gasket materials that have generally been replaced with new non-metallic and spiral-wound gaskets. These new gaskets,

1520

particularly the spiral-wound type, require high compressive loads to become sealed, resulting in stresses that are marginal for some flange ratings and sizes used in current designs. Hence, the increases in the Pressure Vessel Code allowable stresses will have no major impact on the current allowable working pressures for flanges. This must be taken into consideration when existing pressure vessels are evaluated for re-rating. Table 3 shows the pressure ratings, in bar, for flanges supplied in various materials taken from B 16.5, 1996, or calculated using the procedures given in B 16.5 and converted to metric units. Table 3 Maximum allowable non-shock eressure, bar Forged Flanges Flan!Ze Ratin!Z

Class 150 A I 05 (Carbon Steel) 316L (Stainless) 2507 (DuJ2lex Stainless) Inconel625 Titanium Gr 2

Allowable Pressure, bar Temeerature, ac

37.8 19.7 15.9 20.0 20.0

93.3 17.9 13.4 17.9 17.9

148.9 15.9 12.1 15.9 15.9

204.4 13.8 11.0 13.8 13.8

260.0 11.7 10.0 11.7 11.7

16.2

14.1

11.7

9.7

8.6

315.6 9.7 9.7 9.7 9.7 7.2

Forged Flanges

Allowable Pressure, bar

Flan!Ze Ratin!Z

Temeerature,

ac

Class 300

37.8

93.3

148.9

204.4

260.0

315.6

A 105

51.0

46.6

45.2

43.8

41.4

37.9

316L

41.4

34.8

31.4

28.6

26.2

24.8 41.7 41.7

30.3

48.6 48.6 25.2

45.9 45.9 22.4

2507 lnconel625 Titanium Gr 2

51.7 51.7 42.4

51.7 51.7 36.6

Forged Flanges

Allowable Pressure, bar

Flan!Ze Ratin!Z

Temeerature, ac

50.3 50.3

19.3

37.8

93.3

148.9

204.4

260.0

315.6

A 105

102.1

93.1

90.7

87.6

82.8

75.5

316L

82.8

70.0

62.8

56.9

52.8

49.7

2507

103.4

100.3

97.2

91.7

83.4

Inconel625

103.4

103.4 103.4

97.2

91.7

83.4

84.5

73.4

100.3 61.0

50.3

44.8

38.3

Class 600

Titanium Gr 2

Valving Materials. A significant component of the capital cost associated with the installation of a conventional autoclave circuit is the valves, which are typically fabricated in titanium or duplex stainless steel. Any scaling problems must also be addressed in the selection of valves.

1521

The major types of valves and their materials of construction are listed in Table 4.

Table 4 Summary of typical valves in a gold pressure oxidation circuit Description l?luids Body Mat'l Trim Mat'l seats Mat'l Ball Valve Demin Water, Seal Water, ASTM-Al82- AISI 3 16SS PTFE SocketWeld GO mm Quench Water GrF3161BaWStem A/C Oxygen Feed Fe255 Fe255chy Fe255chy Ball Valve Flanged Hardcoat I-Imlcoat Ball Valve A/cSluny ASTM-13348- Ti Gr 1 3 chy Ti Gr 2/3 chy Flanged Drain Polls Gr5 Hardcoatoat Hardcoat Ball Valve A/C Oxygen Feed Fe255 Fe255dw Fe255dw Hardmat Hardcoat Flanged Ball Valve Autoclave Vent Fe255 Fe255dw Fe255dw Hardcoat Hardcoat Flanged Quench Water Ball Valve Quench Water ASTM-A35 1- AlSl3 16SS AISI 3 16SS Flanged process Sluny GICFSM Hardcoat Hardcoat Ball Valve Oxygen,PlantAir ASTM-MSl- SS316 Reinforced Flanged GICFSM BaWStem PTFE Oxygen,PlantAir ASTM-Al82SS316 Reinforced Ball Valve Butt Weld GrF316 BaWStem PTFE Ball Valve A/cSluny Fe255 Fe255chy Fe255dw Flanged Dn20-250 securacoatBall Valve Quench Water ASTM-A182- SS316BIS SS316 Butt Weld (A/C Bleed) GrF3116 Hardcoat Hardcoat Ball Valve Agitatorsealwater ASTM-A182- SS316 Reinforced Butt Weld GrF316 Ball/Stem PTFE Ball Valve Quench Water Fe255 Fe255clw Fe255dw Flanged Autoclave Hardcoat Hardcoat Ball Valve Plant Air Fe255 Fe255dw Fe255chy Autoclave Hardmat Hardmat Flanged ASTM-A351- SS316 Reinforced process Sluny Ball Valve PTFE (DrainPolls) GO8M Ball/Stem Flanged Autoclave Ti45Nb Ti-45Nbcfi-v Ti45Nbchy Ball Valve vent Hardcoat Hardcoat Flanged Check Valve Autoclave Oxygen ASTM-A351- SS616 c/w SS316 Flaneed Feed GICFSM Stellite

Class

moo CL300 CUOO

CL300

CL300 CL300 CUOO

CL800 CL150

m o o CL800

CL300 CUOO

CL300 CUOO CL3oo

Valve companies have developed non-metallic seat material for operations up to 260°C. This type of valve usually costs less than ball valves and so is becoming more popular.

Preheat Circuit Design When leaching laterites or oxidizing ore of low sulfur grade, there is an economic advantage to recovering heat from the slurry discharged from the autoclave and using it to preheat the feed. In most of the recent gold and base metal installations, this has been accomplished by recovering steam that flashes from the discharge slurry as the pressure is reduced and contacting it directly with the incoming feed slurry. In zinc pressure leach plants, the flashed steam is used to indirectly heat the recycled leach solution in a shell-and-tube exchanger while the incoming slurry is left unheated. This latter technique is also used in the alumina industry.

1522

Although indirect heating offers a number of advantages, it has been avoided, particularly where the flashed steam may be dirty and would increase tube fouling rates. Extensive pilot testing would be required to establish relevant design information such as data on heat transfer coefficients and fouling factors. Economic advantages that could be realized include: Less steam addition to the autoclave (hotter feed) Smaller autoclave (less dilution of the feed slurry from condensed steam) Lower temperature feed pumping to the autoclave (pump before, not after preheating) Less acid addition (where a discharge acid concentration is being maintained) Smaller downstream equipment (for base metal leaching). However, if heat transfer coefficients are low, which may be the case with a viscous slurry not amenable to achieving a turbulent flow, or if the heat transfer surfaces have a propensity to scaling, the capital and ongoing maintenance costs of heat exchangers may be excessive. Thus if pilot testing is not undertaken and a proper design applied, any potential economic advantage may well be lost. In recent years, considerable pilot testing has been done with shell-and-tube exchangers for this service; one nickel laterite facility currently in design is using indirect heating for a number of stages. In most recent installations requiring preheat, flash steam is contacted directly with incoming slurry in a series of columns, each with a number of segmental baffle trays that provide a curtain of slurry for steam contact (Figure 6). The baffle trays extend to at least half the diameter of the column and are slightly sloped to avoid the accumulation of solids. Five or six baffle trays are usually adequate for refractory gold installations, while 40% to 50% more are needed for moreviscous feeds such as nickel laterites. For large-diameter vessels, a disc-and-doughnut design (see Figure 7) with sloped trays may be used instead of the segmental baffle design. These designs, though possibly not the most efficient, have proven effective for slurry duty. Other designs, including bubble-cap and packed towers, would be more prone to fouling with deposited solids or scale and are more difficult to clean of such accretions.

Figure 6 Alloy preheat vessel segmental baffle design

1523

0

0-

Noule

with Sch 40 Distribution pipe

Figure 7 Alloy preheat vessel disc and doughnut design Mechanical design of the preheat vessels follows the same principles as those outlined for autoclaves, taking into account the lower temperatures and pressures. Preheater vessels can be free-standing on skirts or supported in steel structures, depending on layout and economics. The number of preheat stages is determined from an overall heat balance for the entire autoclave system that includes temperature restrictions on feed pumps and autoclave operation. Pump Design The operating temperatures of the slurry transfer pumps between preheat tower stages and the autoclave will depend on the approach temperature, i.e., difference in temperature between the incoming steam and discharging slurry. Since the steam condensing coefficient is high, approach temperatures can be close and have been measured as low as 2C". A recommended approach for design is 5C". Consideration must be given in the circuit layout to pump NSPH (net positive suction head) requirements, i.e., the suction head required to prevent cavitation and flow loss in the pump. For acceptable operation and to allow for variable conditions, the available NPSH should exceed the NPSH required by 2 m to 3 m. This head is usually provided by elevating the preheat vessels or, in the case of the autoclave, by adding a centrifugal charge pump ahead of the feed pump. The mode of operation of the preheat vessels must be taken into consideration when selecting the slurry feed pumps and control system. If there is a substantial reduction in heating steam, for instance, and the slurry feed remains constant, the vessel pressure will drop because all the steam will have condensed. Once the head that the pump is pumping against has dropped, the flow from the pump will inherently increase, further reducing the vessel pressure. Typically, a variable-speed drive in combination with mass flow measurement is used to control the slurry feed flow rate; a steep pump performance curve is required for improved control.

1524

The feed pump is normally a positive displacement pump, which incurs an additional energy loss. This is the energy required to accelerate the fluid within the piping, which is a function of the length of suction piping (mass of fluid) and type of pump (duplex single acting, duplex double acting, triplex single acting). Installing a pulsation dampener at the inlet of the pump reduces this loss, making it a function of the piping between the pump and the dampener only, which is relatively low. Discharge dampeners are fitted to reduce pressure pulsations in the discharge piping. The most common type of positive displacement pump used in this service is the piston diaphragm pump. In simple terms, the piston diaphram pump consists of a hydraulically operated diaphragm and two check valves. Unfortunately, elastomeric materials suitable for high temperatures are not available, and so a special design had to be developed for circuits with preheat vessels. The first design for high temperatures was installed at Barrick Mercur in Utah, which had a vertical water-cooled “leg” between the check valves and the diaphragm. More recent designs have incorporated horizontal cooling legs and modified cooling systems. Since check valves include elastomer sealing components that are subject to high wear, the circuit design must allow for maintenance of these items. The current maximum operating temperature for these piston diaphragm is around 200°C. Pressure Relief Pressure relief systems are a critical safety aspect in the design of pressure hydrometallurgy systems. The ASME Code specifies safe practices in design, construction, inspection, and repair of unfired pressure vessels. Pertinent requirements of the code are: The maximum allowable working pressure is the internal pressure at which the weakest element of the vessel is loaded to the ultimate permissible point, when the vessel is assumed to be: a) in corroded condition b) under the effect of a designated temperature c) under the effect of other loadings (e.g., hydrostatic pressure) which are additive to the internal pressure Size of the outlet pipe is to be such that any pressure existing or developing in the discharge line does not reduce the relieving capacity of the protective devices below the requirements for adequate prevention of overpressure.

The design pressure is that at the top of the vessel, and it usually coincides with the relief valve setting unless a rupture disc is fitted in combination with the relief valve. The pressure relief system for early pressure oxidation autoclaves consisted of combined rupture discs and relief valves, with the rupture disc preventing the valve from being continuously exposed to the corrosive autoclave environment. Cheaper materials could therefore be used for the valve. This approach required a higher design pressure for the vessel to provide sufficient margin between the operating pressure and the rupture disc bursting pressure, thereby preventing premature failure of the rupture disc due to fatigue or creep. Unfortunately, the failure of rupture discs has been a frequent problem. Various materials have been tried without success, with one plant even considering gold foil over the disc. Titanium could withstand the high-temperature, corrosive environment but is not suitable for a safety device in the presence of oxygen because of its potential for spontaneous ignition. One approach originally developed in the nuclear industry uses a water-filled U tube fitted to the relief nozzle on the autoclave, with the safety valve at the other end protected from the autoclave environment by the water leg. Although this design has been successful, care must be taken in the design and support of the discharge piping to handle a potential slug of water passing through the system. In addition, the inlet piping to the valve must be sized so that the maximum inlet pressure loss (inlet loss and line loss a maximum flow) to the relief valve is a maximum of 3% of the set pressure.

1525

This is an ASME code requirement. One publication suggests that, as a conservative guideline, the equivalent length-to-diameter ratio (LD) of the inlet piping to the relief valve size should be kept at 5 or less. A set of equipment interconnected by a piping system that does not include valves for isolating individual items can be considered a single unit, requiring a single relief valve(s), but the set pressure should be that of the lowest pressure piece of equipment. Hence, a flash vessel coupled to a preheat vessel without any isolation valves in the interconnecting piping will require only one relief valve or rupture disc. This is usually installed on the flash vessel, which is the source of pressure to the system, but it could be on the piping as long as the set pressure takes into account any losses in the piping between the vessel being protected and the location of the valve. Pressure Let-Down System Design

In all pressure oxidation or leach processes, hot acidic slurry exits the autoclave under pressure and must be let down to atmospheric conditions in a controlled manner. This may be accomplished in one or more stages, depending on whether or not there is a use for the steam evolved. Considerable energy is released in the process, and care must be taken in design to avoid severe erosion of valves and piping. An important feature of the first stage of pressure let down is the ability to control the liquid level in the last compartment of the autoclave. The flow control “equipment” (valve, choke, or combination valve and choke) will sustain a pressure drop, and when the pressure drops close to saturation, flashing will occur. There are a variety of designs for the flash vessel, all aimed at minimizing erosion from the high-velocity, two-phase stream exiting the choke. Most installations control the flow in one stage consisting of a valve located at the top of the vessel, with the high-velocity jet discharging down into the vessel through a large-diameter, ceramic-lined “blast” tube and into the slurry. Energy is dissipated through turbulence of the slurry. The slurry depth is maintained by discharging through a nozzle in the side of the vessel. One of the principal functions of the flash vessel is to remove particulate and liquid carryover from the discharging steam. This is achieved through gravity separation, which requires a quiescent zone for the separation to occur. To provide this, the blast tube is fitted around the jet exiting the valve. The annular area between the blast tube and vessel shell serves as the liquidvapor separation zone. In some installations, the slurry jet has passes through the slurry and wears out the bottom of the vessel. Typically, a ceramic impact plate is installed in the bottom of the vessel to prevent this. Blast tubes have also experienced high wear. A design developed in the geothermal brine energy recovery industry that utilizes one-stage flow control has the control valve located at the bottom of the vessel with the jet projecting vertically into the slurry, causing the slurry to circulate, hence, dissipate the jet energy. This arrangement has been used successfully in the low pressure flash vessel at one of the Western Australian laterite operations. Another design uses a valve and ceramic choke in series, with most of the pressure drop occuring in the choke. The valve allows for adjustment of the upstream pressure to the choke and can thus be used for flow control. Under normal operation the pressure drop across the valve is less than that by which the autoclave operating pressure exceeds the steam vapor pressure. Thus there is no flashing of steam between the valve and choke. If this were not the case, severe erosion of the interconnecting pipe would occur. In this two-stage flow control system, the control valve can be installed at either the top or the side of the vessel, with the choke located inside the vessel at approximately mid height (see Figures 8 and 9). The choke is sized to provide a backpressure on the valve, thereby preventing tlashing in the line between the choke and the valve inside the vessel. The valve trim must be selected so that pressure drop across it, over the normal operating range, is less than the available overpressure at the valve. In other words, this will be the total autoclave pressure minus the steam partial pressure and static head difference between the autoclave and valve location. The overpressure is provided by oxygen and inerts in the case of pressure oxidation systems and by the

1526

addition of air in the case of straight leaching systems. The valve is used to modify the inlet pressure on the choke and thus control flow, with most of the pressure drop being taken across the ceramic choke. With a relatively short and properly designed choke, there is a fan effect rather than a jet, minimizing wear on the bottom of the vessel. A third system, often called enthalpy control, adds coolant to the autoclave discharge slurry before it passes through a choke to control flow. This is a thermodynamic process, not simply one of substituting coolant flow for autoclave discharge flow; its effect can be much greater. The coolant addition modifies the flash point in the choke, resulting in an increase in mass flux (flow) in the choke. This control system may be used in conjunction with the other options to extend the flow range by “retarding” flashing in the valve. The addition of coolant will reduce the amount flashed steam produced in the flash vessels.

Figure 8 Flash vessels

Figure 9 Carbon steel flash vessel with brick lining Agitation Agitator impeller types generally used in pressure hydrometallurgy are the multi-blade, radial disk turbine for gas dispersion and the pitched-blade or hydrofoil type for blending and suspension

1527

applications. Impeller tip speeds range from 4 to 6 d s and are selected by the process designers. The materials of the impellers are the same as or similar to the materials of the internals. The impellers are subject to wear and cavitation, which can be minimized by the proper design of sparging systems for oxygen, air, or steam, and by providing proper baffling to minimize vortexing. Sparge system design for gases is critical, not only from the cavitation standpoint, but to maximize the dispersion of the gas. A minimum gas velocity is required on exiting the sparger; if the velocity is too low, a slug flow of gas is produced and dispersion by the impeller is poor. Mechanical considerations and practices in the specification of agitators consist of:

0 0 0

A drive arrangement with shaft-mounted gearboxes (the most common, as it represents a practical and economical approach) AGMA and DIN standards provide the guidelines for the gear and gearbox designs Bolted shaft couplings are the most common Balanced double mechanical cartridge seals are the standard for autoclave agitators Keyed, clamped hubs are commonly provided so that the disks and/or blades can be readily replaced.

Seals and Seal Water System The choice of materials for the double mechanical seals is critical, and advances have been made over the years that now yield a relatively long seal life. The major components of the seals are the stationary and rotating seal faces. Earlier seal components had to be of different materials because of wear and spalling if the two faces were the same. Typically carbon and chrome oxide were the materials of choice in the autoclave atmosphere. Today both silicon carbide and tungsten carbide seal face materials are used. The silicon carbide materials are brittle and must be handled with care. The tungsten carbide seals require nickel or cobalt binders that leach when exposed to the autoclave environment, however, development and testing of a resistant binder is in progress. Other materials in the seal consist of fluoropolymers for O-rings and stainless steels for the springs and other hardware. One of the advances in seal design is the addition of a loose seal inside the vessel, supplied with its own water source. This flushes away slurry or foam that could enter the seal cavity and thus prolongs seal life. The most critical aspect of seal reliability and life is cooling and lubrication of the seal via the seal water system. The philosophy behind a good seal water system is that it must not only provide the cooling and lubrication, but also contain enough fluid that the autoclave can be safely shut down or continue to operate with one or two damaged seals, and that pressurized water is available in the event of a power loss. Providing only enough fluid to maintain cooling and leakage is an inadequate system for safe autoclave operation. There are two approaches to seal system design. One is to provide individual reservoirs with their own circulation systems fed by a common supply. This is a costly approach, however. A simpler system is to provide all the seal water from one control pumping system, with adequate flow and safety backup. To follow the philosophy of seal operation as outlined above, the seal water system typically consists of a surge/make-up tank, seal water feed pumps (with redundancy and sometimes a separate drive method), filters, an accumulator, and a seal water cooler on the return side of the system. The accumulator is connected to a high-pressure gas source (nitrogen or air), which, upon power failure, provides the requisite pressure on the seals. All materials are stainless steel, and the water supplied must be reverse osmosis quality, or preferably demineralized water, at a maximum temperature of <60°C at the seal discharge. Some installations add glycol (around 10%) to aid in seal lubrication, which considerably increases the cost of the operation.

1528

CONCLUSION Various pressure hydrometallurgical processes are now widely used on a range of ores and concentrates, and equipment has been developed and commercially proven to handle these conditions. While some materials issues are outstanding and other improvements could be made, the designs of these systems, and their safety and reliability, have matured to the point that they can be installed with confidence. As more complex ores continue to be exploited and environmental constraints make other treatment routes less attractive, medium- and high-temperature processing is bound to become more prevalent. REFERENCES ASME Boiler & Pressure Vessel Code, Section VIII Div 1, ASME/ANSI B 16.5 Pipe Flanges and Flanged Fittings. Seminar by Goulds Pumps Canada Inc. Current Trends in Pump Technology. C.C. Smith, D.G. Dixon, M.R. Luque, J.C. Robison, S.R. Chipman. Analysis and Design of Flashtubes for Pressure Letdown in Autoclave Mining Operations. J.G. Banker, J.P. Winski, TitaniudSteel Explosion Bonded Clad for Autoclaves and Vessels.

1529

Barrick Gold - Autoclaving & Roasting of Refractory Ores K.G. Thomas', A. Cole'and R.A. Williams3

INTRODUCTION Over the last two decades several facilities, utilizing autoclaves and oxygenated roasters have been installed for processing refractory gold ores. Two such facilities have been installed within Barrick Goldstrike and this paper reviews the processes, inclusive of associated capital and operating costs.

In this paper the two Barrick Goldstrike refractory processes are described being as follows: Acid autoclaving for sulphide ore Oxygenated roasting for double refractory ore; sulphide and carbonaceous In summary there is a place for both autoclaving and roasting of gold ores, the latter requiring appropriate but significant investment in environmental abatement equipment. BARRICK GOLDSTRIKE MINE' The Goldstrike property is located in the Tuscarora Mountains, Elk0 and Eureka counties in northcentral Nevada, USA, on the Carlin geological trend. The deeper deposits at Goldstrike occur in silty limestones, which have been silicified and argillized. Gold in the sulphide ore occurs mainly as inclusions in fine-grained pyrite and marcasite. Oxidized ore lies in the upper parts of the deposit. Over 75% of the reserves are in the Betze-Post open pit deposit and the remainder in the high grade Meikle mine and Rodeo/Griffin underground deposits. Presently gold reserves are in refractory sulphide or refractory carbonaceous/sulphide ores. This makes autoclaving particularly important to Barrick Goldstrike. As the reserves continued to grow, so did the quantity of carbonaceous/sulphide ore. Half of the ounces that are contained in the current reserves are carbonaceous/sulphide. To treat this growing quantity of double refractory material an oxygenated roaster, at 12,000 stpd capacity, was commissioned in the second quarter 2000.

Besides the 24 million ounces in reserves, over 20 million ounces of gold have been produced. from the Goldstrike property since acquisition in 1987. In 2000 total cash costs were U S 1 7 0 per ounce and total production costs US$220.

'

Managing Director, Mining & Mineral Processing, Hatch (formerly Senior Vice President, Technical Services Barrick Gold Corporation) Roaster Superintendent, Barrick Goldstrike Mines Inc.,Nevada, USA R. A. Williams, Manager Process, Pascua Project (formerly Autoclave Superintendent Goldstrike) Barrick Gold Corporation, Toronto, Ontario.

1530

BARRICK GOLDSTRIKE ACIDIC AUTOCLAVE Figure 1: Detailed representation of the Autoclave Circuit - BARRICK GOLDSTRIKE DETAILED ALKALINE AUTOCLAVE CIRCUIT (Simplified Flowsheet - Pressure Oxidation)

SULPHIDE PRE-TREATMENT

AUTOCLAVING

. ........................... .

NEUTRALIZATION CIRCUIT

Heat Remvery Steam

:

t .....

.....

I

............... ..................... overnow to mill reclaim water lank

+

.....

........................

*....... i

:

1

.........

....

HzSO~ Acidulation/

Surge Tanks

..................................

.....*... .....:

Heat Exchangers (101

Autoclave U 2

-U6

Slurry from the conventional wet SAGhall grinding circuit, at approximately 35% solids and 80% passing 135 microns is pumped to three preoxidation thickeners. Thickener underflow, at approximately 54% solids, is pumped to a train of four acidulation tanks. Sulfuric acid is added to the slurry to destroy sufficient carbonate (C03) prior to entering the autoclave circuit. Process air is also injected into the acidulation tanks to aid in carbon dioxide removal. Carbonate levels are typically reduced to less than 2% in the acidulation tanks. As a rule of thumb, 1% sulphide sulphur (S=)in the autoclave feed destroys 0.9% C03 and typically the autoclave feed ranges from 2.0% to 2.5% S=. The discharge from the high pressure splash vessel is pumped by two positive displacement pumps into the autoclave. The pumps are operated in parallel with individual suction and discharge lines. Each pump can deliver approximately 60% of the required feed rate to the autoclave, the parallel configuration enhances on line availability to 92%. Scheduled maintenance, per autoclave, is a shutdown taken every 12 months, hence the high availability. All six autoclaves have an outside diameter of 15 feet, autoclave No. 1 is 75 feet in total length, and autoclaves No. 2 through No. 6 are 82 feet in total length. Each autoclave is divided into five compartments by brick walls, with each compartment containing an agitator and injection pipes for oxygen, steam and water. Autoclave retention time ranges from 40 to 60 minutes and the vessel operates at approximately 420 psig and 420 to 430°F. The lining of the autoclave vessels is 5/16 inch lead on the carbon steel shell, followed by 1/8 inch fibrefrax paper and 9 inch thick acid brick, 2 to 3 layers, exposed to the slurry. Another key factor for the high on line availability is the correct selection of materials of construction for various parts of the circuit. This is detailed in Table 12. The selection of materials of construction is related to temperature, pressure and process chemistry, especially pH (freeacid). This can be seen by comparing the Barrick Mercur alkaline autoclave materials versus the Goldstrike acid autoclave materials.

1531

Table 1: Materials of Construction for Autoclave Facilities (Wetted Surfaces)*

A typical cut away view of the splash, autoclave and flash circuit is shown in Figure 2. Figure 2: Cut Away of Autoclave Circuit BARRICK GOLDSTRIKE PRESSURE OXIDATIONVESSEL (AUTOCLAVE) ( p/an and efevafion ) 821881

L

! !

_I _ !' ! '

Sulphide sulphur oxidation in the autoclave is typically in the order of 90 to 92%. Residual S= exiting the autoclave is targeted at less than 0.2%. Values greater than 0.25% S= are reflected in lower gold recoveries in the carbon-in-leach (CIL) circuit. Free acid levels in the autoclave discharge are a function of S= and C 0 3 in the autoclave feed but normally range from 10 to 25 g/l.

1532

After the slurry passes through the heat exchange slurry coolers it is pumped to two parallel trains of neutralization tanks where the pH is elevated from approximately 1.5 to about 10.5. Neutralized slurry from the pressure oxidation circuit is pumped to two parallel streams of CIL tanks for gold extraction and subsequent recovery of gold in a conventional Zadra circuit. Suitable ancillary facilities are installed to supply flocculant, sulphuric acid, oxygen, lime, steam and compressed air to the autoclave facility. The control system is a Bailey distributed control system (DCS). The objective of the pressure oxidatiodautoclaving process is the destruction of the sulphides, pyrite, marcasite and/or arsenopyrite, thereby liberating the occluded gold. The gold is then amenable to recovery by the cyanidation process. For an acid autoclave operating at temperatures greater than 350°F and a pH below 2, as at Barrick Goldstrike, the reactions3can be presented by: 2FeS2+702+2H20= 2FeS04+2H2S04

(1)

2FeS04+HzS04+1/202 = Fez(SO4)3+H20

(2)

FeS(S04),+3H20= Fe2O3J+3HZS04

(3)

The highly oxidizing conditions within the acid autoclave is important as ferrous sulphate is converted to ferric sulphate (reaction 2). This reaction is beneficial as ferrous sulphate consumes cyanide, in the cyanidation step of the CIL circuit, which adversely increases operating costs. In acid autoclaving silver is precipitated as a jarosite but can be recovered by a lime boil pretreatment process but is generally not practiced as it is not economical. The key to successhl autoclaving to minimize maintenance and operating costs is to layout the plant for ease of equipment access, by crane, as shown in Figure 34.

1533

Figure 3: Autoclave Plant Plan & Section at High Pressure Flash Tank

I

,Diaphram

Feed

t d

T \

\

Spare

5

Agitatoi

\

\ \ \ \

1

1 Door

J@ -

Pumps

t

L

Laydown

Splash Crane Heating

-

Agitator Removal

i3-m ;pare Agitator

II Diaphram Pumps

-GEHO

1534

Flash Cwling Towers

BARRICK GOLDSTRIKE OXYGENATED ROASTER5 A 12,000 stpd oxygenated roaster facility was commissioned at Barrick Goldstrike in the second quarter of 2000. The roaster treats double refractory ore, carbonaceous and sulphidic, at a temperature of 1000°F using 99.5% pure oxygen in a double "two-stage vertical" roaster. A flowsheet of the process is shown in Figure 4. Figure 4: Barrick Goldstrike 12,000 TPD Oxygenated Roaster ROASTING CIRCUIT

CRUSHING AND GRINDINGCIRCUIT

' i Neutnllzatior

LEACHING &CARBON CIRCUIT

Rotator

As the drilling of the Goldstrike property progressed from 1988 to the late 1990's significant quantities of double refractory ore were added to the reserves, over the original ores classified as a single refractory, that is sulphidic, which are treated with ease in autoclaves. Of the 24 million ounces in reserves at the end of 2000, approximately 9 million ounces were double refractory, requiring treatment by another process. Oxygenated roasters were selected. The roaster circuit at Goldstrike incorporates unique equipment, that is, dry grinding using Krupp Polysius cement technology double rotators as shown in Figure 5, and Freeport McMoRan two stage vertical oxygenated roasters as shown in Figure 6. Crushed ore is fed into the drying chamber of each of two parallel double rotators, followed by coarse and fine grinding, in two separate chambers within the same mill. Ore is reduced from 40 mm to 74 micron, with the aid of static and dynamic classifiers, and the product captured in a series of baghouses before advancement to the roasting facility.

1535

Figure 5: Double Rotator Grinding Mill by Krupp Polysius

Grits

,Return Static Classifier Inlet Duct

Coarse Drying Chamber

Grinding

Coarse Feed & Hot Gas

Diaphragms Bearing Figure 6: Two-stage Fluid Bed Roaster Freeport McMoRan Type WDUNG

SECOW CYCLONE

Two parallel roaster circuits are utilized to oxidize the carbonaceous matter and sulphide to release the gold for subsequent gold cyanidation. Low pressure, high purity oxygen oxidizes the ore, with the majority of the oxidation taking place in the upper bed. The lower bed completes the oxidation of the carbonaceous matter to 80%; sulphide sulphur oxidation is 99%, mainly in the first stage. The process is autogenous but can be controlled by water quenching or coal addition, depending of the calorific value of the ore. The exhaust gas is extensively cleaned, as shown in Figure 4, in a series of process equipment:

1536

0 0

0

Quenching and de-dusting Wet gas condenser Mist eliminator Wet electrostatic precipitator

0 0

0 0

Mercury removal SO2scrubbing CO incinerator N0,removal

Environmental abatement equipment at the roaster was US$40 million out of a total capital cost of US$330 million. As in the autoclave circuit suitable ancillary equipment are installed to supply reagents. Also as in the autoclave facility Bailey distributed control system is installed as a control system. The chemistry of the double reflactory ore oxidation can be represented as follows:

An interesting feature of the roaster chemistry is the fixation of the SO2 by the carbonates contained within the ore, forming gypsum. Between 40% to 70% of the SO2 is fixed by the naturally occurring calcium carbonate, which reduces subsequent processing costs. Fixation is enhanced with the addition of lime to the grinding circuit. A key piece of equipment in the double rotator grinding circuit is the dynamic classifier. A very accurate classifier using a variable speed drive, it cuts the product, with consistency, to 80% passing 74 micron. A cut away view of the dynamic classifier is shown in Figure 7.

Figure 7: Dynamic Classifier:

Following oxidation the ore is slurried, neutralized and advanced to CIL for gold extraction. Gold bearing carbon is transported to the Zadra circuit in the existing autoclave facility for gold recovery.

1537

OPERATING COST COMPARISON The two processes display a marked difference in operating costs as detailed in Table 2 below:

Table 2: Operating cost comparison US$/short ton Cost Item

Crush Grind Roast Autoclave CIL Strip & Electrowin TOTAL

Acid Autoclave (adjusted)* 1.05 4.95 9.40 **

1.90 0.25 17.55

Roaster

1.05 5.25 5.35 ** 1.85 0.30 13.80

* Adjusted refers to grinding to 74 micron; normally 90 microns ** Assumes oxygen plants purchased

The difference between the acid autoclave and oxygenated roaster costs is explained as follows: 0

The grinding costs of wet and dry grinding plants are very similar, although the processes are markedly different. This is due to the advancements pioneered by Krupp Polysius in adopting double rotator cement technology to the mineral industry.6

0

The two stage oxygenated roasters, 99.5% purity oxygen, pioneered by Freeport McMoRan, New Orleans, U.S.A., has allowed an inexpensive alternate to acidic autoclaving to be developed. In the roaster facility acid is not required to remove C 0 3 from high carbonate ores as in the autoclave circuit. Thus saves up to $3 per ton ore, on acid costs for the Barrick Goldstrike relatively high carbonate ore (5% C03).

CAPITAL COST COMPARISONS For a 12,000 stpd oxygenated roasting or acid autoclaving facility, inclusive of comminution and gold recovery, capital costs are similar at US$330 million. As mentioned earlier the oxygenated roasting facility has significant environmental abatement equipment amounting to US$40 million, which is not required to the same extent in an autoclave circuit. PROCESS SELECTION For a double refractory ore, carbonaceous and sulphidic, the process of choice is oxygenated roasting. This may change in the not too distant future, if a thiosulphate process can be developed for treating carbonaceous ore. A thiosulphate lixiviant does not suffer from pre-robbing as with cyanide and accordingly could be part of a gold recovery circuit following autoclaving.

Whether an oxygenated roaster or acid autoclaving process is selected for a sulphidic ore depends on the mineralogy and the gold price. Assume a 50 million ton ore body, 0.25 ounces per ton, US$3 15 per ounce gold and a US$2 per ton operating cost lower in favour of oxygenated roasting. This example assumes a low carbonate level in the ore body, that is below 2% C 0 3 , that is, reducing acidulation costs by US$ 2 to 4 per ton depending on carbonate levels in the ore. Over the life of mine this saves US$lOO million operating costs. But oxygenated roasting, generally has on average 2.5% less recovery. This gives a break-even project between oxygenated roasting and acid autoclaving owing to $100 million less revenue. This simple example illustrates the influence that ore grade, gold price, mineralogy and differential operating cost and recovery has in process selection. If the carbonates are above 2% C03 then the economics are in favour of the roaster owing to the acid costs required for acidification prior to autoclaving.

1538

Since the 1980’s acid autoclaving has generally been the preferred process in treating sulphidic refractory gold ores and concentrates. This is especially the case with high arsenic ores, as the arsenic is fired to environmentally stable ferric arsenate during autoclave chemical reactions. Oxygenated roasting, developed in the late 1980’s has in the last 5 years been developed to a satisfactory level, especially with the utilization of dry grinding double rotators from Krupp Polysius (refer to Figure 5). Accordingly there is now a choice between autoclaving and oxygenated roasting for treating sulphidic ores. CONCLUSION Oxygenated roasting is the process of choice for treating double refractory ore. There is now a choice between acid autoclaving and oxygenated roasting for treating sulphidic ores, using the whole ore process, and process selection depends on several factors, such as gold price, mineralogy and recovery. ACKNOWLEDGEMENTS

The authors would like to thank Barrick Gold Corporation for permission to publish this paper and associates for their assistance over the years.

H.J.H. Pieterse and K.G. Thomas; “Pre-treatment Methods, Autoclaving;” Prepared for the Second International Gold Symposium; Lima, Peru; May, 1996 K.G. Thomas & R.A. Williams; “Alkaline & Autoclaves at Barrick Gold’’, EPD Congress 200, TMS, Nashville, Tennessee, U.S.A., March, 2000. K.G. Thomas; Research, Engineering Design and Operation of a Pressure Hydrometallurgy Facility for Gold Extraction; Technical University of Delft, The Netherlands; May, 1994. K.G. Thomas, “Barrick Gold Autoclaving Process”, World Gold ’91, Cairns, Australia, April 1991. A. Cole, J. McMullen, K.G. Thomas & S. Dunn; “Barrick Goldstrike Roaster Facility Roasting & Gas Handling; Canadian Mineral Processors, 2001, Ottawa, Ontario, Canada; January, 200 1. 6 N.Patzelt; Private Communications; Krupp Polysius; Germany; 1998 to 2000.

1539

Selection and Sizing of Biooxidation Equipment and Circuits Corale L. Brierley' and Andrew P. Briggd

ABSTRACT The first commercial, stirred-tank bioleach plant was commissioned in 1986 to pretreat a sulfidicrefractory gold concentrate, thus enhancing gold recovery. There are now 11 full-scale, stirredtank bioleaching plants employing three different technologies to process sulfidic-refiactory precious metal concentrates and concentrates of pyritic cobalt and chalcopyrite. Bioleaching is also used at commercial scale to heap leach secondary copper sulfide ores and to pretreat sulfidicrefiactory gold ores in heaps. Heap leaching of chalcopyrite ore has been pilot tested and commercial development is anticipated in the near future. Bioleaching of concentrates in heaps is an emerging technology. This chapter describes the design and sizing of aerated stirred-tank and heap operations and discusses the selection of technology and equipment for these plants. Bioleaching, also called biooxidation, employs microorganisms as catalysts to generate ferric iron, an oxidizing agent that degrades sulfide minerals. The microorganisms also oxidize chemically reduced sulfur compounds producing acid. Microorganisms, commercially exploited in bioleaching, are divided into three groups based on the temperature range at which they grow and function. Bioleaching can be accomplished at temperatures ranging fiom about 20" to 95°C; organisms that function at the higher temperature ranges continueto catalyze reactions despite the large amounts of heat generated fiom the rapid oxidation of s lfide minerals in tank and heap reactors. These thermophilic microorganisms are also ne ssary for effective oxidation of chalcopyrite. The fundamental principles of biooxidation are cluded in this chapter as well as a description of the important microorganisms and factors that affect their performance.

,-" l

INTRODUCTION Mineral biooxidation, also called bioleaching, uses certain microorganisms to oxidize sulfide minerals present in ores or concentrates. In this process base metals are released into a dilute sulfuric acid solution for recovery by conventional methods. Precious metals occluded in sulfide minerals are exposed for enhanced extraction by cyanide or other lixiviants. Biooxidation has undoubtedly been employed unwittingly for centuries for the extraction of copper. However, the underpinnings of the unit process commerciallyemployed today had its genesis some 50 years ago when copper was scavenged fiom run-of-mine, submarginal grade materials in large dump leach operations. In the last 20 years process design and equipment availability for biooxidation have achieved a higher degree of sophistication, and bioleaching is now applied in heap leaching of crushed, secondary copper ores and sulfidic-refiactory gold ores and in tank leaching of base and precious metal concentrates. Brierley Consultancy LLC, Highlands Ranch, Colorado Fluor Mining and Minerals, Denver, Colorado

1540

The aim of this chapter is to describe the process design and equipment selection and sizing for biooxidation circuits. Specifically considered are:

0

The principles of biooxidation and incentives for commercial use of the technology, Process design of crushed, sulfide ore biooxidation circuits with emphasis on sulfidicrefiactory gold ores, The design and equipment selection and sizing of aerated, stirred-tank biooxidation circuits for concentrates and high value sulfide materials, Current commercial biooxidation operations and their performance and problems, and New developments in biooxidation processing.

PRINCIPLES OF MICROBIAL OXIDATION OF MINERALS Bioleaching is only applicable to sulfide minerals be they base metal sulfides or sulfide minerals such as pyrite and arsenopyrite that host precious metals. This is because the microorganisms, employed in bioleaching technology, derive energy for their existence and reproduction from the oxidation of ferrous iron and chemically reduced sulfiu compounds much like humans obtain energy fiom the foods we eat. Chemical reactions, catalyzed by the microorganisms associated with the bioleach process, are responsible for releasing base metals into a dilute sulfuric acid solution and making precious metals available for dissolution by cyanide and other chemical lixiviants. Bioleaching is not a process involving a single type of microorganism, but rather involves groups of microorganisms that catalyze different reactions and thrive under different pH, chemical and temperature regimes (Norris et al. 2000). Bioleaching is not a stand-alone technology; rather it is the amalgamation of three separate disciplines - biotechnology, chemistry and metallurgy - whose requirements must converge to achieve a seamless process. This section briefly clarifies the fundamentalsof biotechnology, chemistry and metallurgythat must interrelate in the design and performance of the mineral biooxidation process. Microorganisms Involved in Bioleaching In bioleaching technology microorganisms are divided into three groups based on the temperature ranges at which they grow: 1. Mesophilicbacteria, which actively function in the 15" to 45°C temperature range, 2. Moderately thermophilic ("heat loving") bacteria that oxidize sulfur and iron compounds in the range of 40" to about 6SoC, and 3. Extremely thermophilic microorganisms, which are not bacteria but rather are Archueu, a unique evolutionarygrouping of organismsthat evolved fiom ancient life forms on Earth, and flourish at temperatures fiom about 60°C to nearly 95°C. Although these three groupings of organisms proliferate at different temperature ranges, they all have some features in common. They all derive energy fiom the oxidation of ferrous iron, certain chemicallyreduced sulfur compounds or both:

4 Fe" + 0,+ 4 H'+ 4 Fe3'+ 2 H 2 0

(Reaction 1)

So+ 2 HzO + 02 +4 H'+ SO,'-

(Reaction 2)

All organisms involved in bioleaching need oxygen (0,)and carbon dioxide (CO& both supplied in the gaseous forms; 0,is the electron acceptor for the redox reactions the organisms cany out for their metabolic energy requirements, and CO, is the carbon source fkom which they construct their cellular components, such as proteins, DNA, carbohydrates, etc. They all require ammonium (NH43 and phosphate (PO:-) ions and certain trace elements as building blocks for , ' KI$ etc), are often amino acids, DNA and other constituents. The trace elements required @

1541

abundant in the ore or concentratefeedstock. The bioleaching microorganismsare acidophilic, that is “acid-loving”, and require a pH range of less than 2.5 and preferably greater than pH 1.0. This pH range assures that Fez+is readily available in solution for oxidation by the microorganisms. Mesophilic bacteria. The most common mesophilic bacteria present in sulfide leaching operations are Acidithiobacillusferrooxidans, Acidithiobacillus thiooxidans, and Leptospirillum firrooxidad. niobacillus caldus, technicallya moderatelythermophilic bacterium because of its optimum growth at 45”C, will also grow at mesophilic temperatures. These microorganisms are the workhorses in heap leaching of secondary copper sulfide minerals and during the early phase of sulfidic-refiactorygold ore heap leach operations before heap temperatures exceed the tolerance level of these microorganisms. Many tank leach plants for precious and base metals rely on mesophilic bacteria. Because the mesophilic bacteria are ubiquitous in the environment, they develop naturally within the heap leach environment when the ore is moist and properly acidified, Oz is available and the temperature is between about 15” and 40°C. When heap conditions are optimum these small (approximately 0.5 pm in diameter by 1.0 pm in length), rod-shaped organisms proliferate until they number lo6 to 107 per gram of ore. when concentrate is bioleached in aerated, stirred-tank reactors at mesophilic temperatures, the reactors are inoculated with strains of mesophilic bacteria that have been pre-adapted to the feed and the high metal concentrationsanticipated during leaching. In these reactors the bacterial population numbers lo9 to 10” per ml of solution. Rawlings (1997) has written a more comprehensive treatise on these organisms. Moderately thermophilic bacteria. The moderately thermophilic bacteria are easy to find and culture fiom volcanic areas, acidic thermal pools, warm, acidic mine waters, and sulfidic stock and waste piles where temperatures and conditions support their proliferation. The moderately thermophilic bacteria have not been as well studied as the mesophilic leaching bacteria and therefore the taxonomy (identification and naming of the organisms) is not as well developed. Moderate thermophiles, common to bioleaching operations, include SulJbbacillus thennosulfidooxidans, Surfobacillus acidophilus, Aciabphilus ferrooxidans, and Thiobacillus caldus (Noms 1997). These rod-shaped bacteria, somewhat larger in size (1 pm in diameter by 2-3 pn long) than the mesophiles, appear naturally (without deliberate introduction) in sulfidicrefiactory gold heaps and pyrite-rich copper dump leach operations, as temperatures rise to 4OoC fiom the exothermic oxidation of sulfide minerals. When the maximum temperature limit (45°C) of the mesophilic bacteria is exceeded they die, because their protein structures destabilize, and the moderate thermophiles then dominate. The moderate thermophiles perform the same oxidation reactions with the same degree of efficiencyas the mesophiles. The moderate thermophiles have also been harnessed for industrial application in aerated, stirred-tank reactors. Chalcopyrite bioleaching, a developing commercial process, is accomplished using moderate or extreme thermophiles, because improved copper leach rates and recoveries are observed with these two groups of organisms. Extremely thermophilic microorganisms. The extreme thermophiles are not bacteria, but Archaea, a distinct branch of life evolved tiom ancient life forms on Earth. These organisms, whose existence was discovered in the late 1960s, are very unlike the mesophilic and moderately thermophilicbioleaching organisms in that they are 1 pm spheres lacking a rigid wall surrounding the cells. These “extremophiles”, as they are referred to, proliferate in acid environments rich in chemically reduced s u l k and iron compounds only when temperatures approach 60” to 65°C. At these temperatures the moderately thermophilic bacteria reach their maximum heat tolerance and die. The Archaea, at home in volcanic regions and hot acid springs, not only tolerate but reproduce at temperatures approaching 95°C (Norris 1997). Although these organisms have been cultivated fiom hot, sulfidic coal waste piles, they have not been found to naturally occur in hot copper leach dumps or hot sulfidic-refi-actoryprecious metal heap leach plants. This does not mean, however, that they are not there; it may simply be that nobody has extensively looked for them using the Certain species of the genus Thiobacillus were recently reclassified as Acidithiobacillus

1542

right tools. The best studied of the extremethermophiles are Sulfolobus acidocaldarius,Sulfolobus metallicus, and Acidianus brierleyi. The extremely thermophilic Archaea are intentionally added to heaps used for pretreatment of sulfidic-refiactory gold ores and to aerated, stirred-tank reactors for leaching of chalcopyrite concentrates. The advantages of using Archaea in these applications will be discussed in more detail later. Other microorganisms. In commercial-scale heap and dump leach operations there are a wide diversity of microorganisms present. Many of these are microorganismsthat use organicmatter for their metabolism and reproduction. These '"heterotrophic organisms" include bacteria, fungi and certain members of the high temperature Archaea These acid-loving microorganisms scavenge and oxidize organic matter present in the leach system (Johnson and Roberto 1997). The organic compounds come fiom the degradation of organic matter in the ore, residual organics f?om the solvent extraction process and organic compounds excreted by the leaching organisms, and dead biomass. The exact role of the heterotrophic organisms in leaching is not known, but their presence does influence both the design and performance of the plant. Because these organisms require Oz, their consumption must be considered when designing the aeration system. Heterotrophic organisms, particularly fungi, can be a nuisance as they proliferate to the extent of plugging pipes, disrupting solution flow by growing on the picket fences in the solvent extraction circuit, and forming a crud layer between the organic solvent and aqueous layer in the SX plant. Chemistry of Bioleaching The exact mechanism employed by the mesophilic, moderately thermophilic and extremely thermophilic microorganisms to leach sulfide minerals has been studied and debated for decades. It is well known that many microorganisms present in a bioleach system are firmly attached to mineral surfaces while other microorganisms are suspended in the aqueous phase. Sand and his colleagues (1995) argue that bioleaching of metal sulfides is achieved by way of microbial Fe3' generation and H2SO4. These researchers (Sand et al. 1995; Schippers and Sand 1999) propose that disulfides (e.g. FeS2 MoSz, WSz) are oxidized by Fe3' with thiosulfate (SzO?-) formed as an intermediate product. In contrast, metal sulfides (e.g. ZnS and CuFeS2) are degraded with a combination of Fe3' and H" with the main intermediates being polysulfides (Ss) and So.The degradation mechanisms between disulfides and metal sulfides differ based on differences in electronic structures and solubilities. Sand and his colleagues (1999) further propose that microorganisms, adhering to the surhces of minerals, are encased by a polymeric substance (biofilm). The organisms attach to minerals via electrostatic interaction between the biofilm and the mineral surface. The polymeric biofilm complexes Fe3' iron, which Sand et al. (1999) assert can concentrate to 53 g/L within the biofilm. They propose that the interface between the microbe's polymeric layer and the mineral surface is the reaction zone where metal dissolution fiom the sulfide mineral takes place. This bioleaching mechanism, which was developed using A. ferrooxidans as the test organism, is illustrated in Reaction 3 with oxidation of the disulfide, FeS2 by ferric hexahydrate molecules (Schippers et al. 1996): FeS2+ 6 Fe(H20)63++ 3 H20 + Fez' + s203'-+ 6 Fe(HzO),2' + 6 H"

(Reaction 3)

Fe2' is rapidly re-oxidized to Fe3' (Reaction 1) by iron-oxidizing microorganisms, such as A. ferrooxidans, L. ferrooxidans, and a variety of moderate and extreme thermophiles that are attached to the mineral surhces as well as suspended in the aqueous phase. At the low pH of leach systems the unstable S20: is converted to polythionates (e.g. S402-) and elemental s u l k (SO); these are oxidized to sulfate (SO:-) by attached and unattached, sulk-oxidizing microorganisms, such as A. thiooxidans, T. caldus and other thermophiles in aerated heap and stirred-tank operations. Regardless of the exact mechanism of the bioleaching microorganisms, the crux of bioleaching is redox potential. The microbial oxidation of Fez' increases the Fe3' to Fez' ratio, increasing the redox potential. Sufficient soluble iron must be present that, when biologically

1543

oxidized, the redox potential is high enough to oxidize the target minerals. Oxidation of the target sulfide minerals consumes Fe3", increasing the Fez" concentration and decreasing the redox potential. To avoid a decrease in the redox potential, plant conditions must be .favorable for the immediate re-oxidation of Fe2' by the microorganisms. In stirred-tank reactors, the key is maintaining a very high redox potential at all times so all sulfide minerals oxidize rapidly. The only way to maintain the high redox potential is to ensure that everything the organisms require (02,C02, acidic conditions, nutrients, optimum temperatureconditions, etc) is optimized. In heaps the same is true, but achieving optimum conditions in a heap for the organisms is challenging. Factors Mecting Microorganisms Temperature. Temperature, as discussed above, clearly impacts bioleaching by selecting for the group of microorganisms that will predominate at a specific temperaturerange. Chemical reactions generally proceed more rapidly at higher temperatures; theoretically there is a doubling of the reaction rate for a 10°C rise in temperature, so biooxidation circuits tend to be operated at the higher temperature limit for each microbial group. Higher operating temperatures also provide some benefits in terms of decreased cooling requirements in stirred tank reactors. Chalcopyrite leaching in stirred tanks must be performed at higher temperatures to effectively leach the mineral. There is increasing evidence that operating sulfidic-refkactory precious metal heap leach operations at higher temperatures improves gold recoveries without increasing lime or cyanide consumption. As discussed later, high temperature heap leaching is a developing technology for extraction of copper fkom chalcopyrite ores. pH. All microorganismsimportant in bioleaching are acidophilic and perform optimally when the pH is between 1.2 and 2.3. Above pH 2.5 soluble iron hydrolyzes and precipitates fkom solution. What this means in an operating plant is that the key microbial energy source (Fez'+)and the product of the microbial oxidation and sulfide mineral oxidant (Fe3+)become limited. The higher pH is also not .favorablefor the solubilization of the product metal cations. The consequence of 8 pH higher than optimum is a decline in PLS tenor for base metals and lower than anticipated extraction of precious metals. As pH in an operating plant declines below about 1.2 and acidity increases, escalating selective pressures are placed on the resident microbial population (Norris et al. 2000). The higher the hydrogen ion (H?) concentration in the surrounding solution, the greater the difficulty some leaching microorganisms have in rejecting the transport of H" across the cell membrane. When H' is transported across the cell membrane, the pH of the cell's cytoplasm plummets and the organism dies. Should pH of an operating leach plant decline precipitously, the harsh conditions cause natural selection for those types of organisms able to tolerate the adverse situation. The more adverse the condition the fewer types of organisms survive, limiting metabolic diversity; this in turn decreases metal production. An example is inadvertent selection of the acid tolerant, Acidithiobacillus thiooxihns by operating a plant at low pH. A. thiooxihns can only oxidize chemically reduced sulfur compounds; which means that, in the absence of any iron-oxidizing microorganisms, ferrous iron concentrations increase, redox potential declines and mineral sulfide: oxidation slows and then ceases. Declining pH will also select for Leptospirillum ferrooxidans over Acidithiobacillus ferrooxidans, because L. ferrooxidans is more acid tolerant (Norris 1983). However, because both of these organisms oxidize Fe2', this selection has less impact on the performance of a commercial bioleach plant. Microorganisms intensely dislike abrupt changes and this includes abrupt changes in pH and acidity. Microorganisms are remarkably adaptable, and slow changes in acidity and other operating parameters allow time for microbial populations to adapt to a range of adverse conditions without the loss of metabolically important members of the resident population. Redox potential. The ratio of farous iron to ferric iron also selects for microorganisms in an operating plant (Rawlings et al. 1999). When the redox potential is low and more Fe2' is in solution, Acidithiobacillus ferrooxidans will predominate, because this organism has a faster

1544

growth rate and will build up a larger number of cells in the system. However, as the redox potential increases due to a higher Fe3”:Fe2’ ratio, Leptospirillumfirrooxidans will predominate, because these organisms have a higher a m i t y for Fez’ than does A. firrooxidans.A. ferrooxiduns is also more sensitive to inhibition from high concentrations of Fe3’ in solution. Therefore, in a stirred-tank reactor, in which the redox potential remains relatively constant and is high, L. ferrooxidans is the primary iron oxidizer. Oxygen. Bioleaching microorganisms require 9.4accepts the electrons in the redox reactions catalyzed by the microorganisms. The surest way to cause a production problem in a bioleaching operation is to limit 9.Getting air into the circuits and distributing it efficiently are significant engineering challenges in the design of bioleaching plants. Oxygen utilization efficiencies (oxygen used in reactions as a percentage of that provided) are in the order of 30 to 40% for stirred tanks and 20 to 30% for heaps. Higher figures have been reported, but these could be an indication of insufficient addition. In stirred tanks, the dissolved oxygen level is about 2 ppm in the solution at the top of the tank; this number is higher at the tank bottom due to hydrostatic pressure. Solutions leaving bioheaps should also contain about 2 ppm oxygen. Dissolved oxygen concentrations lower than 2 ppm indicate a shortage of air addition and this Will slow the oxidation rate; concentrations higher than about 4 ppm indicate sufficient or excess air is being added or that microbial activity is minimal for some reason (for example, the heap is depleted in sulfide, and, therefore, the 9 is not being used). Nutrients. The leaching microorganisms have few nutritional requirements: FQ,-: W+and a few trace elements. Trace elements, such as Mg2’ and K’, are generally present in sufficient quantities from the degradation of rack in the acid leach. PO:- and NI&’ are added to stirred-tank bioleach operations, and W’ is occasionally added to heap leach operations. K’ is added in many tank bioleach plants as the hydroxide, sulfate or occasionally as the phosphate. Carbon dioxide. The microorganisms require carbon fbr synthesis of cellular components. They obtain this carbon by reduction of COz in a complex metabolic pathway. Microorganisms expend considerable energy in “fixing” this carbon. C02 is generally available fiom the air added for oxidation or fi.omthe acid neutralization of limestone added for pH control in tank reactors. Energy (fd) source. The bioleaching microbes require a food source and that food source is ferrous iron (Fe2+) for the iron-oxidizing microbes and chemically reduced sulfur compounds such as So for the sulfur-oxidizingmicroorganisms. Microorganisms obey the laws of thermodynamics; they do not perform any oxidation reactions that are not thermodynamicallypossible. Microbes are . referred to as “catalysts” because they speed up certain reactions. For example, the oxidation of Fez’ to Fe3’ in an acid solution is extremely slow chemically; microorganisms increase the rate of this oxidation by some 500,000 times (Lacey and Lawson 1970). The reason the organisms are so good at iron oxidation is because they must oxidize a lot of it to obtain enough energy to fix COZ and synthesize complex proteins, carbohydrates, DNA, etc. Salinity. The microorganisms involved in bioleaching are relatively intolerant to the chloride ion (C1-); the kinetics of Fez+ oxidation by A. firrooxiduns are significantly slowed by 5 g Cl-/L. Attempts to adapt the organisms to higher C1- concentrationshave been unsuccesskl (Lawson et al. 1995) due to the unfettered transport of the ion across the organism’s cell membrane. There are efforts underway in laboratories around the world to find iron- and sulfur-oxidizing microorganisms that are naturally tolerant to C1‘ yet exhibit the same rates of iron oxidation that current leaching microorganisms possess. Finding such organisms is very important fiom a commercial standpoint, because in arid climates, saline and brackish waters are often the only water available for processing. Promising test results were declared by an Australian mining company that evaluated oxidation rates and metal recoveries fiom nickel, copper and zinc bearing ores using microbial strains capable of tolerating salt concentrations nearly six times that of seawater (about 125 g Cl-L) (Titan 2002). No test details have been published. Soluble cation and anion metaYmetalloid concentrations. Leaching microorganisms are tolerant to high concentrations of most heavy metal cations and can be readily adapted to even higher concentrations. In stirred-tank bioleach plants where heavy metal concentrationscan easily exceed 20 or 30 g/L of Cu”, adaptation of the microbial culture to anticipated metal

1545

concentrationsis an important design step. There are some cationic metaldmetalloids,which can be toxic to the organisms. For these substances to be toxic, they must be soluble. Mercury and silver, though toxic, are usually not serious problems, because silver has a low solubility in acidic leach solutions and mercury adsorbs to rock, mitigating its toxic effect. As5+is not toxic, but As3+ is. It is important, particularly in heap leach operations, that the redox potential is sufficientlyhigh to ensure that, when arsenic-Wing minerals such as realgar (ASS), orpiment (As&) and arsenopyrite(FeAsS) are degraded, that As3+is oxidized to AsS+.High concentrationsof A13+have been implicated in toxicity; however, this toxicity may largely be attributable to high TDS (total dissolved solids) in which several potentially toxic cations and anions are collectively at levels that may slow microbial iron oxidation. The nitrate anion (NO;) presents toxicity issues, and concentrations in excess of 200 mg/L slow the rate of Fe2' oxidation. Like other anions, such as Cl-, the mechanism oftoxicity is likely to be disruption of the cell membrane and uncontrolled transport of NO3- into the cell, which suggests that adaptation of the microbes to NO; may not be effective. Discovering leaching microorganismswith innate NO; tolerance may be the answer to this toxicity problem. Process reagents and materials. An important consideration in the design of bioleach plants is assurance that process reagents are not toxic to the leaching organisms. For example, flotation reagents that carry over in the feed to the first stage reactor in a stirred-tank bioleach plant must be evaluated to make sure that they do not inhibit the microbes. All materials, such as rubber linings in tanks, leach pad liners, and all materials that microbes come in contact with in the process, must be evaluated in lab tests to ensure that there are no inhibitory effects. Tailings waters containing traces of cyanide (CN-), thiocyanate ( S C N ) , or cyanate (CNO-) must not be used as process water or make-up water to bioleach circuits. These agents are respiratory inhibitors that deactivate microbial enzymes, and, if they enter the circuit, the result is significant loss in plant performance at best (Bell and Quan 1997) and a total loss of microbial activity at worst. Obviously, the toxicity of cyanide has implications in the treatment of concentratesthat have been previously cyanide leached. Oils, greases, hydraulic fluids, water treatment chemicals, dust suppressors, etc., all substances common to metallurgical plants, are potential inhibitors to the leaching microorganisms. Some of these agents are surfactants, which damage the organism's cell membrane causing the membrane to break open. Little quantitative data are available on the exact concentrations that induce problems. Good housekeeping in metallurgical plants is necessary to avoid contaminating anything in which the microorganisms come in contact with. Biocides are used in cooling circuits to eliminate microbial contamination. For obvious reasons biocides must never be allowed in any part of the circuit in which the leaching microorganisms are employed. Physical Characteristics Affecting Mineral Biooxidation Crystal imperfections and grain boundaries. Microorganisms tend to adhere to locations on mineral surfaces where there are visible imperfections (Sand et al. 1999). These sites are often associated with crystal defects due to dislocations (grain boundaries) and inclusions. Lattice strain in a mineral crystal can be caused by the presence of impurity atoms, such as gold in pyrite and arsenopyrite, cobalt in pyrite, and indium in sphalerite, and these impurities increase the possibility of chemicalhiological attack at that site. Thus, the more ftactured and impure the sulfide mineral, the faster the rate of oxidation is likely to be. For some sulfidic-reftactory ores, the gold particles are often concentrated at pyrite and arsenopyrite grain boundaries. In these situations, not only is the oxidation rate enhanced, but the amount of sulfide that has to be oxidized to attain good gold recovery is diminished, because the majority of the gold is released with oxidation of the sulfide adjacent to the gold grain. Particle size effects. Generally, the smaller the particle size, the more rapid the oxidation. It makes sense that the smaller the particle diameter and thickness, the greater is the overall surface area. However, small particles in a stirred tank system may increase the apparent viscosity of the slurry and decrease oxygen transfer rates. A particle size of 80% passing 53 to 75 p is usually

1546

considered ideal for stirred tank reactors. In crushed ore heaps the size of the sulfide particles cannot be influenced by the ore crushing circuit. Therefore, in heap processes it is important that the crush size is sufficient to liberate the sulfide particle, thus allowing it to be oxidized. Morphology of sulfide particles. Microorganisms and reagents (Fe3+,HzS04and nutrients) must reach sulfide surfaces for oxidation to commence. This may not be possible if there are oxidized surface coatings or flocculant layers. To minimize surface coating effects, a regrind circuit is often employed for flotation concentrates and partially oxidized concentrates. For whole ore heaps the issue does not arise; the material must be oxidized, as is, and this is one reason why oxidation on heaps takes so much longer than in stirred tanks. It should be noted here that during the oxidation of some minerals, surface coatings form which then leads to a rapid decrease in oxidation rate. Galena is an example; initially rapid oxidation quickly ceases, due the formation of an insoluble PbS04 layer. Slurry solids concentration. In stirred tank processes, the solids slurry concentration affxts the oxygen transfer rate fiom the gaseous to the aqueous phase. To avoid limitation in 0 2 and COZ transfer rates, most stirred tank operations run at about 20% solids or less. AERATED, STIRRED-TANK BIOLEACHING The first aerated, stirred-tank bioleach pilot plant was established at the Fairview Mine in South M i c a in 1984 to pretreat a sulfidic-refiactory gold flotation concentrate. The bioleach technology, called BIOXB, was pioneered by GENCOR S.A. Ltd. After two years of suCcessi5l operation, the commercial-scaleplant, which treated 40%of the mine's production, was commissioned. In 1991 the Fairview plant was expanded to 40 tonnes per day of flotation concentrate (Dew et al. 1997). Although Fairview is no longer owned or operated by GENCOR and its successor companies, the bioleach plant continues operation today. The success of BIOX@ spawned bioleach technologies developed by BacTech Environment Corporation and BRGM for treating sulfidic-refiactory gold and base metal concentrates. Collectively the three processes have resulted in 11 tank-bioleach, commercial plants around the world (Table 1). A11 of the plants listed in Table 1 treat sulfidicrefiactory gold concentrates except for Kasese, which oxidizes a cobalt-containing pyrite, and Pering, which is the first industrial-scalechalcopyrite concentratebioleach plant. The aim of this section is to examine the similarities and differences among the three tank, bioleach technologies, consider the design of a tank bioleach plant and present case histories. Emphasis is placed on tank biooxidation as a pretreatment technology for sulfidic-refiactory precious metal concentrates, however, this section will conclude with a discussion on bioleaching of chalcopyrite concentrates in tanks. Proprietary Tank-Bioleachiag Technologies Biooxidation of sulfidic-refiactory precious metal and base metal concentrates in aerated tanks entails essentially the same process flowsheet, regardless of whose technology is used. A typical flowsheet is shown in Figure 1. The ore is crushed, milled and subject to flotation; the tails are dewatered and discharged or may be further treated. The concentrate feed is dewatered, and in some cases the concentrate is reground to improve metal recovery. The concentrate is fed to the biooxidation circuit, which is detailed in the bottom portion of Figure 1. The biooxidation circuit consists of up to four stages, each stage composed of one or more tanks in parallel. The first stage typically has the larger number (usually three) of equally sized tanks in parallel, because the retention time of the feed in the 6rst stage is longer. The purposes of the longer holding time are to establish the microbial population and to allow attachment of the microorganisms to the mineral feed, preventing ashout out'^ (loss) of the organisms fiom the circuit. More sulfide is oxidized -- up to 60% of the total amount requiring oxidation -- in the first stage than in subsequent stages. Blowers are used to supply air to each tank. Large amounts of heat are generated from the oxidation of sulfides, and this heat must be removed to maintain the temperature in the range for the types of microorganisms used in the biooxidation circuit. A cooling system is integral to the

1547

Table 1 Commercial, aerated, stirred-tank, bioleach plants

Plant

Design (t concentrate/day) Initially 10 Expanded to 35 (1991) Expanded to 40 (1994)

Technology Used

Operating Years

BIOX@

1986 - Present

Sao Bento, Brazil

OriginaIIy 150 Expanded in 1994 & 1997

BIOXB Expansion-Eldorado

I990 - Present (BIOX shut down - energy saving)

Harbour Lights

40

BIOXB

1992 - 1994 (Ore depleted)

Wiluna, Western Australia

115 Expanded to 158 in 1995-1996

BIOXB

1993 Present

Sansu, Ghana

720 Expanded to 960 in 1995

BIOXGD

1994 - Present

Youanmi, Western Australia

120

BacTech

1994 - 1998 (Ceased operation; high mining cost)

Tamboraque, Peru

60

BIOXB

1999 - Present

Kasese, Uganda

250

BRGM

1999 - Present

Beaconsfield, Tasmania, Australia

-70

BacTech

2000 Present

Laizhou, Shandong Province, China -100

BacTech

2001 - Present

Pering, South Afiica

BioCOP@

2001 - Present

Fairview, South Africa

(300 m3tanks)

-

-

CONS

FLOAT

DEWATER

I""' I I I

i

I I I

v

--

CONS REGRIND

BACTERlM OXIDATION CIRCUIT

cows STORAGE

-STAGE REACTOR

1st gTAoE REACTOR

Figure 1 Flowsheet for biooxidation of concentrates in stirred tanks

1549

mrrwprr SUPPLY

Id STAGE REACTOR

biooxidation plant. Provisions are made to supply limestone for neutralization of the solution in the tanks to maintain the proper pH for the organisms. After the final stage of biooxidation, the contents of the reactor(s) are subject to solidliquid separation. If the plant is used to oxidize sulfidic-refiactoryprecious metal concentrates, the solids are neutralized with lime and subject to CIPKIL for gold recovery. The solution fiom the solifliquid separation is neutralized with limestone followed by lime and discharged to the tailings dam. However, if the biooxidation plant is used for recovery of base metals, the product of value is in the solution. Therefore, following solidliquid separation of the contents of the h a 1 biooxidation stage, the solids are neutralized and discarded to tailings, and the solution is processed to recover the metal value, for example by solvent extraction to recovery copper. BIOXB process. The BIOXB technology developed by GENCOR, first tested at Fairview anid later applied in other plants listed in Table 1, utilizes a mixed culture of mesophilic bacteria and biooxidation plants are operated in the 40" to 45°C range. In the reorganization of GENCOR, the BIOXB technology relating to precious metals was transferred to Gold Fields Ltd and the base metal biotechnologyshifted to Billiton Ltd. Using the BIOXB technology as a foundation, Billiton then developed similar biotechnologies for recovery of copper (BioCOPB), nickel (BioNICW), and zinc (BioZINCB). Billiton subsequently made advances in the bioleaching of chalcopyrite using thermophilic microorganisms. The technology was piloted in cooperation with CODELCO, a full-scale test reactor was commissioned at the Pering Mine in South Afiica to prove the viability of the process, and the technology is now being commercialized via a BillitonCODELCOjoint venture called Alliance Copper (Craven and Morales 2000). Billiton's interests in biotechnologyare now merged with BHP (Billiton 2002). BacTech process. BacTech (Australia) Pty, founded in the early 1980s, developed moderately thermophilic bioleaching technology researched at Kings College, London. BacTech's first commercial plant was Youanmi, a sulfidic-refkctory gold concentrate bioleach plant in Western Australia (Table 1). This plant operated at about 50°C. In 1997BacTech formed a partnership with Mintek (Johannesburg, South Afiica) to pool bioleaching technology. Subsequent commercial plants at Beaconsfield (Tasmania) and Laizhou (Shandong Province, China) employ BacTecMintek mesophilic bacterial processes. BacTech Enviromet Corporation is a public company with headquarters in Toronto. BacTecMintek, too, are focusing on tank bioleaching of chalcopyrite at thermophilic temperatures. This effort is via a joint venture company, Procescvs Biometalurgicos SA de CV (PMB), with Penoles SA de CV of Mexico (BacTech 2002). BRGM process. In 1989 BRGM, France, initiated a study on bioleaching cobalt-containingpyrite tailings fiom the Kilernbe Mine in Uganda (&Hughes et al. 1997 and 1999). This study culminated in the operation of a 65 m3 pilot reactor onsite in 1993. A full-scale plant design to bioleach some one million tonnes of pyritic tailings for cobalt recovery was completed and the plant was commissioned in 1999 for Kasese Cobalt Company, Uganda (Briggs and Millard 1997). The biotechnology entails leaching the pyrite with mesophilic bacteria. BRGM, also, is focusing on development ot: tank bioleaching of chalcopyrite concentrates using a process the company calls HIOXB (&Hughes et al. 2001).

Selection of the Tank Bioleacb Process Biooxidation is but one process option for the pretreatment of ores and concentrates in which the precious metal values are locked in a sulfide matrix, the other options are pressure oxidation and roasting. Fine grinding followed by cyanide leaching may be suitable for those materials in which the metal is in the metallic form or some other readily cyanide-soluble f m . In some cases, simply making a concentrate and selling that concentrate is an option. All options must be considered when examining a process route for sulfidic-refiactory materials. Tank biooxidation has certain advantages over the competing processes and is selected for these reasons: 0

There are lower capital costs for small to medium sized plants,

1550

The process is flexible, easily controlled and can be managed for only partial oxidation or oxidation of specific minerals, No sophisticatedequipment is required, The process can be operated using relatively unskilled labor, There is the potential for higher metal recoveries, and The process is operated at low temperature and atmospheric pressure; therefore, biooxidation is generally considered safer and a “greener” technology as there are no gaseous emissions. The selection of tank biooxidation over a heap must take several questions into consideration. A few of the questions that must be asked are: Is the precious metal grade high enough to warrant the cost of upgrading the material to a concentrate? If so, tank biooxidation is an option. Is the ore amenable to concentration? If not, tank biooxidation may not be an option unless the ore grade is high. What is the mineable tonnage and life-of-mine? Is there sufficient value present to warrant the cost of making a concentrate and biooxidizing it in aerated, stirred-tanks? Design of an Aerated, Stirred-Tank Biooxidation Circuit Five hdamental criteria are used to design a tank biooxidation circuit: (1) oxygen requirements for air/02 addition, (2) heat balance for temperature control, (3) acid balance for pH control, (4) nutrient requirements, and (5) process control (Slabbert et al. 1992; Dew et al. 1993; van Awegen 1993; Nicholson et al. 1994; Dew et al. 1996; Pinches et al. 2000; do Carmo et al. 2001). Oxygen requirements, heat balance and acid balance are closely related, because the oxidation of sulfide minerals consumes oxygen, generates heat, and either requires acid or produces acid (Dew et al. 1997a). Testing of tank biooxidation usually entails several phases. The first phase involves small bench-scale, stirred (or shaken), batch reactors providing information on (1) amenability of the feed to biooxidation, (2) potential toxicity problems with the feed and site water, (3) ultimate recovery estimates of the precious or base metals, and (4) performance of the different suites of microorganisms. The next test phase is also conducted in the laboratory and involves several stages of continuous stirred-tank bioreactors to simulate the full-scale operation. This phase of testing provides information on the (1) rates of oxidation for various sulfide minerals present in the feed, (2) pH, acid and redox conditions under which the sulfides are oxidized, (3) acid balance data, and (4) the degree of sulfide oxidation achieved in each stage. If the continuous test was performed at a scale larger than “mini-pilot plant” scale, say 20 to 30 kg/day of feed, that would be the final testing before design, construction and commissioning of the full-scale plant. For sulfidic-refiactoryprecious metal ores there is a sufficient knowledge base derived fiom existing operations that minimal laboratory and pilot-scale testing is needed for new operations. This is not the case for base metals; on-site piloting was done at Kasese to evaluate the bioleaching of the cobaltous pyrite feed. However, this is often required to generate sufficient sample for downstream process test-work for metal recovery. Large-scale demonstration plants have also been operated for extended time to evaluate the bioleaching of chalcopyriteconcentrates. Oxygen requirements. The requirement for Ozby microorganisms and the reactions they catalyze were detailed earlier in this chapter. From a metallurgical point of view, it doesn’t matter which reactions require the presence of microorganisms or occur chemically, because the quantity of oxygen reqpired remains the same. Therefore, oxygen requirements are calculated fiom the oxidation reactions (Dew et al. 1997; Miller et al. 1999), the most common being:

Pyrite: 4 FeSz + 15 O2 + 2 H20 + 2 Fe(SO&

+ 2 H2S04

1551

(Reaction 4)

Arsenopyrite: 2 FeAsS + 702 + 2H20 + H2S04 --+ 2 H3A~04+ FeL(SO4)3

(Reaction 5 )

Pynhotite:4 FeS + 9 0 2 + 2 H2S04 -+ 2 Fe~(S04)3+ 2 H20

(Reaction 6)

Chalcopyrite: 4 CuFeS2+ 2 H2S04+ 17 0 2 -+ 4 CuSO4 + 2 F%(S04)3+ 2 H20

(Reaction 7)

Chalcocite: Cu2S+ 512 O2 + H2SO4 -+ 2 CuSO4 + H20

(Reaction 8)

Covellite: CuS + 202-+ Cut304

(Reaction 9)

Bornite: CuSFeS4+ 5/2 H2SO4 + 37/4 O2 --+ 5CuSO4 + % F~(S04)3+ 5/2 H20

(Reaction 10)

Pentlandite: 2Ni9SB+ 33 O2 + 2 H2SO4 -+ 18 NiS04 + 2 H20

(Reaction 111)

Sphalerite: ZnS + 2 O2 + --+ ZnS04

(Reaction 12)

Many other reactions and sub-reactions take place during oxidation of these species depending on which microorganisms are present. These side reactions can affect not only the oxygen requirements, but acid and heat balances as well. To avoid conhsing side reactions, the stoichiometricoxygen requirements for plant design of a sulfidic-refiactorygold bioleach plant are based on data presented in Table 2 (Dew et al. 1997).

Table 2 Process data for oxidation of sulfide minerals associated with sulfidic refractory gold plants (Dew et al. 1997) Mineral

Heat of Reaction Mineral Sulfide

Oxygen Requirement mole kg 0 2 k g 02/mole sulfide mineral 2.25 2.25

HZSO4 Demand &fig mineral) 0.557

Pyritic S Content (% by weight) 36.4

(kT/kg) -11,373

(kJk sulfide) -31,245

Arsenopyrite FeAsS

-9,415

48,036

3.5

3.5

0.301

19.7

Pyrite FeS2

-12,884

-24,173

3.75

1.88

-0.408

53.3

Pyrrhotite FeS

Note that the O2requirements for the plant are typically not determined @omthe lab-scale and pilot plant test work, but rather fkom theoretical calculations. The test work does provide information on what reactions, sub-reactions and secondary reactions are taking place. For calculating process air demand and pressure requirements for stirred tanks, these factors are considered: 0 0

0 0 0

O2 utilization efficiencyfor stirred tanks O2 content of dry air Air SG (specific gravity) Normal (N) temperature Normal (N) pressure

1552

3040% (depending on agitator type) 23.15% by weight 1.293 kglNm3/h 0°C 1 atmosphere

0

Pressure requirement

Hydrostatic head on sparge ring below impeller + 15 H a line loss

Blower power for aerated, stirred tanks is calculated using the Perry and Chilton (1979) formula, adapted to metric units: kW(m) = 1.0195 WT[(P2/P1)0~283 -13 where, k = 1.395 W = air flow in kg/s T = inlet air temp in K PI = inlet pressure P2= delivery pressure (PI & P2 are in the same set of units and are absolute pressure

gauge pressure)

Blower efficiency is taken to be 74 to 78 %. The agitator in stirred tank reactors is mainly for air dispersion, and a first pass estimate of the power required is 27 W/Nm3’h air added. However, this depends on the agitator design. Heat balance: In stirred-tank biooxidation heat is gained fiom reactions involving the oxidation of sulfide minerals (Table 2), from agitation energy, and from air that is added. Heat losses comprise (1) evaporation as the air, saturated with moisture, leaves the reactors, (2) convection through tank walls, (3) heating the incoming slurry, and (4) air expansion (Joule/Thompson effect). Heat balance is determined theoretically, because the lab-scale and pilot-scale tests are so small and the reactors provide such a large surface area to volume ratio that heat generation is not detected. Data from tests showing those sulfide minerals that are leached and the rates of leaching are used in the heat balance determination. Cooling water requirements for stirred tank reactors are derived from heat balance equations and calculated accordingly: Flow rate (m3/h)= Net Heat Load &W) X 3-6 AT X 4.186 AT = Temperature difference (“C) between inlet and outlet water temperature Note that cooling water leaving the reactor will be about 4°C cooler than the reactor temperature. Acid balance: Some oxidation reactions generate acid (Table 2) and increase the sulfate concentrationin solution. There are also secondaryreactions, which are dependent on temperature, pH, ionic strength of the solution and other species in the system; these secondary reactions can also affect the acid balance. Several of these secondary reactions are (Dew et al. 1997): Ferric arsenate precipitation (generates acid): 2 H3As04+ F+(S04), -+ 2 FeAs04&+ 3 H2so4

(Reaction 13)

Acid dissolution of carbonates (consumes acid): CaMg(C03)2+ 2 H2SO4 + CaS04J + MgS041 + 2 CO2t + 2 H20

(Reaction 14)

Precipitation of jarosite (generates acid): 3 F+(S04), + 12 H20+ M2SO4+ 2 M&(sO&(oH)61 H30 where M = K+,Na’, m+or

+ 6 H2S04

(Reaction 15)

+

Lab and pilot plant tests do provide information on the acid balance. These data are coupled with theoretical calculationsto arrive at the acid balance. In some circuits the solution pH is controlled during biooxidation by the addition of lime, Ca(0W2, or limestone, CaC03. The addition of limestone to biooxidation reactors for pH control

1553

also adds COz, required by the microorganisms. Reactor waste products are neutralized with limestone and lime before final disposal. Nutrient additions: Microorganisms involved in biooxidation require nitrogen, potassium and phosphorous, and these nutrients must be added to stirred-tank biooxidation circuits. These are often supplied in the form of (NH4)2S04, KOH and H3P04.Solid (NH4)2SO4 and KOH are mixed with H3P04and water, and these nutrients are pumped to the feed distributor and into the primary reactors with the sulfide feed concentrate. Because of the difficulty in handling these reagents, their price and their availability, different chemicals, including agricultural grade nutrients, are considered as alternatives. Whichever chemicals are used, it is imperative to test them in the laboratory to ensure that they provide a high level of microbial activity and that they do not in any way inhibit the microorganisms. Sometimes there are impurities in the reagents that cause problems. The amounts of NI&+,K' and PO-: added differ somewhat depending on which company's technology is being applied (Table l), but is by and large based on the chemical composition of the microorganisms, CHI.67No.zPo.01400.27, and the number of cells per ml that develop in the solution of the reactor (Pinches et al. 1994). This number is in the lo9 to 10" per ml range. The nutrient concentration is then calculated and reported as kilograms ofN, P and K per tonne of FeSz oxidized or tonne of concentrate fed to the circuit. Table 3 lists examples of nutrient concentrationsthat have been used and also lists the stoichiometricrequirements.

Table 3 Example nutrient concentrations for tank bioleach operations Company CO2 N P (kdt FeSz (kdt FeSz (kdt FeS2 oxid.) oxid.) oxid.) 50 3.177 0.447 Mintek (Pinches et al. 1994)

K (kdt FeSz oxid.)

GENCOR

3.4

0.6

1.8

BRGM (Kasese)

9.3

2.6

3.7

Stoichiometric

0.22

0.32

The disparity observed in nutrient concentrations (Table 3) likely arises because (1) nutrient limitation adversely a M s production, so excess nutrients are added to ensure that no limitation occurs, and (2) the exact number of microorganisms in the biooxidation circuit is not known. The microorganisms in solution are relatively easy to count, but the number of microorganisms adhering to the mineral suriiices is not so easily determined. Therefore, to guarantee that sufficient nutrients are available, excess is often added. In stirred-tank biooxidation, adding excess nutrients is not a serious problem fiom an operating standpoint, but it does increase the operating cost. Excess nutrients added to the system may simply precipitate as FeP04 and ammonium and potassium jarosites and may be wasted. C02 is supplied by limestone added to the biooxidation circuit for pH control and occasionally by providing gaseous COZ.Sometimes sufficient CO2 is provided by the feed, which may contain some carbonate minerals that produce C02 during acid dissolution. Circuit conditions and process controls. The typical circuit conditions for the pretreatment of a sulfidic-refiactoryprecious metal concentrate in an aerated, stirred-tank biooxidation circuit are listed in Table 4.

1554

Table 4 Conditions for sulfidic-refractory precious metal concentrate tank bioleach Element Control Slurry density 15 - 20% solids

Number of reactors

3 primary in parallel 3 secondary in series (some exceptions)

Residence time of sluny in circuit

4 to 6 days (2 to 2.5 days in I‘ stage)

40 - 45°C for mesophiles

Temperature

45 - 5OoC for moderate thermophiles

PH

Approximately 1.2 to 1.6

Dissolved 9

2 PPm

Flotation concentratesare generally dewatered after flotation for density control and removal of flotation reagents that can adversely affect biooxidation rates. Some flotation reagents are toxic to the organisms and may also interfere with the attachment of the organisms to the mineral surface (Huerta et al. 1995). At plant sites where process water quality is a problem, separation of water circuits is necessary. “Medium” quality water is used for flotation and other mineral processing steps. The highest quality water is used for the bioleach circuit or, if limited, is used to dilute saline or brackish waters. When water quality is a concern, good lab testing is essential to determine what site water is acceptable. Tailings return water fiom gold recovery circuits must not be used as make-up water in the biooxidation circuit as cyanide species are toxic to the organisms. The residence time of the slurry in the lirst stage reactors is based on the time required for the microbial population to stabilize. Microorganisms multiply by dividing and in continuousreactors, when conditions are optimum, the growth rate (also called doubling time) is exponential. Different microorganismshave different doubling times. For example, AcidithiobaciZZusferrooxidans may double every 6 to 12 hours, depending on conditions. Therefore, the residence time in the lirst stage reactor should be sufficiently long to ensure that the microbial growth rate is greater than the number of microbes leaving the first stage with the slurry flow. From this discussion it is clear why a condition in the first stage reactor that slows the growth rate of the microorganisms, such as insufficient 4,a nutrient limitation or a toxin, results in “wash-out” of the microbial population; when microbial growth rate is slowed, the passage of slurry through the circuit is faster than the doubling time of the microbes, and the organisms simply wash out of the circuit. Carrying this concept one step further, it is also possible to cause washout of a single type of microorganism. For example, if the pH should decline in the first stage reactors, that population of microorganisms that is inhibited by low pH will washout, leaving other microbes more tolerant to low pH. The remaining microbial population may not oxidize Fez’ in which case production will decline precipitously. In biooxidation circuits recycle of spillage and wash-down solutions is to be avoided. These solutions may contain constituents that are inhibitory to the microorganisms in the biooxidation circuit and result in washout of the microbial population. Excess spillage could lead to washout simply by increasing the volumetric flow-rate through the circuit, thereby reducing sluny residence times. Because oils and greases can also be inhibitory, agitator gearboxes should be equipped with oil spillage trays and spills or drainage fiom the gearboxes should be directed outside the bounded area of the biooxidation circuit.

1555

The control system employed at biooxidation plants is PLC based. Operator communication with the PLC is maintained by PCs equipped with software allowing continuous data acquisition and control instrumentation scanning. Control loops fiom the biooxidation plant include temperature and airflow to each reactor, concentrate, feed and dilution control, nutrient proportional additions, and on the downstream processing, CCD discharge rates, neutralization pH and limestone addition rates. Much data required for indirectly monitoring microbial activity are obtained by routine sampling and analysis for dissolved 02,redox potential, total iron, Fe2+:Fe3+ ratios, arsenic in solution, total dissolved solids, and other metals of importance in the feed. Equipment for Aerated, Stirred-Tank Biooxidation Circuits Materials used in the construction of biooxidation circuits for sulfidic-refkactoryprecious metals and base metals must be able to withstand highly oxidizing and acidic slurries at temperatures of 50°C or higher. The selection criteria for materials of construction are:

Suitability of the material for both the mine site environment and the microbial environment, Chloride levels in the slurry, Non-toxicity of the materials to the microorganisms, Location of the source of the material, Delivery time, including shipping, Cost of the materials, Availability of skilled labor to fabricate and install the selected material, and Ability to repair hbricated equipment and the cost of the repair. Tanks and surrounding area. Most biooxidation tanks used in operations up to and slightly exceeding 50°C are constructed of stainless steel (SAF 2205). At the Wiluna plant Linatrite N50 rubber (Linatex) was selected to line the leach tanks and agitators. Due to commissioning difficulties, cost and maintenance issues, this liner has not been used in subsequent operations. Owing to the corrosive nature of the bioleach solutions, areas surrounding the circuit must be fkee of galvanized materials including cable trays, pipe, pipefittings and light fittings. Any contact of the acidic slurry containing soluble arsenic with reducing metals, such as brass, copper, zinc, and aluminum, will generate arsine gas, which is considerablymore toxic than hydrogen cyanide. Cooling circuit. In designing the cooling water circuit for the biooxidation plant, criteria that must be considered are:

The heat load at design capacity, The wide variation in heat load due to treatment of transitional ores and high-sulfide ores, The airflow and water flow through the cooling towers, and Environmental factors, including dust load, insect control, scale and microbiological control. Some biooxidation plants are constructed using stainless steel cooling coils installed inside of the reactor tanks. Proper anti-scalents must be selected to ensure non-toxicity to the leaching microorganisms in the event of a leak in the cooling system. The design of the coils must allow for easy removal and replacement in the event of circuit failure. A water jacket around the bottom half of the bioreactor tanks is an option, if the amount of heat generated is limited. A water jacket design greatly simplifies the piping. Air supply. The air supply is a critical component of the biooxidation circuit. Because of the wide variation of sulfide contents experienced in the life of an operation, the system must have a large turndown ratio. At existing operations the air is injected into the high shear zone below the agitator impeller via an air sparge ring. The blowers are usually high speed, turbine types with variable inlet and outlet vanes, which give a turndown capability to 40% of rated maximum

1556

capacity of each blower. On this basis air output can be varied over a significant range with the installation of multiple units. Each blower in the circuit is fitted with an aftercooler to reduce air temperature to 50°C to protect the microorganisms eom high temperature zones in the reactors. The materials of construction of the aftercoolers are important, because acidic slurries can be siphoned into the system under some conditions. The reliability of the air is of paramount importance as loss of air for a period of more than 1.5 hours can have serious consequences for the circuit. The loss of air at Wiluna (Table 1) for about 50 minutes in late 1993 resulted in the loss of microbial activity in the primary reactors. It required some seven to eight days to restore microbial activity and 10 days before production levels returned to normal. To avoid this type of problem, at least one standby blower is required which should be operated through a separate power supply and control circuit. Agitators. The agitator design must allow for: Provision of sufficient power to prevent “flooding” of the impeller by the air flow, An oxygen mass transfer rate that exceeds that required for the oxidation reactions at the reactor temperature used, and An agitator pumping rate that will maintain uniform solids suspension at high solid’s specific gravity in dilute slurries. More details of these design criteria are presented in a paper by Batty and Post (1999). A comparison of bioleach tank sizes and installed power in the early plants is presented in Table 5 . The considerablepower addition to the reactors means that all internal structures must be properly secured.

Table 5 Installed power and air addition at several biooxidation plants Plant Tank Volume Air Addition Sao Bento

(primary/secondary) 0.58

@rimary/secondary) 31.7

Wiluna

0.3910.19

27.2130.7

Youanmi

0.15/0.11

27.7150.0

Typical CIL tank

0.03 - 0.05

--

Sulfidic-Refractory Gold Biooxidation Plant Case Studies Two operations are briefly described to illustrate the process design of sulfidiorefiactory gold biooxidation plants. Wiluna and Youanmi are selected (Table l), because they have similar concentrate throughputs, and hence similar sized oxidation reactors. However, their sulfide oxidation levels are quite different and hence air addition and heat removal requirements are therefore different. The oxidation circuits as described for these plants apply to all the concentrate oxidation circuits for gold recovery thus hr constructed, and also to Kasese where pyrite is oxidized for cobalt recovery. Only after oxidation do base metal recovery circuits differ fiom gold circuits. The Bid=OP@ process, developed by Billiton for copper recovery fiom chalcopyrite, is also similar in the oxidation circuit, though operating temperatures and hence materials of construction differ significantly. Wiluna Sulphides, Western Australia. When built in 1993, Wiluna comprised the crushing, grinding and flotation of 400,000 tpa of ore and the oxidation of 115 tpd (42,000 tpa) of concentrate. Oxidation residues and flotation tailings were recombined for gold leaching. The

1557

biooxidation design criteria for the BIOXO process were provided by Gold Fields (formerly GENCOR) (Brown et al. 1994; van Aswegen and Marais 1999). For the original design at Wiluna, the ore grades were 6.0 g/t Au, 1.85% S and 0.7% As, and the concentrategrades were 92.9 g/t Au, 24.0% S and 10.0% As. The milling circuit was standard and will not be described. Flotation at Wiluna is comprised of a rougher scavenger circuit with cleaning and recleaning of concentrates. For biooxidation, concentrates are not required to contain a minimum sulfur grade, as needed for auto-thermal roasting, and thus flotation can be adjusted fix maximum gold recovery into the concentrates. However, it is necessary to control the ratio of acid consuming gangue componentsto sulfide grade to ensure that a positive acid balance is maintained in the oxidation circuit. This justifies the cleaner and re-cleaner circuit. The plant provides a five-day residence time in biooxidation at a slurry density of 20% solids, allowing for sulfide oxidation in excess of90%. The plant operates at 40-42' C, and the waste heat is removed by cooling water through cooling coils. The pH in the reactors remains in the range of 1.2 to 1.6. The oxidation reactors are fabricated &om rubber-lined mild steel. The quantity of air supplied is based on the sulfideoxidation required, an estimated utilization efficiency of the oxygen added is 25%. This leads to a total addition rate of 3 1,300 Nm3/h of air (40.5 tph) or about 8 tonnes of air per tonne of concentrate. The overall heat to be removed fiom the circuit is 11.8 MW. This figure includes the heats of reaction, heat generation fiom the agitators and air blowers, and takes into account heat losses, through evaporation and adsorption by the incoming slurry. Thus the net heat generation is about 2.4 h4Wh per tonne of concentrate or 10.9 hWh per tonne of S2--sulfuroxidized. Waste heat is, removed by a cooling water circuit with cold water circulating through cooling coils in the reactors; shell and tube heat exchangers cool the blower air. Warm water is circulated to a cooling tower and returned to the circuit. The cooling water flow is 936 m3/h (design 1060 m3/h), which is equivalent to 0.87 m3/h per kg of Sz--sulfuroxidized. Note, however, that the size of the cooling water circuit is very dependant on the cold water temperature achieved. At Wiluna warm water (35.6" C) is cooled to just 26" C. If this temperature could be decreased to say 23' C, the water flow could be reduced to 713 m3/hor 0.66 m3/hper kg S2--sulfuroxidized. Power consumption is high in biooxidation circuits. The power is mainly used for air generation in the blowers and air dispersion in the reactors. At Wiluna, the agitator mechanisms were installed with 185 kW drives on the primary tanks and 90 kW drives on the secondary tanks. Each tank is 470 m3 in volume. The difference in agitator power between the primary and secondary tanks is due entirely to the difference in oxygen requirements in the different stages of oxidation. Three operating blowers each with a 400 kW drive provide the air supply. Youanmi Deeps, Western Australia. The Youanmi Deeps project is similar in size to Wiluna. Sulphides, but utilized BacTech technology (Miller 1997) rather than BIOXO. Youanmi was commissioned in 1994, but ceased operation in 1998 due to high underground mining costs. The plant was designed to treat 200,000 tpa of ore, but since the sulfide-sulk content of the ore was over twice as high as that at Wiluna, the concentrate throughput of 120 tpd was similar to Wiluna. The ore content is I5g/t Au, 7.5% S and 1.0% As. The concentratecontent was 60 g/t Au, 28.0% S and 4.3% As. The Youanmi flotation circuit was comprised of a roughing and scavenging stage only; no cleaning of concentrates was required for elimination of carbonates or other acid consumers. The plant provided a five-day residence time at a slurry density of 15-20% solids; this required reactor sizes of 500 m3, just slightly larger than those at Wiluna. The size of these reactors is set by the throughput, slurry density, and residence time, and is unaffected by the quantity of sulfide-sulfur to be oxidized. The reactors were fiibricated fiom S A F 2205, a stainless steel resistant to the higher levels of chloride in the waters on site. The oxidation circuit operated at 45' to 50°C and pH 1.2. Youanmi required only 32% sulfide oxidation for high gold recovery efficiencies, because the gold was associated with the arsenopyrite, which represented only a third of the total sulfidesulfur in the feed. This 32% sulfide oxidation was only about a third of that required at Wiluna, therefore, air requirements and waste heat generation at Youanmi were also a third of those

1558

generated at Wiluna. Air requirements at Youanmi were 11,400 Nm3/h; Wiluna was 31,300 Nm3/h. Because of the lower air demand, the installed power of the blower was 540 kW in two units. Wiluna’s installed power was 1200 kW in three units. Similarly, the design waste heat removal f?om the Youanmi circuit was 4.5 MW, whereas, Wiluna’s waste heat was 11.8MW. Owing to the lower waste heat removal requirements at Youanmi, the cooling circuit could be simplified, and cooling of the reactors was accomplished by a water curtain on the outside of the tanks rather than cooling coils. The agitators at Youanmi operated in similar sized tanks to Wiluna with similar sluny densities, but because they received about a third of the airflow, the installed power on the drives at Youanmi was only 75 kW in the primary reactors and 55 kW in the subsequentreactors. This is compared with 185/90 kW (primaryhubsequentreactors) at Wiluna. The Youanmi and Wiluna circuits both required limestone addition for neutralization of the oxidation products. Wiluna also required limestone addition for pH control in the oxidation circuit. Youanmi required acid for pH control, because the sulfide that was oxidized was predominantly arsenopyrite, whose oxidation is acid consuming. A limestone milling circuit at Youanmi was included in the design with limestone addition to the oxidation products by ringmain.

Figure 2 Youanmi Deeps biooxidation project, Western Australia (photo courtesy of BacTech Environment Corporation)

Other stirred-tank biooxidation projects. Table 6 compares eight stirred-tank biooxidation plants based on their feed and percent sulfide-sulfur oxidized. Provided in this table are the design parameters (oxygen demand, reactor design and volume, and feed residence time) used and the heat generated, total power installed, specific power consumed and sulfide oxidation rate for each plant. The installed power noted in this table includes biooxidation, cooling water and air circuits only; standby units are not included. Sao Bento operates in series with pressure oxidation. The feed rate is determined by overall process economics and is variable. Table 6 illustrates the original design, which had one reactor. Of interest in Table 6 is the relatively constant specific energy demand in terms of k W g of sulfide-sulfur oxidized. This is a very useful number for a rapid economic appraisal of project economics. The figure for Youanmi is high because of the over-sizing of the second stage reactor agitators. See the comparison with Wiluna.

1559

Table 6 Design comparison of aerated, stirred-tank biooxidation plants Item Unit Fairview SaoBento Harbour Lights Concentrate feed tlday 1 Jc J 150 40

Sulfide grade Sulfide feed Sulfide oxidized

vl

8

Specific heat of reaction Specific oxygen demand Total reactor volume (aerated) Primary reactor volume Number of reactors (operating) Residence time Heat of reaction Oxygen demand

Total installed power Average sulfide oxidation rate Specificpower consumption

Wiluna

Sansu

Youanmi

Tamboraaue

Kasese

158

960

120

60

240

%

20

18.7

18.6

22

11.4

28

30

40.9

kfl

292

1,169

310

1,448

4,560

1,400

750

4,172

%

kfl MJACgs2-

89 260 26.7

348 29.2

87 270 29.2

95 1,376 29.8

94 4,286 33.8

32 448 35.3

86 645 70.2

92.3 3,85 1 24.2

kg02/kgS2’

2.05

2.17

2.22

2.27

2.49

2.40

2.57

1.88

m3

913

1374

880

3,391

18,278

3,000

1,336

5,520

m3

1 x 343 4x97 9

2 x 550 1 x427 3

3 x 163

4x471

12x896

3 x500

3 x262

3 x 1 380

6

8

24

6

6 4 4.0 25.8

Days Mw

6.1 2.0

1.o 2.8

5.1 2.2

5.0 11.1

4.4 41.1

5.0 4.0

5.3 6.1

kfl kg/h/m3 kW

540 0.59 798

755 1.30 758

612 0.70 59 1

3,024 0.89 1,797

10,872 0.59 7,323

987 0.33 950

1,656 1.24

7,220 1.31 5,727

kg/m3/day

6.8

14.4

7.4

9.7

5.6

3.3

11.6

16.7

kWh/kg S2-

1.9

1.8

1.9

1.5

1.9

2.3

1.5

Comparative operating and capital costs for the biooxidation leach circuit, counter-current decantation for washing the biooxidized residue, and the neutralization circuit of selected plants were report earlier (Brierley and Briggs 1997). These costs were related to total sulfide-sulk in the feed and the sulfide-sulh oxidized to achieve >90% gold recovery. The cooling circuit adds considerable capital cost, because of the need for stainless steel cooling coils, water supply pipework and cooling tower. Stirred-Tank Bioleaching of Chalcopyrite Concentrates BHP-Billiton (Craven and Morales 2000), BacTecMintek (Miller et al. 1999) and BRGM (#Hughes et al. 2001) are all developing proprietary technologies for stirred-tank bioleaching of chalcopyrite concentrates. Chalcopyrite does not leach well using the mesophilic bacteria, as surface coatings develop on the chalcopyrite mineral phase slowing the leach rate and thwarting copper recovery. In the late 1990s reports surfaced in the technical literature that bioleaching of chalcopyrite was enhanced with mesophilic bacteria, if ferrous iron was added (Hiroyoshi et al. 1997; Hiroyoshi et a1 2000; Third et al. 2000). This revelation led to studies and speculation as to why chalcopyrite leaching is enhanced by ferrous iron addition and controlling redox potential in the narrow Eh range of615 to 645 mV (Standard Hydrogen Electrode) (Timmins and Hack1 1998; Breed et al. 2000). Increasing the temperature and controlling the redox potential in this narrow window further enhances chalcopyrite leaching. A patent (Pinches et al. 2001) was issued and assigned to Mintek, which reveals a process for controlling the redox potential to leach chalcopyrite in both tanks and heaps; the patent includes the concept of using microorganisms to control surfiice potential. The thermophilic microorganisms have been shown to be more effective in leaching chalcopyrite (Dew et al. 1997a; Gericke and Pinches 1999). Although the rationale for their effectivenessis not fblly evident, it is likely that the thermophiles, particularly the Archuea, play some role in controlling redox potential at the chalcopyrite/organism interface. These technical developments in thermophilic chalcopyrite leaching have been transformed into commercial processes. BHP-Billiton disclosed their BioCOPB process for bioleaching of chalcopyrite concentrates with the extremely thermophilic Archeu (Craven and Morales 2000) and B a c T e c W t e k announced their chalcopyrite concentrate leaching technology using moderate thermophiles (van Staden et al. 2000). The engineering principles of aerated, stirred-tank bioleaching of chalcopyrite concentrates at elevated temperatures (60" to 90°C) are fimdamentally the same as described for pre-treating sulfidic-re6actory gold concentrates at 40" to 50°C. Operating the plant at high temperatures is advantageous because of k t e r leach kinetics, however, there are several important considerations that must be taken into account - the effect of high temperature on oxygen mass transfer and materials of construction (Batty and Post 1999; Harvey et al. 1999) and the effect of enhanced evaporation on the water balance. BHP-Billiton pilot tested a new elevated temperature reactor design for chalcopyrite concentratebioleaching at CODELCO's Chuquicamata Mine in Chile and have since commissioned a commercial-scale(300 m3 reactors) plant at the Pering Mine in South Mica. Several patent applications have been filed on the technology (Tunley 1999) and engineering aspects of this technology (Dew et al. 2001; Norton et al. 2001; Basson et al. 2001; Dew et a1 2001a; Dew and Miller 2001). Few details are available on the reactor design at this time, although in contrast to existing bioleach plants, the elevated temperature plant is using 0 2 plant gas as opposed to air to aerate the slurry. Alliance Copper, the BHP-BillitodCODELCO joint venture, is marketing this process. BacTechiMintek in conjunction with Industrias Penoles of Mexico are operating a fblly integrated copper bioleach, solvent extraction and electrowinning demonstration plant in Monterrey, Mexico using moderately thermophilic microorganisms. This technology, which is the subject of several patent applications (Rhodes and Miller 2000 & 2000a; Winby et al. 2000) is being commercialized through a BacTech-Mintek-Penoles joint venture, Procesos Biometalurgicos SA de CV (PBM). No details are yet available on the engineering design of the PBM reactor.

1561

BIOOXIDATION OF ORE AND CONCENTRATES IN HEAPS Heap leaching of oxide copper ores with recovery of the copper by solvent extraction and electrowinning is conventional technology. The fundamentals of copper oxide heap leaching have been integrated with the principles of bioleaching to heap leach secondary copper ores. Copper heap leaching is the subject of another chapter in this book, “Copper Heap Leach Design and Practice” by RE. Scheffel, and the reader is referred to that chapter for details. The emphasis of this section will be on the heap leaching of sulfidic-reffactory gold ores with discussion on the emergingtechnologies of heap leaching chalcopyrite ores and sulfide concentrates. The premise of bioheap leaching is that the heap is the reactor. This means that conditionswithin the heap must be optimized for full participation by a suite of microorganisms that catalyze the oxidation of the sulfide minerals. Heap Leaching Sulfidic-Refractory Precious Metal Ores The heap leaching of sulfidic-refiactory gold ores is similar in many ways to that of secondary copper ores. However, there are some notable differences and this section will only focus on those differences. For more details about heap leaching of sulfides, the reader is referred to R.E. Scheffel’s chapter in this book and an earlier paper by Brierley and Brierley (1999). Sulfidicrefkactory precious metal whole ore heap technology was perfected (Brierley, 1997; Brierley, 2000) and patented (Brierley and Hill, 1993,1994 & 1998) by Newmont Mining Corporation, and is now commercially applied (Bhakta and Arthur, 2001) at the company’s Nevada (USA) operations. Microbiology and chemistry of the process. The fundamentals of the technology are same as for sulfidic-refiactorygold concentrates. Microorganisms catalyze the oxidation of pyrite (FeS2) and arsenopyrite (FeAsS), exposing gold that is locked, or occluded, within these sulfide minerals. The degradation of the sulfide minerals significantly improves the gold recovery over that of cyanide treatment alone. The principal chemical and microbial reactions are the oxidation of pyrite and arsenopyrite with microbiologicallygenerated (Reaction 1) ferric iron FeS2+ 14 Fe3’ + 8 H20+ 15 Fe2’ + 2 SO:-

+ 16 H‘

FeAsS + Fe3’ + 2 H20+ 3 O2 +4 H‘+ As0;- + 2 Fez’ + SO:-

(Reaction 16) (Reaction 17)

and the microbial re-oxidation of ferrous iron (Reaction 1) to perpetuatethe leaching process. Any reduced sulfur species, including elemental sulfur, that accumulates is microbially oxidized (Reaction 2). Crushing and stacking the ore. Laboratory column studies and on-site crib and pilot test heaps are used to confirm the crush size that will provide optimum precious metal recoveries. To ensure good solution and air permeability in the heap, consideration must be given to fines generation and particle size. Newmont’s process entails inoculating the crushed ore with mesophilic, moderately thermophilic or extremely thermophilic microorganisms (Brierley and Hill 1993, 1994 & 1998; Brierley 2001) employing any agglomerating method before the ore is stacked. The microbial inoculum can be prepared in tanks, ponds or can be the effluent fkom the heap. Applying the microbe-containing solution inoculates and acidifies the crushed ore and binds fine material to coarse ore particles. Newmont contends that agglomerating the ore with microorganisms reduces the overall residence time of the feed on the pad, becaw the microorganisms are distributed throughout the heap greatly minimizing the time for the organisms to reach their maximum numbers and performance. Irrigation and aeration. The “on/off or dynamic pad design, discussed in the RE. Scheffel’s chapter on copper heap leaching, is used for sulfidic-refkactorygold heap leaching. Pad liners are constructed of a properly compacted low-permeabilitynatural barrier covered by a HDPE liner. A 200 to 1000 mm thick gravel layer is employed to protect the liner and to allow installation of drain and air pipes below the heap.

1562

Drainaflex pipes are installed at 2-m centers with air injection pipes immediately on top of the drainaflex pipes outside the phreatic zone to prevent flooding. Solution collection ditches are employed on both sides of the cells to direct solution by gravity to the correct pond. The agglomerated ore is stacked, aerated and irrigated. Air is injected into the heap using a set of low-pressure high volume fans or blowers (Salomon-de-Friedberg 1998). The design normally allows for a multitude of "portable" fans to be used, so that new cells on the heap with hst oxidation rates can be provided with more air than older areas where the oxidation is almost complete. A small mobile crane moves fans. Holes (3 mm) are drilled in the bottom of 50 mm diameter air distributionpipes. The density of the holes is dependent on the amount of sulfide-sulk to be oxidized and the oxidation rate. The greater the amount of sulfide-sulk and the h t e r the oxidation rate, the greater is the density of holes. Air distribution networks typically include 500 mm diameter headers and 50 mm diameter laterals at 2-m spacing. Solution irrigation rates vary between 2.5 and 10 Wm2. Rest periods are used to control the heap and solution temperatures. Solution management is a key aspect in successll heap leaching (Schlitt 1984). At a low irrigation rate, liquid percolates downward as a thin film on the rock sur&w while air moves up through the voids in a countercurrent fashion. This promotes good oxygen transfer at the film-air interface. As the irrigation rate increases, flooding occurs at pinch points between voids in the rock. This changes the airflow pattern significantly, and airflow will short-circuit through the leach material via a few large channels. The heap does not have to be completely flooded to be poorly aerated (Schlitt 2002). Modeling of the hydrodynamics of the heap leach process is an emerging discipline (BouffEud and Dixon 2001). Heat balance. Heat gains in heaps include sulfide-sulk oxidation and daytime solar radiation. Heat losses include: 0

0

Evaporation - air addition leaves heap saturated in moisture Evaporation of irrigation solutions Convection fiom heap sides and surfice Radiation at night Heat up of irrigation solutions

Temperature is controlled in the heap by: 0

b 0

b

Heap depth -the higher the heap, the greater the heating Irrigation - cools the heap Rest periods - warms the heap Excess air addition may also cool the heap through evaporation

Heat builds up in the heaps due to rapid sulfide-sulfur oxidation. The temperature exceeds the uppermost limit of activity for the mesophilic, Acidithiobacillus and Leptospirillum bacteria. At about 40°C the moderately, thermophilic bacteria, such as Acidithiobacillus caldus and Sulfobacillus species, increase in numbers and perform the oxidation of iron and sulfur. At about 65°C the extremely thermophilic Archaea microorganisms, such as Sulfolobus and Acidianus species predominate. Sulfidic-refiactorygold ore heaps heat to 70°C or higher (Bhakta and Arthur, 2001). The time required for oxidizing the sulfides in the heap varies fkom a minimum of about 90 days up to 250 days. The oxidation time is ore dependent (Bennett and Ritchie 2002). Iron chemistry. As solutions are recycled, iron concentrations increase in the solution until the solubility of various iron compounds is exceeded. Iron precipitation is abundant in sulfidicrefiactory precious metals. These precipitates are various jarosite compounds, including silverjarosite, and ferric arsenate when arsenopyrite is leached. Because of iron precipitating in the heap, solution iron concentrationsare not a reliable way to assess the degree of oxidation that has

1563

occurred. Sampling of the oxidized solids in the heap followed by bottle-roll cyanidation and analyses of sulfide-sulfiu and total sulfiu are needed for confirmation. Nutrient addition in heaps. In heap leach operations nutrients are usually not required as sufficient amounts for the microbial population are available 6om the ore and ammonium nitrate blasting agents. Occasional analysis of the leach solutions should be practiced to ensure that ammonium ion, in particular, is present at a level of about 2 to 5 ppm. If needed, ammonium ion is added as (N€L,hS04. Downstream processing. When sufficient sulfide mineral has been oxidized, the heaped ore may be rinsed with fiesh water to remove excess acid and ferric iron, which consume cyanide, and the oxidized residue is removed 60m the pad. The leached residue, which still contains the gold and silver, can be agglomerated with lime and re-stacked for leaching with cyanide or limed and placed in a milling circuit with cyanide to extract gold. The latter process, called “bio-milling”, is currently used by Newmont (Bhakta and Arthur, 2001). Pregnant solution fiom the cyanide heap leach or the bio-milling circuit is treated in a carbon adsorption circuit. Carbon loaded with gold (and silver) is desorbed in an elution circuit, regenerated in a kiln and returned to adsorption. Gold is recovered 6om strong eluates by electrowinning and smelting. Some 2.4 million tonnes of ore crushed to 80% passing 1.27 cm and grading approximately 2.7 @tonne gold were processed the first year of Newmont’s Carlin, Nevada (USA) operation; recoveries ranged fiom 5540% for the first three months of operation. Process enhancements are expected to improve fbture recoveries to a b u t 65%. Heap Leaching Chalcopyrite Ores: An Emerging Technology Interest in hydrometallurgical processing of chalcopyrite is increasing, because (1) constructing and operating smelters is less appealing due to capital cost and problems in complying with air pollution standards,(2) many chalcopyrite deposits also host penalty elements, particularly arsenic and bismuth, that smelters are reluctant to accept, and (3) worldwide there are vast resources of chalcopyrite ore that are too low-grade to concentrateand process by conventional routes. Bioheap leaching is particularly attractive for treating low-grade chalcopyrite ores, because of the relatively low cost, simplicity and ability to handle arsenic and bismuth containing minerals. The basics of heap leaching chalcopyrite ore are not well understood, however, redox control (6 15 - 645 mV SHE) and elevated temperature (270°C), as employed in stirred-tank leaching of chalcopyrite concentrates, seem to be key to the process. An aerated, run-of-mine heap leach trial with 960,000 tonnes of chalcopyrite ore grading 0.27% copper at Kennecdt’sBingham Canyon operation in Utah (USA) was considered a success. With just 18 months of operation, 28% of the contained copper was extracted fiom the test heap based on drill assay data. Temperatures in the heap exceeded 6OoC with greater copper extraction noted in areas of the highest temperature (Reamand Schlitt 1997 and 1997a). Chalcopyrite heap leaching is the subject of several patents and applications (Pinches et al. 2001) (Miller 2000). With successll demonstration trials, chalcopyrite heap leaching is very likely to be employed commerciallywithin the next few years. Heap Leaching Sulfide Concentrates The G E O C O A F process, developed by Geobiotics, Inc., agglomerates base- or precious metal concentrates, finely ground ore, or reground tailings onto coarse ore particles or inert aggregate using concentratedsulfuric acid or an acidic, iron-containingleach solution. The coated coarse ore or aggregate is then bioleached in a heap configuration like that described for secondary copper ores and sulfidic-rehctory gold ores (Whitlock, 1997; Johansspn et al. 1999). Heaps can be inoculated with mesophilic, moderately thermophilic or hyper-thermophilic microorganisms depending on the amount of heat generated fiom the suifide oxidation. The GEOCOAT“ process has been pilot tested at several locations, however, there are no commercial applications of the technology at this time. This process is protected by some 15 U.S. patents and numerous foreign patents and patent applications.

1564

SUMMARY Bioleaching of sulfidic-refractory precious metal and base metal concentrates and ores in tanks and heaps is commercially practiced around the world. Since commercial bioleaching practices began in the 1980s, much has been learned about the fundamentals of the technology, criteria for selection of the process, the design of tank and heap bioleach plants and the operation and performance of the plants. This chapter has explored the fundamental principles of bioleaching including the types of microorganisms used, their requirements and factors that affect their performance. Tanks and heap reactor designs were examined and the parameters that must be controlled for the microorganismswere considered. Different suites of microorganisms are used to match conditions anticipated in tank and heap reactors. This development permits greater flexibility in plant design. Microbial leaching processes can now be operated at temperatures approaching 95OC, and this development has led to new tank reactor designs that take into consideration the issues of mass transfer of gases, evaporative losses and materials of construction at elevated temperatures. Technologists are unraveling the mechanisms that microorganisms use to catalyze reactions, and these findings coupled with engineering developments in reactor design are leading to new and important commercial applications such as the tank and heap leaching of chalcopyrite concentrates and ore and the heap leaching of concentrates. The mining industry needs flexible, robust, cost effective, and environmentally acceptable process technologies that exhibit good metals extraction performance. Hydrometallurgical technologies top the list of preferred technologies to meet this need. Bioleaching technology and the engineering design innovations that have been coupled with this technology have m e a long way toward providing an acceptableprocessing alternative for the mining industry.

REFERENCES Batty, J.D. and T.A. Post. 1999. Bioleach reactor development and design. ALTA 2999 NickelKobalt Pressure Leaching & HydrometalluryForum. Melbourne: ALTA Metallurgical Services. BacTech. 2002. www.bactech.com. Basson, P., et al., International Patent Application WO OM8266 (15 March 2001). Bell, N and L. Quan. 1997. The application of BacTech (Australia) Ltd technology for processing refi.actory gold ores at Youanmi Gold Mine. Proc. of the International Biohydrometallurgv Symposium IBS97 BIOMNE 97. Chapter M2.3. Adelaide: Australian Mineral Foundation. Bennett, J.W. and A.I.M. Ritchie. 2002. A proposed technique for measuring in situ the oxidation rate in biooxidation and bioleach heaps Submitted for publication in Hydrometallurgy. Bhakta, P. and B. Arthur. 2001. Heap biooxidation and gold recovery at Newmont Mining Corporation. Presented at the 2001 Annual Meeting of the Society of Mining Engineers, Denver, Colorado. Billiton. 2002. BHP-Billiton Annual Report for 2001. www.bhpbilliton.com. BouEard, S.C. and D.G. Dixon. 2001. Investigative study into the hydrodynamics of heap leaching processes. Metallurgical and Materials TransactionsB 32B3763. Breed, A.W., C.J.N. Dempers, G.E. S&y, M.A. J d e r and G.S. Hansford. 2000. The bioleaching of sulfide minerals: developments in understanding the mechanism and kinetics of bioleachingpyrite, arsenopyriteand chalcopyrite.Proceedingsof the SMEAnnual Meeting, Salt Lake City, Preprint 00-120. Brierley, C.L. and J.A. Brierley. 1999. Bioheap processes - operation requirements and techniques. Copper Leaching, Solvent Extraction and Electrowinning Technologies,ed. G .W . Jergensen, 17-27. Littleton, Colorado: Society of Mining Engineers. Brierley, C.L. and A.P. Briggs. 1997. Minerals bimxidatiodbioleaching:a guide to developing a technically and economically viable process. Aper the Discovery: Proceedings of a Short Course. Prospectors and Developers Association, Toronto, Canada.

1565

Brierley, J.A. 1997. Heap leaching of gold bearing deposits, theory and operational description. Biomining: f i e o v , Microbes and Industrial Processes, ed. D.E. Rawlings, Chapter 5. Berlin: Springer-Verlag. Brierley, J.A. 2000. Expanding role of microbiology in metallurgical processes. Mining Engineering 52(11):49. Brierley, J.A. 2001. Response of microbial systems to thermal stress in biooxidation-heap pretreatment of refiactory gold ores. Biohy&ometallurgy: Fundamentals, Technology and Sustainable Development, Part A, ed. V.S.T. Ciminelli and 0. Garcia Jr., 23-3 1. Amsterdam: Elsevier. Brierley, J.A. and D.L. Hill, U.S. Patent No. 5,246,486 (21 September 1993). Brierley, J. A. and D. L. Hill, U.S. Patent No. 5,332,559 (26 July 1994). Brierley, J. A. and D. L. Hill, U.S. Patent No. 5,834,292 (10 November 1998). Briggs, A.P. and M. Millard. 1997. Cobalt recovery using bacterial leaching at the Kasese Project, Uganda. Proceedings of the International Biohydrometallurgy Symposium IBS 97 BIOMNE 97. Chapter M2.4. Adelaide: Australian Mineral Foundation. Brown, A., W. h e and P. Odd. 1994. Bioleaching - Wiluna operating experience. BZOMINE '94, Chapter 16. Adelaide: Australian Mineral Foundation. Craven, P. and P. Morales. 2000. Alliance Copper: the Billiton-CODELCO strategy for commercializing copper bioleaching. Randol Copper Hydromet Roundtable 2000, 119-126. Golden, Colorado: Rand01 International Ltd. Dew, D.W., D.M. Miller and P.C. van Aswegen. 1993. GENMIN's commercialization of the bacterial oxidation process for the treatment of refiactory gold concentrates. Randol Gold Forum Beaver Creek '93,229-237. Golden, Colorado: Rand01 International Ltd. Dew, D., H. Marais, P. van Aswegen, and C. Loayza. 1996. Bio-oxidation of gold and copper concentrates fiom Peru. Peru: Second International Gold Symposium,243-249. Lima: Comite Aurifero de la Sociedad Nacional de Mineria y Petroleo. Dew, D.W., E.N. Lawson and J.L. Broadhurst. 1997. The BIOX@ process for biooxidation of gold-bearing ores or concentrates. Biomining: %oty, Microbes and Industrz*alProcesses, ed. D.E. Rawlings, Chapter 3. Berlin: Springer-Verlag. Dew, D.W., C. van Buuren, K. McEwan and C. Bowker. 1997a. Bioleaching of base metal sulphide concentrates: a comparison of mesophile and thermophile bacterial cultures. Biohydrometallurgy and the Environment Toward the Mining of the 21"' Century, eds. R Amils and A. Ballester, 229-238, Amsterdam: Elsevier. Dew, D.W. et al., International Patent Application WO 01/18269 (15 March 2001). Dew, D.W., et al., International Patent Application WO OM8268 (15 March 2001a). Dew,D.W.andD.M.Miller,U.S.PatentNo.6245,125(12 June2001). d'Hughes, P., P. Cezac, T. Cabral, F. Battalglia, E.M. Truong-Meyer and D. Morin. 1997. Bioleaching of a cobaltiferous pyrite: a continuous laboratory-scale study at high solids concentration. Minerals Engineering 10507. d'Hughes, P., P. Cezac, F. Battaglia and D. Morin. 1999. Bioleaching of a cobaltiferrous pyrite at 20% Solids: a continuous laboratory-scale study. Biohydrometallurgy and the Environment Toward the Mining of the 21"' Centuty, eds. R. Amils and A. Ballester, 167-176, Amsterdam: Elsevier. $Hughes, P. D. Morin and S. Foucher. 2001. HIOX@ project: a bioleaching process for the treatment of chalcopyrite concentrates using extreme thermophiles. Biohydrometallurgy: Fundamentals, Technologyandsustainable Development, Part A, ed.V.S.T. Ciminelli and 0. Garcia Jr., 75-83. Amsterdam: Elsevier. do Carmo, O.A., M.V. Lima and RM.S. Guimaraes. 2001. BIOXO process - the Sao Bento experience. Biohydrometallurgy: Furdamentals, Technology and Sustainable Lkvelopment, Part A, eds. V.S.T. Ciminelli and 0. Garcia, Jr., 509-524. Amsterdam: Elsevier. Gericke, M. and A. Pinches. 1999. Bioleaching of copper sulphide concentrate using extreme thermophilic bacteria. Minerals Engineering 12:893.

1566

Harvey, P.I., J.D. Batty, D.W. Dew, W. Slabbert and C. van Buuren. 1999. Engineering considerations in bioleach reactor design. BIOMllvE ’99, 88-97, Adelaide: The Australian Mineral Foundation. Hiroyoshi, N., M. Hirota, T. Hirajima, and M. Tsunekawa. 1997. A case of ferrous sulfate addition enhancing chalcopyrite leaching. Hydrometallurgy 47:37. Hiroyoshi, N., H. Miki, T. Hirajima and M. Tsunekawa. 2000. A model for ferrous-promoted chalcopyrite leaching. Hydrometallurgy 57:3 1. Huerta, G., B. Escobar, J. Rubio and R Badilla-Ohlbaum. 1995. Short communication: adverse effect of surhce-active reagents on the bioleaching of pyrite and chalcopyrite by Thiobacillus ferrooxidans. WorldJournal of Microbiol. & Biotechnol. 1 1599. Johansson, C., V. Shrader, J. Suissa, K. Adutwum and W. Kohr. 1999. Use of the GEOCOATTM process for the recovery of copper fiom chalcopyrite. Biohydrometallurgy and the Environment toward the Mining of the 2Ist Century, eds. R. Amils and A. Ballester, 569-576. Amsterdam: Elsevier. Johnson, D.B. and F.F. Roberto. 1997. Heterotrophic acidophiles and their roles in the bioleaching of sulfide minerals. Biomining: Theory, Microbes and Industrial Processes, ed. D.E. Rawlings, Chapter 13. Berlin: Springer-Verlag. Lacey, D.T. and F. Lawson. 1970. Kinetics of the liquid-phase oxidation of acid ferrous sulhte by the bacterium Thiobacillusferrooxidans. Biotechnol. Bioeng. 12:29. Lawson, E.N., C.J. Nicholas and H. Pellat. 1995. The toxic effects of chloride ions on Thiobacillus ferrooxidanrr. Biohydrometallurgical Processing, ed. T. Vargas, C.A. Jerez, J.V. Wiertz and H. Toledo, 165-174. Santiago: University of Chile. Miller, P.C. 1997. The design and operating practice of bacterial oxidation plant using moderate thermophiles (the BacTech Process). Biomining: Theory, Microbes and Industrial Processes, ed. D.E. Rawlings, Chapter 4. Berlin: Springer-Verlag. Miller, P. 2000. International Patent Application WO 00/71763 A1 (30 November 2000). Miller, P.C., M.K. Rhodes, R Winby, A. Pinches and P.J. van Staden. 1999. Commercialization of bioleaching for base-metal extraction. Minerals & Metallurgical Processing 16:42. Nicholson, H.M., G.R Smith, RJ. Stewart, F.W. Kock and H.J. Marais. 1994. Design and commissioning of Ashanti’s Sansu BIOX@ plant. BIOMllvE ’94, Chapter 2. Adelaide: The Australian Mineral Foundation. Norris, P.R 1983. Iron and mineral oxidation with Leptospirillum-like bacteria. Recent Progress in Biohydrometallurgy, eds. G. Rossi and A.E. Torma, 83-96. Iglasias, Italy: Associazione Mmeraria Sarda. Norris, P.R 1997. Thermophiles and bioleaching. Biomining: Theory, Microbes and Industrial Processes, ed. D.E. Rawlings, Chapter 12. Berlin: Springer-Verlag. Norris, P.R, N.P Burton, N.A.M. Foulis. 2000. Acidophiles in bioreactor mineral processing. fitremophiles 4:71. Norton, A., International Application WO 01/18267 (15 March 2001). Peny, RH. and C.H. Chilton (4s.). 1979. Chemical Engineering Handbook, 5b ed., 6-16 (equation 6-22). New York: McGraw-Hill Book Company. Pinches, A., R Huberts, J.W. Neale and P. Dempsey. 1994. The MPNBACTM bacterial-oxidation process. IVth CMMI Congress, vol. 2,377-392. Johannesburg: S A I M M . Pinches, T., J.W. Neale, V. Deeplaul, P. Miller, M. Rhodes and B. Hancock. 2000. The Beaconsfield bacterial oxidation gold plant. Randol Gold & Silver Forum 2000, 169-175. Golden, Colorado: Rand01 International Ltd. Pinches, A., et al., U.S. Patent No. 6,277,341 BI (21 August 2001). Rawlings, D.E. 1997. Mesophilic, autotrophic bioleaching bacteria: description, physiology and role. Biomining: Theory, Microbes and Industrial Processes, ed. D.E. Rawlings, Chapter 1 1. Berlin: Springer-Verlag. Rawlings, D.E., H. Tributsch, and G.S. Hansford. 1999. Reasons why ‘Leptospiril1um’-like species rather than Thiobacillus ferrooxidans are the dominant iron-oxidizing bacteria in

1567

many commercial processes for the biooxidation of pyrite and related ores. Microbiology 145:5. Ream,B.P. and W.J. Schlitt. 1997. Kennecott's Bingham Canyon heap leach program- part 1: the test heap and SX-EW pilot plant. Proceedings of the ALTA 1997 Copper Hydrometallurgy Forum, 20-2 1. Melbourne: ALTA Metallurgical Services. Ream, B.P. and W.J. Schlitt. 1997a. Kennecott's Bingham Canyon heap leach program - part 2 the column leach testwork. Proceedings of the ALTA 1997 Copper Hydrometallurgy Forum, 46 pages. Melbourne: ALTA Metallurgical Services. Rhodes, M. and P.C. Miller. InternationalPatent Application No. WO 00/28099 (18 May 2000). Rhodes, M. and P.C. Miller. International Patent Application No. WO 00/29629 (15 May 2000a). Salomonde-Friedberg,H. 1998. Design aspects of aeration in heap leaching. Copper Hydromet Roundtable '98,243-247.Golden, Colorado: Randol International Ltd. Sand, W., T. Gehrke, R hall ma^, and A. Schippers. 1995. Sulfur chemistry, biofilm, and the (in)direct attack mechanisms - a critical evaluation of bacterial leaching. Appl. Microbiol. Biotechnol. 43 :961. Sand, W., T. Gehrke, P.-G. Jozsa, and A. Schippers. 1999. Direct versus indirect bioleaching. Biohydrometallurgy and the Environment Toward the Mining of the 21"' Century, eds. R. Amils and A. Ballester, 27-49, Amsterdam: Elsevier. Schippers, A., P.G. Jozsa and W. Sand. 19%. Sulfur chemistry in bacterial leaching of pyrite. Appl. Environ. Microbiol. 62:3424. Schippers, A. and W. Sand. 1999. Bacterial leaching of metal sulfides proceeds by two indirect mechanisms via thiosulfate or via polysulfides and sulfur.Appl. Environ. Microbiol. 65: 3 19. Schlitt, W.J. 1984. The role of solution management in heap and dump leaching. Au andAg Heap and Dump Leaching Practice, ed. J.B. Hiskey, 69-83. Littleton, Colorado: SME-AIME. Schlitt, W.J. 2002. Personal communication. Slabbert, W.,D. Dew, M. Godfky, D. Miller and P. Van Aswegen. 1992. Commissioning of a BIOX@ module at Sao Bent0 Mineracao. Randol Gold Forum Vancouver '92, 447-452. Golden, Colorado: Rand01 International Ltd. Third, K.A., R Cord-Ruwisch and H.R Watling. 2000. The role of iron-oxidizing bacteria in stimulationor inhibition of chalcopyritebioleaching. Hydrometallurgy 57:225. Timmins, M. and R.P. Hackl. 1998. Bacterial leaching of chalcopyriteconcentrates - prospects for a commercial process. New Dimensions in Hydrometallurgy Seminar, October 2, 1998, The University of British Columbia, Vancouver. Titan Resources. 2002. www.titanresources.com.au. Tunley, T.H., U.S. Patent No. 5,919,674 (6 July 1999). van Aswegen, P.C. 1993. Bio-oxidation of refkctory gold ores - the GENMIN experience. BIOMNE '93, Chapter 15. Adelaide: The Australian Mineral Foundation. van Aswegen, P.C. and H.J. Marais. 1999. Advances in the application of the BIOX@process for rehctory gold ores. Minerals & Metallurgical Processing 16:61. van Staden, P.J., M. Rhodes, A. Pinches and T.E. Martinez. 2000. Process engineering of base metal concentrate bioleaching. Randol Copper Hydromet Roundtable 2000, 127-129. Golden, Colorado: Randol International Ltd. Wtlock, J.L. 1997. Biooxidation of rehctory gold ores (the Geobiotics Process). Biomining: Theory, Microbes and Industrial Processes, ed.D.E. Rawlings, Chapter 6. Berlin: SpringerVerlag. Winby, R. et al. International Patent Application No. WO 00/23629 (27 April 2000).

1568

Copper Heap Leach Design and Practice Randolph E. Scheffel’

ABSTRACT Copper heap leaching has expanded continuously the last thirty years due to the commercial development of solvent extraction. To ensure both technical and financial success, a heap leach project must follow certain disciplines. These disciplines include: (1) a proper evaluation of the resource; (2) a comprehensive metallurgical test program; and (3) an engineered design and operating plan, all of which result in achieving expected production. The key to developing a successful leaching prospect is determining the actual “leachable” mineral content and then designing, conducting, and interpreting the metallurgical test program. The principal engineering requirement is the selection of the appropriate “scaled-up” leach curve from which to design the leaching area. Additionally, the actual leach design is often dictated by site-specific constraints, and each design requires a different “operating” copper inventory, both in solution and solids. Designing flexibility for increased leach area and volume of solution is critical. INTRODUCTION The objective of this section is to summarize the copper heap leaching development and experience of the last thirty years. Hopefully, guidelines taken from this experience are sufficient to ensure future operators continue the recent level of improving performance. After thirty years of continued improvement, the current status of heap leach design can be characterized more as “educated guess work” than “art” - however, there is still much to learn. One of the misconceptions of heap leaching, particularly with respect to copper, is that it is simple, straightforward,flexible and forgiving. Nothing is further from fact. Design engineers and owners anticipating such a development are well advised to seek all the counsel possible with people who have actual experience. Preferably this includes critical reviews of numerous and uniquely different operations, including other commodities. Further, consulting with investment bankers, and their third party engineers, often proves highly informative as to the cause for many of these projects under-performing initial financial projections.

HISTORY Interest in new technological development in the minerals business generally follows the price cycle of the commodity. Such is the case with heap leaching. While the first reported copper heap or dump leaching may have been at Rio Tinto, Spain, circa 1752, the first commercial modern-day style of heap leaching was probably introduced to the uranium industry in the 1950’s (Merritt 1971). Leaching of uncrushed, low-grade copper ore may have first been practiced in the US at Bisbee, AZ as early as 1923, and heap leaching was discussed by Irving in 1922 (Irving 1922). Vat leaching of finely crushed, oxide ore, combined with direct electrowinning from the impure solutions, was developed as early as 1916 at Chuquicamata in northern Chile (Eichrodt 1930, Rose 1916). This was followed by the New Cornelia Copper Company operation at Ajo, Arizona in 1917 (Tobelmann and Potter 1917). In 1925, Inspiration Consolidated Copper Company, Miami, AZ (Aldrich and Scott 1933) initiated vat leaching of mixed oxide, supergene ore. However, it was not until 1961 that the modern-day practice of heap leaching copper began its development in a significant way. It is believed the Stovall Copper Company (Miller 1967) was the fist to use this process as the primary means of development at the Bluebird Mine near Miami, Arizona. At that time, copper cementation on iron was used to recover the copper, which subsequently required smelting. 1 Consulting Metallurgical Engineer

1571

Gold heap leaching, having benefited from this early work in copper, was initially proven with the Cortez and Bootstrap operations in 1973 and 1974, respectively (McQuiston and Shoemaker 1981). Economic forces helped to continue its growth through the 1980s, contributing significantly to the pool of experience and operating history for later application to copper heap leaching. However, cross-pollination of these two industries was really quite limited during this period, with each separately developing their expertise. In the late 1980s, spurred by improving prices, copper again enjoyed a major expansion in the utilization of heap leaching oxide and secondary copper minerals. However, at the end of the 20th century, the hydrometallurgical treatment of primary copper, e.g. chalcopyrite, was still relegated to dreams of future commercial development. There is no question one of the most significant developments in the copper industry in the last thirty years was the development of the ketoxime, and eventually the aldoxime, chelating extraction reagents. The ketoximes were initially developed by General Mills (House 1981 and Kordosky 1994, 1999) and commercially proven at the Bluebird Mine near Miami, AZ by Ranchers Exploration and Development Corporation (Power 1970) in 1968. The Bluebird was closely followed by Bagdad Copper Corporation’s operation at Bagdad, AZ in 1970, the ammonical, scrap copper leach operation of the Capital Wire and Cable Corporation operation near Casa Grande, AZ in 1970 and ZCCM Chingola in Zambia in 1974. The most rapid expansion of heap leach, SX/EW copper production occurred in Chile during the decade of the 1990s. Both oxide and supergene deposits were developed using fine crushing, generally less than 12-16 mm, acid agglomeration and subsequent acid or biological leaching in heaps primarily constructed via conveyor stacking 6-8 m in depth. The precursor to this development was the “thin layer leaching” concept fust patented (Johnson 1975) by Holmes and Narver Inc. (H&N) for application to uranium and copper (E&MJ 1978 and Mining Magazine 1978). This process was first commercialized by Sociedad Minera Pudahuel (SMP) near Santiago, Chile in late 1980 (Domic 1981 and 1983). The Chilean rights to the H&N patent were subsequently assigned to SMP. Since this was a mixed oxidelsulfide ore, the natural progression of activities over the years resulted in SMP further developing its expertise in biological leaching. As a consequence, Cominco and Rio Algom elected to solicit SMP’s assistance in the development of Quebrada Blanca and Cerro Colorado using a similar process, but stacking the ore to 6-8 m in depth. SWEW, heap leach copper production in Chile expanded from 20,000 tpa in the 1980s to over 1,000,000 tpa by the end of the 1990s. Absent the ability to purify copper from dilute, impure sulfate leach solutions, it would have been impossible to supply the market with the quality copper required by the downstream suppliers of refiied copper products, without further expensive processing. Therefore, the development of the extraction reagents and their continued improvement were the primary reason for the growth in copper heap leaching. Today, SXEW accounts for nearly 20% of worldwide refined copper sales. A further significant improvement in the SX/EW process was the development of copper plating on stainless steel cathode blanks. This was first practiced by Capital Wire and Cable Corporation in the early 1970s and was fully developed into an automated plating and stripping process in 1976 by Mount ISA Mines, in Queensland, Australia.

So successful have been the reagents and subsequent electrowinning practice, that it is a rare occurrence when an SWEW facility does not routinely produce LME Grade A copper, or better. DISCIPLINES FOR THE APPLICATION OF COPPER HEAP LEACHING It is paramount an engineer, developing a heap leach project, understands this unit process requires an inter-disciplinary approach between geologists, mining engineers, and metallurgists. These projects can not be developed successfully with each working without regard for the others. This begins with the initial drilling and the subsequent geologic and mineralogical interpretation. The

1572

metallurgist must understand the geologic and lithologic constraints imposed by the resource. The geologist must understand the inefficiencies and implication of variable ore characteristics on the metallurgical performance of the heaps. The mining engineer must understand that moving tomes at the least possible cost is not always the most profitable approach when total leaching performance is addressed. The metallurgist often considers blending of different rock types to deal with ores displaying poor characteristics, such as highly clay-altered ore. However, open-pit mining generally occurs in distinct levels, and there is only limited ability, without significant stockpiling and re-handling, to conduct such blending. The metallurgist must appreciate these limitations. The author’s opinions and experience in these respects have been presented at the Randol Copper Hydromet Round Tables (Scheffel 1999 and 2000). Dicinoski, Schlitt and Ambalavaner (1998) also discuss the critical elements of developing a copper heap leach, SWEW project. Some of the key required disciplines are: Resource Evaluation Test Work Program Engineering Design Postmortem Analysis Each of these is discussed in detail below.

Resource Evaluation The most critical, and often controversial, aspect of developing a successful heap leach resides with one’s understanding of the resource. One perception of heap leaching is that it is a low-capital approach that can be developed in the shortest possible time in relation to other options. The other options are actually quite limited, e.g., agitation leaching of copper oxides, and most hkely, flotation followed by smelting for copper sulfides. There are very distinct differences in copper mineralogy which make process selection specific to either acid leaching of oxide minerals, femc leaching of secondary sulfide (supergene) minerals, and until a hydrometallurgical process is proven commercially, flotation and smelting of primary copper sulfides. Most likely, leaching becomes the primary option for mixed oxidelsupergene deposits and for medium grade supergene deposits. Therefore, when deciding how best to drill and evaluate a deposit, it is imperative that an early assessment be made of the probability heap leaching will be applicable. This is best accomplished by routinely conducting diagnostic assays, or extensive mineralogy, on the initial drill cuttings or core. This provides an early indication of the potential copper solubility and the type of mineralogy being drilled. If it appears the ore is potentially heap leachable, it will be necessary to consider substantial amounts of diamond core drilling, as opposed to less expensive reverse circulation (RC) drilling. It is critical the physical character, e.g., fracturing and alteration, of the ore be fully appreciated and modeled in the geologic rock model. This can only be accomplished with extensive core logging. Further, the diamond drill coring, in size, depth, and area, should take into consideration obtaining the broadest possible sampling of the resource for metallurgical testing, in addition to resource definition. Under-capitalized companies that develop such projects with less stringent methods often find themselves in trouble. Conversely, well-capitalized companies can spend significant funds for the wrong reasons. For example, spending considerable funds in driving adits to collect large bulk samples for large-scale pilot heap or column testing can be misleading, as these are only “grab” samples, which are most often non-representative. Instead, it is often more cost effective and beneficial to spend exploration dollars on completing significant,

1573

large-diameter core sampling of the deposit, e.g., >50 mm diameter. This ensures all the different lithologies and alteration types are sampled and fully logged. Initial testing can then be conducted on individual rock types versus depth. It is the author’s observation that small columns, with diameters 4 to 6 times the largest particle size, will most often perform like 1 to 2 m square cribs or 1 m diameter column tests. Unfortunately, these will all have relationships to a commercial application that will be quite different, for reasons discussed below. Besides the physical character of the ore, there are two additional items critical to resource evaluation -- “soluble” mineralogy and acid consumption. “Soluble” in this context means the copper minerals that can be expected to dissolve in the acid and ferric ion environment actually achievable in the heap leach system. This leads one to do considerable time-consuming and costly mineralogy on individual drill hole intervals. Another option is to develop “diagnostic” analytical techniques that allow one to estimate the “soluble” copper content quickly and economically. Recent advances in scanning electron microscopy, combined with computer software, may eventually result in economic and timely quantitative mineralogical analysis (Gottlieb et al. 2000). Once the “soluble” copper estimate is determined, it can later be compared to column tests at various crush sizes to determine the general level of recovery of that “soluble” content. It is critical this parameter be followed for prospective development of both oxide and secondary sulfide leach projects. Oxide deposits can contain significant quantities of non-acid soluble copper oxides and other refractory compounds, which can reduce actual leachable content to <50%. Some oxide deposits can actually be poorer performers than secondary sulfide deposits for this reason. Secondary copper deposits can be highly irregular with depth, with some sections being quite thin. Hypogene minerals such as chalcopyrite or enargite can sometimes intrude into the secondary mineralization. Also, secondary deposits can contain a complex mix of oxide and secondary transitional type minerals, including native copper. If enargite is present, this must be identified early in the resource evaluation as its low solubility in acid-ferric medium can be problematic. A diagnostic method, which has shown reasonably broad application in this regard, is called a “sequential” assay. This was used by Inspiration in the early 1980s at its ferric-cure leach operation near Globe, AZ. This technique was further refined and reported by Parkison and Bhappu (1995). This method uses a 20°C (or heated) 5% sulfuric acid digestion for one hour followed by a 30 minute cyanide digestion at 2OoC (or heated) on the acid leached and washed residue. This cyanide leach residue is then digested with aqua regia or a standard four-acid digestion for the remaining insoluble copper. The sum of the three digestions represents a “calculated” copper assay to compare directly against a standard “total” copper assay. The latter is necessary as it represents the only assay that can be reproduced with appropriate accuracy to be used for third party quality control programs required by financiers. In the “sequential” assay, the acid is believed to solubilize the non-sulfide minerals except for native copper and about half of any Cu20. The cyanide is assumed to solubilize the native copper and the secondary sulfide minerals, as well as bornite. The remaining residue is assumed to be the refractory oxide minerals or chalcopyrite.

It is important to understand that diagnostic methods, as well as mineralogy, are only semi-quantitative and strict conformance to procedure between laboratories is paramount, or reproducibility will be impossible. Other diagnostic methods are discussed by Hiskey (1997). It must further be emphasized that diagnostic assays can not be used without extensive initial

1574

mineralogy to substantiate the actual mineral assemblage of the specific deposit. Some copper minerals are soluble in cyanide and not in acid-ferric media and vice versa, e.g., enargite is nearly totally soluble in cyanide and only marginally soluble with ferric iron.

A second analytical parameter important to the initial resource evaluation is to develop an early estimate of the gangue acid consumption. This requires empirical methods and is subject to interpretation. However, it is critical this parameter not be overlooked in early resource evaluation. A method that has some acceptance in this regard is a 24-hour bottle roll test at constant pH of 1.5 to 1.8 and 33% solids. This should be conducted on the assay pulp or the 10 mesh crushed ore rejects from the analytical sample preparation. Bottle roll tests on coarse material, approaching the expected crush size, are too variable due to the degradation that occurs during the test. Some have developed static testing methods to deal with this problem (Fountain 2002). The acid consumption indicated from the fine material will be much higher than will occur in actual practice. Therefore, experience is necessary in relating the results to commercially expected consumption. The actual consumption at coarser particle size, under field conditions, can be only 20-35% of the results obtained on pulp. However, potentially excessive acid consumption will clearly be indicated by this test. Conversely, some ore types can actually show very little acid consumption, which can represent a different problem. If the gangue acid consumption does not exceed the fresh acid addition, and that produced by bacterial activity in the case of sulfide leaching, acid can build in the heaps to the point it interferes with copper recovery in SX. In this situation, the fresh acid addition must be reduced or neutralization of the excess acid may be necessary.

The commercial economic limit for acid consumption is highly variable, depending on other factors such as ore grade, stripping ratio, etc. However, this consumption generally falls within 5-50 kg/t ore. A more reliable means of expressing the acid consumption of a particular project is the “specific” net acid consumption expressed as kg acid per kg of copper recovered. The generally experienced commercial acid consumption expressed in this fashion ranges from 1-7 kg acidkg copper recovered.

Test Work Program Ideally, a properly designed test work program should be developed only after the resource is reasonably well understood. Unfortunately, timing is generally such that the metallurgist must make certain decisions and proceed with a concurrent program. Therefore, initial test work should test samples of individual lithologic character. Once the final details of the rock model are complete, optimized column testing can be conducted on specific composites. These composites should include the major rock types within the deposit with regard to their respective distribution, both in cross-sectional area and with depth. Once the resource evaluation suggests the ore can potentially be heap leached, the principal objectives of the test program should be to determine: the general level of extraction versus lighologic unit, depth, and ore grade; the recovery versus crush size; the potential production of f i e s with crushing; the degradation character of the gangue upon leaching; the acid consumption, with particle size affect; the key solution equilibrium chemistry; and the practical commercial heap height.

1575

Unfortunately, no single column test, or series of column tests, can give all the answers. Therefore, a significant amount of judgment is necessary when interpreting exactly what the collective package of data provides. Scheffel and Reid (1997) and Kaczmarek et al. (1999) describe the process used in two instances to develop and interpret such test work. Specific to developing the “Test Work Program”, the following items are considered important: Sample Collection Types of Minerals Sample Preparation Types of Testing Heap Height Interpretation and Scale-up

Sample Collection. The collection of a proper suite of “representative” samples to be tested is paramount to developing a successful heap leach. To the extent this is not done, the only mitigating option is to be extremely flexible with assumptions on design pregnant solution (PLS) grade, time, and leach pad area. Modem drilling and sampling techniques, including RC drilling, generally give sufficient statistically valid drill results to assess the “total” and “soluble” copper content and allow adequate ore grade modeling. However, only core drilling can provide the information necessary to develop the rock-type model required for developing a low-risk heap leach design. To the extent a company compromises the resource evaluation phase in this regard, the greater the risk of under-performance, or even failure, of the subsequent heap leach design and operation, As mentioned above, the driving of adits into multi-million tonne deposits to acquire large bulk samples can be a near-fatal flaw in many large projects if these are the primary samples tested. These samples are only grab samples, which typically do not represent the actual distribution within the resource.

Types of Minerals. A proper understanding of mineralogy is critical to designing a test program. The metallurgical response, both in extraction time and acid consumption, can be drastically different for oxide minerals as opposed to secondary, transitional, and primary sulfide minerals. Oxides. Oxide minerals, or copper minerals formed in an oxidizing environment, are far-ranging and exhibit a wide range of metallurgical performance (Baum 1996 and 1999). Some oxide minerals, such as malachite ( CU ~ CO ~ (OH)~ and ) azurite (CU~(CO~)~(OH),), can leach extremely fast. However, in practice their rate of dissolution is controlled not by kinetics but by the rate acid can actually be made available from the processing system. Chrysocolla is the next most rapidly dissolved, but can also have a limitation caused by diffusion of hydrogen ion in, or copper ion out, through a non-protective silica layer (Pohlman and Olson 1976). However, for most oxide minerals, the commercial leaching time frame in actual practice can be a factor of 2 or 3 times faster than for secondary sulfides. Some oxide minerals, such as copper in iron compounds and “wad” (neotocite), a manganiferrous copper compound, can have dissolution rates approaching that of secondary minerals. Therefore, these minerals may not reach their ultimate economic

1576

extraction if the heap leach design considers only the fast-leaching copper oxide minerals. Manganiferrous copper compounds actually benefit from reducing conditions, which are not necessarily consistent with the natural leach solution. One of the distinguishing features between gold and copper heap leaching can be the time frame. While many gold heap leaches of 6-15 meter depth achieve their target recovery in 60-150 days, many copper oxide leaches typically require 150-300 days to attain a similar level of extraction of the “soluble” component. Supergene ore can require 300-900 days depending on ore grade. This is generally due to the impact of higher natural fmes (clay) content, degradation of the ore, and higher levels of reagent required in dissolving and displacing more mass. Oxide deposits, by the very nature of their geologic origin, can be more highly altered than supergene deposits. This leads to clay, or fines, which can significantly impact the heterogeneity of the heaps and make scale-up from laboratory column testing even more difficult. (This is not to imply that supergene deposits do not also exhibit significant alteration, as many do.) As a result of the above, oxide deposits can be more problematic than secondary sulfide deposits -- a condition not always fully appreciated.

Sulfides. Sulfide copper mineralogy, especially the secondary copper minerals that lend themselves to possible heap leaching, is very complex. A full range of oxidation products from primary copper minerals such as chalcopyrite and enargite, or naturally occurring secondary minerals, create a complex mineral assemblage. Each mineral has a unique level of potential solubility, with its dissolution kinetics being limited by a multitude of reaction products (Bradley, Sohn and McCarter 1992). Unlike oxide mineral leaching, the dissolution of sulfide minerals is electrochemical in nature, requiring the presence of ferric iron, generally in a sulfate medium. This chemistry can be quite complex. It requires the catalytic oxidizing behavior of aerobic Acidithiobacillus ferrooxidans bacteria at temperatures below 45°C and archaea, or other high temperature or extreme temperature thermophiles, when the temperature reaches 40”-70°C (Rawlings 1997 and Norris 1997). While these reactions are electrochemical in nature, they still benefit from increased temperature, but to varying degree, depending on the mineral. Sulfide minerals also exhibit surface rest potential and can corrode by galvanic action like dissimilar metals (Hiskey and Wadsworth 1981). In commercial secondary copper leaching operations, chalcocite (CuZS) is typically the most abundant mineral. It is generally accepted this mineral leaches in two stages. The first copper is nearly totally removed, leaving behind “synthetic” covellite (CuS) called “blaubleiblender” I or I1 (Hiskey and Wadsworth 1981). The first stage leach exhibits rapid leach kinetics in the presence of sufficient ferric iron. It also has a lesser dependence on temperature, provided there is sufficient temperature to satisfy the bacterial activity necessary to oxidize ferrous iron to ferric under natural biological leaching conditions. Recent commercial experience suggests a threshold temperature near 18°C. Ferric leaching kinetics of the “synthetic” covellite, however, are much slower and exhibit greater dependence on temperature. Naturally occurring covellite and hypogene covellite exhibit even slower leaching kinetics than the “synthetic” form. Therefore, the leaching of chalcocite is characterized by fairly rapid dissolution of 50%60% of the copper under conditions of good bacterial activity (i.e., rapid ferrous to ferric

1577

oxidation), followed by a longer leach time to dissolve the resulting “synthetic” covellite. Therefore, in order to achieve a high recovery from chalcocite, it is necessary to provide the time required to leach the covellite. For any given pad height, as the chalcocite ore grade increases, the amount of covellite requiring dissolution increases, which generally requires a progressively longer leach cycle, provided one wishes to achieve the same level of extraction. It must be appreciated, for the iron chemistry to work as efficiently as possible, both acid and oxygen must be present in addition to maintaining the appropriate temperature. The benefit and need for forced aeration for sulfide leaching is summarized by Schlitt (2000). The author’s experience suggests this is most important for ores containing >0.5% copper as “leachable” sulfides. One must be careful in conducting column tests that are open to atmosphere and concluding that air may not be required. A column can act as a natural chimney, with thermal drafting providing sufficient oxygen to leach +2% supergene copper at a significant rate, which is not possible in actual practice without forced aeration.

Commercial experience with forced aeration has shown, in most cases, about a 3%8% increase in extraction over a 300-500 day leach cycle for chalcocite ore grades <1.2% total copper (CUT) when compared to no forced aeration. In cases of poorer permeability, it might even be of greater benefit. Forced aeration can have even a greater impact on the level of recovery reached in the first 100-150 days during the highly ferric dependent leach stage. Girilambone Copper Company documented a significant increase in extraction on high-grade chalcocite ore (e.g., >1.5% CUT), due, not only to the use of forced aeration, but other factors favoring solution application and fines control (Walsh et al. 1997; Dudley et al. 2000). Dixon addresses some of the heat conservation considerations of sulfide leaching utilizing forced aeration (Dixon 2000). The leaching kinetics of secondary minerals are generally such that leach cycles of 300-900 days are required on a single-lift basis to achieve the economic limit of recovery, depending on ore grade. This has a profound impact on topography and leach pad design issues. Therefore, a comprehensive test program should compare column tests of low and high ‘‘soluble’’ copper content to demonstrate the need for, and impact of, oxygen and time relative to ore grade. Sample Preparation. The preparation of individual rock-type composites, which should initially be tested separately, can be critical to the interpretation and overall results of a copper leaching test program. Testing for potential recovery difference of each rock type with depth in the deposit can also be important. A comprehensive column test program generally requires parallel or duplicate column tests on identical samples to compare changes in certain process parameters. Some of these parameters include: different acid pre-treatment methods; varied solution application rates; crush size; rock type; etc. Duplicate column testing of this type requires the samples to be as close in copper grade and particle size distribution as possible. Often, the recovery difference in “duplicate” columns, where composite samples were collected from a larger master composite by conventional cone and quartering, or riffle splitting, can be >5%. In order to make valid judgment as to whether recovery differences in two comparative columns are statistically significant, the metallurgical balances for individual tests should be within 2%-3 %. This is very difficult, if not impossible, to accomplish without a strict regimen of sample preparation (Keane 1998). The preferred procedure is to first screen the “master” composite into five or more sub-fractions. Then one can split each size fraction separately, collecting a weight for each fraction, so the “master” composite’s natural size distribution is properly replicated. These separate weight fractions are then recombined into the test composite. A separate sample is prepared in the same manner for headscreen assay analysis for the column(s) charge.

1578

It is during this strict sample preparation procedure when opinions and judgment can be made regarding the distribution of copper by size fraction, which can lead to indications of the potential crush size dependency on recovery. Thls depends to a great extent, however, on whether the copper minerals are on fractures or disseminated in the rock mass. A further clue to potential leaching success will be provided with the amount of natural fines occurring at finer crush sizes.

Types of Testing. The organization of a typical test work program is addressed elsewhere in greater detail (Scheffel and Reid 1997; Iasillo and Schlitt 1999). However, a typical recommended approach may be as follows: First conduct mineralogy and diagnostic tests to determine the “soluble” copper content and acid consumption tests on assay pulp to characterize the ore by lithology and with depth and cross-sectional area (a result of the resource evaluation); Conduct bottle roll tests for preliminary characterization of readily soluble species and acid consumption at coarser size (possibly static tests), in preparation for; Mini-column (1.5-2 m) tests in open-cycle to test acid pretreatment options and general solubility at a couple of crush sizes; Then conduct larger diameter, commercial depth column tests in “closed-cycle’’ with solvent extraction to better identify the crush size effect, ore grade versus recovery relationships, the acid pre-treatment scheme, pH control, iron chemistry, impurity buildup, and better overall estimate of acid consumption; and finally Large-scale pilot testing, if deemed necessary. The acid consumption testing and pretreatment schemes rely on “empirical” tests developed by individual laboratories over time. The column test procedures have significant potential for error in actual practice. Therefore, it is critical the laboratory chosen to do this work has a long history of conducting such tests with copper. If a less experienced laboratory is chosen, the ownerlengineer is well advised to have a consultant experienced with this type of testing advise the laboratory and oversee the overall program. The author, over several projects and test programs, has developed the following opinions concerning the testing program and the items that are critical to the design of a successful copper heap leach facility.

“Soluble” Copper and Iron Content and “Net” Gangue Acid Consumption. The initial analytical data package requested on drill hole intervals is critical in defining not only the “total” copper resource, but also the resource’s potential to be heap leached. Obtaining an early understanding of the ‘‘soluble’’ mineral content, the gangue “net” acid consuming character and the acid soluble iron content are parameters of paramount importance to a viable heap leachable resource. These parameters are also required for a workable rock-type model, assuming heap leaching is the preferred approach. Bottle Roll Tests. In the author’s opinion, bottle roll tests at coarse crush size, e.g., >12 mm, have little utility in the overall picture. They are only indicative of the readily acid and, in the case of sulfide deposits, ferric soluble content. In addition, they are generally only good for finely crushed ore, as the term of the bottle roll test is generally much shorter than that required under heap leach conditions. The acid consumption results are still only preliminary and probably not much better than those obtained from the assay regime. Those experienced with both these “empirical” acid consumption tests

1579

have developed methods and judgment as to what the commercial acid consumption might be. (The pH 1.5 or 1.8 acid consumption tests on assay pulp will be 2 to 4 times higher than actually experienced in some heaps.) Another problem with bottle roll tests at coarse particle size is the degradation that can take place, due to attrition, compromises the results.

Mini-column Tests ( 4 - 2 m in depth). These tests are generally used primarily for scoping acid pretreatment techniques (e.g., testing acid cure versus no acid cure, and then only by one experienced in interpretation of the results) and developing a feel for crush size effects. They have the advantage of giving this information more quickly and at less expense than tall column tests. Full Commercial Depth Tests. All testing to develop a successful heap leach should incorporate closed-cycle column testing approaching full commercial depth. Unlike gold, the solution chemistry in copper leaching changes dramatically with ore depth. The relationship of pH, free acid, iron chemistry, and Eh can only be fully appreciated by observing the full depth results with equilibrium concentrations of total dissolved solids (TDS). Generally, the rate controlling reagent concentration limitations are not realized, and therefore not appreciated, if testing stops with the top 1-2 m of a planned 6-8 m commercial depth. If one wishes to observe the actual chemistry with depth, individual column segments can be leached in series. This allows for solution collection and assay at chosen depths, or solution can simply be removed by selective sampling. This can be particularly interesting if one anticipates run-of-mine (ROM) leaching, where acidification in agglomeration is not practical and placed ore depth is generally greater. Also, there is better history in scaling up from column tests to field heaps when the heights of each are equal. Such scale-up is the most critical objective of the metallurgical development program. There is sufficient experience within certain analytical laboratories and consulting groups to use empirical acid consumption and agglomeration tests to adequately hit a proper target of acid addition and go straight to tall columns after minimal preliminary testing. It is preferable to test more commercial depth columns across a broad range of rock types, depths, and reagent doses than to spend this money on preliminary testing, which may have limited utility to the final result. With respect to acid consumption and leach kinetics, it is best to conduct column tests in closed-cycle with solvent extraction of the copper and recycle of the raffiate. It is also important to simulate the expected equilibrium composition of impurities which occurs in actual practice. This can be accomplished by simple digestion tests and chemical analysis of the constituents that go into solution or by using the leach solution from a similar mine. One of the more difficult impurity levels to replicate is the iron chemistry, both total iron content and the ratio of ferric to ferrous iron. Generally, column tests ultimately reach higher levels of EMF (ferric to ferrous ratio) than is experienced in practice, at least with finely crushed, high-grade ore. This may be due to lower gaseous porosity in field constructed heaps. It must also be appreciated in batch-wise column testing the reductant, e.g., chalcocite, is not replenished as it is being dissolved. In operations, there is always fresh ore being placed under leach.

Larger Scale Pilot Heap Tests. There may be reasons, or circumstances, that warrant constructing large-scale pilot heaps of 5,000-50,000 tomes. However, one should

1580

understand the limitations of such tests. They do not always provide any better information than a properly planned column test program, due to problems such as: sample representativeness; inability to construct the heap in the same manner as a commercial heap; better attention to solution application; better oxygenation; and greater attention to details. Some advantages of large-scale pilot testing may include: operator training; testing under actual climatic conditions; ROM testing requires large tests; testing under actual conditions of ore placement method(s); and potentially, better “bankability” for large projects. For on-going operations, large-scale test heaps can be much more timely and cost effective for testing alternative operating conditions, taking into consideration all the compromises discussed above. However, given the significant expenditure required for such large-scale tests, if no commercial operation is available, this money is likely better spent in providing more detail to the column test program. This detail should include more extensive resource sampling, allowing a larger number of variables to be tested.

Heap Height. What is the proper heap height? There is no easy answer to this question. For a single lift of ore, it is best answered by assessing the implications for a given material with respect to its mineralogy, available reagent, and hydraulic character. It is believed, for example, that supergene material, which is force-aerated and relies primarily on ferric generation by bacterial activity, may be leached at greater depths than oxide material, where acid can not be generated in situ at depth. Commercial supergene operations have successfully leached single lifts of ore at depths of 10-12 m. Multiple-lift leach designs generally leach in 5-10 m lifts stacked to greater than 45 m in some instances. However, there are compensating factors to be appreciated in all cases, as discussed below. The uniformity of particle size distribution, which crushing and acid agglomeration provides, can extend the workable depth by improving the percolation character. If the reagent can be provided in situ, e.g., bacterial oxidation of pyrite and ferrous iron with the use of forced aeration, combined with good hydraulic behavior, a greater depth can be tolerated. High-grade oxide (2%-4%) copper, especially for malachite and azurite, may need to be leached in shallower lifts to control the pH to maintain the copper in solution. These minerals consume acid at a significant rate, nearly depleting all free acid. Excessive fines content can be another factor limiting the depth that can actually be leached, regardless of the reagent availability, due to poor solution flow characteristics. The commercially viable depth for crushed ore heaps appears to be between 2-12 m depending on the trade-offs mentioned. Both application rate and ore depth can have profound effects on the resulting PLS grade, especially if the system is not reagent limited.

1581

One key to understanding the benefits, or disadvantages, of different heap heights is to appreciate the actual field extraction rate and level of extraction when considering the longer residence time allowed by placing ore at a greater depth. Inefficiencies in wetting and reagent availability, which may come with greater depth, may not be offset by the longer leach time. Interpretation and Scale-up. As mentioned above, it is nearly impossible to combine and incorporate into any single column test the equilibrium conditions that will be achieved in the field. Additionally, changes in ambient conditions, such as temperature and altitude, can not be attained unless the test work is conducted on site. And even then, the ambient conditions within a column are not what are experienced under actual leach conditions.

Many have attempted to mathematically “model” the heap leach process with little success. The knowledge of the chemistry included in these models is generally excellent, but the models fail primarily because they do not properly address the various heterogeneity factors, which are specific to each application. However, empirical models based on actual experience have generally proven to be a reasonable planning tool. It is critical one understand the typical inefficiencies that exist between actual field heaps and columns. It is the author’s experience that small diameter columns perfom very close to larger diameter column or “crib’ tests - both of which perform better than actual field experience. Figures 1 through 4 show typical laboratory column test results compared to actual field experience for gold, uranium, oxide copper and sulfide copper. urm1urn

Fig. 1 Typical Uranium Recovery

Qold

Fig. 2 Typical Gold Recovery

Chalcocita

Fig. 3 Typical Chalcocite Recovery

Fig. 4 Typical Cu Oxide Recovery

1582

These inefficiencies are typical for single-lift leaching. Additional inefficiency, or inventory implications, must be added to this when leaching through multiple ore lifts, as discussed below. Most of the inefficiencies are related to limitations of reagent in contact with minerals. This can be related to the reagent strength in contact with the ore, simply nonwetting, or long diffusion paths due to the non-homogeneity of solution flow through unsaturated medium (O’Kane 1999 and Lupo 2000). The ability to properly design a heap leach scheme with sufficient leach area and flow rate is solely dependent on estimating a proper field extraction rate from the column testing results. In practice, a heap leach is always working against negative factors, all of which are cumulative in their impact on overall results. Therefore, a successful design must anticipate a full range of inefficiencies and uniformity factors and provide sufficient time to deal with this variability. As illustrated in Figures 1 through 4, operating experience for four different mineral types show similar inefficiencies. This suggests there are limitations with respect to the physical conditions and uniformity of actual heap parameters, regardless of the mineral. Some of the parameters where striving for more uniformity can be beneficial are:

ore grade; acid consuming character; solution application, especially with differing ore depths; near surface distribution and wetting; uniform permeability; uniform acid addition, via agglomeration, if crush size allows; and particle size. Figures 5a and 5b indicate the author’s recent experience with the range of recovery that can be expected for commercial leaching of 6-8 m heaps for oxide and supergene ore of about 1% “soluble” copper content. 4% Acid Soluble Cu I d m or*dopfh. 12-iBnm.AddCurnd

Fig. 5a Typ. 1% Ac. Sol. Cu Rec. Rate

Fig. 5b Typ. 1% Fef3 Sol. Cu Rec. Rate

The upper and lower curves bound where the author believes 90% of commercial operations will fall, depending on the fines content or other parameters that limit wetting or reagent availability, e.g, acid availability for oxides and oxygen restriction due to poor air permeability in the case of supergene leaching. The upper curve is the most optimistic one should expect if the ore is free draining and the fiies content is 4 0 wt% -150 micron (100 M) and <5% -74 micron (200 M). The lower curve is what the metallurgist hopes to avoid if at all possible. However, if fiies or degradation can not be prevented, or

1583

removed, the lower limit must be analyzed relative to its impact on potential capital and operating costs and recovery. Recent commercial experience, employing conventional heap stacking methods, suggests that for each 2 wt% increase in -150 micron sized particles, between 8-20 wt%, the recovery time to achieve a similar terminal extraction can be extended by 20%-30%. Experience at Chuquicamata on the ripios (vat tailing) re-leach project shows the importance of understanding the relationship of particle size and application rate to ensure non-saturated flow (Chahbandour et al. 2000 and Guzman et al. 2000). It is with this area of hydraulic understanding where the greatest improvements in heap leaching may eventually be developed.

Ramp-up to Full Production. A critical aspect of the final field extraction curve is the amount of ore that must be on the pads under leach to reach full production. This is not always fully appreciated, but it is very simple. If little ore is on the leach pad at the start of the SWEW commissioning, full production will not occur on a daily basis under equilibrium operating conditions until the full ore inventory is placed under leach. For example, if the recovery curve to 90% of the “soluble” copper content is the target, and the scale-up curve suggests this will take 500 days, then there must be 500 days of ore under leach. Until this happens, production will be less than initially projected. If one wishes to start-up at full production, once the SX/EW plant is commissioned, and then maintain that rate of production, there must be 30%-40% of the 500 day inventory of ore on the pad before any leaching begins. This issue has not been fully appreciated in the past, and therefore, full production took longer to achieve than thought. If the actual field leach curve is not what was initially projected, then this problem is further exacerbated.

Engineering Design With the support of a comprehensive test program, one can begin to consider the primary engineering design aspects of the heap leach. Some of the principal areas of engineering discipline requiring attention are: Conceptual Leach Design Pad and Liner Design Heap Construction Method Solution Application Method SXIEW Design Engineering and Project Implementation Closure Issues While all of the above disciplines must be considered in conjunction with the other, they are discussed separately below.

Conceptual Leach Design. Once one appreciates the implications of a proper scale-up from a comprehensive test program, the mass balance becomes clear and the options for the various means of stacking and leaching ore can be ascertained. It is important to emphasize that most heap leach projects are only as successful as the initial design is correct. It is very difficult to correct inadequate recovery response if increased leach pad area can not be readily constructed due to constraints imposed by topography or poor pad geometry. Also, one must make sure there is a long-term mine

1584

plan available, which supports the selected plant s u e and this plant size is compatible with allowing flexibility in operations. There is no clear or definitive method with which to guide one in planning the conceptual leach pad design. The topography, the size of the ore reserve, operating philosophy, and environment all have to be considered in arriving at the proper choice. In general, for higher-grade ores (i.e., generally ore grade >1%), where the leach residue can be affordably removed, the odoff, or “dynamic”, leach pad is becoming the preferred choice. When the required horsepower in lifting both solution and ore, as well as the added copper inventory, are properly addressed, the odoff pad is generally favored over the permanent multi-lift leach pad. This is regardless of whether ore is leached in a single lift with intermediate liner, or leached through all lifts. The individual advantages and disadvantages of each are discussed from a conceptual design perspective below. On/Off (“dynamic”) - Single Ore Lift. The primary advantage of single-lift leaching is a higher “operating” recovery, which can approach that of the column tests for similar crush size, as long as the data has been properly scaled.

One of the initially considered disadvantages of the odoff pad is the cost of rehandling the leach residue plus the associated cost of that disposal. If a lined area for the leach residue is required for reasons of environmental isolation, then the cost may favor stacking ore on a permanent pad. A more important hidden cost of this approach is the loss of copper production if the initial scale-up is incorrect. Historically, this has been the case with the majority of these operations. Many have resorted to leaching the residue, but this is not always practical, or possible, due to the residue being placed on un-lined areas or placed in a manner, and at a depth, where the permeability is marginal. The choice of the type of heap leach method can be dictated solely by topography. For example, an area of initially limited flat surface containing deeper valleys or canyons can begin as a “valley-fill” operation (discussed below) and then change to an odoff operation once the proper flat area is developed.

Permanent Pad. Until recently, the most often practiced form of heap leaching was to stack ore in multiple lifts on top of one another, all over a single “permanent” liner. As mentioned above, one of the primary disadvantages of this concept is considered to be the added cost of operation with time, due to the power required in elevating both ore and water. However, an often overlooked feature of this design is the fact the base footprint of the pad must be sufficient to contain the entire resource and the liner system must be substantial enough to contain a high loading factor. In addition, the topmost lift of ore must provide sufficient “functional” leach area for the full leach cycle, plus the “dead” area required of the stacking equipment. If the base footprint is not sufficient, production in the later years can be severely hampered, An additional complicating factor is the total resource is not always fully known, requiring additional area on which to expand the pad. Another distinguishing feature of a permanent pad, multiple-lift leach design, where solutions are allowed to flow through the underlying lifts, is the copper inventory tied up in the solution. This is discussed further below. An offsetting advantage of this method is the ore is under leach for a much longer time, until the heap reaches its full height, or an intermediate liner is installed. At this time, the copper in solution inventory can be recovered.

1585

Once a permanent pad is decided upon as the primary design concept, there are two modes of operation - either through single lifts of ore, or through multiple lifts. SinPle Lift Leaching on a Permanent Pad. Choosing to leach through a single lift of ore on a permanent pad generally means the ore is marginal in its permeability character. This may be caused by an excessive fines content, either initially or after significant degradation. Another reason to leach through only a single lift can be concerns about reprecipitation or excess solution inventory. The primary disadvantage is similar to the odoff pad, where under-estimation of the necessary leach cycle will result in having to stack over ore when recovery is below expectation. The operating costs of this design are also higher than might initially be expected, as one has to replace all drainage piping (and forced aeration piping, if required) with each lift of ore. An added cost is also necessary to provide the intermediate liner, either in compaction of the natural leach residue or laying of synthetic geomembrane. Further, with each additional lift, the pumping and stacking costs typically increase due to the added lifting of both ore and water. (Engineered studies of the trade-off in capital and operating costs for the permanent pad versus the odoff pad generally show that the increased operating costs of the permanent pad can actually exceed the life-of-mine total costs of re-handling leach residue for the odoff concept.) Again, other factors, such as topography, can sway the decision. Multiple-Lift Leaching on a Permanent Pad. Leaching through multiple lifts was popularized in large copper and gold ROM leach operations, especially those of low relative ore grade. This method has the advantage of being able to keep ore wetted for very long periods of time. However, it has significant disadvantages with the increase in copper values tied up in the heap moisture, and potentially, lowered permeability, due to compaction with increased load. The author refers to this type of heap leaching as “upside-down” in that fresh ore is placed over leached ore. This reverses the chemical potential for displacement of the soluble values. This can lead to a soluble copper inventory representing from 4%of the expected “soluble” copper (that actually dissolved) to as much as 30%. The actual amount of copper tied up in the inventory depends on the combination of “soluble” ore grade, designed PLS grade and the ore’s natural water retention character. This combination of PLS grade and moisture has been the primary factor in the underperformance of expected production, in many instances, where this style of operation was used. As mentioned above, this copper is recovered once no more fresh ore is stacked on top of old ore and the full leach column can be washed with raffinate or water. Unlike gold, copper solvent extraction plants are generally more capital intensive and nearly directly proportional to the flow rate. Therefore, the tendency is to maximize the PLS grade to keep capital cost low. This can be exactly the wrong thing to do with low-grade ore utilizing a multiple-lift leach design. One of the errors with many multiple-lift leach designs is planning for too little recovery from the top lift, expecting to get this in an underlying lift, and then losing operating control in the lower lift. This is a tricky decision. There are many interesting questions concerning the in situ conditions in lower lifts of ore, which must be anticipated and a judgment made regarding the lower ore lifts. For oxide ore, there is the question of increased incremental acid consumption that may be uneconomic relative to the marginal copper dissolved at depth. For sulfides, the

1586

Next Page

chemistry is much more complex. There can be advantages with increased depth regarding temperature development, but there can be acid balance or precipitation issues that are very complex. A flexible design is probably one that achieves an economic level of extraction in the top lift and leaves the “marginal” recovery, from say larger particle sizes, to the lower lifts. Intermediate liners at some point can be used if adverse solution chemistry or permeability problems are encountered. However, one problem with the latter option is that at the time of the placement of an intermediate liner, copper production will decline by the percentage of copper that was being dissolved in the lower lift(s), unless a system of irrigation can be provided for the lower lifts. The latter option is most lkely quite problematic, and it is not known if this has been commercially proven. The initial mining plan for cash flow estimates must recognize the added ore that must be mined on a current basis to address the inventory required of the “reversing” of the chemical potential. This increases the copper in solution inventory with each additional ore lift by a nearly constant amount. This copper can be considered “delayed” production. It can be displaced once the total ore depth is reached, generally at the end of the mine life, if the pad is monolithic in design. Historically, this additional inventory has not been provided for in advance, which causes shortfalls in production. To deal with this increased inventory and subsequent dilution effects, there can be a requirement to recycle as much as 30% of the total leach solution flow. This can also require a proportionally larger surface area - something not always fully appreciated initially. Valley-fill Leaching. A hybrid of the above two leaching designs is the valley-fill design as practiced at Compania Minera Cerro Verde, Arequipa, Peru. If adequate topography is available, where the slopes of the hills or sides of broad canyons can be adequately lined, leaching can progress while ore is being placed in a fashion that ultimately fills the canyon. One of the advantages of this concept, besides probably lowest cost in liner per tonne of ore stacked, is the leach area can increase with time. This more naturally matches the needs of flexibility in area for a maturing multiple-lift leach design. This is exactly opposite of the permanent pad concept, where the base footprint is a fixed size and leach area is lost with each successive lift. However, one must anticipate the ever-expanding drainage volumes that the initial pipe works may have to handle. Flooding of the ore must be prevented, as it can cause a significant increase in copper values tied up in solution and may result in ore instability. An additional advantage of a valley-fill design, in special circumstances, is that if a dam is placed downstream of the toe of the heap, the natural valley can allow for storage of solution within the heap. This has advantage in colder climates and has been used mostly in gold and silver heap leaching. However, the added hydraulic head on the liner system in the area storing the solution requires a more robust liner design than typical.

Pad and Liner Design. The following discussion concerns the basic geometric design aspects and options for evaluating and designing a functional leach pad system. It does not address the actual environmental standards and options of the pad liner itself, which is a critical and necessary component of a properly designed pad. The ability to construct a “permit approved” leach pad is of paramount importance. In fact, it can be the deciding factor in the final design when the local topography is taken into consideration. The actual engineered construction of the leach pad liner system is a subject unto itself. This paper addresses primarily the mechanics and systems unique to the chemistry and physical parameters to meet the production goals. Others address the subject of proper pad liner design and stability (Breitenbach 1997, 1999,2001).

1587

Precious Metal Heap Leach Design and Practice Daniel W.Kappes'

ABSTRACT Heap leaching of gold and silver ores is conducted at approximately 120 mines worldwide. Heap leaching is one of several alternative process methods for treating precious metal ores, and is selected primarily to take advantage of its low capital cost relative to other methods. Thirty-seven different heap leach operations with a total production of 198 tonnes of gold per year (6,150,000 ounceslyr.) were surveyed to determine operating practice. These operations together produce 7.4% of the world's gold. When mines not surveyed are taken into account, it is likely that heap leaching produces 12% of the world's gold. Heap leaching for silver is conducted using the same principles and operating practices as for gold, but heap leach operations produce only a small fraction of world silver production.

INTRODUCTION Heap leaching had become a fairly sophisticated practice at least 500 years ago. Georgius Agricola, in his book De Re Metallica (publ. 1557) illustrates a heap leach with a 40-day leach cycle (Figure I), which could pass in many ways for a modern heap leach. The Agricola heap leach recovered aluminum (actually alum) for use in the cloth dying industry. Copper heap and dump leaches in southern Spain were common by about 1700. Gold and silver heap leaching began with the first Cortez heap leach in 1969. While many projects have come and gone, Cortez is still going - their new 63,000 tonnelday South Area leach is scheduled to start up in 2002. The largest U.S. precious metal heap leach is the Round Mountain, Nevada, operation with over 150,000 tonneslday of ore going to crushed or run-of-mine heaps, at an average grade of 0.55 grams goldtonne [This chapter follows the North American convention of "ton" for short ton and "tonne" for metric ton]. Worldwide, Newmont's Yanacocha, Peru, operation holds the record, with a 2002 target of nearly 370,000 tonneslday, at an average total reserve grade of 0.87 grams gold per tonne. On the other end of the scale, some very high grade ores - up to 15 grams per tonne (0.5 odton) - are being successfully processed at rates of several hundred tonneslday (Sterling, Nevada; Hassai, Sudan; Ity, Ivory Coast). A cursory worldwide summary in late 2001 was able to identify 78 active precious metal heap leaches worldwide, of which 34 were in the U.S. (22 in Nevada). The survey no doubt missed many operations, so the worldwide total is certainly over 100. To provide a basis for this chapter, technical andlor cost data were gathered from 37 of these operations. Because many operations impose restrictions on the release of detailed data, composite results are presented. Nevada was the "birthplace" of modern gold heap leaching in the late 1960's, and is only now giving up its dominance of this technology. Other very large gold districts - notably the preCambrian shield areas of Canada, Australia and South Africa - show relatively few heap leaches. There are several reasons for this geographic concentration, but the primary reason is that Nevada gold deposits tend to have been created by low-energy geologic processes - near surface hot

1 Kappes, Cassiday & Associates, Reno, Nevada

1606

springs and moderate depth, hydro-thermal systems that deposited low grade gold in permeable rocks. Besides aiding gold deposition, the permeable nature of the rocks allowed uniform and deep oxidation that liberated the gold from its sulfide and carbonaceous host minerals. Shield deposits have generally had a more complicated history, which has resulted in coarse gold contained in poorly-permeable rocks. Often these ores can be successfully heap leached only after weathering has completely destroyed the rock matrix.

Figure 1. "The rocks are . . piled in . . heaps fifty feet long, eight feet wide and four feet high, which are sprinkled for forty days with water. The rocks begin to fall to pieces like slaked lime, and there originates a . . new material". Drawing and text from De Re Metallica, Herbert Hoover translation, published by Dover Publications, Inc.

WHAT IS HEAP LEACHING? To those of us in the gold industry, the question "What is Heap Leaching?" seems to have an obvious answer. In the simplistic sense, heap leaching involves stacking of metal-bearing ore into a "heap" on an impermeable pad, irrigating the ore for an extended period of time (weeks, months or years) with a chemical solution to dissolve the sought-after metals, and collecting the leachant ("pregnant solution") as it percolates out from the base of the heap. Figure 2 is an aerial photograph showing the typical elements of a precious metals heap leach operation - open pit mine, a heap of crushed ore stacked on a plastic pad, ponds, a solution process facility for recovering gold and silver from the pregnant solution, and an office facility. For a small operation such as the one illustrated here, very limited infrastructure is required. In a more complex sense, heap leaching should be considered as a form of milling. It requires a non-trivial expenditure of capital, and a selection of operating methods that trade off cost versus marginal recovery. Success is measured by the degree to which target levels and rates of recovery

1607

Figure 2 Heap leach installation at Mineral Ridge, Nevada. The open pit mine is shown on the left. On the right is a two million ton heap of crushed, conveyor-stacked ore placed on a plastic-lined leach pad. Pregnant and barren solution storage ponds are located downslope from the heap. Buildings include process plant, laboratory, maintenance shop and administration offices. Photo courtesy of Tom Nimsic, American AuIAg Associates. are achieved. This distinguishes heap leaching from dump leaching. In dump leaching, ores are stacked and leached in the most economical way possible, and success is achieved with any level of net positive cash flow. The bibliography of precious metals heap leaching is quite extensive, and because of time limitations a very limited bibliography has been compiled for this chapter. However, the following publications are good places to start a literature search: "Global Exploitation of Heap Leachable Gold Deposits", by Hausen, Petruk and Hagni, February, 1997 "The Chemistry of Gold Extraction" by Marsden and House, 1992 "World Gold '9 l", Second AusIMM-SME Joint Conference, Cairns, Australia, 1991 "Introduction to Evaluation, Design and Operation of Precious Metal Heap Leaching Operations", by Van Zyl, Hutchison and Kiel, 1988. Special recognition and thanks should be given to Hans von Michaelis of Randol International, Denver. Between 1981 and April 2000 Randol organized four major symposia followed by four published studies of the gold industry, and several minor meetings with their own proceedings. The combined Randol literature occupies nearly 40 volumes covering six feet of shelf space. Most modern heap leach operations are referenced.

1608

WHY SELECT HEAP LEACHING AS THE PROCESSING METHOD? Gold and silver can be recovered from their ores by a variety of methods, including gravity concentration, flotation, and agitated tank leaching. Methods similar to heap leaching can be employed: dump leaching and vat leaching (vat leaching is the treatment of sand or crushed ore in bedded vats with rapid solution percolation). Typically, heap leaching is chosen for basic financial reasons - for a given situation, it represents the best return on investment. For small operations or operations in politically unstable areas, it may be chosen because it represents a more manageable level of capital investment. Some interesting examples that illustrate this issue of choice are presented below. Capital Risk Several years ago, the author's company was advising on a project in which the ore reserve was a few million tonnes at a grade of 7 grams of gold per tonne (0.22 odton). Heap leach recovery was about 80%, well below the 92% that could be achieved in an agitated leach plant. Financial considerations strongly favored milling, and the owner was financially strong. However, the operation was located in an undeveloped country with unstable politics and socialist leanings. The owner concluded that he might lose control of a high-capital investment, whereas he could maintain control of a heap leach with an implied promise of a future larger capital investment. The operation has been running successfully and very profitably. Lack of Sufficient Reserves The Sterling Mine in Beatty, Nevada (Cathedral Gold Corporation) began life as an underground mine, with a reserve of 100,000 tons of ore at a grade of 11 grams goldkonne (0.35 odton). Over a fifteen year period, it mined and processed nearly one million tons, but never had enough ore reserves to justify a conventional mill. Fortunately, the Sterling ore achieves excellent heap leach recovery - the original heaps reached 90% from ore crushed to 100 mm. Equal or Better Percent Recovery Comsur's Comco silver heap leach at Potosi, Bolivia, showed the same recovery in both a heap and an agitated leach plant. However, the silver ore leached very slowly and residence time of up to 4 days was needed in an agitated leach plant. Although the heap leach took several months to achieve the same recovery, the economics favored the heap. At the joint AIME/AusIMM Symposium "World Gold '91", T. Peter Philip of Newmont presented a paper "To Mill or to Leach?" in which he evaluated the decision of Newmont to build the Carlin No. 3 mill. He concluded that the mill recovery was over-estimated and the heap leach recovery underestimated, and that the decision to go with milling may have been incorrect. Differential Recovery Is Not Sufficient To Justify Added Investment A recent review (Kappes, 1998) concluded that for a "typical" Nevada-type ore body with ore grade of 3.0 grams goldtonne (0.088odton), the mill recovery would have to be 21% higher than the heap leach recovery to achieve the same return on investment - and this is very seldom the case. Of the 37 operations surveyed for this chapter, four have a head grade below 0.65 grams gold per tonne and half are between 0.65 and 1.50 grams per tonne. At these gold grades, it is usually impossible to justify the investment in a conventional agitated leaching plant. Capital Is Very Difficult Or Expensive To Raise Heap leaching has often provided the route for a small company to grow into a large company. A good example is Glamis Gold Corporation, which has gone from total assets of $12 million in 1984 to $112 million in 2001, based largely on its low grade heap leach projects at Picacho and Randsburg, California.

1609

At the time this is being written in early 2002, the precious metals mining industry is experiencing a severe capital shortage and a consolidation of producers into a few large corporations. This will open up an opportunity for the creation of a new generation of junior mining companies to exploit smaller deposits, and heap leaching will play a key role in this process.

TYPE OF ORE Heap leach recovery is very dependent on the type of ore being processed. Some typical examples are discussed below. Carlin-Type Sedimentary Ores These ores consist of shales and “dirty” limestones, containing very fine (submicroscopic) gold. Oxidized ores leach very well, with low reagent consumption and production recovery of 80% or better. Ores are typically coarse-crushed (75mm) but may show recovery of 70% or better at runof-mine sizes. The largest of the northern Nevada heap leaches (Carlin, Goldstrike, Twin Creeks) treat this type of ore. Unoxidized ore contains gold locked in sulfides, and also contains organic (carbonaceous) components, which absorb the gold from solution. This ore shows heap leach recovery of only 10 to 15% and is not suitable for heap leaching. Because of the different ore types, the northern Nevada operations (for instance, Barricks Goldstrike Mine) may employ roasters, autoclaves, agitated leach plants and heap leaches at the same minesite. Crushing is usually done in conventional systems (jaw and cone crushers) and ores are truck stacked. Low Sulfide Acid Volcanics Or Intrusives Typical operations treating this type of ore are Round Mountain, Nevada, and Wharf Mine, South Dakota. Original sulfide content is typically 2 to 3% pyrite, and the gold is often enclosed in the pyrite. Oxidized ores yield 65 to 85% recovery but may have to be crushed to below 12 mm (1/2 inch). Usually the tradeoff between crush size and percent recovery is a significant factor in process design. Unoxidized ores yield 45 to 55% gold recovery and nearly always need crushing. At Round Mountain, Nevada, approximately 150,000 tons per day of low grade oxide ore is treated in truck-stacked run-of-mine heaps, 30,000 tons per day of high grade oxide ore is treated in crushed (12mm), conveyor-stacked heaps, and 12,000 tons per day of unoxidized ore is treated in a processing plant (gravity separation followed by leaching in stirred tanks). Crushing is done using jaw and cone crushers; fine crushed ore contains enough fines that conveyor stacking is preferred over truck stacking. Oxidized Massive Sulfides The oxide zone of massive sulfide ore deposits may contain gold and silver in iron oxides. Typically these are very soft and permeable, so crushing below 75mm often does not increase heap leach recovery. The Filon Sur orebody at Tharsis, Spain (Lion Mining Company) and the Hassai Mine, Sudan (Ariab Mining Company) are successful examples of heap leaches on this type of ore. Because the ore is fine and soft, the ore is agglomerated using cement (Hassai uses 8 kg cemenvtonne), and stacking of the heaps is done using conveyor transport systems. Saprolites / Laterites Volcanic- and intrusive-hosted orebodies in tropical climates typically have undergone intense weathering. The surface “cap“ is usually a thin layer of laterite (hard iron oxide nodules). For several meters below the laterite, the ore is converted to saprolite, a very soft water-saturated clay sometimes containing gold in quartz veinlets. Silver is usually absent. These ores show the highest and most predictable recovery of all ore types, typically 92 to 95% gold recovery in lab tests, 85% or greater in field production heaps. Ores are processed at run-of-mine size (which is often 50% minus 10 mesh) or with light crushing. Ores must be agglomerated, and may require up to 40 kg of cement per tonne to make stable permeable agglomerates. Many of the West African and

1610

Central American heap leaches process this type of ore. Good examples are Ity in the Ivory Coast, and Cerro Mojon (La Libertad) in Nicaragua. When crushing is required, one or two stages of toothed roll crushers (Stammler-type feeder-breaker or MMD Mineral Sizer) are usually employed. Conveyor systems are almost always justified; ore can be stacked with trucks if operations are controlled very carefully.

Clay-Rich Deposits In some Carlin-type deposits, as well as in some volcanic-hosted deposits, clay deposition or clay alteration occurred along with gold deposition. The Buckhorn Mine, Nevada (Cominco, now closed) and the Barney's Canyon Mine, Utah (Kennecott) are good examples. These ores are processed using the same techniques as for saprolites, except that crushing is often necessary. Because of the mixture of soft wet clay and hard rock, a typical crushing circuit design for this type of ore is a single-stage impact crusher. Truck stacking almost always results in some loss of recovery. Agglomeration with cement may not be necessary, but conveyor stacking is usually employed. Barney's Canyon employs belt agglomeration (mixing and consolidation of fines as it drops from conveyor belts) followed by conveyor stacking. The new La Quinua operation at Yanacocha employs belt agglomeration followed by truck stacking. Silver-Rich Deposits Nevada deposits contain varying amounts of silver, and the resulting bullion may assay anywhere from 95% gold, 5% silver to 99% silver, 1%gold. Silver leaches and behaves chemically the same as gold, although usually the percent silver recovery is significantly less than that of gold. Examples of nearly pure silver heap leaches are Coeur Rochester and Candelaria in Nevada, and Comco in Bolivia. CLIMATE EXTREMES The ideal heap leach location is a temperate semi-arid desert location such as the western U.S. However heap leaching has been successfully applied in a variety of climates:

0

Illinois Creek, Alaska, and Brewery Creek, Yukon are located near the Arctic Circle and experience temperatures of minus 30°C for several months per year. Several heap leaches are located in the high Andes of South America (Comco at Potosi, Bolivia; Yanacocha and Pierina, Peru; Refugio, Chile) at altitudes above 4000 meters (13,000 ft). Although oxygen availability at these altitudes is only 60% of that at sea level, gold heap leaching proceeds at rates similar to that at sea level (oxygen is required for the process, but is not usually rate-limiting in a heap leach operation). At another extreme, Hassai, Sudan, is in the dry eastern Sahara fringes. This operation experiences normal daytime temperatures that routinely exceed 50°C in the summer, with annual rainfall of less than 20mm. One of the advantages of heap leaching over conventional cyanide leach plants and gravity recovery plants, is that heap leaching consumes very little water. With good water management practices, water consumption can be less than 0.3 tonnes water per tonne of ore. In tropical wet climates, rainfall of 2.5 meters per year can add over 5 tonnes of water to the leach system for each tonne of ore stacked. As discussed in a later section, this amount of water can also be handled with good water management practices.

CHEMISTRY OF GOLD/SILVER HEAP LEACHING The chemistry of leaching gold and silver from their ores is essentially the same for both metals. A dilute alkaline solution of sodium cyanide dissolves these metals without dissolving many other ore components (copper, zinc, mercury and iron are the most common soluble impurities).

1611

Solution is maintained at an alkaline pH of 9.5 to 11. Below a pH of 9.5, cyanide consumption is high. Above a pH of 11, metal recovery decreases. Many heap leachable ores contain both gold and silver. Of the 28 mines that reported bullion assays, five produce a dor6 (impure gold-silver bullion ) bar that is greater than 70% silver. Another five produce a bullion greater than 30% silver. Only five produce a bullion with less than 5% silver. Deposits in western Africa and Australia tend to be very low in silver, while those in Nevada are highly variable, ranging from pure gold to pure silver. Silver is usually not as reactive with cyanide as gold. This is because gold almost always occurs as the metal, whereas silver may be present in the ore in many different chemical forms some of which are not cyanide-soluble. Reported heap leach recoveries (32 operations) averaged 71% gold, and ranged from 49% to 90%. Reporting run-of-mine heap leaches averaged 63%. Typical recovery for silver is 45-60%, although when silver is a minor constituent, its recovery may be only 5 2 5 % . The level of cyanide in the heap onflow solution ranges from 100 to 600 ppm NaCN, and averages 240 ppm for the 28 operations reporting. Forty-five percent of the operations reported cyanide strength below 200 ppm, 25% were above 300 ppm. Heap discharge solution (pregnant solution) averages 110 ppm. Cyanide consumption, via complexation, volatilization, natural oxidation or oxidation by ore components, typically ranges from 0.1 to 1.0 kg per tonne of ore. Price of sodium cyanide is currently at a historical low of $1.00 per kg. Cement and/or lime consumption ranges from 0.5 to 40 kg per tonne of ore. Several operations use cement for alkalinity control (instead of lime) as well as for agglomeration. The price of cement or lime is $60 to $100 per tonne, $160 delivered to remote African locations. Other leaching agents - thiosulfate, thiourea, hypochlorite, bromine - have been experimented with as an alternative to cyanide, but cyanide is by far the most effective and the most environmentally friendly leaching agent. A good discussions of the process chemistry can be found in "The Chemistry of Gold Extraction" by Marsden and House, Ellis Horwood Publishers, 1992.

LABORATORY TESTING & CONTROL As with any processing method, it is very important to base the design on the results of a comprehensive program of laboratory testing. During the production operation, laboratory tests including column leach tests should be continued on a regular basis, since the initial ore samples are seldom representative of the entire orebody. For a heap leach, the key parameters that are defined in the laboratory are crush size, heap stability, permeability versus heap height, cyanide strength and consumption, the need for agglomeration and the amount of agglomerating agent (usually Portland cement) required, leach time, and percent recovery. Derivative parameters such as the height of individual lifts and the method of stacking are also determined. Heap leaching has inherent risks that can be largely eliminated if the operating practices follow the results of initial and on-going laboratory testing. The risks result from the nature of the operation. The results of the process are usually not known for several weeks or months after the ore is stacked, and at this point it is not economical to reprocess the ore. Mistakes made in the initial plant design or operating practices, for instance by not crushing finely enough, or by not agglomerating or stacking properly, can result in cash flow problems that might persist for up to a year after the problem is solved. An on-site laboratory is an important part of the infrastructure at a heap leach operation. It should include an analytical section (for ore control) and a metallurgical testing section that regularly runs column leach tests on production samples. For a small operation processing less than 5,000 tonnes of ore per day, staffing is 2-3 technicians for sample preparation and assaying and one metallurgist to conduct process tests. Large operations may have a laboratory staff of ten to fifteen people.

1612

HEAP PERMEABILITY & FLOW EFFICIENCY The key element in a successful heap leach project is a heap with high, and uniform permeability. In any heap there are three zones of different flow regimes:

9

coarse channels, which allow direct short-circuitingof solution from top to bottom highly permeable zones, in which solution is efficient at contacting the rock and washing the gold downward in "plug flow" zones of low to zero permeability where high grade solution or unleached ore may be trapped.

Efficiency Of Solution Displacement If the heap were "ideal" - moving in true plug flow - then when one displacement volume of solution was placed on top of the heap, it would fully replace the solution in the heap. This would be 100% wash efficiency. In practice, the "best" heap leaches exhibit a wash efficiency of about 70%. At 70% per displacement, three displacement washes are required to achieve a recovery of 95% of the dissolved metals. A fourth "displacement" is required initially, to saturate the ore. Since a typical heap contains 20% moisture, 95% recovery (of the dissolved goldsilver content) requires that 0.8 tonnes of solution must be applied to each tonne of ore. Typical practice is to apply 1.3 tonnes of solution per tonne of ore during a 70-day primary leach cycle. This suggests two things: (a) most heap leach operations are able to maintain reasonably good permeability characteristics, yielding at least 50% wash efficiency; and (b) a high percentage of the gold is solubilized early in the 70-day leach cycle. Drainage Base A drainage base of crushed rock and embedded perforated pipes is installed above the plastic leach pad and below the ore heap. The importance of this drainage base cannot be overemphasized. Solution should percolate vertically downward through the entire ore column, and then enter a solution removal system with zero hydraulic head, If the drainage base cannot take the entire flow then solution builds up in a stagnant zone within the heap, and leaching within this stagnant zone can be very slow. To put this in context, a "typical" heap might run 500 meters in an upslope direction. All of the onflow solution in a strip one meter wide by 500 meters must flow out at the downslope edge of the heap through the drainage base, which is typically 0.65 meters thick. The design horizontal percolation rate through the drainage base is therefore nearly 800 times the design rate of the heap itself. This is not a difficult engineering accomplishment since flow is carried in pipes within the base. At one Australian copper heap leach operation, three adjacent leach panels were built. The two flanking panels had a good installed drainage base but the center panel did not. Recovery in the center panel was depressed 20%. A similar effect has been seen but not quantified at gold heap leach operations. Recovery Delay In Multiple Lift Heaps As subsequent lifts are stacked, the lower lifts are compressed and the percentage of low permeability zones increases. The first solution exiting an upper lift may have a gold concentration of up to three times that of the ore. If impermeable zones have developed in a lower lift, high grade solution may be trapped, causing a severe reduction in recovery rate and possibly in overall recovery percentage. The highest heap leaches currently in operation are 120 meters high, with about ten lifts of ore. Hard ore, crushed or run-of-mine, can withstand the resulting pressure without significant permeability loss. Many softer ores can be agglomerated with enough cement so that they can perform under a load of 30 meters; some agglomerated ores perform satisfactorily

1613

to 100 meters. These properties can be properly evaluated in advance, in laboratory column tests, which are run under design loads. The delay in recovery as lifts are added to the heap is partly a function of the impermeability of the lower lifts, and partly a function of the wash efficiency discussed earlier. The net effect is that average recovery is delayed as the heaps get higher, and overall pregnant solution grade decreases (requiring more solution processing capacity). Of 22 operations reporting, 18 indicated that they see a delay in time of average recovery, which ranges from 3 to 30 days per lift. The most common figure was 7 daysflift. This cash flow delay must be allowed for. Also, this number implies that for each extra lift, the capacity of the process plant should be increased in response to the decrease in gold content of pregnant solution. Intermediate Liners

If impermeability of lower lifts becomes a serious problem, it is possible to install intermediate liners. Four of the surveyed operations reported that they install plastic intermediate liners on a regular basis. One operation reported that it regularly installs a clay intermediate liner. There are two problems with installing an intermediate liner: (a) the heap below the liner is compressed as the upper lift is placed, resulting in differential settlement and tearing of the liner; and (b) the ore below the liner cannot be washed with water, which is sometimes required as a part of final heap closure.

SOLUTION APPLICATION RATE AND LEACH TIME With regard to sprinkling rate, the timing of gold recovery is a function of five factors: 0

0 0

0

the rate at which the gold dissolves, Coarse gold particles dissolve very slowly, and may not fully dissolve for several months in a heap leach environment. the percentage of gold that exists as free or exposed particles the rate of diffusion of the cyanide solution into rock fractures, and of gold cyanide back out of the rock fractures. Where the gold occurs on tight fractures or in unfractured rock, the rock must be crushed into fine particles to achieve target rates and levels of recovery. the effect of chemical reactions within the heap, or within rock particles, which destroy cyanide and alkalinity or which consume oxygen the rate of washing of gold off of the rock surfaces and out of the lift of ore under leach. This is a complex issue, which depends on the overall permeability of the lift and the local permeability variations due to segregation and compaction as the lift is being constructed.

The above factors cause wide theoretical differences in the response of various ores to leaching. However, in practice most heap leach operations apply solution to crushed-ore heaps within a fairly narrow range of flows: Of 19 operations reporting, application rates for crushed-ore heaps ranged from 7 to 20 l/hr/sq. m (0.003 to 0.009 gpdsq. ft) with an average of 11 Vhr/sq. m (0.048 gpdsq. ft). Only three applied solution above 10 I/hr/sq. m (0.0044 gpdsq. ft), and only four were below 8 Vhr/sq. m. (0.035 gpdsq. ft). For 17 run-of-mine heaps, the average application rate was 8.3 l/hr/sq. m (0.037 gpm/sq. ft), with only two operations above 10 Vhr/sq. m (0.044 gpdsq.

ft). Laboratory columns always respond much faster than field heaps, for two reasons: the ore is placed in the lab column much more uniformly so that percolation is more effective; and the solution-to-ore ratio (tonnes of solution per tonne of ore in a given time frame) is generally higher in lab columns than in field heaps. For some field heaps, notably where the ore is fine crushed and the ore leaches quickly, the so1ution:ore ratio is a more important factor than overall leach time. However, for the majority of heap leaches, time seems as important as specific application rate. For the 32 operations reporting, average single lift height was 8.9 meters (yielding 14.2 tonnes of ore per sq. m of top surface) and average irrigation time was 70 days during the primary leach

1614

cycle. This yields a specific solution application rate of 1.30 tonnes solution per tonne of ore. Operations with low cycle times tended to have higher application rates, suggesting that the ratio of 1 to 1.5 tonnes of solution per tonne of ore is a universal target. For ores with very slow leaching characteristics, an intermediate pond and a recycle stream may be added to the circuit, so that each tonne of ore sees two tonnes of leach solution during an extended leach period. The process plant treats only the final pregnant stream - one tonne of solution per tonne of ore. Of 36 operations reporting, 16 had only one leach cycle, 16 had two cycles, and four had three cycles. The use of multiple cycles is good operating practice for single-lift heaps of high grade ore. However, for multi-lift heaps this is not the case. Heap modeling indicates that once the heap attains a height of three lifts, the intermediate solution contains almost as much gold as the pregnant solution. Recycling results in a significant build-up of dissolved gold within the heap, causing a slight overall recovery loss and a cash flow delay. For multi-lift heaps, it is often possible to justify an increase in the size of the recovery plant so that only fully barren solution returns to the heap. Some successful single-lift heaps achieve a high percentage recovery in the first leach period, but continue the leach for much longer. Sterling, with very high grade ore, leached the same ore for 18 years. The initial Cortez heaps were leached intermittently for ten years. Cost to intermittently leach old heaps may be as low as $0.10 per tonne per month. It is extremely important to design a heap leach system so that the ore can be leached for a very long time. Unlike an agitated leach plant where the ore can be ground to a fine powder and intensively agitated, heap leaching is not a very energy-intensiveprocess. Once a heap is built, one of the most significant variables, which the operator can employ to solve design or production problems, is leach time. Successful projects employing on-off (reusable) leach pads have been undertaken, but this is a risky practice. Some operations (such as Round Mountain, Nevada) utilized on-off pads to achieve rapid first-stage recovery, then transfer the ore to long term heaps to complete the process.

DESIGN FOR AMBIENT WEATHER CONDITIONS Design For High Ambient Temperature In regard to ambient temperature, high temperature is not a direct problem. In very hot desert areas where drip irrigation is used, sunlight will significantly heat the solution. Even then, because of the effect of cool night time temperatures, it is unusual to see heap effluent solution temperatures above 15°C. Mesquite at one point covered the ponds to prevent evaporation, and the resulting recirculating solution was above 30°C. Hot leach solutions dissolve less oxygen than cold solutions and this could affect the rate of gold recovery in oxygen-starved heaps. However, usually there is sufficient oxygen present, and the higher overall activity due to the higher temperature more than offsets the oxygen effect. No operating heap leaches have reported a direct problem due to high temperature of the rock or the leach solution. Design For Low Ambient Temperature Low temperature can be a problem. Many Nevada heap leaches report a significant recovery decrease in winter, which is offset the following summer. When a cold weather project is anticipated, column tests should be run under cold conditions. There are several reasons for a reduction in recovery rate with lower temperatures: Rate of reaction is slowed as solution tempeature approaches freezing. Comparative laboratory column tests show that recovery rate drops significantly when the heap temperature drops below 5°C. Solution viscosity increases significantly as temperature drops. This affects both the heap and the process plant.

1615

Solution flowing slowly through the normally unsaturated heaps flows via the meniscus on the surface of particles, and the thickness of this meniscus is a direct function of viscosity. Thus, cold heaps tie up more process solution (and more gold inventory) than warm heaps. In carbon columns, the rate of fluidization can be significantly affected. PICA USA Inc. (one of several activated carbon suppliers) has generated a graph of solution temperature versus percent fluidization, which is shown as Figure 3. As the graph shows, bed expansion of carbon can increase from 15% to 33% as the solution temperature decreases from 20°C to 5°C. This same viscosity effect will alter the ability of the solution to flow through the heaps. BED EXPANSIONv TEMPERATURE PICA 210 CARBON

TEMPERATURI? 0

5

0

'0

x

'5

A

20

c

VELOCITY OF WATER (Whr)

Figure 3 Effect of Temperature on Bed Expansion of Activated Carbon Solution surface tension drops, although not as fast as viscosity. Surface tension can affect the flow of solution through the heaps, and also it affects the ability of the solution to penetrate tight fractures within the rock. Table 1 shows the effect of temperature on the viscosity and surface tension of water. In very cold climates, it is possible to create a frozen wedge of solution within the heap. This occurred at Summitville, Colorado. For this reason, Brewery Creek (Yukon) stacks ore only in the summer, although they leach all year.

I

Temperature 0°C 5°C 10°C 20°C 40°C

I

I

Surface Tension, dyneskm 76 75 74 73 70

I

I

Viscosity, centipoises 1.79 1.52 1.31 1.oo 0.66

Table 1 Surface Tension and Viscosity as a Function of Temperature

1616

1

Water Balance Since many heap leach operations occur in desert areas where water is scarce, and others occur in environmentally sensitive areas where water discharge is not acceptable, the balance between water collection and evaporation is important. Fortunately, by adjusting the method and scheduling of solution application, it is usually possible to meet the local requirements. Evaporation of water, regardless of its mechanism, requires a heat input of 580 calories per gram (8300 BTU/U.S. gallon) of water evaporated. A heap leach gets this heat input from three sources: direct solar heating on heap and water surfaces; latent heat in the shroud of air within the "sprinkler envelope"; and latent heat in the air that is pulled through the heap by convection. Average 24-hr incident solar radiation on a flat horizontal surface ranges from 12,000 BTU per sq. m per day (central US.) to about 30,000 BTU (equatorial desert), which could theoretically evaporate 5 to 12 liters of solution per day. With a typical heap application rate of 10 Vsq. m /hr, incident solar radiation could account for an evaporation rate of 2 to 5% of applied solution when using sprinklers. Evaporation would be somewhat less when using drip irrigation (1 to 4%) because some of the solar energy is re-radiated from dry areas on top of the heap. This same heat input would result in pond evaporation of 5 to 13 mm per day. Use of sprinklers rather than drips may result in the loss of up to 30% of solution pumped. This is because the sprinkler droplets trace an arc through a shroud of air, which is very seldom at 100%humidity. A gentle breeze of 3 km (two miles) per hour will replace the "saturated shroud" on a typical 500 m long heap with unsaturated air every 10 minutes, and the pumping action of the sprinkler droplets will cause additional rapid air replacement. A good discussion of evaporative sprinkler losses is presented in Univ. of Florida Cooperative Extension Service Bulletin 290. Typical sprinkler evaporation at operations using coarse-drop sprinklers in Nevada-type climates (arid, temperate) is up to 15% of solution pumped on summer days and 2-4% on summer nights, averaging about 7% annually. Overall evaporative losses include the sprinkler losses, convective loss from air flowing through the heap, and losses due to heatinglevaporation from ponds and from other areas not sprinkled. These have been determined at several Nevada operations to be up to 20% of total solution pumped in summer months and 10% annually. Thus, direct sprinkler loss accounts for about 60% of the total. Use of drip irrigation can reduce but not eliminate evaporative loss. In tropical climates, noticeable losses occur even during the rainy season. KCA's in-house experience on several tropical heap leach projects where rainfall is seasonal and up to 2.5 meters per year, is that overall annual evaporative loss from all sources, when using wobbler-type sprinklers operated 24 hourdday, is about 7% of solution pumped. Typical heap application rate is 10 l/hr/sq. m, or 88 meters per year (9 inchedday). Thus, evaporative loss of 7% is equal to 6.2 meters per year on the areas actually being sprinkled. If the heap and pond systems are properly designed, the active leaching area can be up to 40% of the total area collecting rainfall; it is therefore possible to operate in water balance when rainfall is 2.5 meterdyear. For these operations, very large solution surge ponds are required: at Sansu, Ashanti, Ghana (rainfall 2.5 meterdyr.), for a 3000 tonne/day heap leach, total pond volume was 60,000 cu meters. Where rainfall is high and evaporation rate is low, some operations (such as at Yanacocha, Peru, altitude 3500 meters) cover sideslopes with plastic to minimize rain collection. Others (Rio Chiquito, Costa Rica - Mallon Minerals COT) have tried to cover the entire heap during the rainy season, but this has not worked very well because of the mechanical difficulties of moving the cover. In West Africa and Central America it is acceptable practice to treat and discharge excess solution during the rainy season. Typically, excess process solution is routed through a series of ponds where cyanide is destroyed using calcium hypochlorite or hydrogen peroxide, followed by adjustment of pH to near neutral. The INCO SO2 system, using copper-catalyzed hyposulfite to destroy cyanide, is also employed for this purpose. Cyanide-free solution is further treated in controlled wetlands (swamps) to remove heavy metals prior to discharge.

1617

The worst water balance situation occurs in cool, damp climates such as high altitude operations (for instance the Landusky-Zortman operation in Montana, now closed). In such climates, rainfall and snowfall may be significant and evaporation is minimal. Generally such heaps can stay in water balance with an aggressive program of summer sprinkling. Arctic heap leaches (Brewery Creek, Yukon; Illinois Creek, Alaska) have been able to stay in water balance because precipitation is lower than the total water requirement needed to saturate the ore.

SOLUTION APPLICATION EQUIPMENT A variety of solution application methods have been employed, but for mainstream heap leaches the following equipment has become standard:

0

0

0

Drip Emitters. Drip emitters, which issue drops of water from holes every 0.5 to 1.5 meters across the heap surface, are very common. Drip emitters are easy to maintain and minimize evaporation. The main drawback to drip emitters is that they do not provide continuous drip coverage. Thus the top one meter of the heap may not be leached very well until it is covered with the next lift. Other problems are that emitters require an intense (and expensive) use of anti-scalant, and they require the use of in-line filters. Wobbler Sprinklers. Wobbler Sprinklers are used at a large number of operations. Their main advantages are that they issue coarse droplets, which control but do not eliminate evaporation, and that they deliver a uniform solution distribution pattern, which ensures uniform leaching of the heap surface. The coarse droplet size has another advantage cyanide is readily oxidized by air and sunlight, and the wobbler-type sprinkler minimizes this loss (but not as well as drip systems). Wobblers are typically placed on a 6 x 6 meter pattern across the heap surface. A disadvantage of all sprinklers is that they require continual servicing, and personnel spend extended periods working in a "rainstorm". Occasional skin contact with cyanide solution does not pose a health problem, but an environment that encourages repeated skidsolution contact is not recommended. Sprinkler maintenance personnel wear full rain gear to eliminate any exposure problem, but the working environment (especially in cold weather) is not as nice as with drip emitters. Reciprocating Sprinklers. Reciprocating sprinklers shoot a stream typically 5 to 8 meters long of mixed coarse and fine droplets. They are not considered ideal for heaps, but often find application for sprinkling sideslopes since they can be mounted on the top edge to cover the entire slope. If emitters and wobblers are used on sideslopes, they must be installed on the slope, which is a difficult and sometimes dangerous place for personnel. High Rate Evaporative Sprinklers. High rate evaporative sprinklers typically operate at high pressures with an orifice designed to produce fine droplets and shoot them in a high trajectory. Evaporative blowers using compressed air to atomize and launch the droplets can also be used. This type of equipment is not normally used at heap leach operations, but it will become more common as more heaps enter the closure mode where rapid evaporation is needed.

For the 37 operations responding, solution application methods are summarized in the list below. 0 0 0 0 0

0

13 use only drip emitters 5 use only wobbler sprinklers 19 use both drip emitters and wobbler sprinklers 10 bury the drip emitters 5 use buried drip emitters between lifts all five "tropical" leaches - rainfall above 1500 mm per year - use wobblers.

1618

Regardless of the systems used for solution application and management, capital and operating costs for solution handling are usually small. On the heap, header pipes up to 400 mm diameter are located every 30 to 60 meters across the heap. Material of choice for these pipes is usually HDPE (High Density Polyethylene), but sometimes it is mild steel. Distribution pipes of PVC or UVstabilized PVC, usually 75 mm to 150 mrn diameter, take off from the header pipes and are placed on similar (30 to 60 meter) spacing. From these, individual drip emitter lines up to 60 meters long cross the heap on 1.0 meter centers, or sprinkler manifold pipes (25 to 50mrn PVC) up to 60 meters long cross the heap on 6 to 8 meter centers. Total piping cost including header pipes (installed) is about $0.60 per square meter, or $0.05 per tonne of ore leached.

POWER COST FOR PUMPING For the average two-cycle leach, two tonnes of solution are pumped to the heap and one tonne to the recovery plant, for each tonne of ore leached. Typically on-heap pressure for pumping barren solution is 100 psi at the pumps, and in-plant pressure for pumping pregnant solution is 30 psi. Thus, for two tonnes of solution per tonne of ore, power for pumping is 1.8 kWhrltonne of ore and cost is $0.14/tonne. Where heaps are very high or where evaporation is required, power consumption can approach 4.0 kwhrltonne. LEACH PADS AND PONDS The leach pad below the heap is a significant element of a heap leach design. The ideal location for the heap is a nearly flat (1% slope), featureless ground surface. Usually some earthwork is required to modify contours, but it is not necessary to eliminate all undulations. It is only necessary that all solution will flow across the surface towards collection ditches on the base or sides of the heap. Where the slope exceeds 3%, the front edge of the heap (30 to 50 meters) should be graded flat to provide a buttress to prevent heap failure. Heaps can be placed in fairly steep-walled valleys with side slopes up to 20% (12 degrees). For long slopes above lo%, careful choice of pad material is necessary. LLDPE (Linear Low Density Polyethylene) offers a good choice because it has the ability to stretch but also has a high tensile strength, and it can be heat-welded to HDPE in flatter areas. Valley Fill Heap Leach A "Valley Fill Heap Leach" is a heap leach that has been built upslope from an earth dam. The containment area of the dam is filled with the stacked ore. The voids in the ore provide solution containment, and this volume serves as the pregnant solution storage pond. The ore stacked in the containment area behind the dam is usually a small part of the heap, which continues upslope and above the containment area. Thus, the dam might be ten meters high and the heap 50 meters high. A good example of a Valley Fill heap leach is Rochester, Nevada, shown in Figure 4. Valley Fill heap leaches are used where terrain is steep and the ore must be placed in a narrow valley. They are also employed in arctic or high altitude environments as a method of keeping the process solution from freezing. In normal leach pad construction, best design practice is to spread the solution out across the liner and to minimize solution hydraulic head to a few inches in any area. With a Valley Fill design, solution flow is concentrated and hydraulic head is high. The leach pad immediately upslope from the dam (in the solution storage area) must be built very carefully, usually with extensive sub-base preparation, double liners, and extra leak detection. PAD CONSTRUCTION COST A typical pad consists of several layers of material, listed from bottom to top. The ideal padsite begins as a uniformly sloping area with a grade of 0.5% to 2.0% in the direction of the process ponds. However, orebodies often occur in mountainous areas. It has been general practice to place the heaps within one or two kilometers of the orebody even if this requires extensive pad

1619

Figure 4 Valley Fill Heap Leach - Silver Heap Leach at Rochester, Nevada. Much of the process solution is stored within the heap, behind the dam which can be seen at the downslope (left) edge of the heap.

area earthworks. The Tarkwa operation in Ghana employs a three kilometer long overland conveyor to move crushed ore to the padsite. Overland conveyors of ten kilometers or longer are common in other segments of the mining industry, and can be profitably employed to move ore to a heap leach site. It is not necessary to grade the padsite to a uniform grade of one or two percent. Internal hills and valleys within the padsite can be accommodated, as well as internal slopes up to 20 percent, provided that internal drain pipes can intercept the solution and direct it downhill to the process ponds. The cost shown in Table 2 are typical installed cost per square meter of pad surface for a padsite requiring minimum preparation. If complicated earthworks are required, these may add up to $5.00 per sq m to the costs shown in Table 2. Ponds are installed downslope from the heap to provide storage of process solution. Usually there is a pregnant solution pond, a barren solution pond, and an overflowlstorm water pond. There may be one or more intermediate solution ponds (sometimes solution is recycled from older to newer heaps to build up the gold content before processing). Ponds are sized to permit storage of sufficient process solution so that the operators do not have to closely watch the pond levels. In addition to this "operating capacity", ponds are sized to contain solution during a several-day power outage and a major rainstorm event. Pond construction is similar to leach pad construction, except there is usually a second plastic liner with leak detection between the liners.

MINING, ORE PREPARATION & STACKING Mining of ore for heap leaching employs the same techniques and equipment as mining of ore to feed any other process method. Where uncrushed ore (run-of-mine ore) is placed on the leach pad, ore may be blasted very heavily in order to reduce rock size and improve gold recovery. In highrainfall environments when processing clay-rich material, it is very important to practice a mining routine that minimizes the amount of rainfall absorbed by the ore.

1620

0

0

0

0

0

0

CONSTRUCTION ELEMENT Preliminary earthworks - removal of topsoil, building of edge berms and collection ditches. Cost assumes minimal alterations to topography. Sometimes it is necessary to do extensive site preparation, at a cost of several dollars per square meter. 150 to 300 mm of compacted clay-rich soil, engineered to a permeability of cdsec. Limited leak detection, usually embedded small-diameter perforated pipes placed near the lower edge of the heap and in areas of solution concentration. These "daylight" to collection sumps at the front of the heap. Leakage is usually permitted up to a certain small limit provided the area is not extremely environmentally sensitive. Plastic liner, usually 0.75-1.00 mm (30-40 mil) thick PVC, or 1.50 to 2.00 mm thick HDPE or LLDPE. The liner is delivered in rolls up to 2000 sq. meters each, and field-welded to form the total liner. The initial installation for a "small" heap leach may cover 100,000 square meters; large installations may install 500,000 sq. meters each year. An HDPE liner of 2.00 mm thickness has sufficient strength and puncture resistance to support a heap up to 150 meters high. Geotextile Cover may be placed above the plastic to prevent damage of the plastic by rocks in the drainage layer. The use of the geotextile is an economic tradeoff with the crush size of the gravel. Drain pipes, usually 75-100 mm perforated flexible tubing, are placed on 6 meter centers above the plastic. Where solution does not drain directly out the front of the heap, large collector pipes may also be embedded in the drainage layer. Gravel cover, up to lOOOmm thick, is placed next to protect the pipes and the liner, and to provide a permeable base below the heap. Cost may be low if the gravel can be produced from the ore.

COST, $/M2

$1 .oo

$1.00 - $3.00

$0.50

$3.00 - $5.50

$1.50

$0.50

$0.50 - $5.00 $8.00 - $17.00/M2

TOTAL INSTALLED PAD COST

Table 2 Leach pad component costs Ore preparation varies widely. Run-of-Mine (ROM) ore may be hauled from the mine and dumped directly onto the heap, as shown in Figure 5. Nineteen of 36 operations surveyed had ROM heap leaches. Of these, twelve had only ROM leaches and seven had both ROM and crushed ore leaches. At the other extreme from ROM leaches, Comsur's Comco silver heap leach at Potosi, Bolivia, crushes and then dry grinds all ore prior to agglomeration, with a grind size of 50% passing 150 microns (100 mesh). Three operations (Ruby Hill, Barney's Canyon and Castle Mountain), grind high grade ore and reblend it with the ore stream going to the heap leach (at Ruby Hill and Castle Mountain, the high grade is leached in agitated tanks to partially recover the gold). Ores high in clay (such as saprolites) are typically processed by two stages of crushing using toothed roll crushers, then agglomerated in drums and stacked using a conveyor stacking system. Such a system is shown in Figure 6. Many ores are crushed and then either truck-stacked or conveyor-stacked without agglomeration. For these harder ores, crushing is usually done using a jaw crusher followed by one or two stages of cone crushing.

1621

Figure 6 Agglomerating drum and conveyor stacking system with 6 meter high heap at Ity, Ivory Coast.

1622

Agglomeration The term "agglomeration" means different things to different operators.

At the simplest level, the ore is hard but contains a large percentage of fines. Agglomeration means simply wetting the ore with water so the fines stick to the coarse particles, and do not segregate as the heap is built. At the next level, the ore contains amounts of clay or fines that begin to plug a heap of untreated ore. Belt Agglomeration may be employed. In this technique, cement and water are mixed with the ore at a series of conveyor drop points, and the mixture tends to coat the larger rock particles. The primary goal is stabilization by mixing and contact. A typical conveyor stacking system involves 10 or more drop points, so Belt Agglomeration may occur as a normal part of the process. Operations that intentionally employ drop points or slide chutes are Barney's Canyon and the La Quinua operation of Yanacocha. Where ores are nearly pure clays, such as the lateritelsaprolite ores in tropical climates, Drum Agglomeration is usually employed. The ore is first crushed finely enough (typically 25 to 75 mm) to form particles that can be a stable nucleus for round pellets. Cement and water are then added and the ore is sent through a rolling drum. The fines and the cement form a high-cement shell around the larger particles, and the rolling action of the drum compacts and strengthens the shell. Drum size and throughput are a function of several factors, but typically a 3.7 meter diameter, 10 meter long drum can process 750 tonnes of ore per hour. A 2.5 meter diameter drum can process 250 tonneshr. At the Tarkwa mine in Ghana, two 3.7 meter drums are installed to process up to 20,000 tonnes ore per day. Of the 24 crushed-ore operations responding, 11 use drum agglomeration, 5 use belt agglomeration, and 8 do not agglomerate. Fifteen use conveyor stacking systems, the remainder stack with trucks. All the operations that use drum agglomeration stack with conveyor stackers.

Truck Stacking Where rock is hard and contains very little clay, it is possible to maintain high permeability even when ore is crushed and dumped with trucks. Truck dumping causes segregation of the ore - the fines remain on the top surface, and the coarse material rolls to the base of the lift creating a highly permeable zone at the base. To control the degree of this segregation the ore may be partially agglomerated (wetted to cause the fines to stick to the coarse material) prior to placing in the trucks. Short lifts also result in less segregation. At Sterling, Nevada the problem was avoided by stacking the ore in 1.5 meter (5 ft) lifts but leaching in 6 meter (204) lifts. Truck dumping can also result in compaction of roadways on top of the heap. Several studies have indicated large trucks noticeably compact ore to a depth of two meters. To mitigate this problem, most operations rip the ore after stacking (but prior to leaching). Number of ripper passes is important; usually it is four passes in a criss-cross pattern. Some operations (Candelaria, Nevada) practised building an elevated truck roadway that was then bulldozed away. However this requires substantial bulldozer traffic on the heap surface, which can lead to permeability problems for some ores. Stacking the ore with trucks can result in the tie-up of a large tonnage of ore below the truck roadways. This is a bigger problem on small operations than on large ones, because the roadway width is nearly the same regardless of the daily production rate. For a heap leach of 5000 tonneslday, the roadways on the heap can tie up one month's ore production, with a value of $1.8 million. A conveyor system that stacks ore from the base of the lift can reduce unleached inventory to a few days' production. Because of this inventory reduction, at smaller operations where the ore is crushed, it is usually less capital-intensive to install a conveyor stacking system. For operations of 100,000 tonneslday, truck stacking is more flexible and may be less capital intensive than a conveyor system.

1623

Conveyor Stacking Conveyor stacking systems commonly include the following equipment:

One or more long (overland) conveyors that transport the ore from the preparation plant to the heap. Typically these consist of conveyors up to 150 meters long. At Tarkwa, Ghana, a 3 km overland conveyor is used. A series of eight to fifteen "grasshopper" conveyors to transport the ore across the active heap area. Grasshoppers are inclined conveyors 20 to 30 meters long, with a tail skid and a set of wheels located near the balance point. A transverse conveyor to feed the stacker-follower conveyor A stacker-follower conveyor, typically a horizontal mobile conveyor that retracts behind the stacker A radial stacker 25 to 50 meters long, with a retractable 10 meter conveyor at its tip. Wheels, discharge angle, and stinger position are all motorized and are moved continuously by the operator as the heap is built. Figure 7 shows a typical stacker system in operation. Stackers are usually operated from the base of the lift (as shown in the figure) but may be located on top of the lift, dumping over the edge. Inclined conveyors can be installed up the sides of the lower lifts, and the stacking system can be used to build multiple-lift heaps. Stackers for this purpose should have very low ground pressure tires and powerful wheel drive motors to cope with soft spots in the heap surface. Stacking systems like the one shown in Figure 7 can be used for heaps processing up to 50,000 tonnes of ore per day, but beyond that the size of the stacker (and the bearing pressure that is exerted by the wheels) becomes prohibitive. Typical cost of a complete stacker system with a 900mm (36-inch) wide belt for a 10,000 tonne/day heap leach operation, including the stacker and

Figure 7 Stacking system for capacity of 10,000 tonnes ore per day. Elements include stacker with extendable stinger; follower conveyor; cross conveyor; and several grasshopper conveyors.

1624

follower conveyors, and ten grasshopper conveyors, is $1.5 million (delivered and installed at a typical developing-country heap leach site). Three hundred meters of overland conveyor connecting the stacking system to the crusher/agglomeration system cost an additional $500,000.

Figure 8 Rahco stacker building a 12 meter lift by tripping the ore over the advancing edge. The stacker can climb ramps and turn sharply to fit project requirements. For operations stacking very high tonnages, large stackers can be mounted on caterpillar tracks to reduce ground pressure. Rahco International, Inc. (Spokane, Washington) makes a unique stacker, which is ideally suited to building large heaps at high tonnage rates. The stacker, shown in Figure 8,has individual drive adjustments so that it can climb up ramps to the next level and make sharp radius turns.

RECOVERY OF GOLD AND SILVER FROM HEAP LEACH SOLUTIONS Other chapters in this book cover the details of recovery plant operations, so this section will be limited to a brief summary of heap leach plant operating results. Basically, gold and silver can be recovered from solution by contacting the solution with granular activated carbon in columns (CIC), followed by stripping of the carbon using a hot caustic solution. This caustic solution is processed in electrolytic cells or a zinc dust precipitation vat to recover the metal, which is then melted to produce a dor6 (impure bullion) bar. A CIC plant is shown if Figure 9. Where the ore is high in silver, typically with a recoverable silver content of more than 10 grams per tonne (0.3 odton) of ore, Merrill-Crowe zinc precipitation is used instead of carbon adsorption. In this process the solution is clarified and de-aerated, then contacted with zinc dust to precipitate metallic gold and silver. This precipitate is then melted to produce bullion. Of 34 operations reporting, 28 use carbon in columns (CIC) for adsorption of gold and silver from leach solutions, and six use Merrill-Crowe zinc precipitation plants. Three of the six using zinc precipitation reported at least 9:l si1ver:gold in the bullion. Another, with 2.6 silver to 1.0 gold, produces leach solutions that are very high grade in both gold and silver content, thus justifying the choice of zinc instead of carbon. The other two process low grade gold solutions more typical of CIC plants. Average pregnant solution gold content at the 28 CIC operations is 0.70 grams gold per tonne of solution. For these operations, loaded carbon averages 3900 grams gold per tonne of carbon, with six above 5000 grams gold per tonne of carbon. Six of these regenerate 100% of the carbon

1625

after each strip cycle, eight regenerate only 50% of the carbon per cycle, three do not regenerate. Three "high grade" CIC operations, all in Africa, reported pregnant solution grades of 3.5, 3.0 and 11.O grams gold per tonne. These operations reported carbon loading of 8000, 6000 and 28,000 grams gold per tonne respectively. Stripped carbon from all operations averages 90 grams gold per tonne with 50% reporting in the range of 50 to 150 grams gold per tonne.

Figure 9 Five-stage carbon adsorption column plant (CIC plant) at Glamis Gold's San Martin, Honduras project. There are two parallel column trains (one is behind the other in this view). The plant can process up to 900 cu m of solution per hour, and is sized for an operation that processes up to 20,000 tonnes of ore per day.

DESIGN CONSIDERATIONS FOR RECLAMATION AND CLOSURE Once the heap leaching operation is completed, the facility must be closed in accordance with local environmental requirements. Closure activities are highly variable depending on the environmental sensitivity of the site, and on the regulatory regime. In general, heaps are washed for a short period of time (commonly three years), during which time one tonne of wash water or recycled treated process solution is applied. Heaps are then capped, and ponds are filled and capped. The easiest heaps to reclaim are single-lift heaps because the older heaps are abandoned early in the life of the operation and can be washed while production operations continue. In "Valley Fill" heap leaches, nearly all the ore ever placed on the pad is situated directly under active leach areas. Thus, washing of the entire heap must wait until operations are completed. Larger operations may have two or more "Valley Fill" leach areas, and can appropriately schedule closure activities. Environmental regulations usually applied in the United States call for reasonably complete washing of the heap to reduce pH, to remove cyanide, and to partially remove heavy metals. Cyanide is fairly easy to remove since it oxidizes naturally, but pH and heavy metals are more difficult to control. Regulators are recognizing that a better approach is to conduct a "limited" washing program and then to cap the heap with a clay cover andlor an "evapotranspiration" cover of breathable soil with an active growth of biomass. These covers are designed to prevent infiltration of water into the heap. After several years of active closure activities, the flowrate of the heap effluent decreases to a manageable level (or to zero in arid environments). Once the

1626

flowrate is an acceptably low level, heap closure is accomplished by installing a facility for recycling collected effluent back to the heap. A relatively small "cash perpetuity bond" is maintained such that the interest on the bond covers the cost of maintaining and operating the intermittent pumping facility as long as is necessary. A two million tonne heap of ore covering 90,000 sq. meters (average thickness 14 meters), located at Goldfield, Nevada, was recently closed with a clay/soil cap. Heap effluent gradually and steadily declined to 2.0 literdminute after 36 months. Periods of intense above-average rainfall did not affect effluent rate. While this is a small and not very high heap, scaleup of this data should be applicable for preliminary design purposes. Worldwide practice ranges from simple washing and abandonment, to the more complex procedure described above. Environmental design is an industry unto itself, and the simplistic concepts discussed here may not be applicable in other situations. Heap closure needs to be addressed in the feasibility stage of the project. Typical cost of closure, including three years of heap washing, is $0.50 per tonne of ore stacked. This can be accumulated as a deferred operating cost. However, for U.S. heap leaches, regulators may require a closure bond to be put up at the beginning of the project. The amount of the closure bond is calculated using "government-defined" guidelines that typically result in a bond of $1.00 per tonne of total ore to be placed. This adds a generally prohibitive line item to capital cost, which is one of the reasons why new project activity has declined in the U.S. in recent years. (This item has not been included in the capital cost summary presented in Table 3).

CAPITAL COST Capital cost for a small "basic" heap leach (3000 tonnedday) with minimal infrastructure at a developing-country leach site is typically $3500 to $5000 per daily tonne of ore treated, with the higher cost attributed to high logistics expenses at remote sites such as central Asia. Larger operations (15,000 - 30,000 tonnes per day) cost $2000 to $4000 per daily tonne, but may commonly reach $6000 where "corporate culture" calls for process redundancy and infrastructure. Use of a mining contractor and/or a crushing contractor is common, and may eliminate the capital costs for these line items. Capital costs for some recent installations are shown in Table 3. Glamis Gold's San Martin heap leach (built 1999) had a published capital cost of $27 million (Glamis 1999 Annual Report) and began operations at 13,000 tonnes of ore per day (equal to $2,100 per daily tonne). Ore was crushed, agglomerated and conveyor stacked. Mining equipment was transferred from another operation at nominal cost; the operation was designed with excess capacity to allow for rapid expansion to 20,000 tonnes per day. Canyon Resources Briggs Mine in Southern California, built in 1996, cost $29.9 million for 9,500 tonnedday (Marcus, 1997). This cost included $4.2 million for permitting, and a flowsheet that included 3-stage crushing. Mining equipment was leased. Adjusted for inflation to year 2002, Briggs' capital cost equaled $3,600 per daily tonne. Anglo American's Yatela Project in Mali started up at an annual rate of 7,000 tonnes per day, and cost about $8,000 per daily tonne (actual published capital cost was higher, but included extraordinary items). Capital cost breakdown is shown in Table 3 for "typical" developing-country,remote sites with minimal infrastructure and minimal redundancy. Each operation, of course, will have a unique mix of capital cost line items.

OPERATING COST Table 4 shows the breakdown of direct cash operating costs for the 27 operations that reported results for this chapter. (Direct Cash Operating Cost as used here includes all site costs including site and local office support costs, property taxes, import duties and fees. It excludes income and

1627

severance taxes, finance costs, royalties, product marketing costs, and depreciatioddepletion). These operations had an average production rate of 15,800 tonneslday. Average mining cost per tonne of material moved was $1.16, and average waste:ore ratio was 1.68: 1. Labor rates varied widely, with seven operations reporting costs below $2.00 per hour, and thirteen above $15.00 per hour. Heap leaching is not a labor-intensive process, and where labor costs are low, logistics costs are usually high. Therefore there is not an obvious correlation between labor cost per hour and total operating cost per tonne.

I

site office and service facilities) Owner's preproduction cost EPCM (engineering, procurement, construction management) Import duties / IVA Equipment I materials transport Initial operating supplies Working Capital TOTAL CAPITAL COST PER DAILY TONNE

1,700,000 700,000

I I

900,000 800,000 600,000 300,000 1,200,000 14,OOO,OOO US$ 4.700

7,500,000 2,800,000

US$

2,000,000 7,000,000 2,100,000 1,500,000 3,000,000 53,600,000 3.600

Table 3 Heap Leach Capital Costs

REPORTED DIRECT OPERATING COSTS I Mining, $/tonne I Other, $/tonne I Total 27, average I 3.11 I 4.00 2.50 1 0.88 1 Seven lowest, avg I Six highest, avg 5.90 I 8.17 I

I

I

Total $/tonne 7.71 3.38 14.07

Table 4 Direct cash operating costs for the 27 operations which reported costs for this chapter. Operating cost is not very sensitive to the size of operation. Published direct cash operating cost for Barrick's Pierina mine (85,000 tonneslday) is $3.93 per tonne, including $0.87 for mining. A recent study of an on-going operation in Africa concluded that increasing production from 4,300 tonneslday to 13,000 tonneslday would decrease costs (excluding mining) from $5.80ltonne to $5.lO/tonne.

1628

Cement for agglomeration (10 kgltonne) Cyanide, lime, other reagents Environmental Reclamation/ Closure General & administrative, support expenses TOTAL SITE CASH OPERATING COST

1.oo

1.oo

-.oo

0.30

0.30

0.30

0.50

0.50

0.50

1S O 11S O

0.50 8.30

0.30 5.20

Table 5 Typical Heap Leach Operating Costs Average operating costs for "typical" heap leaches can be broken down as shown in Table 5. Costs are shown for ores that need crushing, agglomerating and conveyor stacking. Not all items in the list are appropriate for all operations; the right-hand column shows costs, that are more typical of a 30,000 tonnelday, coarse-crushed, unagglomerated Nevada heap leach.

TRADEOFF BETWEEN LEACHING IN HEAPS AND IN AGITATED TANKS The alternative to leaching of ore in heaps is to grind the ore to a fine pulp, and to leach it as a water slurry in agitated tanks. Where a large amount of cement is required for agglomeration or where the ore needs to be fine-crushed, the operating costs of agitation leaching are not necessarily higher than for heap leaching. Heap leaching normally has significant capital cost advantages, so it is favored over agitation leaching where operating factors are identical. Combined flowsheets are also utilized. Of 37 operations reporting, four use some form of grinding or grindinglagitation leaching for part of the ore stream going to the heaps. Homestake's (now Barrick's) Ruby Hill, Nevada, operation partially leaches high grade ore in agitated leach tanks, filters the tailings and combines them with crushed ore going to agglomeration and heap leaching. Castle Mountain (Viceroy Gold) uses a similar flowsheet. Barney's Canyon (Kennecott) wet-grinds part of the ore stream, but does not leach it before adding it to the agglomerator feed. Good discussions of these combined flowsheets can be found in Lehoux (1997) and Jones (2000). As presented in the paper by Lehoux on Ruby Hill, direct operating cost of the grindtleach portion of the operation was $4.98/tonne. Analysis of the capital and operating costs presented in the paper indicate that the heap-leach-only option may have been more economic. Comsur's Comco silver heap leach dry grinds the entire ore stream to minus 105 microns (150 mesh) prior to agglomeration. In its third year of operation, Comco switched to wet grinding but it could not control water balance in the agglomeratingdrum, so it switched back to dry grinding. Six of the heap leach operations reporting also have agitated leach plants for oxide ore that run as separate "stand alone" facilities. Ore is diverted from one to the other depending on grade (or in one case, depending on sulfide content). Average nominal "cutoff grade" to the mill for these operations is 2.30 grams goldltonne (0.067 ozlton), and cutoff grade to the heap is 0.41 grams goldtonne (0.012 ozlton). In practice, the cutoff grade to the mill is a function of the ore available on a daily basis - the agitated leach plant is fed to its capacity provided the ore is of reasonable grade.

1629

CONCLUSION Although the concepts of precious metals heap leaching are simple, the practices have substantially evolved over the past 35 years. Early choices for pad materials, sprinkling systems, and stacker designs have been discarded under the pressure of operating experience and cost-reduction factors. Overall operating costs have continually declined as "superfluous" activities and controls have been eliminated. In spite of the apparent simplicity of the heap leach process - or perhaps because of it - there were many failures in the early years. There is now a large resource of successful operations from which to draw the experience needed to optimize the process. Heap leaching is expected to maintain its place as one of the principal tools for extracting gold and silver from their ores for both large and small deposits. The challenge for the future will be to remember and apply the experiences of the past.

REFERENCES Agricola. Georgius. 1556. De Re Metallica. Translated by Herbert C. Hoover and Lou Henry Hoover. 1912. New York: Dover Publications, Inc. 1950 edition. Hausen, D. M., Petruk, W., and Hagni, R.D. 1997. Global Exploitation of Heap Leachable Gold Deposits. Warrendale, PA: The Minerals, Metals & Materials Society (T MS). Jones, A., 2000. Pulp Agglomeration at Homestake Mining Company's Ruby Hill Mine. Randol Gold & Silver Forum 2000. Golden, Colorado: Randol International. Kappes, D. 1998. Heap Leach or Mill? Economic Considerations in a Period of Stable Gold Prices. Randol Gold & Silver Forum '98. Golden, Colorado: Randol International. Lehoux, P., 1997. Agglomeration Practice at Kennecott Barney's Canyon Mining Co. Global Exploitation of Heap Leachable Gold Deposits. Warrendale, PA: The Minerals, Metals & Materials Society (TMS). Marcus, Jerry. 1997. The Briggs Mine: A New Heap Leach Mine in an Environmentally Sensitive Area. Engineering & Mining Journal. Sep. 1997. Marsden, J., and House, I. 1992. The Chemistry of Gold Extraction. Chichester, England: Ellis Horwood Ltd. (div of Simon & Schuster Intl. Group). Philip, T.P. 1991. To Mill or To Leach? World Gold '91. Second AusIMM-SME Joint Conference. Victoria, Australia: The Australasian Institute of Mining and Metallurgy. PICA, USA, personal communication, Ken Thomas. website: picausa.com. Randol International. Various Symposia Proceedings. Golden, Colorado: Randol International, Inc. vanZyl, D., Hutchison, I., and Kiel, J. 1988. Introduction to Evaluation, Design and Operation of Precious Metal Heap Leaching Projects. Littleton, Colorado: Society of Mining Engineers, Inc. (SME). World Gold '91. Gold Forum on Technology & Practice. 1991. Second AusIMM-SME Joint Conference. Victoria, Australia: The Australasian Institute of Mining and Metallurgy.

1630

Agitated Tank Leaching Selection and Design Kathleen A. Altman', Mike &hafie?, StuartMcTavish3

ABSTRACT Design of the agitated tank leach circuit begins with metallurgical testwork to determine optimum leacbg conditions for specific ore types or expected ore blends from the orebody. Metallurgical test results are used along with production requirements and site specific idormation to develop a Circuit layout that includes tank sizes and arzangement. Equipment selection ( i l d i n g agitator design, tank desigs tank confugwatiosair and power requirements) is generally based on past experience with similar ore types and application. With new ore types or unusual applications, scale-up testwork must be conducted to determine equipment specifications.

INTRODUCTION In metallurgical applications, leaching is the process of dissolving a soluble m i d or metal h m an ore (Lapedes, 1974). All metals can be solubilized, or leached, in some manner or another. However, the leaching process requires a variety of lixiviants and operatiug conditions, which are dependent on the mineralogy of the ore to be processed. For example, oxide gold and silver ores can be easily leached at ambient conditions in an alkaline cyanide solutiOn. However, rehctory gold ores generally require pretreatment in an acidic pressure leach or biological leach Circuit Copper leaching normally takes place in an acidic environment. Uranium leaching is done in eithex acidic or alkaline leach solutions depending on the specific mineralogy of the ore to be treated,such as the carbonate content. Other ores may the use of ammonia or chloride solutions in order to solubilize the metals of interest For the purpose of this paper, agitated tank leaching is defined as leaching of an ore under ambient operating conditions using a recovery methd that does not incorporate extraction of the metal in the same unit operation. Using this -on excludes carbon-in-Pup,(CIP) and carbon-in-leach (CIL) procesSing. Although CIP and CIL are oftentimes considemi gpecialized cases of agitafed tank leaching, they are included in other chapters of this section that are dedicated specitilcallyto them,which is the reason for the exclusion here. The number of new mining projects that utilize agitated tank leaching, as defined here, has declined over the last few decades. This decline is due to the reduced demand for uranium,the advent of CIP and CIL processes for gold recovery and the limited number of hi#-@ oxide copper deposits that have been developed. currerdly, the most prevaleut use of agitated tank leaching is the recovery of precious metals in conjunction with counter current decantation (CCD) thickeners and the Merrill-Crowe zinc @pitation recovery process. This recovery technology is usually preferred when large amounts of silver are present either alone or along with gold However, CCDMedl-Crowe is an established recovery process that has been used sucaxMly for the treatment of ores that contain pmiominmtly gold, too. For some companies, agitated tank Ieaching/CCD/h4erriU€rowe is still the selected gold reoovery prous, particularly at remote, third world mine sites. Even though the most prevalent current application is the recovery of precious metals h m alkaline Cyanide leach solutions, agitated tank leaching is a recovery method that is important in a number of apPiications. This section deals with a generic discussion of agitated tank leaching. The discussion includes a description of the process; an outline of how specific processes are selected for new projects; and details regardugthe design of agitateajleach circuits 1 Independent consultanf Denver, Colorado 2 Coeur Rochatex, Inc., Lovelock, Nevada 3 SNC-Lavalin Engineers & Constructors, Inc., Toronto, Ontario

1631

The process of selecting and designing processing circuits should be a collaborative effort between the operating company, metallurgical testing hcility, engineering company, and equipment suppliers. Process selection and engineering design is but part of the iterative process that takes place during the development of new projects. During the prelimimy stages of the development of a new projed, cost data that were generated for similar projects in the past is modified, as required, and incorporated into a financial model in order to select the option that is most attractive from a financial perspective. It is important to understand that selection of processing alternatives is seldom done on a “stand alone” basis. AlI of the costs associated with the project are developed and compared in order to make a f d selection For example, the pre- and posttreatment options may be quite merent depending on the recovery process that is selected. A heap leaching process may or may not require a crushing circuit, whereas agitated tank leaching circuits generally requireboth crushing and grinding circuits. The cost differences plus the impact on both t k total metal recovery and the time of recovery must be incorporated into the financial model in order to make decisions based on accurate information Project decisions are generally based on successively more accurate data as a project proceeds through the development stages. Initial selections may be based on order of magnitude estimates with an accuracy level as low as plus or minus 50 percent, or less. As m01p: detailed information is acquired, deckions are generally based on data that has been developed to successively higher levels of accuracy. The final decision to proceed with a project is normally basedon a feasibility study that has accumulated sufficiently detailed data to complete an economic evaluation to a level of accuracy in the range of plus or minus 10 to 15 percent At this stage, a process flow diagram has been selected and froze4 equipment has been selected and fum, fixed quotations have been solicited from equipment mamfhcturers.

PROJECT DEVELOPMENT Ideally, the process of choosing an extraction method begins in conjunction with the process of delineating an ore body. Numerous stages of assessment are conducted along the way to operating a mining project successfully (Altman, 1999). The speci6ic process of developing a new project, including the selection of an appropriate metal recovery process, is highly dependent upon the corpomte philosophy of the company completing the development and their internal procedures and evaluation criteria If a company relies on outside financing or listings on a public stock exchange, the development process typically becomes more rigorous and it is subjected to external reviews and legal requirements in addition to corporate requirements. Multidisciplinary project teams are required to develop mining projects. Key disciplines include geology, mining and metallurgy. While an ore resource may be large and interesting h m a geological perspective, it is not an ore body until it is proven that the material can also be mined and processed economically. Development of mines is an iterative process. Each assessment stage requires the acquisition and evaluation of data culminating in a specific decision to proceed with the project; delay the project development; or abandon the project. Each decision point is based primarily on an economic evaluation that is based on the available data. Early assessments may be based only upon amlytical information a d historical cost data for similar Projects that have previously been evaluated Typically, each time a decision is made to proceed with a new project, additional funding is allocated so that moce detailed information can be generated and evaluated It is imporQnt to recognize that process development does not occuf in isolation As metallurgical data is gemted, it is entered into an economic model along with other data specific to the project. Routinely this is a discounted cash flow @CF) model anda decision is based on the internal rate of return (IRR) or the net present value (”V) or both HISTORY In the early 1890’s, C.F. Brown developed an air-agitated tank used to leach gold ores in New W a n d (Hen 1985). This tank later became known as a Pachuca tank because of its popularity in Mexico. In the early 19OO’s, Pachuca tanks were the staodard leaching vessels in the mining industry. By the 196O’s, mechauical agitation had proven to be more economic than air agitation and exjsting PaChuGi tanks were graduaUy remfiaed with agitators while new plants were designed with mechanically agitated tanks. Over the last several decides, the use of agitated tank leaching has become much more sophisticated. The knowledge of agitator function has increased to the point where specific agitators are designed to

1632

improve the efficiency of the various agitator functions including air &ear, slurry pumping, e g , and suspension of solids. These improvements have made choosing the proper agitator for a specific ore body a more onerous task, which often requires expertise from consultants, engineering firms, and agitator manufachum. Experience has shown that a properly designed leach Citcuit will operate for years with little maintenance or attention whereas a poorly designed circuit will have many operational and mechanical problems that are costly to resolve.

AGITATED TANK LEACHING - PROCESS DESCRIPTION The concept of agitated tank leaching is fairly simple: An ore is prepared (this typically includes crushing, grinding,and pH conditioning); placed in an agitated leach tank with a leaching agent; and the metals are allowed to leach from the ore into the solution. Once leaching is complete, the “pregnant” solution is separated from the slurry with filters or CCD thickeners and procesed through a metal recwery system. Merrill Crowe is typically used to recover gold and silver whereas solvent exsdcton and electro-winning are used to recover copper. Although the process seems simple, the design engineer must have a solid understanding of the ore body, the ore characteristics, and the leaching process in order to design a successful c i d t Some primary operating parameters, which directly affect the performance of the agitated tank leach citcuit include: Grind - The ore must be ground to a liberation size thatexpwes the desired mineral to the leaching agent, as well as a size that can be suspended by the agitator. Grinds coarser than 65mesh tend to cause excess abrasion and wear due to the degree of agitation required for suspending the coarser particle. Ores that can be successfully leached with a grind coarser than 65mesh are typically good candidates for heap leaching. Slurry density - The slurry density Wrcent solids) determines retention time. The settlingrate and viscosity of the slurry are functions of the slurry density. The viscosity, in tum, controIs the oxygen mass transfer and the leaching rate. Number of tanks- Agitated tank leach circuits are typically designed with no less than four tanks and preferably more to prevent short-circuitingof the slurry through the tanks. Dissolved oxygen - Air or oxygen is often injected below the agitator to obtain the desired dissolved oxygen level. Reagents - Adding and maintaining the appropriate amount of reagents throughout the agitated leach circuit is critical to a successfuloperation. Adding insu€licient q d t i e s of reagents reduces the metal recovery but adding excess reagents increases the operating costs without recovekg enough additional metal to cover the cost of the reagents.

PROCESS SELECTION General Consideratiom In the case of agitated tank leaching, the initial screening may be based entirely on analytical data. For example, the ratio of the value of a cyanide atomic absorption (AA) assay to a fire assay or an oxide copper assay to a total copper assay indicates whether the ore type is amenable to leaching or not. It also pmvides

an indication of the recovery that can be expected in a leaching circuit. The main extraction methcds that are considered for non-refhctory, i.e. amenable to leaching, ore bodies are agitated tank leaching, heap lac-, CP,CIL, and gravity concentdon. Process selection is not only related to the metallurgical performauce of the ore contained in an ore body but also the overall size of the ore body. Drill samples that are used to delineatethe ore body can also be used for preliminary metallurgical testing.Following an analytical detmmah‘011 that the ore body is amenable to leaching, the first step in a metallurgical test program, for ore bodies that are contemplating some type of leach recovery, is usually bottle roll tests. These tests begin to provide information about the relationship between particle size and recovery and time versus recovery in addition to information about reagent consumptionrates. As a rule of thumb, small ore bodies cannot support the highex capital costs associated with milling circuits. Therefore, l a r g e r d e testing of samples may be Limited to heap leach column tests until such time as suflticient ore reserves are delineated to support the cost of a milling cinxit, such as an agitated tank leaching circuit At this time, additional tests are conducted to determine if the ore is amenable to gravity

1633

concentration and to ascertain the liberation size, leach retention time, and reagent addition rates, as well as crushing and gmdmg characteristics, settling rates, $ltration characteristics, etc. It is impoaant to remember that accurate test results mean nothing unless the samples that are tested are representative of the ore that will be processed For this reason, care must be taken to ensure that all expected ore types are sampled and tested. This is one d v i t y that must be closely coordinated between the geologists, mining engineers and process enginem. One of the W steps in a metallurgical test program for a project that has selected an agitated tank leaching circuit is generally the operation of a continuous pilot plant that incorporates all of the unit operations that will be used in a full-scale plant. These tests are used to confirm and refine the test results that have been generated to date. Results for the operating parameters required in the engineering design and for the economic analysis are generated during thistest campaign. The cdmimtion of this testing is the tbat will be used for a hll-scale plant. selection of the process flow diaThe iinal selection usually includes consideration of project elements that are not easily quantifiable on an economic basii, such as environmental and political risks, the corporate financial position, and the company’s experiences and preferences with regad to operating circuits. Therefore, the final process selection may or may not be the option with the highest NPV or IRR

Heap Leaching versus Agitated Tank i.ea&ing Process Description. Modem heap leaching was first used in the North America uranium industry in 1950. The methods were adapted to the copper iradustry in the 1960’s and the gold industry in the 1970’s (Scheffel, 2002). In heap leaching, ore is placed on an impemeable liner and a leaching agent is percolated through the ore and recovered on the liner. The ore may be placed on the pad as run-of-mine, crushed,or crushed and agglomerated merial.The pregnant solution f b m the heap is treated through a metal recovery plant, and the barren solution that exits the recovery plant is re-circulaied back to the pad. Capital Costs. The capital cost for a heap leaclung circuit is generally less than the capital cost for an agitated tank leachmg circuit. This is because a heap leach operation typically requires less equipment and infrastructure. A typical heap leach operation consists of a leach pad, crusher, metal recovery plant, truck shop, officecomplex, and solution distribution and collection pumps. A typical agitated leach tank plant consists of a m h e r , grinding mills, thickeners, leach tanks, filters or CCD thickeners, metal recovery circuit, tailings faciky, truck shop,and an office complex. Operating Costs. Heap leach operating costs are generally lower per tonne of ore processed than the operating costs associated with an agitated tank leach circuit due to the r e d u d equipment and manpower requirements. In addition, heap leaching is typically conducted at a coarser size fiaction than agitated tank leaching, thus reducingthe power requirement and reagent consumption Recovery. Although a heap leach operation is less expensive to build and operate, metal m e r i e s are typically lower than recoveries in agitated tank leaching circuits due to the coarser particle sizes and less eflicient contact of the leaching agent with the ore. In addition, the time required to process the ore is greater for heap leaching, which, in tum, impacts the inventory of metal in the circuits and the rate at which the metal is recovered. A typical gold heap leach will have an average leach cycle of 90-180 days to reach the targeted recovery, and copper heap leach cycles can be anywhere from 180-360 days to yield targeted recoveries. Generally less than 72 hours are required to reach targeted recoveries in agitated tank leaching circuits.This recoveryast relationship makes k i p leaching the best alternative for low-grade ore bodies. When the head grade is higher, the increased recovery and rate of recovery in addition to a lower metal inventory generally makes agitated tank leaching the better option Combinations of recovery circuits may also be selected. It is common for large ore bodies with both high and low grade ore to maximize the economics by instalIing both heap leadung and agitated tank leaching recovery circuits. In recent years some smaller ore bodies, including Kennecott’s Barney‘s Canyon Mine and Homestake’s Ruby Hill Mine, have been developed with an integrated mill and heap leach flowsheet. In this flow sheet the leached mill tailingsare agglomerated with low-grade ore and placed on a leach pad. This improves recovery, allows the economic recovery of lower grade ore, and eliminates the need for a tailings pond Additional Considerations. Besides the obvious economic reasons, several environmental and site specific concerns should be considered when selecting between heap leaching and agitated tank leaching. These include: 0

Tenain - Is there a suitable area and location for a heap leach pad or mill tailings facility?

1634

0

0

0

0

Ambient tempetature - Will the project experience extreme cold or heat that causes freezing or excessive evaporation? A n n 4 precipitation - What is the annual precipitation and precipitation cycle? Periods of excessive precipitation can cause dilution to the solution and may make it diflicult to contain the i n d solution volume within the processing circuit without &stamdl. y increasing the size. Location - Is the ore body near a national park or forest, or metropolitanarea? Wildlife - What wildlife exists and what measures must be taken to protect it? Local and federal mining regulation - Are there any Werences in regulations for heap leach or agitated tank leach pmcessing circuits? Local and federal permittingregulations - Are there any differences in permitting requirements for heap leach or agitated tank leach processing circuits?

CIP-CIL vems Agitated Tank Leacbmg Pmess Description. The first industrial CJP plant was installed at Homestake’s South Dakota operation in 1973 (Fleming, 1998). A CIP circuit utilizes the same flowsheet as an agitated leach circuit up to the point where the slurry discharges from the final agitated leach lank. At this poinf, the precious metal values are recovered directly from the slurry onto granulated activated carbon in agitated CIP tanks. The carbon is retained in the tanks by any one of several Merent types of screens through which the slurry discharges. CIP circuits are typically designed with at least four tanks to prevent short-circuitingof slurry and allow sutlticient retention time for recovery of all the metals. Although agitated leach tanks are used before CIP tanks,CIP and CIL circuits are considered as a separate process in sections D-1-4 and D-1-5. CIL circuits are similar to CIP circuits with the exception that leaching and extmction OCCUT simultaneously in agitated leach tanks that also contain cabon and are equipped with &n retention

screens. The evaluation of agitated tank leaching verses CIP and CIL circuits is not as complex as the heap leach-agitated tank leach analysis. CIP and CIL circuits generally have lower capital and operating costs for gold ore bodies than agitated tank leach circuits.Silver ore bodies show better economics with agitated tank leach-MeniU Crowe circuits. This is because the volume of carbon that would have to be processed to recover economic levels of silver would inaease the capital and operating cost of a CIP or CIL circuit above that of an agitated tank leaching/CCD/MeniU-Crowe circuit.

PROCESS DESIGN Design Procedure Once agitated tank leaching is selected, and a decision is made to proceed with the project, a detailed, project-specificengineering design is completed. Tank Design

Time R e q u i d (Tank Sizing). For any leaching process, including agitated tank leaching, the first. and foremost process design criterion is the amount of time required to complete the leaching process. The leach time required for a new circuit is determined by leach tests conducted on representative samples of the ore. If less than the optimum time is specified, the amount of metal dissolved may be insutticient On the other hand, if too much time is spe&ed, the incremental recovefy increase may not justifv the cost of largertanksand agitators Slurry Density. The optimal slurry density is defined by metallurgical testing. The density of the slurry directly affects the retention time for a circuit The higher the percentage of solids, the smaller the volume required to achieve a specified retention time. Ideally,sluny produced in the @ding circuit for an agitated tank leach circuit will report to a thickener ahead of the agitated leach tanks. The thidcener heips to e m that the feed to the leach circuit is consistat, which, in turn, eosures optimal operation of the leaching process. The optimum slurry density is directly affected by the grind size and the viscosity of the slurry. Plant Throughput. Obviously, the tonnage of ore to be processed in a plant has a direct correlation with the size of the tanks that are required The size of a processing plant is normally determined in conjunction with geologists who are familk with the size of the orebody and mining engineers who have

1635

helped to select an appropriate mining method for the orebody. This information is input into appropriate financial analyses in order to select the optimal mining and processing rate for the project. Number of Stages of Leaching. In general, a single, large tank is less expensive to fabricate than a number of smaller tanks with a combined volume equal to the volume of the large tank (von Essen and Ricks, 1999). However, the agitation process is such that short-circuiting is likely to OCCUT ifa single tank is used. As a rule of thumb, the use of four or more smaller tanks minimizes the short-circuiting problem. However, the number of leach tanks incorporated into a process design is also dependent on the profile of the leach solutions and solids in the tanks. A final important consideration in the design p m m s is that an “extra” leach tank may be included in the design so that any of the agitated leach tanks may be bypassed for maintenance without reducing the residence time. Sample Calculation. The nominal required tank volume is calculated by multiplying the volumetric flow rate of slurry through the circuit by the amount of retention time required. This volume is then divided by an Effective Volume Factor, which makes an allowance for the volume associated with a d o n , settling, agitation, etc. Additionally, an allowance for &board may be made in the tank height that is used to provide the requml tank volume. For a 1,OOO t/d plant with an ore specific gravity of 2.8 operated at 50% solids, and req-g a 24-h retention time, the nominal volume required is determined with the following equations. (tonnes / day / 24 hours / day) = tomes / hour (1,OOO t/d) / (24 h/d) = 41.7 t/h slurry volume / hour = (solidsm3h0ur + water m3/hour)or

(solids t / h / sg ore) + (((solids t / h / % solids) - solids t / h) / sg water) [(41.7 thY(2.8 t/m3)] + [((41.7 t/h)/0.5)-(41.7 tih)/(l t/m3)] = 56.5 m3/h tank volume = slurry m3 h * retention time (56.5 m3/h) * (24 h) = 1,357 m3 An appropriate Effective Volume Factor is 0.92, which brings the required volume to 1,475 cubic meters [(1,357 m3Y(0.92)]. Assuming that a total of five leach tanks provides an acceptable leach profile, the volume per tank is 271.4 cubic meters. [(1,475 m3)/(5)] Tank Height to Diameter Ratio. In general, the mining industry uses cylindrical tanks with a tank height to diameter ratio of 1:1. That is, the tank height, h, equals the tank diameter, T, i.e. h = T.

tank volume = d h

(T/2)%

= 271.4 m3

‘IC [(15)/4] = 271.4

m3

T=7m In addition to this, the tank height may be i n a w e d to 7.6 m to allow for additional freeboard. It should be kept in mind that this calculation method determines the “average” residence time for the slurry in the tank. Oldshue (1%3) reminds us that all particles sizes are not necessarily retained for the same residence time. The size of a solid particle in slurry may &ect the residence time. The impact of this phenomenon will be discussed further in the section on agitation requirements and design. Tank B a i n g Generally, agitated leach tanks must be b a f n e d Standard batning includes four equally-spaced baf€les that are one-twelfth of the tank diameter (Oldshue, 1%3). However, baffle requirements are related to mixer torque, so they are dependent on the selected agitator (Sakman, et al. 1983).

1636

Tank Configuration.Agitated leach tanks are usually designed so the slurry can flow by gravity from one tank to the next Building each tank taller than the subsequent tanks and putting a false bottom in the tank to maintain the required tank volume is a common design If the site topography allows, the tanks m y be built into a hillside to ammoodate gravity flow. Materials of Construction.The materials of comtrudion for the equipment, including tanks, agitator shafts, and impellers should be selected for the environment of the leaching process. Unlined mild steel tanksare suitable for the alkaline, cyanide leach solutions used in gold leaching for short term operations; however a chemical resistant liner should be considered for longer term operations. The impellers may be rubber-bed to minimize the abrasion due to the particles in the slurry. A leach process that utilizes sulfuric acid may require stainless steel tanksand agitator components.The choice of the materials of construction impacts the longevity of the equipment and the maintenance requirements in the future. Reagent Additions The reagent addition systems are dependent on the specific reagents that are required and whether the reagents are added to the slurry prior to the time it enters the leach circuit or whether the concentrations of reagents need to be trimmed as the s h q advances through the leach circuit. For example, in a gold cyanide leach circuit, the pH of the slurry may be initially controlled through the addition of lime on the feed belt to the g i m h g circuit a d the grinding may be done in sodium cyanide solution. In this case, it is good engineering practice to provide for trim additions of slaked lime and cyanide in the leach circuit so that the reagent additions can be optimized The determining factors to consider in designing reagent addition systems are the sensitivity of the leaching process to the concentration of the reagents and the amount of variability that is anticipated in the ore that will be -P In many leaching p r o a ~ it~is~neGessary , to add oxygen, generally in the form of air,to the leach circuit. This may be done with air blowers or with low pressure air compressors depending on the volume of air and the delivery air pressure that are r e q d In some cases,it may even be necessary to inject oxygen into the leaching process, which d t s in an even more sophisticated gas production andor storage and injection system.

Agitation Requirements and Design Miring Requiremeots and Peaformaace. Agitators are used as a mixing device in the agitated tank leaching process. Mixing can be divided into five Merent areas (Oldshue and Todd,1981): 0

0

0

Liquid-solid dispersion Gas-Iiquiddispersion Liquid-liquid dispersion Blending of misC%le liquids production of fluid motion

The performance of mixers is evaluated by: 0 0

The physical uniformity of the contents of the tank Mass W e r or chemical reaction

The mixer design can be divided into three elements. First, the prooess design, which includes the fluid mechanics of the impellers, the fluid regime required by the process, scaleup from laboratory or pilot scale to plant scale operations, and s k k i t y . Second, the impeller power characteristics that relate power, speed, and impellerdiameter. Third, the mechmical design that includes impeller design, agitator shafts, the drive assembly and the support structures (Oldshue and Todd,1981). In agitated tank leachingthe main requirements of an agitatorare: 0 0 0

Solids suspension Oxygen transfer by gas dispersion Maintaining the chemical composition and physical uniformity (Fraser, d al. 1993).

1637

Choke of Impellers Impellers are divided into two basic categories. These are radial flow impellers and axial flow impellers. Radial flow impeller designs include the Curved Blade Turbine (CBT), Flat Blade Turbine (FBT), Smith Mine, and Rushton turbine (Guide to impeller selection for maximum process results). As the name implies, the fluid is discharged in a horizontal or a radial direction to the tank wall (Oldshue, 1987). The flow pattern created by radial flow impellers rotates around the impeller. If lowviscosity liquids are agitated in un-baftled tanks,a vortex is formed and the liquid swirls around the vortex (Rushtonand Oldshue, 1953). Axial flow impellers include the marine propeller and the Pitched Blade Turbine (Oldshue, 1983) in addition to a number of hydrofoil designs (Guide to impeller selection for maximum process results). As the name implies, axial flow impellers create a flow pattern that is p a d e l to the impeller shaft (Oldshue, 1983). Radial flow impellers have been used traditionaLly to dispefse gas, which meets the agitated tank leaching requirement for oxygen transfer. Axial flow impellers, then, produce higher, more efficient fluid flows than radial flow impellers (Olderstein, e t al., 1989). The flow, or pumping, capability of a@tators is responsible for meeting the other process requirements of agitated tank leaching, i.e. maintaining solids suspension and the physical and chemical uniformity in the leach tanks Shear is the impeller charactexistic that is responsible for gas dispersion. Because the process of agitated tank leaching reqairesampoxtents of both flow and shear, it is important to understand the basics of good agitator design in order that a reasonable compromise is achieved. The design must optimize the capital cost of the agitation system with the operating costs resulting from power consumption, yet assurethat the process results meet the required criteria. In order to achieve the requiredresults,the mixer design must incorporate a balance between shear and flow (Kubera and Old.shue, 1992). The design of agitation equipment has advanced dramatically over the past half century, or so (Kubera and Oldshue, 1992; Oldshue, 1989). With the m n t advent of the inkmet, agitator suppliers have made technical data and inqujl sheets, technical and reference information, aad design calculations instantly available from our desktops. This eases and enhances coordination,cooperation and understanding between the various individuals involved in the design proas. As stated earlier, Pachucas were initially used to meet agitation rqukments. Shaw (1982) pmvides a good description Air lift agitation was commonly used in Pachuca tanks.A Pachuca tank has either a 900 or a 600 conical bottom and a height to diameter ratio between 2.5 and 3. An air lift tube of approximately 10%of the tank diameter is placed in the center of the tank and air is injected into the bottom of the air lift tube.This produces a flow of slurry up the tube at high velocities As processing rates increased and equipment capacities became larger, Pachuca systems became expensive due to the cost of cone bottom tanks, c o m p ~ r s and , piping. They also use a high amount of energy for the level of agitation that is produced. The alumina industry used Pachuca tanks for precipitation in the Bayer process. In the 196O's, the alumina industry began to investigate the design of mechanical agitation systems. Mechanical draft tube circulators are the results of these efforts. The concept s p d from the alumina indushy to a variety of other imlustriesincluding gold leaching (Shaw, 1982). Subsequently, the role of micro- and macfoscale mixing has continued to evolve and the development suchas composites, has removed the limitation of forming impellers of advanced materiaisof co-on, from only flat stock (Oldshue, 1989). These advances have enabled a new generation of impeller design. Wow Relationships. Oldshue (1983) explains that the power (P) applied to a mixer produces both flow (Q) and impeller head 0.The head is proportional to the velocity head of the slurry, or fluid that is being mixed NomaUy, the total flow is e x p d as mass flow, e.g. t/hThe total power is proportional to the total flow and the impeller head, just as it is in a pump. PocQH

However, an agitated tank is not a confined channel such as a pump and piping system Therefore, the determioation of impeller flow and head for an agitation system are not as quantitative as it is for a pumping system At constant power, the flow to head ratio is proportional to the impellet diameter @) and the tank diameter (T) mtio. The equation used to express this relationship is:

1638

The impeller head is also directly related to the square root of the fluid shear rate. Therefore, it is a measure of the flow to fluid shear rate around the impeller. Practically speaking,these relationships mean:

0

A large impeller operating at a slow speed produces high flow and a low fluid shear rate A small impeller operating at high speed produces a high shear rate and low flow.

Walter (1%8) provides further explanation of some additional relationships. Power @) is proportional to the speed (N) and the diameter of the impeller @) according to the following relationship:

P ~ N W

(3)

An4 the pumping capacity (Q is proportional to the impeller speed (N) and the impeller diameter @) amrding to:

QaW

(4)

Combining equations (3) and (4) provides the following relationship:

These equations show that for a constant power level, a large diameter impeller has a higher pumping capacity than a smaller diameter impeller and it will run at a slower speed. Also, the larger the impeller, the lower the power required to achieve a specified pumping capacity. However, even though the power is reduced and the speed is decreased, the torque increases. The capital cost of an agitation system is d i d y related to the torque requirement. Obviously, the power requirement of the system is primarily an operating cost. Therefore, an economic evaluation is needed to determine the optimum solution to be used for a specific application. Assume that two options are available to produce identical process results. They are a 37kW (50-hp) agitator and a 30 kW (40 hp) agitator. The 30-kW unit has a larger impeller and it operates at a lower speed Thdm, it needs a more expensive drive due to the higher torque that is prod~~ced. Assuming a 24 hld, 7 d/wk operation and an operating availability of approximately 95°C~the selected unit will operate for about 8,300 b/yr. For simplicity, assuming full power draw, the largex unit d c o m e roughly 58,250 kwh of additional electrical energy per year. [(365 d/y)(24 h/d)(O.95)(37-30 kW)] If a power cost of $O.O8kWh is anticipated, the difference in annual power costs is $4,660. This increased operaling cost then needs to be used as an input to the projects economics. For a 2 year payback, the smaller, less expensive agitation unit will be attractive as long as the difference in the initial installed capital cost is more than $9,320 per unit. Solids Suspension. In gold recovezy processes, the suspension of solids predominates the agitation application. Solids Suspension is a flowcontrolled application, which means that the process results are directly p ~ d o n atol the pumping capacity of the impeller (Kubera and Oldshue, 1992). Oldshue (1963) presents a comprehensive review of solid suspension in hydrometallurgy. Solid suspension processes can be separated into two areas:

-

Hinderedsettling Freesettljngsolids.

If the slurry density is high enough that the particles interfere with each other during settling, it is considered hindered settling. Hindered settlingincludessettlingvelocities of less than 0.3 dmin (1 fUmin). Generally, this occur^ in slurries with greater than 50% solids by weight but it is also a function of particle size and fluid viscosity. In M e m i settling appkations, it is more appropriate to consider the process criteria as fluid blending or fluid motion. Although this may be the case for agitated tank leaching, it is more likely that the application will be free settlingsolids.

1639

In applications governed by free settlimg solids, the equipment design is highly dependent on the settling velocity of the solids. Percent suspension is an important consideration in solids suspension applications. It takes into account variations in the content of the leach tanks.

Percent Suspension =

mint Wt % solids at the -1e Wt % solids in the tank

x loOo/o

It may be necessary to consider three or four separate size fractions if the particle size distribution for the process is very large. The required process criteria for solids suspension can be subdivided into a variety of categories, including: 0 0 0

0

Completeuniformity Complete off bottom suspension Complete motion on the tank bottom Filleting permitted but progressive fillet build-up not allowed A specified height of the suspension.

Complete uniformity implies that percent suspension is 1W/o at any point in the tank. However, practically spealang it is very difficult to achieve 100% suspension in the upper layer of liquid in a tank.If the settling velocity of the particles is greater than 1.8 d m i n (6 Nmin), this dficulty is confined to the upper three percent of the tank because the horizontal flow patterns that are created by the agitator cannot keep the solids with a high settling velocity suspended. If a sampling point is placed near the top of a tank,a m e of power (kW or hp) versus percent suspension can be plotted. On this curve, there is a breakpoint where, even though there may be a large increase in the power, there is little improvement in the percent suspension This breakpoint on the curve can serve as a practical definition of Complete Uniformity. It should be kept in mind that this point also depends on the depth of the sample point. The other process criteria listed are relatively self
1640

Mass Transfer. Leaching is the p r o a s of dissolving a specific component, or metal, out of a mixture of ’ components in the solid. Therefore, leaching is the mass transfer of that component from the solid state to the liquid state. It is critical that the particles are small enough to e m that the liquid has access to the metal of interest. The rate of leaching is equal to the sum of the Concentration driving force (AC), the exposed interfacial area (A) and the mass transfer coefficient &).That is: Leaching Rate = AC k, A The mass transfer coefficient is a function of the mixing system. The percentage of the specific component leached fkom the solids is leaching efficiency. Stage efficiency is the amount of the component that is leached compared to the amount that it is theorebically possible to leach assuming that equilibrium is reachedThe initialparricle size determines the interfacial area (A) unless the particle size changes dramatically due to a large percentage of the solids being leached. The concentrations of the reagents and the concentrations of the leach components in both the solids and the leach solution determine the concentration driving force (AC). The power (P), the impeller diameter to tank diameter ratio (Dmand the impeller type all have an effect on the mass transfer coefficient &)The leaching rate is generally determined by batch laboratory experiments, at least M y . In batch leaching experiments, the residence time for all particles is the same. As discussed previously, this is not necessarily true in a continuous operation because the feed is rapidly dispersed throughout the tank volume in agitated tanks. Even though a high degree of nonconformity may be present in the solid suspension, the process result is the important outcome and good leaching results can still be achieved. Running batch experiments on individual particle size fractions can piwide an estimate of the individual mass transfer rates (Oldshue, 1963). Gas Dispedon. In leaching operations, oxygen is sometimes required to complete the chemical reaction for the process. Oxygen is often supplied by the injection of air, however, oxygen gas may be used in some instances. Other gases may also be required in some instances.This process area falls into the category of gas-liquiddid mass transfer. Leaching applications, which rely on gas dispersion, include ammoniacal leaching of nickel, the leaching of iron out of reduced ilmenite in synthetic mtile production, and neutral leaching of zinc calcine in addition to the leaching operations that have been mentioned previously (Fmser, 1992). Traditionally, radial flow impellers have been used for gas dispersion. Wal flow impellers typically have high shear characteristics (Optimizingand scaling up mixingsystems). Once again, Oldshue (1983) provides a comprehensive discussion of gas dispenion in kpkk. Dispersion of gas in a Quid is most sensitive to the design of mixing systems. This is because the mixer design innuences both the interfacial area and the mass transfer coefficient. The relationship between the mixer power and the isothermal expansion power of the gas determines the type of dispersion that is achieved- If the mixer power is less than the gas power, the gas rises unhindered and “blurps” on the surface. In this case, the mass transfer rate may be high enough even though the gas is not dtspersed throughout the tank. As the impeller power incmseq the gas dqersion improves. When the two power levels are about equal, minimum dispersioa occurs,but the flow pattern is still gasantrolled To achieve mixercontrollled flow, the mixer power must be two or three times higher than the energy of the gas. When the mixer power is greater than three times higher than the gas power, intimate gas dispersion is achieved. To drive the gas bubbles to the bottom of the tank,the power must be two or three times greater than the power required for intimate dispersion At this point,the mass transfer coefficient is controlled by both the agitator power and the gas flow. The impeller controls the bubble size and the i n t e f i d area, so the design of the sparge ring is not a major concern The numbef of holes, the size of the holes, and the direction of the holes are not important. E v a an open pipe can be used ifplugging occursor corrosion is a problem. It is not appropriate to use upward pumping axial-flow impellers. An open radial-flowimpeller without a disk allows the gas to enter the low-shw area around the hub. The gas, then, passes through the impeller without going into the high shear mne at the tips of the impeller and the recitculatonof the gas occurs at the tips. Ifa sparge ring is used, it should be about eight tenths the diameter of the impeller so the gas can enterthehigh-shearmne.

1641

The k-factoris a measure of the ratio of the power drawn by the impeller with the gas turned on, to the power drawn with the gas turned off. An impeller operating in a liquid without gas draws more power than it does when it operates at the same speed in liquid with gas because the 5uid density changes with the addition of gas to the liquid. If the power draw does not decrease when gas is added to an agitated tank, it indicates that the gas is not being effectively dispersed in the tank (Rushton and Oldshue, 1953; Oldshue, 1983). The k-factor varies with the gas rate. In order to compensate for the significantly Merent power demand with and without gas sparging, the agitator should be designed with an interlock to prevent operation without gas sparging or a twospeed motor should be used The motor rating is selected for the lower power that is required when the gas is on Then, a switching gear changes to the lower motor speed if the gas supply stops (Oldshue, 1983). A combination of impeller typesmay be used to achieve the desired process results. Agitation in liquid-gas applications, such as agitated tank leaching, also involves inass transfer as discussed previously. Commonly, the design basis would be the mass transfer rate, which has units of mol/[(m3)(s)J. In some simple systems the gas dispersion requirements may be adequately expressed as a quantity of gas per unit of time. When the gas phase and the liquid phase are mixed well, the concentfation driving force (AC) is the difference between the partial pressure of the gas leaving the tank and the concentrationof the dissolved gas in the liquid leavhg the tank (Oldshue, 1983). Oldshue (1983) goes on to show the following, which impact the mass transfer: 0

0 0 0

0 0

There is an optimum impeller diameter to tank diameter ratio (DK)for different ratios of gas flow to mixer power. The optimum ratio of sparge ring to impeller diameter is approximately 0.8. Impeller positioning has very little impact As the viscosity of the liquid kreaseq the bubble size must increase in order for the bubbles to flow upward at a specilied rise velocity. Higher viscosity reduces the mass transfer coefficient. Tall, thin tanks require less power to sparge a specilied volume of air than short, squat tanks with an equal volume.

Lmpeller Shear. Even though the impellers used for gas dispersion typically have high shear characteristics, shear is not a particular concern in agitated tank leaching. It is, however, a major consideration in recovery processes that utilize activated &n. For this reason, the discussion on shear is included in Section D-1-5, CIP/ClL/CIC Adsorption Circuit Equipment Selection and Sizing. SUMMARY AND CONCLUSIONS Agitation requirements for tank leaching requixe characteristics of both axial flow and radial flow impellers. Agitator manufacturers have developed mixing systems that are specilically designed to meet the process requirements for the s p e d & application (von Essen and Ricks, 1999; von E s n , 1998; Fraser, et. al., 1993; Fraser, 1992; Kubara and Oldshue, 1992; Olderstein, et. al., 1989; Lally, 1987; Salzman, et. al., 1983; Shaw, 1982; Oldshue et. al, 1988). Since agitated tank leaching is an established process, the design can generally be based on previous successful projects and existing data (Oldshue, 1%9). However, it is impoaant to properly define the fluid properties and to c o d y specify the desired process results. A proper balance between flow and shear is requiredto achieve the correct design (Fraser, 1992). As discussed previously, imp&& considerations in the design of an agitated leach circuit include the effects of particle size on the actual residence time achieved in the circuit and the relationship between the mass transfer rate and the quantity of gas that must be dispersed in a given time. The current understanding of agitation technology allows the design engineers to select the optimum design based on an economic comparison The best choice is the correct balance between higher power agitators that have a lower initial installed cost and lower power units that have a hi@ initial installed cat. REFERENCES Altman, K A . 1999. Model for Developing a MetallurgicallySuccessful Project. PhD. diss, University of Nevada,Reno, Nevada

1642

Fleming, C.A 1998. Thirty Years of Turbulent Change in the Gold I n-. CIM Technical Paper November/ December, 58. Fraser, G.M. 1992. Gas dispersion and mixing for m i n d oxidation reactors. conference on Extractive Metallurgv of Gold and BaseMetals The Australian Institute of Mining and M d w a . 293-301. Fraser, G-M, P.M. Kube~a,MD. Schutte, and RJ. Weetman. 1993. Process/mechanical design aspects for Lightnin A3 15agitators in mineral oxidation Proceedings of Randol Gold Forum. 247-253. Guide to impeIler selection for maximum process results. Philadelphia Mixers. H e q N 1985. SME Mineral Proassing Handbook Historical Developments in Milling of Gold Ores. 18-4. Kubera, P.M., and J.Y. Oldshue. 1992. Advanad impeller technologies match miXing performance to process needs.ProceedrigsofRmdo1 GoldForum. 1-15. Lally, KS. 1987. A315 axial flow impleler for gas dispersion. Mixing Equipment Company, L I G H T " Technical Reprint Lapedes, D. Ed 1974. McGraw Hill Dictionary of Scient$c and Technical Terms. New York City, NY.St. Louis,MO. San Francisco, C k McGraw-Hill Book Chmpany, p. 831. Olderstein, A.J., T . k Post, T.J. Klimasewsla -_ 1989. New impeller pmvides more efficient mass transfer in gas-liquidalid systems. Proceedings I U S Annual Meeting. 1-13. Oldshue, J.V., 1963, Solid-liquid mixing in hydrometallqzy~Unit Processes in Hydrometallurfl,P. 1-24 Oldshue, J.V. 1969. How to specify mixers. Hydrocarbon Processing. Oldshue, J.V. 1983. Fluid mixing technology and practice. Chemical Engineering. 83-108 Oldsuhe, J.V. 1987. Fluid Mixing. Emylopedia of Physical Science and Technology, Vol. 5 Oldshue, J.V. 1989. Fluid mixing in 1989. Chemical Engineering Progres. 33-42. Oldshue, J.V., T.A. Post, RJ. Weetman, and C.Coyle. 1988. Comparison of mass transfer characteristics of radial & axial flow impellers. Proceedingsfiomthe 6European Confirence on Mixing. Oldshue, J.V., and D.B. Todd. 1981. Mixing and blending. Encyclopedia of Chemical Technolo@. ed. Kirk-othmer, 3d ed., v. 15, New York: John Wiley & Sons. Optmuzhg and scaling up mixing systems. Philadelphia Mixers. Rushton, J H , and J.V. Oldshue. 1953. Mixing present theory and practice, part J.I. Chemical Engineering Progress. 161-168. Salzman, RN., C. Chyle, RJ. Weetman, and J.C. Pharamond. 1983. High efficiency impeller for s h y storage. Proceedings@om the Eighth International Technical Con@rence on Slurty Transportation. 305-309. Scheffel, R E. 2002 Copper Heap Leach Design and Practice. Mineral Processing Plant Design. Control and Practice, A. L. Mular, D. N. Halbe, D. J. Barratt, Eds., Society for Mining, M d u r g y and Exploration, CanadianMind Processing Division of the Canadian Instituteof Mining Shaw, J.A. 1982. The design of draft tnbe circulators. Z%eAustralian Institute of Mining and Metallurgv. 47-58. Von Essen,J.A. 1998. Gas-liquid-mixer correlation Chemical Engineering. VonEssen, J.A., and B. Ricks. 1999. Design agitated slurry storage tanks to minimize costs. Chemical Engineering Progress. Walter, T.E.1%8. Take care choosing agitation equipment.Pulp and Paper.

1643

CIP/CIL/CIC Adsorption Circuit Equipment Selection and Design KathleenA. Altman’, Stuart McTavishz

ABSTRACT Design of a CIP/CIL/CIC circuit begins with metallurgical testwork and data acquisition The metallurgist who conctucts this testwork must work closely with the process engineer to develop

an optimum design. This design must also include secondary factors such as plant capacity, site. and environmental considerations. Equipment selection is based on speCi6c geographic-oc standard plant practice unless there is reason to believe the feed will be substantially different from the norm. In this case, scale-up tests are conducted in order to obtain more rigorous design data, and the design is modified based on this information.

INTRODUCTION The design of carbon-in-pulp (CIP) and carbon-in-leach (CIL) circuits is closely related to the design of agitated leach circuits with a few additional considerations. The primary consideration is the use of granulated activated carbon (GAC) to recover gold fiom the leach circuit. This means that carbon attrition is a mjor concern in the selection of the equipment and additional equipment is required to haudle the carbon that is utilized in the circuit Carbon losses due to attrition, or deteriorationof the GAC, are important not only because of the cost of the carbon itself, but also because the fine carbon that results fiom the attrition continues to adsorb precious metals, and then, cannot be contained in the recovery circuit. The associated loss of fine carbon that is precious metals-laden results in reduced recovery of the precious metal, which detrimentally S e c t s the profitability of the operation. In CIP/CIL adsorption circuits, the slurry that is being leached and the carbon being loaded with precious metals flow counteramnt to one another. This includes a “forward flow” of the pulp, or slurry, and a “reverse flow” of the carbon (Rogans, et. al. 1998). The sluny generally flows by gravity from the jirst tank to each SucceSSive tank just as it does in an agitated leach circuit. The carbon, then, is advanced fiom the last tank towards the first in the countercucrent fashion. The slurry flow is continuous and the carbon is typically advanced on an intermittent schedule using pumps. The selection of an appropriate recovery process is based on an economic evaluation of the proposed project. In order to quam@ the capital and operating costs to use in these evaluations, a certain amount of engineering is required, including the selection and basic design of appropriate equipment for each option under consideration. As a general rule, CIL circuits provide better project economics and are, therefore, the preferred treatment option, particularly when preg-robbing ores are leached. This is due primarily to an increase in gold recovery. In theory, at least, the GAC has a higher aijinity for the gold in solution than the &MC~OUS material that is contained in the ore itself. Thus, the gold losses in the solid residues, or tails, are reduced and the overall recovery is increased, which results in am economic advantage. CIL circuits generally have larger carbon inventories and larger elution circuits as a consequence. Carbon-inalwnn (CIC) circuits are typically used to recover precious metal from solutions, as opposed to slurries. This recovery method is most commonly used in precious metal healp leaching applications, although it is sometimes used to recover gold (andor silver) fiom thickener overflow solutions. 1 Independent Consultant, Denver, Colorado 2 SNC-Lavalin Engineers and Constructors, Inc., Toronto, Ontario

1652

As with most recovery processes, there have been dramatic improvements in the design of equipment over the course of time. For example, agitators have been specifically designed for CIL and CIP applications, and carbon handling concepts and designs have evolved from EPAC and vibrating screens to Kambalda, NKM and MPS interstage screens and, most recently, to pump cell technology. PROCESS DESIGN

Design Procedure The process of designing CIP/CIL/CIC adsoqtion circuits is similar to the procedure used for agitated leach circuits. It requira a collaborative effort between the operating company, the metallurgicaltesting fhcility, the engineeringcompany, and the equipment suppliersjust as it does for the agitated leach circuits. Please refer to section D-1-3, Agitated Tank Leaching Selection and Design, for a more detailed discussion. The number of stages and carbon movement are determined based on the solution feed grade, desired barren solution grade, design loaded carbon grade and precious metal - carbon adsorption characteristics of slurry (solution). The capacity of the elution circuit is then selected based on the calculated carbon movement. CIP/CIL Circuits Tank Design. The tank design is nearly the same as it is for agitated leaching. This includes: Tank sizing Slurry density Plant throughput Number of stages of leachinghdsorption Tank height to diameter ratio Tank bafning Tank configuration Materialsof constructioa

The slurry density in a carbon adsorption circuit (CIL and CIP) is usually lower than it is in an agitated leach circuit. It must be conmlled carehlly in order to keep the GAC properly distributed throughout the tanks.A range of 35 to 50 percent solids is typical, although it is dependent on the slurry characteristicsof the ore being processed. If the slurry density is too high, the carbon floats near the surface of the tank.If it is too low, the carbon sinks to the bottom of the tank. In both cases, mass transfer of gold to the carbon particles is impaired, and the interstage transfer of carbon is constrained Tank sizing for the leach section of a CIP plant and for the 1eacWadsorptionvessels of a CIL plant are based on the slurry flow rate, the retention time required to leach the gold from the ore and load it onto carbon, and maintaining the requirednumber of stages. The retention time in leach and CIL circuits is typically 24 to 48 hours, and in CIP circuits is typically 0.75 h to 1.0 h per stage, usually with 5 to 7 stages. Reagent Additions. In CIP/CIL cyanide leach circuits, the pH of the slurry may be initially controlled through the addition of lime on the feed belt to the grinding circuit and, particularly in CIP circuits, the grinding may be done in cyanide solution. In this case,it is good engineering prack to provide for trim additions of slaked lime d o r caustic soda and cyanide in the leach Circuit so that the . .reagent mnsumptions can be optimized, i.e. reagent costs minimized and metal recoveries maxLI11l2Rd. The determining factors to consider in designing reagent addition systems are the sensitivity of the leaching process to the concentration of the reagents and the amount of variability that is anticipated in the ore that will be processed When ores contain high mncentmtions of copper, high Eree cyanide levels are maintained to mhimize adsorption of copper onto the carbon.

1653

In CIP and CIL carbon adsorption processes, it is necessafy to add oxygen, generally in the form of air, to the circuit. This may be done with air blowers or with low pressure air compressors, depending on the volume of air and the delivery air pressure that are required. In some cases, either liquid or gaseous oxygen may be introduced into the leaching process, which results in even mote sophisticated gas production and/or storage and injection systems. By definition, CIP is for adsorption only- No air is normally added to ClP tanks for leaching, although some leaching does occur in CIP circuits. The amount of carbon added to the circuit is dependent on the plant throughput and the quantity of recoverableprecious metal mntained in the ore. The amount of carbon moved through the adsorption circuit is also coordinated with the capacity of the carbon elution and carbon regeneration circuits. For a 1,OOO t/d plant operated at 45y0 solids, the solution flow rate is 1,222 t/d [(l,OOO W0.45) - 1,000 t/d)]. If the recoverableprecious metal grade in the ore is 5 glc the total amount of metal recovered daily is 5,000 grams (loo0 t/d * 5 g/t) and the solution grade would be 4.1 g/t [(1,OOO t/d * 5 g/t)/1,222 t/d]. Using the calculation method presented in the agitated tank leaching section, each tank would be approximately 343.3 m3 based on the criteria presented in the table below. These are design criteria for CIL.

Table 1: Sample CIL Design Criteria Plant throughput ore specific gravity Sluny density Retention time Effective volume factor Number of stages Tank volume

1,000 tpd 2.8 45% solids 24 h 0.92 5 343.3 m3

The concentration of carbon required to achieve a certain target gold extraction efficiency can be determined by metallurgical testwork and mathematical modeling. If the carbon concentmtion is 10 @, the total amount of GAC contained in each tank is 3.4 t for a total inventory of 17 tonnes p 4 3 . 3 m3*1,000 u&)*(io gn)~(ioooglkg)~(i,ooo kg/t)]. Assuming that a carbon processing rate of 1 t/d has been selected, the anticipated carbon loading would be about 5,000 glt [(5,000 gld)/(l t/d)]. This is well below the maximum amount of precious metal that can normally be loaded on GAC. However, it is important that the carbon loading be checked against the t h e ~ d c a lprecious metal loading at the design operating conditions. The selected carbon gold loading has to be consistent with a reasODable retention time in the CIP/CIL circuit to minimke carbon poisoningfrom prolonged loading Agitation Requirements and Design. The agitation requirementsand design for CP/CIL circuits are the same as they are for agitated leach circuits with the additional emphasis on mkimiz@ carbon attrition The main criteria include the following, which are explained in detail in section D-1-3 0 0

0 0

0 0

Mixing requirements and performance Choice of impellers Flowrelatidps Solidssuspension Masstransfer Gasdispersion.

Impellers used for gas dispersion l y p i d y have high shear characteristi&. Shear is a major consideration in ClP and CIL recovery circuits because it is an underlying cause of attrition of GAC. However, some shear is required for leaching to take place.

1654

Shaw and McDonough (1982) provide a comprehensivediscussion of mechanical agitation in the CIP process. The impeller type, impeller diameter and speed, as well as the geometrical relationship between the impeller and the tank will determine the split between flow energy and shear energy. Large impeUers NDning slowly convert more energy into flow energy than shear energy. Or, another way of looking at it is that large impellers can produce the same flow as smaller impellers while using less energy. However, larger impellers also have a higher capital cost. It is possible, with carelid design, to operate at higher speeds and, hence, reduce the capital cost without d e r i n g from a large shear penalty. Shear rate is a change in velocity with distance.It has the units of wm/sec.n Several Merent factorsare important in agitation systems: 0 0

0 0 0

Maximum shear rate in the impeller zone Average shear rate in the impeller zone Maximumshearrateinthetank Average shear rate in the tank Frequency of circulation through the high shear zone.

These hctors are related to impeller diameter, mtational speed and impeller type. Although impeller tip speed is sometimes used as an indication of shear rate, Werent impeller types can have dramatically merent shear rates even though they have the same tip speed. Maximum shear rate results fiom velocity gradients generated in the liquid by the impeller. The maximum shear rate oftenoccurs some distauce f h m the impeller itself. Shear stress is dependent on both the shear rate and the viscosity of the liquid. It is shear stress that causes particle damage. Although it is diaicult to quantify carbon attrition, Shaw and McDonough (1982) provide some guidelines. 0 0

Draft tube agitation systems provoke less carbon attrition than open impeller SySkXW. Aerofoil impellers provoke less carbon attrition than flat plate impellers in both Open impeller and Qaft tube agitation designs.

Both of these reductions appear to be due to the velocity gradients generated by each of the respective designs. In general, draft tube circulators have higher capital costs but consume significantly less power than open impeller agitator designs. Since draft tube circulators are preferable in some cases,a description of key design factorsis warranted. Shaw (1982) provides details about the design of draft tube circulators.Historically, draft tube circulators were first designed for the alumina indust~~, which had previously used Pachuca systems, just as the gold industry had. A draft tube circulator includes a flat-bottomed tank and a draft tube. The diameter of the draft tube is on the order of 20- to %percent of the tank diameter. An axial flow impeller is positioned near the top of the draft tube. The impeller pumps down the draft tube and this flow, in turn, produces an upwad flow in the annulus. Two distinct velocity criteria are considered in the design of draft tube circulators: 0

The velocity down the draft tube is determined by the bottom-sweeping design, which is a function of the particle size distributionof the slurry, the pulp rheology and the shape of the tank bottom. The average upward velocity in the annulus must be greater than a minimum, which is dependent on the free-settljng velocity of the largest particle to be circulated.

Power requirements are the lowest at the ratio of draft tube diameter to tank diameter where both of these velocity criteria are met

1655

The operation of a draft tube circulator is similar to the operation of an axial flow pump. Enem requirements depend on the hydraulic efficiency and the power-flow relationships. Also, the stability of the draft tube circulator is dependent on a head versus flow curve. Draft tube circulator designs may be very simple, almost primitive, or quite sophisticated. Although more sophisticated systems are generally more expensive, they may be justified due to lower energy consumption. Trash and Safety Screens. CIP and CIL circuits rely on screens to control the movement of carbon. In order to prevent the interstage screens from being impacted by wood chips or plastic, the feed to a CIL or CIP circuit is first passed through a trash screen. This screen can be a vibrating screen or a hear screen that is the same mesh size as the interstage screens. On the discharge of the circuit, the slurry is generally passed over a safety screen. This screen can also be either a vibrating screen or linear screen and is usually one mesh smaller than the interstage screen. The safety screen recovers any carbon that has been attritioned to a small enough size to pass through the interstage screen and recovers any full size carbon should the last interstage screen develop a hole. Carbon Control and Handling.Screens and pumps are the main components of the carbon handling systems associated with CIP and CIL operations. Screens are used to contain the carbon in the slurry tanks while the slurry is allowed to gravitate through the screens on a continuous basis. Then, some sort of pumping is required to transfer carbon from one stage of the circuit to the next stage upstream. Initially airlifts were used for inter - tank slurry transfer. However, this method was quite inefficient, and in some cases was found to result in excessive carbon abmion. In the mid - 1980’s fecessed impeller pumps replaced airm for slurry transfer. Vertical d impeller pumps are used in CIP/CIL tanks while horizontal recessed impeller pumps are used to transfer carbon within elution/regeneration circuits. Interstage screening has been one of the biggest problems to overcome in carbon adsorption circuits (Rogans, et. al., 1998). One reason for the difficulty is that the particle size distribution is small for both the solids in the slurry and the carbon. Plus, the size of the carbon particles is continually reduced due to the aitrition mechanisms. Because interstage screening has been a major concern, it has gone through a continual development process since the inception of CIB and CIL processing. Early circuit designs utilized equalized-pressure air cleaned (EPAC) screens or external vibrating screens (Rogans and Cartner, 1996a). EPAC screens had high costs that resulted in high operating costs.The screen clothes were the subject of frequent failures. One of the first developments in interstage screen technology was the replacement of air swept screens with mechanically swept screens. Derrick in - tank vibrating screens replaced the EPAC screens in North America and were the standard in the late 1980’s to the mid 1990’s in large CIL/ClP plants. In South African gold plants, Kambalda screens replaced many of the EPAC screens.However, Kambalda screens still had major difiicuties. In order to perform maintenance on the screen, the pulp level had to be lowered, which required bypassing of the tanks.Kambalda screens also had high power requirements because they used horizontal mechanical pulse blades. This led to the development of vertical, mecbanidy swept interstage screens, which were developed for the Daggafontein CIL plant in South Afiica. The idea came from a similar screen in Australia so they were named North Kalgwrlie Mines (NKM) scnxm. NKM screens are formed in the shape of a basket. They have rotating pulse blades around each screen surface. These blades dislodge the carbon that is retajned in the screens. The NKM design provided a higher pulp flow through the screens than previous designs. Rogans and Cartner (1996a) provide a comparison of throughput rates for the various screen designs.

1656

Table 2: Interstage Screen Throughput Rates

Screen Type Vibratory EPAC Kambalda NKM

Pulp Flow Rate m3/m2/h 54 20 to 40

40 70

Even though NKM screens dramatically improved interstage screening in CIP and CIL plants, they, too, were not without problems. The internal conical wedge was subject to plugging, which results in loss of open area and a subsequent increased flow rate, as well as associated increased wear rate that reduces screen life. Also, pbase separation occurs when the flow is reduced or

stopped because the dampening effect of the screen on the pulse blades is high enough that there is no agitation in the annular area Then, even after normal flow is resumed, the NKM screen still operates in a clogged or partially clogged condition. Again, this results in increased wear rates caused by higher flow rates in the open areas of the screen. This resulted in the development of the M i n d Process Separating (MPS) and the h4ineral Process Separating (Pumping) [MpSOI screens. The M P S screen is used in conventional, gravity flow CIP/CIL circuits. The MPS(P) screen incorporates an up-pumping mechanism that lifts the pulp inside the Screen and places it in a launder above the pulp level in the tank. Rogans and Cartner (1996a) and A.AC. Mineral Process Separating Screens, describe details of the design of MPS and M p S ( P ) screens. The MPS screens eliminate the problems encountered with NKM interstage screens. Addition of the pumping mechanism in M p S ( P ) screens allows the leach tanks to be placed at the same elevation, which reduces the capital cost of the leach tanks due to red& civil and conslNction costs. Benefits of using MPS and M p S ( P ) screens include the following, as outlined by Rogaus and Cartner (1996a):

Increased throughput rates, e.g. 70 to 100 m3/m2/h Reduced power requirements, i.e. 5.5 k W per 500 m3/h for MPS screens versus 21 kW for the same throughput through Kambalda screens Good tolerance for surging flows Cleaning can be accomplished in a matter of minutes without reducing the slurry level in the leach tank MPS(P) screens can also transfer highly viscous pulp through the interstage screens,even though this material cannot normally be processed in conventional CIP/CIL plants. One final advantage of the M p S ( P ) screen is that it has allowed the development of the Anglo American Corporation (AAC) Pumpcell Adsorption circuit, which relies on rotating the pulp feed and tailings discharge positions through a series of contactors that are all placed at the same elevation. This design results in the countercurrent movement of &n without physically moving the carbon from one contactor to the next (The AAC. Pumpcell Adsorption Circuit). Rogans, et. al. (1998) use the results of simulations to quantify the advantages of pumpcell carousels as compared to more traditional carbon adsorption circuits. Rogans and Cartner (1996b) and Schoeman, et. al. (1996) compare pumpcell technology with CIL and CIP plants and describe installations at two South Afiican gold mines. Operating data from five pumpcell installations and two CIP plants is presented by Macintosh, et. al. (2000).

CIC c I R m The design and equipment selection for CIC circuits is based on the required solution flow rate and the solution grade, which directly af€ects the carbon loading concentration that can be achieved. Calculating the quantity of gold to be recovered on a daily basis and dividing it by the

1657

grade of the loaded carbon determine the &n capacity required in a CIC circuit. The quantity of carbon to be advanced daily is nominally the capacity of each cell in a carbon column. This is also generally matched to the capacity of the carbon elution circuit- Cascading columns are the most prevalent design for CIC circuits, although alternative designs such as stacked vertical columns have been used successfully. The columns are operated in an up-flow mode, and the design is such that the solution velocity provides fluidization of the carbon bed in the cell. Obviously, the sizing of the carbon cells must allow for this bed expansion. Published curves are available, which show the relationship between fluidization and velocity for carbon of various size ranges. Stationary screens are provided to ensure that the carbon does not escape fiom the cells of the carbon columns in the event of surges in the solution flow rate. Carbon is advanced from cell to cell in CIC circuits, counter current to the solution flow rate, similar to the process used in CIP and CIL circuits. The movement can be accomplished using eductors. However, carbon attrition is an important consideration in the selection of carbon handling equipment, and recessed impeller pumps have been shown to reduce the rate of cadmn athition in CIC circuits. Therefore, the carbon movement design is similar to that used in CIP/CIL circuits.

SUMMARY AND CONCLUSIONS CIP/CIL/CIC circuits are common gold recovery circuits. Clop circuits include a separate agitated leach circuit and a carbon adsorption circuit, while a CIL circuit provides simultaneous leachmg and carbon adsorption. The design process for CIP and CIL circuits is similar to that used for agitated leach circuits, with the added considerationsrequiredwhen granulated activated carbon is used in the fecovery circuit. Tank designs are dependent on the residence time required for leaching and adsorption, and the number of stages required to achieve an appropriate leach profile. Successful agitator designs use either open impeller or draft tube circulator designs. Interstage Screen designs have, perhaps, changed the most since the initial development of carbon-in-pulp and carbon-in-leach pracessing for the recovery of gold Current design technology minimizes operating difticultiesand reduces operating costs. The newest technology for interstage meens has the added advantage of allowing the development of pump-cell carbon adsorption

circuits. CIC circuits are similar to CIP and CIL. circuits except that they recover solubilized gold from solutions instead of slurries. This allows a simpler design that depends primarily on the solution flow rate and the gold tenor in the solution. CIP/CIL and CIC are all proven technologies that W e now been successfully used for a number of years. Therefore, there are a number of operating plants that can serve as a design basis for new plants.

REFERENCES A.A.C. Mineral Process Separating Screens. Kemix. The A.A.C. P u m p 1 1 Adsorption Circuit- Kemk Macintosh, A., D. McArthur, and R.M. Whyte. 2000. Process Choices for Carbon Technology, Proceedings Randol Gold & Silver Forum. Vancower, B.C. Canada. Rogans, E.J., and W.N. Cartner. 1996% The Development of Mineral Process Separating Interstage Screen (MPS) from the NKM Interstage Screen, Proceedings Randol Gold Forum. Squaw Creek, CA, USA Rogans, E.J., and W.N. Cartuer. 1996b. The Pump-cell Adsorption Circuit for In Pulp Applications, Proceedings Randol Gold Forum, Squaw Creek, CA, USA Rogans, E.J., A. J. Macintosh, and N. Momson. 1998. Carbon-In-Pulp and Carbon-In-Leach Adsorption Circuits - Optimisation of Design Using the Carousel System, Kemix Technical Paper, 23 pp.

1658

Schoeman, N., E. J. Rogans, and A..J., Macintosh. 19%. AAC Pump-cells: A Cost-effective Means of Gold Recovery from Slurries, Proceedings SAlMM Hidden Wealth Conference,.Jobannesburg,S.A. p. 173-179. Shaw, J.A., 1982, The Design of Draft Tube Circulators, Proceedings of the AusPulasian Institute of Mining andMetalluqy, No. 283, p. 47-58 Shaw, J.A., and RJ. McDonough. 1982. Mechanical Agitation in the Carbon-in-Pulp Process. Technical Reprint. LIGHT".

1659

Zinc Cementation - The Merrill Crowe Process A . Paul Hampton,P.Eng.

ABSTRACT The classical method for the recovery of precious metals from cyanide leach solutions is cementation using zinc powder, the Merrill Crowe process, whch was developed in the 1890’s and used almost exclusively until the introduction of carbon adsorption processes in the 1970’s and 1980’s. The unit operations of the Memll Crowe process include solid liquid separation, clarification, vacuum deaeration in packed towers, zinc addition and filtration of precipitated gold and silver using pressure filters. This paper discusses Menill Crowe process application and circuit design including solution chemistry, preliminary process design parameters, equipment selection and sizing, process performance and relative costs.

INTRODUCTION Zinc cementation, which later became known as the Merrill Crowe Process, was the original method chosen for the recovery of precious metals from solutions generated by the cyanidation process, during its developmental stages. The zinc cementation process was first applied in the 1890’s and was used, with subsequent improvements, until the introduction of carbon adsorption processes in the 1970’s and 1980’s. This paper discusses the application and design of the Merrill Crowe process beginning with a brief history of the process, the basic chemistry and the development of basic process design criteria for the development of a Merrill Crowe process flow sheet. The Merrill Crowe process consists of the following main steps, which follow cyanide leaching:

0

Solid liquid separation, using CCD thickeners, or vacuum filtration; Clarification of pregnant solution to approximately 1 ppm suspended solids; De-aeration of the clarified solution using packed towers (Crowe) under vacuum; Addition of powdered zinc using a mixing cone; Precipitation in the pipeline between the press feed pumps and the filter presses; Filtration of the precipitated metals in a filter press; Smelting of the precious metal precipitate in the refinery to form dore;

Figure 1 is a simplified flow sheet of the Merrill Crowe process for low temperature cyanide leach solutions, requiring deaeration. A Merrill Crowe flow sheet designed to treat high temperature carbon elution streams, would be similar, but would not require the deaeration step (Marsden, 1990).

History Cyanidation was first applied in the extraction of gold and silver from low grade ores by 3. S. McArthur and Doctors Robert and William Forrest of Scotland in the mid 1880’s. A British patent

1663

was issued to the group in 1887 and U.S. patents were issued in 1889 for the process, which included agitation of pulp in the presence of air followed by the precipitation of gold with zinc in a separate solution. The first mining application of cyanidation for gold recovery was at the Crown Mine in New Zealand in 1889. The process was subsequently applied in South Africa, the United States and mining districts around the world. The zinc cementation process was introduced in 1890 and became an integral part of the cyanidation process (Dorey, van Zyl and Keil, 1988; Fleming, 1998). The first application of the zinc cementation process in the United States was by C.W. Merrill at the Homestake Mine in Lead South Dakota in 1897 (Chi, 1992). )Y GMNDlNG

cull

d

GRlNOlffi

cull

Na.1-6

No. 6

i

r

PROCESS WATER

MUHIER-CURREHI DEUNlAmN THCKENERS Nos. 1 - 6

ANNOSPHERE

I

Figure 1. Simplified flow sheet for a typical leaching, CCD and Merrill Crowe circuit. Zinc cementation was initially performed using long sloping boxes filled with bundles of coarse zinc shavings. Gold bearing solutions were passed through sand filters to remove suspended solids and then through the zinc boxes for metal precipitation. Vertical plates were installed in the boxes forming chambers to direct the flow of solutions through the beds of zinc shavings to improve the contact of the solutions with the zinc. The method proved to be effective but inefficient due to coating of the coarse zinc surfaces with deposited metals or insoluble zinc hydroxide. (Atwood, 1985; Wood, 1996) Lead salts were introduced in 1894 to address the passivation problem. The bundles of zinc shavings were dipped in solutions of lead acetate before placement in the zinc boxes (Wood, 1996). The lead deposits on the zinc surfaces formed cathodic areas for preferential precipitation of precious metals, leaving the adjacent anodic zinc surfaces exposed for dissolution (Chi, 1992).

1664

Clarification was found to be a very important stage, which affects both metal recovery and precipitate grade. When suspended solids were present, the rate of precipitation was found to decrease, which lowered recoveries. The zinc boxes acted as sand filters trapping fine materials that would potentially blind the beds of zinc and dilute the grade of the precious metal precipitate. (Atwood, 1985) Through operating experience, it was also identified that some sands trapped in the zinc boxes, such as pyrite and marcasite, tended to improve precipitation and to lower zinc consumption. The improvement was attributed to the reaction of these sulfide minerals with oxygen and cyanide resulting in the consumption of dissolved oxygen. The reduced oxygen was found to improve the efficiency of the process by both reducing the passivation of the zinc surfaces and also by reducing the direct leaching of zinc, which requires oxygen (Atwood, 1985). The next significant step in the development of the zinc precipitation process was substitution of zinc powder for zinc shavings in 1907-1908 by C.W. Merrill, while studying the effect of surface area on the rate of precipitation. Menill added zinc dust (fume) to leach solution and then pumped the slurry through a filter press. The precipitate and zinc remained in the filter press and the barren solution was recycled to the process (Wood, 1996). The result was an increase in rate and recovery and improved zinc utilization. The increased surface area and reactivity also resulted in higher zinc consumptions as the rates of the passivation and zinc dissolution reactions increased, making it all the more important to remove oxygen from the system. In 1916, the vacuum deaeration tank was introduced by T.B. Crowe and incorporated into the Merrill Crowe process. (Atwood, 1985; Chi, 1992; Wood, 1996). The Merrill Crowe process has generally remained in tact since 1916. The main improvements in the process since that time have been in the design and efficiency of the equipment and automation systems; clarifiers, vacuum towers with modem packing and filter presses. There is also a better understanding of the process chemistry, aiding in optimization of reagent usage and consumption resulting in reduced operating costs.

PROCESS SELECTION - CHOICE OF MERRILL C R O W VS CARBON ADSORPTION The two main processes currently in use for the recovery of precious metals from cyanide leach solutions are zinc precipitation and carbon adsorption. Fleming reports that the carbon in leach process has, in most cases, proved to be more efficient and to have 20 to 50 percent lower capital and operating costs than Merrill Crowe. Currently over 70 percent of world’s gold production is recovered using carbon adsorption processes (Fleming, 1998). The Merrill Crowe process has an advantage over the carbon adsorption process in cases where the metal concentrations are high, such as ores containing significant amounts of silver; high silver to gold ratios. Zinc cementation can also be applied to the recovery of precious metals from carbon strip solutions as an alternative to direct electrowinning. Carbon in leach has the advantage over Merrill Crowe when ores contain significant levels of organic carbon, high base metal concentrations and when the ore contains high clays, which are difficult to filter. Examples of operations, which utilize zinc cementation include:

0

Newmont Mining Company - Minera Yanacocha mine in Peru Barrick Gold’s Pierina Mine in the same district in Peru

1665

0 0

0

Placer Dome’s La Coipa Mine in Chile Goldfields Operating Company, Chimney Creek- Carbon Strip Circuit FMC Paradise Peak - Carbon Strip Circuit Equity Silver Mine, British Columbia - Carbon Strip Circuit

PROCESS CHEMISTRY Cyanide Leaching Leaching of precious metal ores in alkaline cyanide solutions yields metal bearing solution containing relatively low concentrations of gold, silver, copper, zinc, and other metal cyanide complexes depending on the composition of the ore and the type method of leaching employed. The dissolution of gold and other metals in alkaline cyanide solutions is generally accepted to be an electrochemical reaction composed of two half-cell reactions. The oxidation of gold from Au’ to Au’ represents the most common anodic reaction. The cathodic half cell reaction consists of the reduction of oxygen and water. The reactions are corrosion type and occur in adjacent areas on the surfaces of gold particles. The gold (I) or aurocyanide complex is very stable and was reported by Finkelstein to be the predominant species formed in cyanide leach solutions. The stability of aurocyanide, Au(CN),, is such that it remains stable in the absence of free cyanide and at very low pH. (Hedley, 1958; Finkelstein, 1972) Gold is typically present in gold ores in elemental form or as an alloy with silver; electrum. The overall reaction for the dissolution of gold is represented by Eisner’s equation (Hedley, 1958): 4Au + 8 0 4 - + 0

2

+ 2H20 = 4Au(CN)y + 4(OH)- .

The anodic half reaction, the oxidation of gold, is represented by: 4Au + 8CN- = 4Au(CN);

+ 4e-

The cathodic half reaction, the reduction of oxygen and water, is represented by:

O2 + 2H20 + 4e- = 4(OH)Cyanide soluble silver minerals, typically found in gold-silver ores, include electrum, an alloy of gold and silver, and argentite, a single sulfide mineral. More complex ores contain minerals such as tetrahedrite, which contain varying amounts of antimony and arsenic, can also be present. The overall reaction representing the dissolution of Argentite, silver sulfide, is represented by the following (Hedley, 1958): AgzS + 3CN- + 0 2 + 2H20 = 2Ag(CN)y

+ CNS- + 4(OH)-.

The base metal reactions are of similar form. The dissolution reaction for chalcocite is as follows: Cu2S+ 7CN- + O2+ 2H20 = ~CU(CN)~’+ CNS- + 4(OH)-. Some of the important metal complexes produced during the cyanide leaching of gold and silver ores include:

1666

Antimony and arsenic ,,rm soluble oxide compounds in alkaline solutions and are significant consumers of oxygen.

Cementation Chemistry The zinc cementation reaction in alkaline cyanide solutions is an electrochemical displacement reaction, which involves the reduction of gold and silver, which occur as cyanide complexes, Au (0 and ; Ag(CN)2- ,the reduction of oxygen and water and the oxidation and dissolution of the zinc metal which forms a cyanide complex, Zn(CN)42- . The gold and silver metal forms coatings on the surface of the zinc particles and the zinc in turn corrodes and dissolves into solution. The reactions are driven by the differences in electrochemical potential between the more noble precious metals and zinc. Metals, which are more electropositive than zinc, will reduced to their metallic states, while zinc will be dissolved. Table 1 contains a selection of reduction potentials for metals typically found in cyanide leach solutions.

Table 1 Electrochemical Series - Standard Reduction Potentials (Vanysek, 1984, Finkelstein, 1972, Fang, 1992). Reaction

E",V

2H++ 2e- = H2 2H20+2e- = 20H- + H2 Au' + e- = Au' Au(CN), + e- = Auo + 2CNA U ~ ++ 3e- = AUO A E + + ~=AE' Ag(CN);+e- = Ago +2CNCU+ + e- = CUO cu2++ 2 e- = CUO Cu2++ 2 cN-+ e- = Cu(CN) 2Zn" + 2 e- = Zno Zn(CN)t-+ 2e- = Zno + ~ C N O2+ 2H20+ 4e- = 40HPb2++ 2 e- = Pbo

0.00 -0.828 1.692 -0.473 1.498 0.8

-0.269 0.521 0.342 .~ 1.103 -0.762 - 1.260 0.401 -0.126

~~

The relationship between the change in Gibbs free energy, AGO, of an electrochemical reaction and the standard reduction potential, Eo, is represented by: AGO

=-

ZFEO,

where, z is the number of electrons transferred and F is the Faraday constant. To combine half cell reactions, the fiee energy for each reaction is determined and then summed. The fiee energy of the combined reactions can then be used to determine the new standard potential, Eo = - [AGo/zF].

1667

The reaction will proceed spontaneously if the free energy is negative (Moore, 1983). The cell potential for a system, which is not at unit activity, is determined using the Nernst equation,

E = Eo+ 2.303 RT/zF log a oxlared where E is the cell potential, R is the gas constant, T is the absolute temperature, z is the number of electrons transferred, F is the Faraday constant and a ox and a red are the activities of the of oxidized and reduced species. The resulting cell potential is then used to calculate the change in Gibbs free energy for the reaction,

AG = - zFE. The reduction of gold by zinc metal is represented by the following overall reaction (Finkelstein, 1972), 2Au(CN); + Zn = 2Au + Zn(CN),

'-

Assuming that free cyanide is available and a direct transfer of cyanide ions is not necessary (Marsden, 1990),

'-+ 2CN-

Au(CN) 2- + Zn + 4CN- = Au + Zn(CN)4

The half-cell reactions representing the reduction of gold and the oxidation of zinc are: Au(

+ e- = Au + 2 c "

Water and dissolved oxygen are also reduced by zinc: 2H20+2e- = 20H- + H2 0'

+ 2H20 + 4e- = 40H-

The efficiency of precious metal cementation is dependent on the effective dissolution of zinc, however if sufficient oxygen and free cyanide are available, zinc will dissolve independently, resulting only in increased zinc consumption. If the concentration of free cyanide becomes low, due to insufficient cyanide addition or excessive zinc addition, zinc will react to form zinc hydroxide, which may passivate zinc surfaces and plug filters. Competing reactions include (Finkelstein, 1972; Fang, 1992): 2Zn + 8CN- + O2+ 2H20= 2Zn(CN);Zn + 4CN- + 2H20 = Zn(CN);-

+ 4(OH)',

+ 2(OH) - + H2

2Au(CN) 2- + Zn + 3(OH)- = 2Au + HZn02-+ 4CN- + H2O zn2'

+ 20H- = Zn (OH)'

+ 2e-

1668

Base metals including copper are precipitated along with the precious metals, though in practice it has been shown that copper precipitation can be minimized by maintaining and excess of free cyanide.

Kinetics The rate of the cementation reaction is first order with respect to the concentration of aurocyanide complex, Au(CN) and directly proportional to the surface area available for reaction as represented by the following equation (Finkelstein, 1972):

where: Co = initial concentration of aurocyanide Ci = concentration of aurocyanide at time t k = reaction rate constant A = zinc surface area available t =time The rate is controlled by the diffusion of aurocyanide to the surface of the zinc particle under all conditions. The extent of reaction is limited by the amount of available cathodic surface. The anodc reaction is the limiting step only when the zinc surface is blocked by insoluble precipitate (Finkelstein, 1972).

Effects of Operating Parameters on Zinc Precipitation The zinc precipitation reactions are affected by the following parameters: PH CN- concentration Suspended solids concentration Dissolved oxygen Temperature Metal concentrations Lead addition Scale forming compounds such as calcium sulfate and sodium silicate The effects of each of these parameters with respect to operating efficiency follow:

The pH must remain high to prevent the formation of volatile HCN gas. An increase in pH, in the absence of precious metal cyanide complexes, will tend promote the formation of insoluble zinc hydroxide. (Finkelstein, 1972; Fang, 1992)

CN- Concentration The concentration of cyanide must be optimized along with the addition of zinc powder, Excess free cyanide will reduce the tendency for zinc hydroxide formation, but will lead to an increase in zinc consumption through independent zinc dissolution. Very low CN- concentration results in the formation of Zn(OH), on zinc surfaces inhibiting cementation.

1669

Oxygen Concentration

0

The rate of precipitation decreases with an increase in oxygen concentration; The rate of independent zinc dissolution reactions increases with an increase in oxygen concentration; The potential for redissolution of gold increases as the oxygen concentration increases and the zinc concentration decreases;

Temperature

0

An increase in solution temperature results in an increase in the rates of reaction for both precious metals precipitation and the dissolution of zinc, however the reactions are very fast at ambient temperatures; As the temperature increases, the solubility of oxygen decreases, which reduces the need for dearation; Deaeration is not required when applying zinc precipitation to hot carbon strip solutions (Marsden, 1990);

Zinc Concentration The amount of zinc required, to precipitate the precious metals, increases as the precious metal concentration decreases. The rate controlling step is the diffusion of metal cyanide complexes to the zinc surfaces (Fmkelstein, 1972); Excess zinc addition must be optimized, as it will consume cyanide, which will allow formation of zinc hydroxide. Metal Concentrations The efficiency of cementation increases with an increase in metal concentration (Finkelstein, 1972); Base metals lnhibit the precipitation of precious metals by forming coatings on the zinc particles; Base metals are large zinc consumers; Base metals precipitated in Menill Crowe will have to be removed in the refining stage of the process. PbNO3 Addition Lead nitrate additions of 10 to 15 parts per million increase the activity of zinc by forming a lead zinc couple. The lead precipitates on portions of the zinc surface forming cathodic areas and adjacent anodic zinc areas. Precious metals are preferentially precipitated on the cathodic lead surfaces and zinc dissolution occurs at the exposed zinc surfaces. Excessive lead addition will result in complete coating of the zinc surfaces, inhibiting cementation (Fmkelstein, 1972; Chi, 1992; Fang, 1992); Lead increases the activity of zinc as described previously and reduces the tendency to form passivating layers of zinc hydroxide as cyanide concentrations decrease. If the cyanide concentrations fall too low, the lead will not prevent zinc hydroxide blinding (Fang, 1992); 0

SEM work by Fang shows a change in the crystal structure of silver deposits on the zinc particles when lead is added. Without lead, a smooth coating of silver effectively covers

1670

the surface of the zinc particles. With lead, the silver deposits form dendrites or clusters of crystals allowing contact of reagents with the zinc surface. Gold was found to form smooth deposits both with and without lead addition. (Fang, 1992); Zinc Particle Size The rate of the cementation reactions increases with a decrease in zinc particle size. The rate of zinc dissolution increases as the zinc particle size decreases, Filtration becomes more difficult as the size of the zinc particle decreases. Scale Formation Zinc dust will act as a seed particle for the precipitation of scale formers such as calcium sulfate and calcium carbonate, which will encapsulate the zinc particle. CHARACTERISTICSOF COMMON CYANIDE LEACH SOLUTIONS

Cyanidation is applied to a variety of ore types, which require different leaching methods and produce pregnant solution chemistries, which range from simple to very complex. The Memll Crowe process can be applied to all of these systems, however there are those in which carbon adsorption, the chief competitor to Merrill Crowe, would prove to be more economic. The most commonly used cyanide leaching systems include heap and vat leaching of coarse (greater than 8 mm) crushed ore, agitated tank leaching of finely ground (80 percent passing 74 microns) ore and flotation tailings, and agitated tank leaching of sulfide concentrates, which may be very finely ground (80 percent passing 45 microns). Zinc precipitation can also be applied to the recovery of metals from high temperature eluant from carbon stripping circuits (Marsden, 1990). Solution characteristics for each system will differ due to the differences in grade, mineralogy, preparation and type of solid liquid separation applied. Solution quality will differ in suspended solids load, types of suspended solids, concentrations of precious and cyanide soluble base metals, solution temperature, free cyanide concentration, the presence of flocculants and scale forming compounds. Carbon, strip solutions are characterized by high temperature and high precious metal concentrations. Vacuum deaeration is not required (Marsden, 1990). PROCESS DESCRIPTION AND DESIGN PARAMETERS This section provides a more detailed description of the Merrill Crowe process. The flow sheet assumed for the purpose of this paper is given in Figure 1, and example ranges of process design

criteria are presented in Table 2. Development of actual process design criteria during engineering will require substantiation though metallurgical testing. Final equipment selection and sizing should be determined using the metallurgical test data and information from equipment manufacturers. Solid Liquid Separation

Pregnant leach solution reporting from the solid liquid separation step will flow to a pregnant solution storage tank. The pregnant solution storage should provide sufficient volume to allow operation of the plant steadily and independently from the rest of the mill. The main issue will be scheduled and unscheduled mill down time. The pregnant solution storage tank also provides additional time for the settling of coarse solids which may carry over from the CCD or filtration sections. If flow is lost to the precipitate filters, the cake will be dropped and should be removed before restarting the plant to prevent blinding.

1671

Table 2 - Merrill Crowe Process - Order of Magnitude Process Design Criteria DESCRIPTION Solid Liquid Separation Pregnant solution storage tank Clarification Filter type

Specific solution flow rate Number of filters Operating/Standby Filter cloth type Filter aid types Suspended solids in feed Suspended solids discharge Deaeration Vessel types Packing types Tower specific flow rate

UNIT hrs

m3/h/m2

mg/L mg/L

m3/h/m2 m3/h/m2 Tower aspect ratio ht to dia. Vacuum required for deaeration mm Hg Pa o2conc. in pregnant solution mg/m3 O2 conc. in barren solution mg/m3 Precip filter feed pump type Zinc Addition Zinc feeder type Zinc addition rate Stoichiometric Gold g Znfg Au Silver g Znfg Ag Mercury g Z d g Hg Copper g Znfg c u Excess zinc addition rate Typical Metal Concentration vs. Excess 100 ppm 5 ppm 1 PPm Zinc induction Very high grade solution Lead nitrate addition Precipitate Filter Type

Specific Flow Rate Operating cycle time Precipitate Composition Gold + Silver Zinc Lead Copper Mercury Insol Barren Set Point

VALUE/SPECIFICATION CCDRiltration 4 Horizontal pressure, US Vertical pressure, Funda Vertical tubular, Stellar 1.5 - 2.0 2 1/1 Polypropylene Diatomaceous earth or Perlite 100 - 300 1 Packed tower Rings or tellerettes 50 - 85 70 2:l -3:l 500 67,500 6 1 Vertical centrifugal

REFERENCE Maintenance

Atwood, 1985 Minimum

Atwood, 1985 Design

Varies WJ T and P Design Target

Variable speed auger 0.33 0.61 0.33 1.03 150 - 300%

Calculated Calculated Calculated Calculated Design

10% 200% 1,500%

Atwood, 1985 Atwood, 1985 Atwood, 1985

20&300ppm ppm

Agitated mixing cone Dry zinc to reaction tank 10- 15

Mansanti, 1989 Fang, 1990

m3/h/m2 days

Plate and Frame Horizontal leaf (Funda) Tubular (Stellar) 4.5 7

% % % % % %

g Au/t

30 - 80 5-30 0.2 - 2 0.1 - 2 0-2 5-15 1.7

1672

Typical Typical Ore Dependent

Mansanti, 1989

Clarification The pregnant solution, typically thickener overflow solution will be pumped from the pregnant solution storage tank through pressure filters, to reduce the solids concentration from a typical 100 to 300 part per million range to less than 1 part per million. Clarification is achieved by the use of pressure filters containing either horizontal or vertical leaves, which can be automatically washed. At least two pressure filters should be available which can be operated in parallel with one operating and one on standby to allow continued operation through wash cycles. The pressure filters consist of horizontal tanks, which contain a series of filter disks mounted on the solution dischargepipe, which extends through the center of the tank. The filter elements or disks are covered with a polypropylene cloth, which is must be precoated with a layer of fine silica to create a bed of filter media to trap very fine particles. The fine silica consists of either diatomacem earth or perlite and can be purchased with varying particle size distriiutions. The blend of precoat materials used for a given operation is selected to provide the required solution clarity, the highest flow rate, lowest pressure drop and longest filtration life. The optimization of these variables will be unique to each operation and procedures will be developed based upon each specific ore type. In addition to precoating the filters, a continuous addition of the diatomaceous earth may be added directly to the pregnant solution as body feed, when the suspended solids are very fine or clay rich. Pregnant solution is pumped through the pressure filters until the pressure drop increases to a preset l imitbased on the pressure rating of the filter and the flow rate through the filter, which will decrease as the pressure increases. At the end of the filtration cycle, a second parallel filter is placed on line to maintain the desired process flow rate, and the filter to be cleaned is taken off-line. The filter tank is drained along with filtered solids that will slump by gravity from the filter clothes when the process solution flow is discontinued.The filter cloths are then washed with built in high presswe sprays which remove filter cake that may stick to the cloths. The film discs rotate past the sprays. Upon completion of the wash cycle, the vessels are closed and filled with barren solution and the precoatingprocess is initiated. The precoat system consists of a precoat mix tank and circulation pumps. The appropriate amount of precoat is added to the precoat mix tank and the precoat sluny is then circulated in a closed loop h u g h the filter vessel until the returning solution becomes clear. The filter is then put back into operation by introducing pregnant solution while closing the precoat circuit, being careful to maiutain differential pressure on the filters so that the layer of-precoatwill remain in place. If ~ ~ t pressure i a l is lost across the filter the precoat will fall to the bottom of the vessel and the procws must be repeated. The sizing of the filters is based on the flow rate of pregnant solution and the amount and type of suspended solids contained in the solution. Typically the filters are designed for a specific flow rate in the range of 1.5 cubic meters per hour per square meter of filter area. Blinding of the filter cloth is the leading problem, which affects the performance of the clarification filters and may determine the ultimate cyck time. Some of the sources of blinding include: 0

Flocculent from the CCD circuit; Very fine suspended solidq clays; Calcium carbonate precipitation or scaling can cause premature blinding of film. Antidants are used to prevent the scaling p m b h

1673

Precipitation of sodium silicate gel on the filters can be a problem if significant amounts of sodium hydroxide are used in the leaching circuit. Potential sources of sodium hydroxide in the leaching circuit include cyanide addition, especially when using low concentration, recovered cyanide, and final pH control. Precipitation of iron hydroxides. Deaeration - Crowe Tower Deaeration of the clarified pregnant solution is typically accomplished using a packed tower under vacuum. Clarified pregnant solution discharging the filters flows to the top of the deaeration tower, where the solution is distributed over a bed of packing, which provides surface area for thin film formation and release of dissolved oxygen. A vacuum pump is used to reduce the pressure within the vessel to approximately 500 mm Hg or 67,500 Pa. Evolved gases including oxygen and ammonia are exhausted through the vacuum pump. Deaerated solution discharges from the tower, by gravity, through a vertical discharge pipe directly into the suctions of the product filter press feed pumps. The height of the discharge pipe, which remains full of solution (the barometric leg), is determined by the amount of head required to overcome the vacuum developed by the vacuum pump and to provide the required net positive suction head for the filter press feed pumps. Various types of towers have been used for deaeration, fiom the original Crowe tower, which contained triangular section, wooden boards, to modem packed towers that use efficient polypropylene rings, saddles or tellerettes. In addition to packed towers, bubble cap tray or baffle towers could be used. Baffle towers consist of plate arrangements within the towers. The solution will cascade over the edge of one plate to the next forming a thin curtain of droplets providing surface area for transfer. Baffle tower arrangements include the disk and donut type and opposing inclined plate. The advantage of baffle towers is that they are less susceptible to blockage by carbonate or gypsum scale than the more efficient dumped packing (Fair, J.R., 1984). In sizing a packed tower, the diameter is determined by the liquid flow rate, the capacity of the packing and the gas flow rate, which moves countercurrent to the solution. These rates along with the characteristics of the liquid and gas streams can then be used to estimate the pressure drop and predicted flooding points for a packed tower using the pressure drop correlation by Eckert (Fair, J.R.,1984). The pressure drop correlation helps to define the acceptable operating range of a tower and provides a means to optimize the tower diameter. The height of the packing within a tower can be determined fiom the overall volumetric liquid phase mass transfer coefficient, &a, for the process, which is found experimentally (Edwards, W.M., 1984). The relationship between the mass transfer coefficient and the packing height is given by the equation:

where nA is the overall rate of transfer of solute A, hT is the total packed depth in the tower, S is the tower cross sectional area, and AX*lm is the log mean concentration difference given by: Ax*lm= (x* - x)2 - (x*- X)I/I~[(x*-x)~/(x*- X)I] where subscripts 1 and 2 designate the bottom and top of the tower respectively. x* is the mole fraction of gas in the solution which is in equilibrium with the bulk gas concentration, and x is the mole fraction of gas in solution (Edwards, W.M., 1984). The equilibrium gas concentration in the solution at the specified pressure can be determined using Henry’s Law,

1674

where H is the Henry’s law constant, PA is the partial pressure of A in the gas phase and XA is the mole fraction of A in solution. If A is assumed to be air and pA is the total pressure, the concentration of oxygen can be estimated from the chemical composition of air (Edwards, W.M., 1984). The desorption of oxygen, hydrogen and carbon dioxide from water was studied by Sherwood and Holloway and correlations were developed for the mass transfer coefficient and the height of a transfer unit for the system. The correlations can be used along with the generalized equation by Cornell to determine the packmg height for various types of packing (Fair, J.R., 1984). The best source for mass transfer data including the overall mass transfer coefficient would be from existing commercial operations. Performance data on various packing types and materials can be obtained from the packing and tower internals manufacturers. The specific flow rate for preliminary design of packed, vacuum, deaeration towers is 3 2 approximately 70 to 85 m /h/m . The height to diameter ratio for reported towers ranges from 1:1 to 3:l. The 3:l ratio is more common and is thought to be more efficient as long as the tower is not allowed to operate in a flooded condition. The vacuum pump is sized for the maximum amount of gas to be transferred and the target absolute pressure of 67.5 kPa.

Cementation Zinc powder is metered into the deaerated pregnant solution using various types of feeders. Some commonly used feeders include, variable speed auger type feeders and rotating disk. In most cases it is necessary to use a vibrating feed hopper to prevent bridging. Zinc powder is typically fed into a small agitated mixing cone containing barren solution, which is positioned above the suction of the filter press feed pumps. The zinc slurry will flow by gravity, through a control valve, into the pump suction. The zinc cone should be installed on a platform adjacent to the deaeration tower and above the liquid level in the tower to allow gravity flow into the pump and prevent overflowing when the filter press, feed pumps are shut down. A steady head tank will be provided to maintain a constant level in the mixing cone. The zinc addition rate will be the stoichiometric amount of zinc required to precipitate the precious and base metals in solution plus an excess amount, which will be dependent on the metal concentrations of the solution. The excess zinc required with respect to metal concentration in solution has been reported to be ten percent for solutions containing 100 parts per million metal, 200 percent for solutions containing 5 parts per million metal and 1500 percent for solutions containing 1 part per million metal (Atwood, 1985). (Marsden, 1991) reported that five to 10 times the stoichiometric requirement would be needed for carbon eluants. The cementation reaction occurs very rapidly and sufficient retention time is available for the reaction to take place in the pipeline between the filter press feed pumps and the filter presses. The key issue in the design of the filter press feed pumps is the prevention of air ingress. Leakage of air through the pump shaft seal will allow oxygen to enter the system. The standard method of coping with this problem is to use inline vertical centrifugal pumps, which are submerged in barren solution above the shaft seal. Operating problems reported in adding zinc slurry to the system include wetting of the zinc, control of the zinc solution feed rate, plating of metals on the wetted parts of pumps and build-up of zinc and precipitate in the pumps and pipelines. Extreme cases are reported in the treatment of

1675

very high grade carbon strip solutions (Mansanti, 1989). In the Chimney Creek case, the plating and build-up problem lead to the addition of a separate pumping system to inject the zinc slurry into the press feed pump discharge line. One recommendation to minimize the plating problem or at least contain it was to add dry zinc to an agitated reaction tank to permit the precipitation reaction to occur before entering the piping system (Mansanti, 1989).

Precipitate Filtration Filtration of the precipitate has been accomplished in various types of pressure filters. A common type is the plate and frame filter press. Others include enclosed, automated, horizontal leaf filters and candle or tubular type filters. Plate and frame filter presses can be fitted with canvas or polypropylene filter clothes. In some cases a combination of canvas filter clothes covered by filter paper has been used. The filter presses are typically precoated with diatomaceous earth or perlite at the beginning of the filtration cycle to prevent blinding. Zinc can be added along with the precoat to insure that the gold is completely precipitated during the start-up of the precipitatiodfiltration cycle. During steady state operation, the presses will contain excess zinc. The filters will be operated for specified periods of time determined by the maintenance schedule, the accounting schedule, the time it takes for the pressure to reach the maximum recommended operating level or the flows decrease to an unacceptable level. At the end of the selected operating cycle, the filters are taken off line and drained. Compressed air is blown through the filters to force as much of the liquid out of the filter cake as possible. The presses are then opened and the filter cake is dropped into carts, which can be transported to the refinery. This operation is sensitive to security and is typically performed under the supervision of the refinery personal and plant security staff. The filter clothes are scraped and washed, the presses are reassembled and the system is put back in operation. The precipitate is then transported in the same cart to the refinery area for smelting.

Clean-up and Smelting There are different approaches to the treatment of Memll Crowe precipitates in the refinery depending on the composition of the precipitate and the types of equipment available. The three main types of furnaces referred to in the literature include gas or diesel fired reverberatory, submerged arc and induction type furnaces. Induction furnaces are typically not the best choice for the smelting of zinc precipitate unless the precipitate is acid washed prior to smelting. Induction furnaces only heat the metal and so the slag is heated indirectly. The high quantities of fluxes result in reduced crucible life. The reverberatory furnace has been used successfully in operations requiring high flux to charge ratios. The reverberatory furnace is either gas or diesel fired and provides heating directly to the top of the melt allowing the use of higher temperature slag mixtures. The precipitate clean-up and smelting processes at the Nerco DeLamar Mine and at Coeur Rochester were similar. The precipitate at DeLamar in the late 1980s and early 1990’s was filtered using plate and frame filter presses. The filter cloths were covered by filter papers, which were removed along with the precipitate during cleanup. The wet precipitate was mixed with fluxes and loaded into the 2000 lb copper reverberatory furnace. The smelting process was performed in a single step and dore bars were poured. It should be noted that the precipitate at DeLamar was relatively free of base metals. Treatment at the Rand refinery was much more involved. The precipitate was washed from the Stellar filters into vats in which sulfuric acid was introduced to dissolve the zinc. After acid washing, the remaining gold leach residue was filtered using plate and frame, drum or belt filters

1676

and the filter cake was packed into calcining trays. The residue was calcined at a temperature ranging from 550 to 700 degrees C for 16 hours. The resulting calcine was then smelted in submerged arc furnaces. The treatment process at Paradise Peak included an initial sulfuric acid leach step followed by filtration and retorting of the filter cake in order to remove and capture the mercury. The retorting process was performed over 24 hour period at a temperature of 730 degrees C (1350 F) (Mansanti, 1989). Equity Silver acid leached the precipitate using hydrochloric acid. The residue was filtered, drying and fired in a 225kg induction furnace. The buttons from the first pour were then remelted in a 45 kg induction furnace. Initially, sulfuric acid was used to leach the precipitate, however, sulfates and sulfides remaining in the leached residue resulted in the formation of significant sulfide matte layers during melting (Semple, 1987). OPERATING COST VARIABLES

Consumables The operating costs associated with the Memll Crowe process are associated with operating and maintenance labor and consumable items. The preliminary design criteria presented in Table 2, include estimated unit consumptions for some of the consumable items. The following is a summary of the key operating costs

Manpower Operating Labor Maintenance Labor Consumables Power Cyanide Filter aid Filter cloth Propane

zinc Lead nitrate Antiscalant Filter paper (if used) Freight

Actual operating costs for a given project should be estimated based on the results of metallurgical test work, local labor costs, local power and quotes for all consumables including freight. In very remote regions, the availability of operating and maintenance supplies may dictate the type of plant constructed.

CONCLUSIONS The zinc precipitation or Memll Crowe process has proved to be a flexible, time tested method for the recovery of precious metals from cyanide leach solutions and can be applied to the majority of ore types except those containing organic carbon. The key parameters for the successful operation of zinc precipitation include: Low concentrations of suspended solids Low concentrations of dissolved oxygen Low base metal concentrations Optimized free cyanide concentration during precipitation Optimized zinc addition rate 10 to 15 ppm lead nitrate addition

1677

The carbon adsorption processes have steadily improved over the last twenty years and are now reported to be more efficient and less costly than zinc precipitation for many applications. Cases in which zinc precipitation remains more cost effective than carbon adsorption include the treatment of ores containing high silver to gold ratios and ores containing significant mercury concentrations. Carbon adsorption is required for processing ores containing organic carbon and carbon adsorption has an advantage over zinc precipitation when high base metal concentrations are present and where high clay is present preventing effective solid liquid separation. To obtain the benefits of both processes when treating high silver ores, operators have applied a combination of Merrill Crowe and carbon in leach. Initial pregnant solution or No. 1 thickener overflow solution is treated in a Merrill Crowe circuit to remove the majority of the silver and the thickener underflow and remainder of the slurries and solutions are treated using either carbon in pulp or carbon in leach processes.

REFERENCES 1. Atwood R.L. and R.H. Atwood. 1985. Design Considerations for Merrill-Crowe Plants. Society of Mining Engineers of AIME. Preprint Number 85-353. 2.

Chi, G. 1992. Study of Merrill Crowe Processing: Solubility of Zinc in Alkaline Cyanide Solution. Master of Science Thesis, University of Nevada-Reno, Nevada.

3. Dorey, R., D. van Zyl and J. Kiel. 1988. Overview of Heap Leaching Technology, In Introduction to Evaluation, Design and Operation of Precious Metal Heap Leaching Projects, Chapter 1, ed. van Zyl, D.J.A., I.P.G. Hutchison, and J.E.Kie1, Chapter 1. Society of Mining Engineers Inc. 4. Edwards, W.M., 1984. Mass Transfer and Gas Absorption, In Perry’s Chemical Engineers ’ Handbook, ed. Perry, H.P., D.W. Green, and J.O. Maloney, Sixth Edition, Section 14. New York: McGraw-Hill Book Company. 5. Fair, J.R., D.E Steinmeyer, W.R. Penney, and B.B. Crocker. 1984. Liquid-Gas Systems, In Perry’s Chemical Engineers’ Handbook, ed. Perry, H.P., D.W. Green, and J.O. Maloney, Sixth Edition, Section 18. New York McGraw-Hill Book Company.

6. Fang, M. 1992. Solubility of Zinc in the Merrill Crowe Process. Master of Science Thesis. University of Nevada-Reno, Nevada. 7. Fmkelstein, N.P. 1972. The Chemistry of the Extraction of Gold from its Ores, In Gold Metallurgy in South Africa, ed. Adamson, R.J., Chamber of Mines of South Africa, Johanesburg. 284-347. 8. Fleming, C.A. 1998. Thirty Years of Turbulent Change in the Gold Industry. Proceedings 30thAnnual Operator’s Conference of the Canadian Mineral Processors. Paper No. 3.

9. Haggerty, S.1991. An Overview of Corona’s Nickel Plate Mine. Proceedings 21fhAnnual Meeting of the Canadian Mineral Processors. Paper No. 16. 10. Hedley, N. and Kentro, D.M. 1945. Copper Cyanogen Complexes in Cyanidation. The Canadian Institute of Mining and Metallurgy Transactions, Volume XLVIII, 237-25 1. 11. Hedley N. and H. Tabachnick. 1958. Chemistry of Cyanidation. In Mineral Dressing Notes Number 23. New York. American Cyanimid Company.

1678

12. Lindeman, D. and R. M. Nendick, 1992. A Comparison of South African and North American Practices in the Extractive Metallurgy of Gold. 24'h Canadian Mineral Processors Conference,Ottawa, Ontario. 13. Mansanti, J.G. and M.F.Gleason.1989. Funda Filters for Zinc Precipitation, Start Up and Operation. 2"dAnnual Intermountain Mining & Processing Operators Symposium, Elko, Nevada. 14. Marsden, J.O. 1990. Practical Aspects of the Chemistry of Zinc Precipitation from High Temperature Carbon Eluates, Randol, Phase IK Innovations in Gold and Silver Recovery, Volume 11, Chapter 33, pages 6357- 6361. 15. Milligan, D. A., O.A. Muhtadi, and R.B. Thomdycraft. 1988. Metal Production. In Introduction to Evaluation, Design and Operation of Precious Metal Heap Leaching Projects, Chapter 9, ed. van Zyl, D.J.A., I.P.G. Hutchison, and J.E.Kie1, Chapters 1. Society of Mining Engineers Inc.

16. Moore, W.J. 1983. Electrochemical Cells. In Basic Physical Chemistry, Chapter 17. New Jersey: Prentice Hall, Inc. 17. Muhtadi, 0. A., 1988. Metal Extraction (Recovery Systems), In Introduction to Evaluation, Design and Operation of Precious Metal Heap Leaching Projects, Chapter 8, ed. van Zyl, D.J.A., I.P.G. Hutchison, and J.E.Kie1. Society of Mining Engineers Inc. 18. Nendick, R.M. 1983. An Economic Comparison of the Carbon-in-Pulp and MemllCrowe Processes for Precious Metal Recovery. Process Economics International, Vol. II1,No. 4. 19. Peter, E.L., R.C. Reid, and E. Buck. Physical and Chemical Data. In Perry's Chemical Engineers' Handbook, ed. Perry, H.P., D.W. Green, and J.O. Maloney, Sixth Edition, Section 3.New York: McGraw-Hill Book Company. 20. Shank, R. 1976. Leaching Chalcocite with Cyanide. Engineering and Mining Journal, October. 21. Semple, P.G. 1986. Equity Silver Mines Scavenger Circuit. Proceedings lgh Annual Meeting of the Canadian Mineral Processors. Paper No. 7. 22. Vanysek, P. 1983-1984. General Chemistry, Electrochemical Series. In CRC Handbook of Chemistry and Physics, ed. Weast, R.C., M.J. Astle and W.H. Beyer, 6 4 Edition, ~ Section D, Florida: CRC Press Inc.

1679

Selection and Design of Carbon Reactivation Circuits Dr. Joerg von Beckmann’ and Paul G. Sempl2

ABSTRACT This paper deals with the selection and sizing of various components in carbon reactivation circuits. Carbon reactivation includes the acid washing and the thermal reactivation of activated carbon. The acid wash circuit design addresses the flow sheet design, process control and materials of construction aspects of the acid wash circuit. The thermal reactivation circuit includes design consideration for feed bins, quench tanks, screens, carbon feeders, and carbon regeneration kilns. Comparative operating costs of Merent energy sources and kiln efficiencies are discussed. The need for off-gas particulate scrubbing and mercury vapour handling is addressed. In addition to the design of the carbon reactivation circuits, some procedures for monitoring the performance of the circuit are discussed. INTRODUCTION The carbon reactivation circuit is series of unit processes designed to restore the activated carbon’s ability to recover precious metals from cyanidation circuit solutions. Since each circuit will be treating a unique solution, which will result in unique carbon fouling problems, the reactivation circuit design must consider several variables, which includes the preference of the plant operator. The main unit operations within the reactivation circuit are acid washing, elution and thermal reactivation. This paper will focus on the design considerations in the acid wash and thermal reactivation circuits, although it should be noted that all three operations must be operated efficiently to ensure proper carbon reactivation and the resultant low soluble gold losses from the cyanidation circuit. A typical flow-sheet for a carbon reactivation circuit is shown in Figure 1. ACID WASH CIRCUIT The first stage in the carbon reactivation circuit is the acid wash circuit. Early carbon circuit designs often considered acid wash after carbon elution but recent designs have typically been based on acid washing prior to elution. Acid washing prior to elution has become prominent due to the following:

Hydrochloric acid has become the predominant acid used and precious metal loss, especially silver, when nitric acid was considered has been minimized, 0 The acid wash tank can be used to measure loaded carbon batch sizes, 0 Most operators believe that removing scale build-up prior to elution improves the overalll elution efficiency.

0

Several operators, especially in Australia, utilize a common acid-wash elution tank.In this instance the loaded carbon is recuvered in a holding tank and then transported to the elution tank for further treatment. This practise is less common in North America, and the design considerations for this alternative are discussed later.

’ Lochhead Haggerty Engineering. & Mfg. Co. Ltd., Delta, B.C., Canada Penguin Automated Systems Inc., Oakville, On
1680

The objective in the acid wash circuit is to remove scale build-up on the carbon, thereby opening up the activated carbon micro-pores. The step is important to maintain the carbon's ability to recover additional gold as well as maximize the surface exposure that will improve the overall gold elution efficiencyin the downstreamprocessing. Early test-work in acid washing also considered the removal of base metal contaminants during acid wash. The acid washing circuit will remove some base metals but the overall efficiency is low since activated carbon has a high affinity to base metals at a low pH. Once these mechanisms were understood, it became apparent that base metals, especially copper, could be more effectivelyremoved in the elution circuit. Process Flowsheet As previously mentioned there are two predominant options for acid washing. The first option utilizes a stand alone acid wash tank while the second option utilizes the elution column as the acid wash vessel. The selection of the option depends on the operator preference, and the second option is predominantly used in Australia. In either case the circuit will include an acid wash pump, pump box and acid metering pump. The loaded carbon will be recovered from the loaded carbon screen and transported to the acid wash tank.The acid, typically hydrochloric acid, will be metered into the acid wash pump box and the acidic solution pumped through the acid wash tank.The overflow solution from the acid wash tank will return to the acid wash pump box where additional acid will added and the solution is recycled through the acid wash tank. Once the acid wash cycle is complete the acidic solution is drained back to the acid wash pump box where it is either neutralized and discarded to tailings or retained as the solution for the next acid wash cycle. Typically hydrochloric acid is utilized in the acid wash circuit although nitric acid has been used at several operations. Although the use of nitric acid did simplify the materials of construction, there were claims made by some carbon manufacturers that the use of nitric acid &d attack the cell structure of the carbon and lead to excessive carbon fines losses after numerous

1681

cycles through the acid wash circuit. These claims have caused most operators to utilize hydrochloricacid in the circuit. Process Design Criteria Acid Wash Tank Carbon Bulk Density Carbon Batch Size

440 m i t e r

1-10 tonnes, defined by carbon loading and movement rate 4 hours day One cycle per day TypicaUy >4: 1

Cycle Time Acid Wash Cycle Tank H:D ratio Carbon Bed Expansion Freeboard Tank Bottom

50-100% 0.5-1.0 meter conical

Acid Wash Pump Box Geometry H:D Ratio Volume Pumping Rate

Flat bottomed tank Typically 1:1 Equal to acid wash tank 2 Bed Volumes per hour

Acid Addition Acid Used Acid Consumption Acid Pump Acid Storage Neutralizing Agent

Hydrochloric 25-50 kg/tonne carbon Metering or barrel pump Barrels Caustic

Materials of Construction The use of hydrochloric acid limits the use of stainless steel alloys, and most acid wash circuits utilize either plastic or FRP tanks. The Australian design where the elution column is used for acid washing incorporates a butyl rubber insert liner into the elution culumn. This design was also initially used in South African gold plants but has been largely replaced with stand-alone acid wash tanks. The main issue around the liner in the elution column was maintaining the liner when exposed to high acid concentrations and subsequent high temperature conditions. The Australian operators tend to accept this maintenance aspect and replace the column after the h e r fails and the structural integrity of the column is compromised. Most North American operators will not accept this inconvenience and acid wash circuits are designed as stand alone unit operations. Process valves within the circuit must be selected based upon their location and the duty to which they are exposed. Several options of introducing the solution into the acid wash circuit have been utilized.Each option has Merent requirements for the material of construction of valves and screens if required. Since screening material under acid conditions typically requires expensive acid resistant materials most recent circuit designs attempt to avoid screens when ever possible. Process Control Once the acid wash circuit is ready to be operated the process control is typically limited to a metered addition of acid into the acid wash tank and circulation of acid through the column for a fixed period of time, typically 4 hours. Some operations control the acid flow into the system based upon a pH reading from the acid wash tank and control the solution pH in this tank at a setpoint of 1.0. Other operations monitor both the pH in the acid wash tank and the pH of the tank overflow solution. In this instance the circuit is operated until the two pH readings converge indicating that no additional acid is being consumed and the acid wash benefit is diminishing.

1682

Containment The acid wash circuit is located within a facility that contains numerous solutions containing cyanide and care must be taken to avoid accidental mixing of these solutions. The acid wash circuit will be designed with its own separate containment area that is typically lined with a polymer coating. This coating is required to prevent accidental spills from attacking the concrete and degrading the floors and curbs. The volume of the containment must be sufficient to hold the contents of all tanks containedwithin the circuit. Since the loaded carbon screen may be located over the acid wash circuit the screen undersize must be directed to a pump outside the acid wash area and within its own containment ELUTION CIRCUIT The design of the elution circuit is beyond the scope of this paper, but it is important to note that proper elution plays a significant role in carbon reactivation. In addition to removing the precious metals, thereby re-establishing the carbon’s ability to recover additional precious metals, the elution circuit can be utilized to remove base metals, especially copper. Base metals such as copper can be removed during a pre-soak with a high cyanide solution. This approach has proven to be extremely effective and is much more efficient than attempts to remove copper in the acid wash circuit. The elution process must be monitored and operated in a manner where the maximum precious metals and base metals are removed from the carbon. There have been operations that were not effectively removing copper from the carbon during elution and this copper was contributing to poor carbon reactivation and ultimately high soluble loses. This can be especially important when silver is present since inefficient copper and silver removal cau lead to alloys of copper-silver building up on the carbon during thermal reactivation. These alloys, once formed, will not be removed in the next carbon elution cycle and will lead to escalating stripped carbon assays and poor carbon activation. In summary, elution does play an important part in carbon activation, and during operations it is important to survey the carbon assays to make sure the elution circuit is not only effectively removing precious metals but also ensure that the maximum base metal removal is also being achieved. THERMAL REACTIVATION A number of factors dictate the selection and sizing of equipment for the thermal carbon reactivation circuit, including the frequency of operation, the most economical energy source and the presence of mercufy in the circuit. The selection of type of kiln,whether horizontal, vertical or other, often depends on a combination of past experience, capital cost, life expectancy of the equipment, and personal preference. Horizontal kilns have proven to be the most reliable and trouble free and are the most prevalent in the industry today. The main focus of discussion will be with reference to horizontal thermal carbon reactivationkilns. There are several papers (Avraaides, Miovski, and Van Hooft 1989, Urbanic, Jula and Faullcner 1985) discussing the optimum environment for the reactivation process. The generally accepted conditions for reactivation are carbon temperatures in the range of 650 to 75OOC for a period of 15 to 30 minutes, in a non-oxidizing environment. Steam is the most economical inert atmosphere to blanket the carbon and prevent oxidation of the carbon at the higher temperatures. Whether the steam is produced in the kiln by the introduction of wet carbon, or whether it is produced separately and injected into a kiln with dry carbon, is from an energy consideration, academic. However, there is evidence suggesting that the expulsion of steam from within the carbon particle is beneficial in the reactivation process, hence it is better to produce the steam from wet carbon than to inject steam into the kiln.

Feed Bins, Quench Tanks and Screens Carbon is transported throughout the circuit using water, either by pumping or educting. The carbon concentration in the slurry is low, and the excess water must be removed by a dewatering screen before being introduced into the kiln.

1683

The optimum moisture content of the carbon as it enters the kiln depends on the type of kiln, the condition of the seals which prevent air from entering the kiln (or steam leaving). Successful reactivation can be achieved with moisture contents as low as 20% by weight with good seals and without the injection of supplementary steam. Most mechanical methods of dewatering can achieve moisture contents in the range of 50% to 35% without the use of preheating the carbon. It is important to keep in mind that the mechanical dewatering of carbon is much less costly than thermal dewatering. Any excess water which is not required for reactivation or to produce the inert atmosphere, should be removed before it enters the kiln. It takes 0.63 kWh to convert 1 kg of water to steam and a carbon circuit reactivating 100 kg carbon per hour at 50% moisture content (100 kg water + 100 kg carbon) would save the equivalent of 21 kW by reducing the moisture content from 50% to 40%. At $O.OS/kWh, for example, this would equate to $9,20O/year savings and it would be more economical to spend these initial savings on a high quality dewatering screen to guarantee 40% moisture. The feed bin should be sized equivalent to the carbon elution batch or larger. Excess water that can be drained through lower screens will provide additional savings to the cost of evaporating. If the dewatering screen is capable of reducing the moisture content to 40% or less, there will be minor savings, while at 50% moisture the carbon will continue to drain in the feed bin. Quench tanks are often undersized, when they are batch operated. The quench tank needs to have carbon storage capacity equal to the production rate multiplied by the time between emptying cycles. Whether the quench tank is operated on a batch or continuous process, the rate of water removal from the tank must be equal to the water entering the tank.The level in the tank must not drop below the depth of the discharge chute, else the air seal is lost, and the carbon drops into an empty tank at over 600°C. When water is added to the tank,quantities of steam dangerous to personnel will be produced for a short period of time. Another consideration in the quench tank is that carbon adds 0.24 kWh/kg carbon to the quench tank water. This heat must be removed through the introduction of cool water, or the circulating water volume must be sufficient to absorb this energy without overheating. In many installations, building height is limiting in the design of the feed bin and quench tank. Often the feed bin is sized for a full batch, but the quench tank size will be reduced since insufficient height is available. Feed bins having included angles of 60 degrees or greater, and quench tanks having included angles of 45 degrees or greater will empty completely. Carbon Feeders Feeding wet carbon into a kiln requires that the feeder act as a seal, or air lock, between the feed bin and the kiln inert atmosphere. At the same time, carbon attrition is a consideration for all mechanical handling systems. An inclined feed chute will not work by itself, since there is no way to control the feed rate to the kiln. A knife gate is not suitable, since it tends to bridge and not provide uniform flow. An inclined feed screw or auger having variable pitch and incline backwards in the direction of flow provides an excellent seal, and with additional dewatering screens below the screw, the screw will provide supplementary water removal. The screw should not be operated at high speed. Feed screw diameters vary from 50 mm to 200 mm.

Energy Sources Available and Cost The most common sources of energy for operating kilns are electric power and fossil fuels. Included in the common fossil fuels are natural gas, diesel fuel oil, and propane. In arriving at a decision as to which energy source is most economical, the unit price of the energy source is not the only consideration. An electric furnace and a fossil fuel furnace do not have the same thermal efficiencies,and this must be factored into the decision. The major differencebetween electric and fossil fuel furnaces, is that the electric furnace does not require the combustion of air in order for heat to be released or transferred. An electric furnace is refractory lined with resistance type electric elements which surrounds the kiln shell, and heat is transferred to the shell though radiation and free convection. There is no air flow requira and

1684

there is no air required entering or leaving the furnace for the heat transfer process to be completed. The fossil fuel kiln.although having the same configuration as the electric furnace, generally has a series of tangentially aligned burners firing to the underside of the kiln shell. Heat is released through combustion, and combustion requires a continuous supply of air being introduced into the furnace to oxidize the fuel. The hot gases produced lose much of their heat as these products of combustion come in contact with the cooler kiln shell. Here the heat transfer process is partly through radiation and mostly through forced convection. Combustion of air, without enhanced oxygen, will under stoichiometric concentrationsof air and gas (or oil) produce flame temperatures in the order of 1900°C. The temperature of the combustion products as they leave the furnace in conjunction with the flame (or hot gas furnace) temperature determines the efficiency of the furnace. A furnace having a stoichiometric flame temperature and an exiting flue gas temperature of 15°C would have an efficiency of 100%. That is, all the heat that was released by the combustion of fuel has been removed, and the exiting products of combustion have the same temperature as the entering combustion air. Boilers almost achieve this efficiency, since they are never allowed to run out of water (the heat sink) and have flue discharge temperatures which approach the inlet combustion air temperatures. One key to this efficiency is the constant water load. A boiler which runs out of water and continues to operate, rapidly turns to molten metal. A kiln operating under stoichiometric conditions risks a similar fate. Kiln furnaces cannot be operated at boiler efficiencies since the carbon load cannot be guaranteed at all times and the maximum flame temperature cannot, from a practical consideration, exceed the melting point of the metal shell. Since the heating of the kiln shell is indirect, the furnace temperature must be equal to or exceed the process temperature. In this case, the flue gas temperature cannot be much less than 700°C if the product temperature is expected to be 650°C. With a hot mix temperature of 12OO0C, the efficiency of this furnace, neglecting heat losses, is approximately 40% (Figure 2). In other words, 60% of the heat is which is generated through combustion cannot be recovered in the process. An electric kiln, by comparison would have a 100% efficiency. It is assumed in this comparison that the radiant and convective losses through the furnace walls for both the electric and fossil fuel furnace can be considered equal, since each operates at the same temperature rnerentials and insulating properties. Using these figures, the electric fuIllace is two and a half times more efficient in converting the source energy to heat. For cost comparison, therefore, the unit energy cost of electricity would need to be two and a half times more expensive than the fossil fuel before one would select fossil fuel as the preferred energy source. For example, if electric power is available at $O.OS/kWhr, diesel fuel must be available at $0.2 1L. Heat recovery techniques are available that allow the preheating of combustion air or preheating of the carbon, but few of these alternatives will achieve the energy efficiency of the electric kiln. Both the electric kiln (Figure 3) and the fossil fuel fired kiln (Figure 4) have a process stack that discharges the steam and organics (flotation reagents, cyanide, oils, etc.) liberated by the reactivation process. For an electric kiln, this is the only stack. The fuel-fired kiln has in addition to the process stack a furnace products of combustion stack. This extra stack may require a separate emissions permit, and the initial permitting and maintenance of this permit may influence the decision. The most common sources of energy for operating kilns are electric power and fossil fuels.

1685

Figure 2 Available heat from fuel fired furnace (taken from North American Combustion Handbook, Volume 1, 3rdEd., 1986). Example: If the maximum flame temperature is 1200"C, corresponding to 100% excess air curve, follow the curve up to flue gas temperature?where the curve intersects with 700°C. Read left to 40% available heat. Regeneration Kiln Selection There are a number of different designs of carbon reactivation kilns available, and these may be divided into three main groups - vertical, horizontal rotary and horizontal rotary W d r y e r combinations. In all cases, the basic heat transfer mechanism is through conduction, radiation, or both.

1686

+ PROCESS GASES

F

CARBON

Y

KILN ELECTRIC WRNACE

WENCH TANK

Figure 3 Typical Horizontal Electric Kiln

ROCESS GASES

KILN FEED BIN

CAR!ON

1

7!

KILN GAS FURNACE

QUENCH TANK

Figure 4 Typical Standard Horizontal Fossil Fuel Fired Kiln

1687

Vertical Kilns The vertical kiln (Figme 5 ) has been used successfully in some applications, and not so successfully in others (Semple 1987). The kiln consists of a vertical shaft, annulus or series of tubes through which the carbon flows by gravity. The heat is applied to the outer surface of the carbon transport mechanism, usually supplied by a separate furnace attached to the hln assembly. Carbon can also be directly heated using electrical current passing though the carbon bed, using electrodes. The carbon is fed into the top of the kiln, usually after some form of predrylng from furnace off gases. The transport of the carbon is a plug flow. Some designs have used spoiler devices within the tubes or annulus to cause mixing of the carbon, to allow more contact with the heated surfaces.

Figure 5 Typical Vertical Kiln In the vertical kiln, the heated gases flow countercment to the carbon, and the carbon is tightly packed in the tubes or annulus. The flow of steam and organics released in the reactivation process is also countercment to the flow of carbon, with the hot steam and organics coming in contact with the coldest carbon. The net effect is a column type distillation where compounds evaporate only to be recondensed M e r up the column, resulting in fouling of the carbon, or a buildup of compounds in a section of the kiln. Removal of the steam and organics is not immediate, since the gases must flow through the packed carbon bed.

Horizontal Rotary Horizontal rotary kilns require that the carbon be transported along the inside of a cylindrical tube by rotating the tube. This rotary shell is normally positioned at a slight incline, or slope, to facilitate the movement of the carbon from the feed to the discharge end. Heat is supplied to the outside of the rotary shell, though a furnace surrounding the major length of the shell. The shell extends beyond the furnace at both the feed and discharge ends, and is supported on machined supporting rings fixed to the shell. The rings, or tires, are supported on two smaller trunnion

1688

wheels located under each tire. The kiln rotation is achieved usually through a chain drive, since kiln rotation speeds are quite slow, normally in the range of 1 - 6 rpm. At each end of the rotating drum a rotating seal is required to seal the process from the ambient air, and prevent steam and organics from discharging into the plant. The rotary shells must be manufactured from high temperature alloys, containing nickel, which have large linear thermal expansion coefficients. Between ambient and operating temperature of a 10 m long kiln, the change in shell length can be 100 mm. The seals must be able to accommodate this change in length. Normally, the shell expansion is taken up at the discharge end of the kiln, with the dnve being located at the feed end, where the feed end tire position is fixed between two thrust rolls. The shell is allowed to expand only towards the discharge end. The furnace has two (sometimes three) heating zones, each controlled by a separate control loop. Thermocouples or infrared sensors located in each furnace section sense the air or shell temperature, and provide a constant temperature in each furnace section. Carbon bed temperature can be monitored using a thermocouple projecting into shell from the discharge end of the kiln, but due to thermal lag between the furnace and the carbon bed this does not provide a good opportunity for process control. The feed end of the kiln is the “cold” or drylng zone and the discharge end the “hot” or reactivation zone. Furnace setpoint temperatures are higher in the reactivation zone and vary between 100 and 200”C, with the feed end being set at say 550°C and the discharge end set at 700°C. The first half of the kiln acts as a dryer and preheater of the carbon where the steam and low boiling point organics are removed from the carbon. The major heat load is at the feed end, where the water evaporates. The lower setpoint temperature at the feed end allows the process to be more uniformly distributed along the length of the kiln. If the temperature at the feed end is set too high, the feed end may never achieve setpoint, and will be operating at 100% output constantly, while the reactivation zone will be operating at around 15% output. The kiln diameter and length are designed to provide optimum heat iransfer to the carbon, and provide the necessary residence time at reactivation temperatures. Carbon volumetric shell fillage the kiln is in the range of 5-100/, and typical length to diameter ratios are in the order of 7:l to 1O:l.

The steam being produced and organics being liberated in the horizontal kiln are readily removed from the carbon bed, since the dryer end of the kiln has lifters attached to the shell to both increase heat transfer surface area and mix the carbon. The large free volume in the kiln shell, 8590% which is not filled with carbon, allows the steam and organics to be quickly removed without recombining with the carbon. There is some debate as to whether the steam should be removed at the feed end or the discharge end. Although it would seem advantageous to remove steam and organics as soon as possible, at the feed end, a steam blanket is required over the entire carbon bed, and if steam is removed at the feed end, additional steam injection may be required. Furthermore, the problem of low boiling point organics coming in contact with colder parts of the kiln or the cold carbon results in the same problems found in the vertical kilns. Since the temperatures increase along the shell length from the feed to discharge end, the steam becomes superheated and the organic vapors remain at high enough temperatures not to condense in the process, if the gases are allowed to travel the length of the kiln and are removed at the dwharge end. Once the carbon leaves the furnace section, it passes though a short cooling zone. The cooling will reduce the carbon to below a visible color temperature, after which the carbon drops directly into the quench tank water. The steam and organic vapors are drawn from the kiln by an induced draft fan which maintains a slight negative pressure (0.2 - 0.5”WC) in the kiln. Although negative pressures risk ingress of air into the kiln, due to poor seal maintenance, positive pressures risk steam, orgamcs or mercury vapor discharginginto the work area,which is even less desirable.

1689

Horizontal Rotary KildDryer Combination Direct heat rotary dryers have been used for decades to dry many products from sand to pharmaceuticals. The rotary dryer uses hot gases (air) in a cocurrent or a countercurrent configuration to dry materials by making direct contact between the particles and the hot gases. The rotary dryer showers the product across the cross section of the cylindrical drum to produce a curtain of material through which the hot gases pass, resulting in a high thermal drying efficiency. This process is much more efficient than the indirect heating drylng process in the carbon reactivation kiln. A recent introduction of a patented piggy back combination of a rotary predryer and a fossil fuel fired horizontal carbon reactivation kiln (Figure 6) uses the furnace gases from the kiln to dry the carbon in a cocurrent directly heated dryer. The 650 - 700°C furnace gases make contact directly with the cold 40-50?! moisture carbon. The carbon discharges from the dryer at 15-20940 moisture, suflticient to produce the inert atmosphere in the kiln, and is preheated to about 90 “C. Most of the steam and lower boiling point organics are removed in the dryer, and do not find their way into the reactivation kiln. The predryed carbon is fed by gravity to the carbon reactivation kiln below the dryer, and the kiln and predryer atmospheresare separated by a rotary gate valve. The valve allows the carbon to flow through without damaging the carbon, since the gate valve fillage is less than 5%. The dryer is operated at a negative pressure, similar to a krln, however the setpoint is slightly more negative (-0.75”WC) causing any leakage between the two atmospheres to favor movement of gases from the kiln to the dryer. The kiln section is no different from the standard kiln, except that fillage can be as high as 15%, since the heat transfer is reduced. Typically, a standard rotary kiln designed for 100 kg/hr would process 200 kghr carbon if fitted with a rotary predryer. Table 1 compares the thermal efficiencies of the electric kiln,the fossil fuel fired kiln and the kiWdryer combination. The capital cost of the kiWdryer combination is higher W’a standard kiln, and the capitd cost must be weighed against the operating cost of the conventionalkiln. Capital cost recovery for the higher efficiency is typically 6 months on a 125 kg/hr kiln operating 24 Wday.

Process Emissions Common to all emissions will be those products released by the reactivation process, which include steam, organic vapors, carbon fines, and perhaps mercury vapor if found in the ore. In addition, fossil fired kilns will have products of combustion, including C a , NO,, CO, SOz and water vapor. In a standard kiln the process gases and the products of combustion have separate stacks. Although permitting may be required for both stacks, treatment of the products of combustion will not generally be necessary. Treatment may be required for the process gas. An electric kiln will have no products of combustion, and therefore the only emissions will be those produced during reactivation. The W d r y e r combination may require all of the gas to be treated, since it comes directly in contact with the carbon in the dryer. Particulate scrubbing may be required, depending on the hardness of the carbon. Softer mbon, or as previously mentioned, Ntric acid washed carbon, can produce more fines. Particulate scrubbing involves the removal of carbon fines, which can be achieved with a number of common scrubbers. Consideration must be given to the high temperam of the steam, 30O-35O0C, and the fact that the majority of gas flow from the standad kiln is superheated steam and not air. Wet scrubbing will remove many of the organics. Residual mercury on the carbon will be liberated in the carbon reactivation kiln since reactivation temperatures are much higher than the vaporization temperature of the mercury (357°C at sea level). The mercury vapor will condense in the ducting, exhaust fan and stack if not immediately removed once it leaves the kiln. Mercury can be removed by condensing the vapor in a wet scrubber followed by treatment in sulfur or iodine treated carbon columns to bring concentrations to below acceptable levels. Other methods of condensing mercury can be used; however, these alternativestypically requirepre-treatmentto remove particulates.

1690

+ PROMSS

KIU4 FTED

GASES

n

KtLN GAS FURNACE

Figure 6 Fossil Fuel Fired Kiln/Dryer Combination

Table 1. Net heat input required for different kilns, in kW/kg dry carbon and relative thermal efficiencies. Heat input based on 50% carbon feed moisture. Fuel units kW. Net heat is the actual operating requirement, and does not include reserves.

r------

1.46

P@=----

n Efficiency relative to Std. Fossil Fuel Efficiency relative to KiMD ercombo

I

I

Heat Recovered in Efficiency relative to

I

Std. Fossil Fuel Fossil Fuel Kiln/Dryer Horizontal Combo 0.72 1.46 1.46

Electric Horizontal

3.65

1.80

2.19

1.08

1.08 100%

40%

81%

250%

100%

203%

123%

49%

100%

1691

Carbon column design is controlled by two variables; free stream velocity and contact time. Free stream velocity determines the cross section of the columns and contact time determines the column height. The resulting volume determines the weight of carbon needed to fill that space. If carbon volume is selected on the basis of adsorption capacity of the treated carbon only, the columns will be undersized. Typical free stream velocities and contact times range from 100 - 200 d s e c and 8 to 20 seconds, respectively (Calgon Carbon COT.). Presswe drops across the carbon columns must be taken into account in the selection of the exhaust fan.

QUALITY CONTROL For each batch of carbon sent through the carbon reactivation circuit, grab samples should be obtained at each stage. The samples should include loaded carbon, acid washed carbon, stripped carbon and thermally reactivated carbon. These samples will be used for both precious and base metal analysis to determine an approximate material balance as well as a quality control for the reactivation circuit. The standard procedures developed for testing the adsorption characteristics of activated carbon are based upon highly controlled laboratory conditions and the modeling of the results to provide a loading profile that approximates a Freundlich isotherm relationship. In general these procedures are too time consuming to be practically applied to the routine testing of carbon quality and should be utilized if sipficant problems arise with carbon activity. Several companies have developed simple carbon testing procedures that allow the relative carbon activity to be measured and monitored for each batch of carbon. These tests are typically used to measure the carbon activity as a percentage of fresh carbon activity, allowing the relative efficiency of each stage to be gauged. The tests developed are usually based upon contacting activated carbon samples with a stock cyanide solution obtained from the leach circuit. A fixed amount of carbon,typically 10 grams, is added to a one-liter leach solution sample and the carbon slurry is contacted either mechanically or on a bottle roll for one hour. The solution is assayed for precious metals and the gold recovery is expressed as a percentage of the activity of the results from a control test using fresh carbon. This simplified procedure will provide information on the relative carbon activity for each stage of the circuit. It is also important to note that carbon specific gravity will change sigmiicantly during the various stages of the carbon circuit. Although a fixed carbon sample size is desired, it is important to utilize a fixed volume of carbon, thereby eliminating the specific gravity effect. Typically a graphical representation of carbon volume versus fresh carbon weight is developed and the carbon sample size is based upon the initial graphical relationship. In addition to the carbon activity throughout the circuit, this procedure can be used to monitor the quality of all new carbon added to the circuit. The quality of activated carbon is dependent on numerous factors and in certain instances it may be prudent to routinely test the quality of fresh carbon added to the circuit to ensure it performs to the level indicated during the initial evaluation and selection. REFERENCES 1.

Aguayo S., S., Dim G., H. “Fundamentalsof activated carbon”, First English Ed. 2000, Universidad de Sonora, Hermwilo, Sonora, Mexico.

2.

Avraaides, J., Miovski, P. and Van Hooft, P. “Thermal reactivation of carbon used in the recovery of gold from cyanide pulps and solutions”, Research and Development in Extractive Metallurgy - 1987; The Aus. 1.M.M; Adelaide Branch, May 1987.

3.

Calgon Carbon Corporation,Product Bulletin, HGR-LH impregnated activated carbon.

1692

Sem le, P., Equity Silver Mines Scavenging Circuit, paper No. 7, Session B, Proceedings

4.

- 19P Annual Meeting of the CanadianMineral Processors Convention, 1987.

5.

Urbanic, J. E., Jula, RJ., and Faulkner, W.D. “Regeneration of activated carbon used for recovery of gold”, Min. and Metal Process.,2(4) 193-198, Nov. 1985.

1693

Selection and Sizing of Elution and Electrowinning Circuits Paul Hosford’ and John Wells’

ABSTRACT This paper will discuss the different aspects of gold elution and electrowinning. In terms of elution, it will review the two main approaches to elution, namely “AARL,” and “Zadra”, and discuss the main operating variables, such as temperature, time, pressure and column dimensions. Electrowinning will be discussed, with particular emphasis on circuit design and operating parameters. INTRODUCTION The commercial use of carbon in gold recovery plants (CIP and CIL) became widespread from the late 1970s, early 1980s, and within the space of two or three years had become the technology of choice for most of the many new gold projects built during that period. The Merrill Crowe process, based on the precipitation onto zinc of gold and silver from clarified solutions, thereafter became generally restricted to projects with high silver values, or for high tonnage, high dore production heap leach projects. The first generation of carbon plants were usually small tonnage operations, and the quantity of loaded carbon produced was typically 1-2 tonnes per day. Thus the frst generation of carbon elution and electrowinning plants were based upon relatively small equipment sizes. However, these initial plants had to overcome many developmental problems in all aspects of the process. As these problems were resolved, and with the increasing acceptance and confidence in the carbon based processes, the plants became exponentially larger, requiring elution systems that could treat 20-30 tpd of carbon, or more, and electrowinning plants that could treat increasingly larger volumes of pregnant solutions. ELUTION There have generally been two approaches to the elution of gold from carbon, that have stood the test of time from both the commercial and technical points of view. These are generally known as: The Anglo American Research Elution Process The Zadra Elution Process THE ANGLO AMERICAN RESEARCH (AARL) ELUTION PROCESS Development of this elution process commenced in the late 1970s in response to the growing awareness of the capital and operating cost advantages of the carbon process compared to the conventional Menill Crowe process. Anglo American Corp. investigated a process that would “decouple” elution and electrowinning, as compared to the Zadra process, where the two unit operations run in series, in a continuous manner. The AAlU system is based on batch elution, over typically a 6-10 hour period. Initially, the AARL elution cycle incorporated acid washing and elution in a single elution column. However, due to the necessity for special materials of construction, the general practice nowadays is to use separate acid wash and elution vessels. The loaded carbon from CIP or CIL is typically transferred to a loaded carbon pulp dewatering screen, where the carbon (-6 + 16 mesh or similar) is separated from the slurry (or solution in

’ Hatch Associates Ltd., Vancouver, B.C. Canada V6G 1A5 Hatch Associates Ltd., Vancouver, B.C. Canada V6G 1A5

1694

the case of CIC). These screens are usually fitted with molded polyurethane panels, with apertures of 14-18 mesh. The carbon is washed with water sprays and then falls by gravity into the column feed bin, or into the acid wash tank, if separate acid wash and elution vessels are used.. The column feed bin is usually up to 1-2 times the volumetric capacity of the column, allowing surge capacity between absorption and elution. The mild steel bins have typically a circular, vertical segment with a cone, (cone included angle at 90"). It is a simple operation to flush this carbon from the bin into the elution column, which is normally placed directly beneath the bin. A simplified diagram of the AARL Elution System is shown in Figure 1 (Stanley)

Figure 1: Simplified flowsheet of the AARL elution process

In the early days of development of the process, contaminants in the loaded carbon such as slimes, wood fiber and plastic material (particularly prevalent from hard rock underground mines such as those in South Africa) caused severe problems in elution, particularly blocking of the internal screens and nozzles. Elutriation was attempted, with some degree of success, to remove this trash material. With the advent of better trash removal ahead of CIP/CIL using trash screens such as horizontal belt filters and more efficient screening and washing of loaded carbon, this problem has been more or less eliminated. After loading the column, the carbon is usually first acid washed with one bed volume of dilute (3%v/v) hydrochloric acid, followed by 1-2 bed volumes (BV) of water. The temperature in the acid wash process varies from ambient to an elevated temperature of typically 5O-9O0C, (the latter is technically more efficient and is probably cost effective). The temperature of the water rinse raises the system temperature to 110" - 120°C. The spent acid is usually discharged to the absorption circuit. In the original AARL plants, the acid wash and gold elution were done in the same column. This required special column liners and construction materials to handle the range of operating conditions (both acid and alkaline) and

1695

temperatures. The circuit development quite rapidly moved to a two-column approach, one for acid washing and one for elution. The authors are aware of operations that have not applied the acid wash step on a regular basis, prefemng to acid wash every third batch for example. This appears to allow acceptable elution, but is probably only effective for particularly clean ores, and is generally not recommended. When using a separate vessel for acid washing, FRP is usually a satisfactory material of construction although rubber lined mild steel is also appropriate. The acid washed carbon is then pre-heated with !4 - 1 BV of a strong caustic-cyanide solution (typically 2% NaOH, 3% NaCN) at 110°C for thirty minutes. Some recent experience suggests that the solution can have a much lower cyanide content. After this soak the gold and silver is then eluted with 5-6 BV of high quality softened water (less than about 300 g/t sodium), at 120°C and at a flow rate of about 2 bed volumes/hour. Thus, each batch elution can be completed in an 8-hour operating shift, using an automatic control sequence of reagent pumps and valves. The gold is eluted in a high pH solution and is typically recovered by electrowinning as described later.

OPERATING VARIABLES IN THE AARL ELUTION PROCESS Acid Washing Initially it was considered that the acid wash (generally with dilute, 3% hydrochloric acid, but nitric acid is sometimes used) is more effective at an elevated temperature. Cold acid removes calcium and zinc from carbon, while hot acid at about 60°C - 90°C effectively removes calcium, zinc, nickel, and iron. However, common practice today is not to acid wash at elevated temperature. For highly contaminated carbon, the benefits of a hot acid wash in terms of subsequent elution efficiency and carbon activity could be justified. Water Quality for Elution The initial development of the AARL process focused upon elution with water with low ionic strength. This required the installation of water softeningldemineralization equipment. Subsequent work and plant results suggest that following efficient acid washing the elution water quality is not as critical, and good quality regional water may be suitable. Column Geometry and Design The fKst generation of carbon plants was designed for daily carbon production of typically 1-2 tpd, or less. At a carbon density of 0.5 t/m3, this represented a carbon volume of 24m3/day. Handling this small quantity of material was relatively easy, and a high column height to diameter ratio could be maintained without the column height becoming impractical. Early development work had illustrated the higher elution efficiency that could be achieved with a H:D ratio of between 6: 1 and 10:1 and a ratio of about 8: 1 became the desired value. Thus, in these early plants, column dimensions of about 1 metre diameter and heights of 6-8 metres were selected. As the plant sizes increased, and the daily carbon movement increased accordingly, it became more difficult to maintain these column H:D ratios. With a lower H:D ration, the importance of good solution distribution became more critical. The flow in a column can be either: Upflow,or Downflow Upflow is generally preferred as it dilutes (expands) the bed, rather than compacting the bed as would occur in a downflow mode. Solution distribution across the bed is achieved by a series of nozzle caps. These have generally been replaced by the use of tubular wedgewire screens that can be removed, cleaned and replaced without opening up the column. Four (or more) of these screens are placed on a manifold to distribute the incoming eluting solution. The use of the manifold system provides a major advantage in that eluted carbon can be removed from the bottom of the column, as compared to side exit above a nozzle distributorplate. The initial AARL process contemplated a single column for both acid wash and alkaline cyanidecaustic elution. The advantage of this was a simpler layout, and less carbon movement. The disadvantage was the need for more exotic (and high cost) materials of construction, such as Hastelloy, or liners of butyl

1696

or ebonite type rubber. The industry quite rapidly adopted a two column approach. Acid washing is generally carried out in a rubber linedmild steel or FRP vessel, and elution vessels are typically fabricated from mild steel, rubber lined mild steel or stainless steel. In the design of the original plants in South Africa, various methods were considered for moving carbon through the process, from absorption to regeneration. Eductors were the early method of choice. These required fairly precise sizing and were prone to choking. These have normally been replaced by more simple, and easy to operate systems based upon recessed impellor pumps, compressed air or pressurized water. Temperature Initial development concluded that elevated temperature in the 115-125°C range would allow elution to be completed (95% or more precious metal recovery) with as little as 4-6 BV’s of eluate. At temperatures below this and particularly below 100°C the number of BV’s increased rapidly (10-12 BV or more). Most AARL operations have elected to run at elevated temperatures. The columns use pressure rupture discs under these conditions. A few operations used or considered low temperature/ambient pressure elution, particularly in remote areas with a need to use simple technology. However, the high pressure/temperature elution system is essentially simple and robust, and can be controlled using low cost PLC type systems. Temperature is usually achieved and controlled by means of heat exchangers (see Figures 1 and 2). The early generation of plants used low cost plate and frame heat exchangers, which were adequate for the relatively low volumetric flowrates. However, they were susceptible to scaling and choking and more recent, larger plants have preferred shell and tube heat exchangers. Heat to the process can be provided by electrical power or burning gas. Thermopacs (Davidson and Schoernan 1991) can be used to heat a thermic oil ring, which feeds the shell and tube heat-exchangers. ELUATE SOLUTION The initial plants used a soak solution of 1-2% NaCN and 2-3% NaOH. The eluate (EW feed) was therefore a dilute solution of NaCN/NaOH. Recent plant experience has reported lower reagent concentrations yielding acceptable elution results, and in some cases the NaCN has been eliminated. ZADRA ELUTION As stated previously, Zadra elution is different to the AARL system in that the elution and the electrowinning operate simultaneously by continuous circulation of the eluate through the column and the electrowinningcell(s) in series (as shown in Figure 2). The eluate solution is typically 0.2-0.5% of NaCN and 1-2% NaOH. All of the operating variables that apply to AARL are also valid for Zadra elution. A question ofien asked by gold metallurgists is which elution system is best, AARL or Zadra? (along with other common carbon technology questions such as CJP vs. CIL vs. and Carbon vs. Merrill Crowe). Several technical papers have attempted to answer these questions. In the case of Zadra vs. AARL, the general consensus is there is not a great deal to choose between the two and it will frequently come down to operator experience. Zadra has tended to predominate in North America where it was initially developed, whilst AARL quite clearly dominates (but not exclusively) in the Southern Hemisphere.

1697

Figure 2. Simplified flow diagram of the Zadra elution system. Advantages of AARC are its relative speed, and the decoupling from electrowinning may have benefits (although these are somewhat difficult to quantify). It also requires less heat input. Advantages of Zadra are a (perceived) more simple flowsheet, and less need for high quality water. One paper (Costello et a1 1988) claimed that while Zadra had overall economic benefits, the AARL system could treat 2-3 batches of carbon per day in a single column, (as long as sufficient tanks and EW cells are provided). Whichever system is used, it is important to elute efficiently, and to obtain an eluted carbon value of 50 gh Au or less. This carbon is returned to the final absorption stage, and to ensure low soluble losses of gold, the equilibrium conditions in this stage require a carbon with less than 100 g/t Au throughout the tank, (Davidson and Schoeman 1991). Eight operations were reported (Richards and Wells 1987) where loaded carbon was eluted from 5500 g/t Au (average) to 150 g/t Au (average), representing an elution gold recovery (efficiency) of 97.3%. The authors of this paper would suggest that there is little to choose between the two processes. Both are relatively simple, rugged and fairly forgiving of operational problems. Automatic valve sequencing using a PLC can be equally effective for both systems, requiring only minimal operator interface once the

1698

column is loaded. Elution represents a small part of the overall gold plant costs, certainly less than 5%. Hence the impact on overall total project capital (and operating) costs of selecting Zadra or AARL is negligible. ELECTROWINNING Principles and Chemistry Like any other electrowinning process, oxidation reactions taking place at the anode generate electrons, which are consumed at the cathode to deposit the metal. The following electrode reactions take place during electrolysis of an alkaline gold cyanide solution: 0

Au(CN); + e-

Cathode: Anode:

2H20

+

Au + 2 CN' 4Hf+0z+4e-

In cyanide solutions, gold is present as a stable auro-cyanide complex anion with a comparatively high cathodic potential (E,,). This cathodic shift demands higher cell voltage and consequently, other cathodic reactions like the evolution of H2 by discharge of H'and the reduction of O2 can also take place. These additional reactions consume current and reduce the current efficiency of the gold electrowinning process. The equilibrium potentials for various metal cyanides are listed in Table 1 below. The cyanide complexes of Hg, Pb, and Ag are nobler than that of gold, so that these metals will deposit preferentially to gold. The concentration of metal in eluates has an influence on electrowinning performance although the cyanide complex of copper is less noble than gold, it must be appreciated that lo4 moll of dissolved copper is only 6,3 g/t. Every lox increase in concentration causes the equilibrium potential to shift positively by 0,06/e V. The equilibrium potential for 630 g/t dissolved copper will therefore be -0,63V, which is the same value for that of gold at 20 g/t. As a result, copper will also deposit with gold if the concentration is sufficiently high. The remaining ions in Table 1 (i.e. Fe, Ni, and Zn) will not normally COdeposit with gold. However, if their concentrations are extremely high in relation to that of gold, a small amount of iron and possibly nickel may be deposited. Table 1. Equilibrium potentials for the reduction of various metal cyanide ions (metal ion concentration = lo4 moYI; 0,2% NaCN)

Pb(CN):-

+

Pb

-2

-0,38

4dcnr);

-3

Ag

-1

-0,45

Au(CN); Cu(CN):-

+

Au

-1

-0,63

+

cu

-2

-0,75

Fe(CN):-

+ +

Fe

-4

-0,99

Ni

-2

-1,07

-+

Zn

-2

-1,22

Ni(CN):Zn(CN):0 2

2H20

-+ -+

2OHH>+2OH

1699

p H = 13 pH=3

0,45 -0,78

Also listed in Table 1 are the equilibrium potentials at a pH value of 13 for the reduction of dissolved oxygen (0,45 V), and water to hydrogen (-0,78 V). The reduction of dissolved oxygen is the most favourable reaction, and consumes more than 50 percent of the total cathode current under normal conditions. As the reduction of water to hydrogen is not limited by mass transport, the production of hydrogen consumes a considerable proportion of the remaining current. Current efficiencies for the production of gold are therefore very low, and values of 0.5 to 20% are typical. It is important to also note that the evolution of hydrogen and reduction of oxygen at the cathode results in a localized increase of the pH at the cathode surface, whereas the oxidation of water at the anode results in a localized fall in the eluate pH. Types and Features of Electrowinning Cells Because the reactions occurring in any electrowinning process are heterogeneous, involving the exchange of electrons between a solid electrode and ions or molecules dissolved in solution, the rate of any reaction will depend upon the electrode potential, the electrode area, and the rate of mass transport of electroactive species to the surface of the electrode. If the electrode potential is sufficiently negative that all electroactive species undergo reaction as soon as they reach the electrode surface, the overall rate of reaction is determined only by the available electrode area and the hydrodynamic conditions in the electrolyte. (Paul, Filrner, and Nichols 1982). In order to increase the rate of an electrowinning process it is necessary to increase the electrode area, or the mass transport characteristics of the electrolyte, or both. The standard rectangular electrowinning cells, which are well proven in Industry, exploit the very high surface area of steel wool or stainless steel mesh as the active cathode material. Other cell designs based on steel wool cathodes include the cylindrical Zadra and membrane AAIU cells, which have since become obsolete in the Industry. The rotating tubular bed reactor, the impact rod reactor, and the EMEW cell have been developed for the recovery of metals from the rinse waters of electroplating operations, and are in use commercially. It has been suggested that the EMEW cell could also be used for recovery of gold from carbon strip eluates (ElectroMetals 2001). Developments in cell design include features to facilitate operational convenience. The Kemix sludge reactor allows both electrowinning and cathode wash cycles to be performed in an enclosed vessel. Currently, five units are in commercial operation at two mines in South Africa (Proudfoot 2002). The Summit Valley sludging cell allows cathode washing to occur with the cathodes in place. Precious metal sludge is collected in the cell bottom and is then pumped to a filter. ( Weldon 2002). STANDARD ELECTROWINNING CELL Features The typical standard cell consists of a rectangular, stainless steel tank containing typically between 14-33 cathodes and 15-35 anodes, for cells ranging in nominal capacity from lm’ to 3.5m3. Cathodes and anodes are mounted alternately along the length of the cell, so that the cathodes are “sandwiched” between the anodes. The design concepts for this type of cell dates from work camed out by Mintek in South Africa and the USBM in the mid-1970’s to late-l980’s, and later modified and commercialized by a number vendors in the USA, South Africa and Australia. A schematic of a typical standard electrowinning sludging cell is shown in Figure 3. The cells are fitted with stainless steel lids, which are usually connected to a gas extraction fan to remove hydrogen, ammonia, and oxygen gases evolved during the electrowinningprocess.

1700

Operation Originally, most cells were designed and operated to electroplate the precious metals onto the cathodes. The cathodes were periodically removed and either digested in hydrochloric acid to remove most of the steel wool or simply calcined and smelted. The gold deposited onto the cathodes is very fine grained and usually appears almost black in colour. These cells are equipped with stainless steel mesh anodes and cathodes consisting of a polypropylene frame, or basket, packed with steel wool. Cathode loadings of 4 to 5:l metal to steel wool mass are typical but ratios of 18:l can be attained. The cathodes closest to the feed inlet experience the highest solution grades, and load the most rapidly. Typical operating procedure is to remove these highly loaded cathodes for harvesting, advance the following cathodes and replace the fresh cathodes at the discharge end of the cell. The harvesting and maintenance of these cells can take 4 to 24 hours of operation time per week, depending on the gold production rate. Presently, most new operations favour operating cells so as to promote the deposition of gold as sludge rather than plating. The electrodeposited gold forms as fine grains on the cathode surface and are readily dislodged by the co generated hydrogen gas bubbles and the velocity of the eluate solution and accumulates on the bottom of the tank below the cathodes as a black sludge. The bottom of a sludging cell is sloped to a drain point. Periodically, the cell is taken off line and drained to expose the cathodes. The cathodes are washed in place using a pressure washer to dislodge the deposited gold. The gold sludge is typically filtered through a filter press or a sock filter, dried and smelted. These cells generally use punched plate stainless steel anodes and stainless steel mesh cathodes, so called basketless cathodes. The harvesting and maintenance of these cells typically takes 30 minutes to 1 hour of operation time per week, depending on the gold production rate and the extent of metal “bonding” to the cathodes. The main reasons for this change to sludging operation are significantly decreased operator time required to service and maintain the cathodes and reduced operator handling of the cathodes. This minimizes operator exposure to toxic metals if these are present in the strip solution (i.e., Hg, Cd, AS) and reduces the potential security risk. ( Weldon 2002)

1701

Figure 3: Standard Electrowinning Sludging Cell Schematic, Courtesy Summit Valley Equipment and Engineering, SLC, USA

The main variations in operating parameters between these two modes of operation are current density and fluid superficial velocity. Operation in sludging mode generally requires hgher levels of both7as The higher current density promotes a random growth of electrodeposits which loosely adhere to the cathodes and are readily dislodged by the greater evolution of hydrogen gas. The higher fluid superficial velocities in the cell reduce the metal ion depleted zone in the immediate vicinity of the cathodes, and reduces the thickness of the electrical double layer of the cathodes (Weldon 2002). If the superficial velocity is too low, the reduction reaction is limited by the diffusion rate of the metal through the electrical double layer. Typical curves for extraction efficiency (percent) vs. effective cell retention time (Rf Em, minutes) are shown in Figure 4 (Courtesy of SVEE), and demonstrates the general improvement in extraction efficiency attained as the cell design and operation of the cell has been improved. Current design and innovation includes a new cell that is completely enclosed and can be automatically washed down (Weldon 2002), which should provide both operating and security benefits.

ELECTROWINNING CELL PREDICTED EXTRACTION EFFICIENCY 100 C U M U

L A T

90 80 70

I V

E

z E X T R A C T

I

I

I

; / I

60

I

1 1 '

f

50

I

J

I I

I

I

1

TRADITIONAL BASKET CATHODES

I

.

I

/

J

2 '

CURVE @ 1-3 GPM/FT

,-APX.

I

I

40 --------------J

A

I

I /

30

I / /

SUMMIT VALLEY EQUIPMENT & ENGINEERING

I /

20

I / I 1 I

10

I

I

E-CELL EFFICIENCY AS A FUNCTION OF E / W RATE CURVES

I

0 N 0

(Note: 1 gpm/@

= 2.4

mfi)

Figure 4: Courtesy of Summit Valley Equipment & Engineering, SLC, USA

1703

ELECTROWINNING CIRCUIT DESIGN AND OPERATION The process variations to be considered in electrowinning circuit design and operation relate to flowsheet configuration, the design of the cells and the chemistry of the eluate solution. Configuration of StripIElectrowinning Circuit Two main stripping procedures are commonly used as discussed in the previous section of the paper, the Zadra and AARL processes. Each places different constraints upon the electrowinning circuit. In the Zadra process the electrowinning process is integral with the stripping process. The electrowinning cells must be able to cope with an electrolyte whch is very caustic (pH above 13), which contains relatively low concentrations of gold (less than 50 g/t) for most of the elution period, and which is at a temperature of 80 to 90°C. It is desirable that the cell should have a single pass extraction as h g h as possible in order to ensure rapid elution of gold from the loaded carbon. Strip and electrowinning feed flowrates are typically in the order of 2 bed volumes per hour. Stripping using the AARL method requires a separate electrowinning circuit, independent of the strip process. The eluate is circulated continuously through the electrowinning cell(s) until the cell effluent is acceptably low. The concentration C, of an electroactive species at any time, t, after the start of a multi-pass electrowinning operation can be calculated from the equation (-U.E.t) V Where, Co is the initial concentration of species in the reservoir at the start of the operation, ppm U is the eluate feed flowrate to the electrowinning cell, m3/h E is the single-pass extraction, percent t is the electrowinning cycle operating time, hrs V is the eluate volume to be processed, m3. C, = C, exp

In contrast to the Zadra elution procedure, whch requires a high value of E to ensure rapid elution, the electrowinning cells employed for treatment of AARL eluate do not necessarily require a high single pass extraction provided that they can be operated at a h g h circulating flow rate; i.e. it is the product UE in equation (1) which will determine the overall rate of gold recovery. However, in practice, by balancing eluate flowrate and operating time, single pass cell extraction efficiencies in the range 90 to 97% range are readily achievable with standard cells. Sizing of the electrowinning circuit requires consideration of a number of factors, which are discussed below. It is prudent for the design metallurgist to discuss in detail with the selected cell vendor, the duty requirements for each particular unit. 1. Cell hydraulic loading. Typical standard hydraulic loadings for optimal cell extraction efficiency are in the range of 0.1 to 0.35 m3/min/m2 (dmin), with a maximum of 0.4. d m i n . The cells shown in Figure 4 have an approximate submerged cross sectional area. of 0.7 m2 and hence, based on the hydraulic loading discussed above, are suitable for flowrates of up to approximately 14 m3/h per cell. If electrowinning at higher flowrates is required, it is necessary to utilize a number of cells operated in parallel. 2. Cell extraction efficiency. The size of cell required for a particular duty is estimated from the cathode retention time for the type of cell considered. For example, Fig 2 shows that for a basketless sludging standard cell with a retention time of 10 minutes, where the cell hydraulic loading is between 7 - 19 m/h (3-8 gpm/ft2), an extraction efficiency of about 97% can be attained. However, efficiency is influenced by site specific strip solution chemistry, such as concentration of competing metals and pH. 3. Metal electro-deposition rate. Rectifiers are initially sized on the theoretical current requirements for the mass of metals to be electrowon in a particular time period, and a current efficiency estimated from experience of similar or typical strip solution chemistry. Current densities ranging from 20 - 70 A/m2 are typical.

1704

4.

Cell production capacity. Operating experience with SVEE standard cells has indicated that metal loadings in the range of 9000 to 18000 oz/yrlm2of cathode area can generally be attained. The number of cathodes per cell can be estimated initially from the required metal production capacity of the circuit and the current requirement for the system.

Operational Variables Other important operating variables that significantly affect the performance of the electrowinning cells are: Eluate temperature - significantly higher extraction efficiencies are achieved at eluate temperatures in excess of 70"C, likely due to a combination of lower dissolved oxygen content, reduced solution viscosity and increased ionic mobility at the higher temperatures. Previous versions of cells operated at lower temperatures due to problems with warping of the polypropylene cathode baskets, whereas the present all metal cathode cells can operate at temperatures up to 90°C. Solution chemistry - generally, at the elevated pH ranges typical of AARL and Zadra elution processes, solution conductivity is not a problem. It is important to maintain eluate pH in the range of 12-13 to achieve adequate ionic mobility and electrolyte conductivity. However, some operations have experienced problems with silicate dissolution during elution at high pH's, which has lead to subsequent severely reduced extraction efficiency in the electrowinning circuit. The silicates originated as sand entrained in the loaded carbon fed to the strip column. This serves to emphasize the need to consider the elution and electrowinning circuits as an interrelated system. A survey of the system and operating parameters for a number of gold mines worldwide are presented in Figure 5 , to illustrate the range of operating practices in the Industry.

Practical Considerations The electrowinning plant designer has also a number of specific practical issues to consider: Provision of adequate ventilation to minimize the fire and explosion hazard resulting from the generation of hydrogen gas in operating cells. Typically, a metal extraction hood encloses the top of the cell and is ducted to an extraction fan. Materials of construction. Initially, cells and hoods were fabricated from polypropylene or polyurethane, but the practice has ceased resulting from a number of accidents caused by fire and explosion in these cells. Presently, most standard cells are fabricated from stainless steel. Plant layout issues. Consideration has be given to: 9 Height above the cell in order to remove cathodes 9 A hoist above the cell in order to remove loaded cathodes 9 Sludge collection and handling equipment 9 The location of the rectifier, which should be as close as possible to minimize the size of the cables, yet far enough away from the cell for accessibility and safety due to accidental wetting. 9 Personnel safety from accidental exposure to hot eluate pipelines.

1705

e of anode

Strip Process Location

Survey of Current Gold Industry Electrowinning Practice 1 Zadra I Zadra I Zadra

[

S.America

I

S. Africa

I

S. Africa

I I

S. Africa

20

I

24

24

1

Zadra

18

I Zadra I Zadra I Zadra I N.America I N.America I N.America 20 4 -2 3.54 40.9 31.8 102.2 9.4 90.8 91 Stainless

Steel Wool

3.54 15.9 13.6 171.4 27.4 95.0 82 Punch Plate SS wool

2.12 6.8 4.5 308.6 1.7 98.0 82 Punch Plate SS wool

Punch plate & screen

Punch plate & screen

High press

press

steel Stainless Steel

Stainless Steel

Steel

Steel

press

Stainless Steel Punch Plate Pressure wash 85 3.1 1150 0.1 1 6.68

I 15.58

High

[

28.05

REFERENCES 1. Davidson, R.J. and N. Schoeman. June 1991 The Management of Carbon in a High-tonnage CIP Operation. Journal SAIMM, 2. ElectroMetals Technologies Ltd. November 2001 Private communication., Australia.

3. M. Proudfoot. March 2002 Private communication. Kemix Ltd., South Africa. 4. M.C. Costello et al. 1988 Carbon Absorption, Elution, and Electrowinning of Gold Ores with up to 4: 1 Silver to Gold Ratios. Perth Gold.

5. Paul R.R., A.O. Filmer, and M.J. Nichols. 1982 The Recovery of Gold from Concentrated Aurocyanide Electrolytes. Hydometallurgy - Research, Development and Plant Practice. Proc. 3rdInternational Symposium Hydromet, Metal Voc. AIME 6. Richards, R.H. and J.A. Wells. September 1987 Contemporary Practices and Innovative Features of Gold Recovery Installations in Canada, American Mining Congress. 7. Ritson, G. May 1998 World Gold Survey, Technical Report. Hatch Associates. Canada. 8. Stanley, G.G (Edited by). The Extractive Metallurgy of Gold in South Africa, Vol. I. 9. Weldon, T. March 2002 Private communication. Summit Valley Equipment and Engineering Inc. (SVEE), USA.

1708

SELECTION AND SIZING OF COPPER SOLVENT EXTRACTION AND ELECTROWINNING EQUIPMENT AND CIRCUITS Corby G. Anderson’, Mike A. Giralico’, ThomasA. Post, Tim G. Robinson4 and Owen S Tinkle?

ABSTRACT Since the late 1960’s copper solvent extraction (SX) coupled with electrowinning (EW) has been a growing technical application for production of copper metal. The industrial scale copper solvent extraction process is a counter-current multi-stage contacting operation. This usually consists of an extraction and a stripping section involving the use of kerosene based solvents and sulfuric acid to increase the copper concentration and purity of a copper rich solution. Typically, copper recovery from the purified copper solutions is done by electrowinning. This paper will outline this technology and the fundamentals involved in the selection and sizing of copper solvent extraction and electrowinning circuits. INTRODUCTION , Feed solutions to SX circuits typically arise from leaching processes such as heap leaching. The complete solvent extraction chain typically involves two extraction steps (El & E2). Copper is extracted from the pregnant leach stream (PLS) into the organic stream. Then one or two stripping steps (S1 & S2) follow where copper is stripped from the organic phase into a depleted copper solution bearing sulhric acid from the electrowinning tankhouse. Copper is recovered by electrowinning, which is precipitation of a metal by electrolytic reduction on a cathode while using an inert insoluble anode. THE MCCABE-THIELE DIAGRAM The solvent extraction process, as applied to metals extraction, can be defined as: “The selective transfer and concentration of metal ions from an impure aqueous phase via an organic phase to a second pure aqueous phase from which the desired metal can be recovered in a usable form.”

Figure 1 illustrates a typical solvent extraction and electrowinning circuit. Oxime based reagents typically form the basis of copper extraction and they are illustrated in Figure 2. The fimdamental reversible chemical reaction for solvent extraction is noted as equation 1.

~ R +Hcu

++

+ so4-2 <--->R ~ C U+ 2Hc + s04”

(1)

To design a solvent extraction circuit and to predict its performance we require the use of McCabe-Thiele stage construction methodology. This allows us to make predictions on extraction and stripping performance under imperfect conditions. In order to construct a McCabe-Thiele diagram, distribution curves or equilibrium isotherms must first be generated in the laboratory. For the extract isotherm (Fig 3), aliquots of feed liquor are mixed with a measured amount of organic reagent, such that a range of say 6-9 samples is

’

Center for Advanced Mineral and Metallurgical Processing, Montana Tech of The University of Montana, Butte, Montana * LIGHT”, Rochester, Mew York ConsultDrPost, Rochester, New York CTIANCOR, Scottsdale, Arizona Avecia, Phoenix, Arizona

1709

Figure 1. Typical industrial copper solvent extraction and electrowinning circuit.

1710

CSHlS

COHlS

( 8 ) Unloaded

(b) Loaded

Figure 2. Nonyl salicylaldoxime molecules and copper complex. Two salicylaldoxime molecules complex with a Cu* cation and release two H'ions (Biswas and Davenport, 1992). 10 vol% Acorga M5640 TM, PLS 3 g/l Cu, pH 2

s

0.000

1.000 2.000 3.000 Cu in Aqueous phase

Figure 3. Typical extraction isotherm.

1711

4.000

available representing various O/A ratios. The ratios typically cover a range from around 2:l to 1:10.Mixing should be vigorous and for about15 minutes, to ensure that equilibrium is reached in each case before the phases are allowed to separate for analysis. The exact reagent concentration required to match the concentration of copper in the feed solution will depend on factors such as solution pH, circuit configuration (i.e. the number of extraction and stripping stages) and electrolyte composition. For the purpose of a preliminary evaluation, however, the approximate reagent concentration required (expressed as percentage by volume) can be calculated simply by multiplying the concentration of copper in the feed (expressed in g/l) by three. Current commercial reagents load from 0.52 0.58 g/l copper per volume percent (do) reagent and are able to transfer around 60% of this available copper during the stripping process. A strip isotherm (Fig. 4) is generated in a similar fashion by contacting loaded organic phase with various proportions of the strip liquor (electrolyte). In this case 4-5 points are adequate, typically in the range 5: 1 to 1:1. The isotherms, of the types shown in Figures 3 and 4,form the basis of preliminary design for any solvent extraction plant. They are used in conjunction with each other to produce a McCabe Thiele construction, the metallurgical profile of an SX circuit. An example is shown in Figure 5.

-

10 vol% Acorga M5640 TM, Lean Electrolyte 35 gA Cu 180 g/I Sulfuric Acid

8

R

v)

..

1.200

1.600

2.000 2.400 Cu in Organic phase

2.800

Figure 4. Typical strip isotherm. The first principle to consider is that of the operating line. The slope of the operating line defines the organic to aqueous flow ratio required to achieve a desired recovery e.g. if the volumes of organic and aqueous were to be equal, then an increase in organic tenor of 1 gpl copper would be matched by a decrease in the aqueous tenor also of 1 gpl; if say we had twice the volume of organic then it would only need an increase of 0.5 gpl in the organic to effect a decrease of 1 gpl inm the aqueous tenor. In the first case, the operating line has a slope of 1 and in the other it has a slope of 0.5. Therefore, it is obvious that by choosing the O/A ratio, the concentration in the organic layer will be varied for any aqueous feed tenor. The origin of the operating line is important to locate and this is defined by the stripped organic level and the raffinate level, which latter is obtained by trial and error (if done by hand), or iteration (simulation software will be discussed later). In the example of Figure 5, an O/A ratio of 1:l is illustrated so that the slope of the operating line is defined accordingly, passing through the stripped organic value of 2.34 gpl copper and a

1712

raffinate level of that required. In order to predict stage performance, a vertical line is drawn from the aqueous feed tenor to meet the operating line, and then horizontally to meet the equilibrium isotherm. This point on the isotherm represents the gpl copper in the aqueous and organic phases leaving the first stage of extraction.

Extraction section

Stripping section

m

0.00

0.75 1.50 2.25 Copper in Aqueous phase

1.00

3.00

3.04 4.06 Copper in Organic phase

2.02

Figure 5. McCabe-Thiele diagrams for extraction and stripping. In practice, as stated, the reaction will not reach equilibrium exactly and say 90-98 percent efficiency can be assumed for both extract and strip. In Figure 5 , a stage efficiency of 95% was used in the extract section and 95% in the strip section. Graphically this is represented by the diagonal of the rectangle completed about the first stage, relative to the extended diagonal touching the isotherm. When done manually, this naturally is also a matter of trial and error. The example shows that the first (El) stage of extraction would result in a loaded organic of 5.08 gpl copper and an intermediate aqueous raffinate of 1.59 gpl copper. Continuing the stepping-off process to a second extract stage (E2), again allowing for a stage efficiency correction, it can be seen that the aqueous copper tenor drops from 3 gpl in to 0.26 gpl, while the loaded organic rises from 2.34 gpl to 5.08 gpl. Naturally, if this is done by hand, it will be unlikely that the chosen raffinate level will be achieved in an integer number of stages and therefore a trial and error method must be used, changing the raffinate and hence the origin of the operating line. Alternatively, a different O/A ratio could be chosen to change the slope of the operating line. Figure 6 illustrates a typical industrial solvent extraction 2x1 circuit derived from the McCabe Thiele methodology. Clearly this stepping-off procedure, allowance for stage efficiency, and iteration regarding raffinate levels and O/A ratios, can be very tedious to other than a skilled technologist and with this in mind, computer modeling programs have been developed by engineers and suppliers such as Avecia. The benefits and descriptions of a computer model will be described in the following section.

1713

5.08

! !

!

I

f

!

!

A!

!

!

Figure 6. Industrial solvent extraction 2x1 circuit with typical copper tenors noted in brackets ( Biswas and Davenport, 1992).

1714

Avecia Computer Modeling Software The software can operate in design mode or performance mode. In either case the extraction and strip isotherms are needed as input and then the computer will either decide the O/A ratio, number of stages, etc., to achieve a chosen raffinate, or alternatively will estimate the best achievable raffinate for a given set of plant parameters. The End User Module (MEUMTM) enables users to manipulate data as rapidly as binary file data exchange, via the Internet. The software package includes expert system technology and database and plant control capability. Features include: 1. Isotherm construction from laboratory-generated equilibrium data using the curve-fitting facility integral to the software package. Isotherm prediction. 2. Prediction of copper recovery and generation of corresponding mass balance data for a range of plant configurations. 3. The ability to vary the number of extraction and strip stages employed and arrange them in conventional or "parallel/series" mode as required by the design engineer. 4. Automatic recalculation of isotherm data as the reagent concentration is varied. 5. Automatic analysis of plant performance data and calculation of operating stage efficiencies and O/A ratios from metallurgical data supplied by the user. 6. Display of output data in the form of mass balances, circuit mimic diagrams or constructed McCabe-Thiele plots. 7. Local and "Master" database storage and retrieval of isotherms i.e.: flowsheets. 8. Spreadsheet links to allow plant design data to be calculated directly from flowsheet results. SX Testing The metallurgical engineer who has been commissioned to develop a metal recovery process flowsheet using SX technology will typically complete a laboratory study to assess the feasibility of the envisaged treatment route as the first step in the process. This study will identify possible problem areas. The scale on which pilot plant testing should be carried out is dependent on many factors. Some of these are:

1. The quantity of aqueous feed solution available. 2. The data required including whether this data will be required for SX plant design purposes. 3. Funding available. 4. Waste disposal facilities available. It should be noted for a pilot plant trial in which the SX operation is required to be integrated with the leach and EW operations that the scale of these latter operations will determine the size requirements of the SX pilot plant. For laboratory scale pilot plant testing, suitable equipment is readily available from a variety of sources. One common design is the so-called "Bell rig" which consists of mixer-settler equipment fabricated in clear UPVC or Polycarbonate to allow inspection of dispersion bands and solids activity at the O/W interface. "Bell" units typically have small square mixer boxes of 125 150 ml capacity and a settler design which largely resembles that of commercial plants. However, the dispersion distribution system tends to be fairly basic compared to that on a commercial plant. It has been our experience that this scale of equipment is best used along with good quality metering pumps to better control flow rates and prevent "phase flipping" in the mixers. The purpose of carrying out tests on this scale is to increase confidence generally in SX process technology and more specifically in the effectiveness of the flowsheet under development following the original bench tests. The power input to the mixers on this type of equipment is far in excess of that required. For this reason, testwork on this scale will not provide reliable information for design purposes on either the flow rate per unit of settler area (Specific Flow), the entrainment characteristics or the stage efficiency of the system under test. Useful data, however, may be obtained on performance factors such as metal recovery, selectivity or basic problems e.g.: phase separation under dynamic operating conditions

1715

For the production of more accurate data for preliminary plant design purposes a pilot plant capable of handling 10-40 liters per minute of combined aqueous and organic flow is considered advisable. Industry experience has shown that pilot plant scale test work can be used effectively to evaluate reagent performance, confirm the suitability of the envisaged flowsheet and define more closely the physical operating characteristicsto be expected. The features of this size of pilot plant are: It allows collection of operating information on a scale which can be used for plant design. 2. The settler walls can be fabricated from transparent plastic to allow operators to assess visually what is happening in the dispersion band and how solids are being controlled. The structure of the plant is sufficiently strong to withstand all but the very worst weather conditions. 3. Standard pumps, valves and fittings are readily available in sizes suitable for this type of operation. 4. It can be integrated easily into a commercial plant circuit to carry out investigational work on problems arising in existing operations. 1.

Where long term effects are being studied, it is preferable that this type of pilot plant operates continuously, although most of the useful data is likely to be gathered during daylight hours. An important part of the information to be gathered from this scale of operation is entrainment data. Although not strictly representative of the values likely to be obtained on the full-scale plant entrainment values determined will indicate whether the system under evaluation is likely to have major problems in this area. Certainly any improvements observed on this type of pilot plant should translate to the full scale plant. Samples of exiting phases for entrainment measurement should be collected in equipment specially modified for this purpose. These are designed to fit in a laboratory centrifuge and the amount of entrained phase measured directly on a marked scale calibrated to give results expressed in ppm. The quality of the information which can be obtained on this scale with regard to % metal recovery, selectivity, band depths, entrainment and solids distribution against variables such as staging, phase continuity, phase recycle ,temperature, specific settler flow (settler sizing), mixer retention time (mixer sizing) and impeller speed (power input) has been proved to be very reliable. Care should be taken, however, in scaling up from some of this data as solution linear velocities, for instance, on this type of pilot plant tend to be higher than on commercial plants due to differences in settler dimensions. SOLVENT EXTRACTION MIXERS AND SETTLERS While the concept of solvent extraction has existed for centuries, it is the engineering design and application which brought it to prominence in the copper industry. The backbone of this development has been the successful full scale practice of phase mixing and phase disengagement. As illustrated in Figure 7 this is generally accomplished in so called mixer-settlers. The design and operational concepts behind these unit operations is discussed in the following sections of the paper. SOLVENT EXTRACTION MIXING CONCEPTS Since the late 1960s, the Holmes&Narver pumper design has been the standard, serving the SX-market faithhlly. Then the Davy BB, with its characteristic draft tube, made some inroads in the late 1970s. For nearly 3 decades nothing changed until the requirement came for reducing organic entrainment and increasing throughputs. In 1986 L I G H T " introduced the A3 10 hydrofoil as an efficient mixer in the auxiliary mix boxes. Then they came up with a new curved bladed pumper called the R320 in 1995. Since then numerous modificationshave been made to reduce entrainment,decrease power and increase pumping capacities, while maintainingthe head requirement of the system. These new designs greatly

1716

Figure 7. Conventional solvent extraction mixer and mixer - settler (Biswas and Davenport, 1992).

1717

improve the hydraulic efficiency of these open pumpers. Most of the mix boxes now consist of both pumper and hydrofoil combinationswhereas the hydrofoils in the auxiliaries are now up-pumping. The mix boxes aren’t even straight sided anymore. The best designs now call for cylindrical tanks. Figure 8 illustrates a modem mixing system design. These optimized impeller designs reduce aqueous and organic entrainment losses significantly,reduce or eliminate air entrainment and crud formation, enhance mass transfer and maximize the overall copper yield. When compared to the earlier pumper designs, these new designs require less installed power and better process results. The auxiliary hydrofoil impellers further assist the curved bladed pumpers by maintaining a good dispersion and mass transfer, without increasing the head requirement. The designer should select impellers that require minimal installed power while maintaininggood dispersion and mass transfer to minimize entrainment and maximize copper production. Both the pumper and auxiliary impellers can be fabricated in any material including Derakane 470 Vinyl Ester FRP Composite. Currently, these new impeller designs are operating worldwide in over 60 installations with flow rates fiom 50 to 22,000 GPM and developed heads up to 50 inches. Process Requirements of SX Pumpers and Auxiliary Mixers The pumpers are required to provide the desired head and flow through the plant, while creating the initial liquid-liquid dispersion. The auxiliary mixers are intended to maintain this dispersion and extend the stage residence time to improve mass transfer stage efficiency. Once plant design conditions are met, fine-tuning can make a conventional plant run much more efficiently, in terms of operating cost and lower entrainment losses. Pump-Mixer Requirements SX-pumpers must develop a head high enough to achieve the desired flow rate through the SX plant. If the pumpers fail to provide this key requirement, all other features of the SX-plant are meaningless. This flow rate must also be variable, due to the changing concentrations of dissolved copper in the PLS. The best way to vary the flow rate is with variable speed frequency drive controllers, not by adjusting pipe valves, which increase turbulence. It is also desirable that provisions be made during the design phase that would allow future increases in the flow rates, since most mines will investigate increased production rates. The pumpers must also create a liquid-liquid dispersion. Ideally, the droplet distribution is uniform; creating droplets small enough to promote mass transfer but large enough to inhibit entrainment losses. Not only must the pumpers create the dispersion, they must maintain a stable dispersion in the primary stage. Too much shear and turbulence creates a wide distribution of droplet sizes. A large percentage are fines; so small that they end up in the other phase and as entrainment. Too little shear creates unstable large droplets that tend to separate quickly into phases. Phase separation in the mixing stages causes less than optimum copper recovery and limits production capacity. The droplets are created by the bursting mechanism. This means that large droplets or agglomerates of one phase burst into many smaller spherical droplets in localized areas behind the blades where the pressure drop is the greatest. Elongated (stretched) droplets do not exist in copper SX. Excess power dissipation, as described by power per unit volume or P N , is responsible for this bursting mechanism. Depending on the local P N , this bursting of droplets can cascade downward to create very fine and sometimes unrecoverable droplets as entrainment losses. A minimum P N is needed to maintain a stable dispersion. The local P N in the shedding vortices is typically an order of magnitude higher than the overall P N and this inhomogeneity creates the droplet distribution. Inefficient pumpers create very high local P N and high overall PIV, which causes a wide distribution in droplet sizes, fines, haze, and entrainment losses. Ideally, the pumper should not operate so violently as to cause air entrainment through vortices and rough air-liquid surface motion. Air that reaches the pumper blades will also disperse into fme bubbles. They attach themselvesto aqueous droplets and ruin the function of the settlers and increase entrainment

1718

Figure 8. Example Solvent Extraction 3 Mixer Tank Design and Layout.

1719

losses. If silicates are present, air-liquid-solid mixtures create crud, which is another leading cause of entrainment loss. Foam has also created problems with organic carryover to the auxiliary stages. Air entrainment should be avoided at all costs. Comers in mix boxes are notorious for initiating vortexes. This is one reason to move away fiom mix boxes and toward stages made of cylindrical tanks. Also, baffles protruding through the liquid surface form vortices behind the baffles. Carefd design of the baffles is also important. Auxiliary Mixer Requirements Auxiliary mixers need to maintain the dispersion created in the primary stage. The auxiliary box provides more residence time and increases the stage efficiency based on the isotherm of the extractant. Proper placement and operation of the auxiliary impeller can reduce the head requirements for the pumper. These mixers create a mixing flow superimposed on the flow through the system created by the pumper. This additional flow creates a better contacting pattern for the two phases, so that maximum concentration gradients can be achieved. This enhances mass transfer and stage efficiency. To achieve a well-mixed dispersion, a CSTR (continuous stirred tank reactor) model can be achieved if the flow generated by the auxiliary mixer is at least three times the flow created by the pumper. To achieve the best results, the auxiliary mixer should also be equipped with a variable speed device. As the flow rate is varied in the pumper, the developed flow of the auxiliary should be changed, too.

A minimum overall PN must also be obtained in the auxiliary mix stage to avoid phase separation. Radial turbines require similar power levels as the radial pumpers, because they are designed to transform the power into turbulence and shear and not into flow. To get sufficient flow in all areas of an auxiliary mix box, the PN of radial turbines must be relatively high. This causes further decreases in the droplet size population and creates emulsions or phase inversions. Axial turbines concentrate their power into developing flow and not turbulence and shear. Thus, axial turbines require less power to activate all areas of the mix box. Properly designed hydrofoil impellers are similar to airplane wings, which reduce the drag (turbulence) and create more lift (flow). Installed in an up-pumping mode in an SX-plant, they can reduce the minimum overall P N to an order of magnitude less than the pumpers and still achieve 3-13 times the flow rate. Variable speed devices should be installed so the PN can be adjusted, to allow the droplets to coalesce into larger stable sizes. This is ideal, because it preconditions the dispersion before entering the settler. Another reason to install variable speed devices on the auxiliaries is to account for changes in temperature and the so-called viscosity effect of these dispersions at cold temperatures. As the temperature gets colder, the viscosity-effect increases the separation times. Thus, the likelihood of phase separation in the mix stages decreases and less P N is required to maintain that dispersion. The same is valid when discussing the difference between organic continuous and aqueous continuous operations. Typically, a conventional plant is designed for the worst case, i.e., organic continuous operation, when determining settler dimensions and aqueous continuous operation, which is overkill on power demands for the mixers most of the time. With variable speed, the entire unit can be fine-tuned. Like the pumpers, excessive power should be avoided in order to reduce air induction through vortices or surface splashing. The reasons are the same as for the pumpers. Surface splashing can be minimized by careful design of the tanks internals, tank liquid depth to diameter ratio, and proper mixer to tank diameter ratios. Advantages of Conventional SX Mixer-Settler Designs Conventional SX mixer-settler designs have been around since the late 1960s. As commercial trends push towards larger plants and higher production rates, these designs have shown definite weaknesses in the generation of excessive shear and air entrainment at the higher power levels required to meet production targets. This has left some to believe that the only way to improve performance is to design completely new technologies. Now that the requirements for the mixers have been clearly stated, reassessing the value of the conventional design is important.

1720

The greatest advantage of the conventional design is in its simplicity. Most of the piping is easily accessible as well as the mixing stages and settlers. If a pumper needs maintenance, a complete pumper-shaft-gearbox unit can be exchanged with a replacement in less than 20 minutes. The down time is reduced. Splash walls, covers and tank lids do not need to be removed first. An auxiliary mixer can be replaced without any down time at all, if required. A change in performance can be easily detected. Where sight glasses in the newer technologies give a very limited view (the dispersion is essentially black) the mixing stages and settlers of the conventional units are open and extend the view over the entire surface. Changes can be detected visually. Detecting surface organic pockets is very easy, indicating the onset of phase disengagement. The effect of changing the mixer speeds of either the pumpers or auxiliaries on the dispersion and air entrainment can be easily monitored. Thus, if a cooler weather period approaches, the impeller speed of the auxiliary mixers can be lowered to deliver the minimum power required to maintain the dispersion. If the pumper speed is lowered to decrease the overall flow rate, the speed of the auxiliary mixers can be increased, since the increased residence times require more power to maintain the dispersion. The open architecture allows immediate visual response of the changes made. Similarly, the open architecture allows easy access to the fluids being mixed. Samples can be easily taken to detect such things as O:A ratios, fluid continuum, and separation or split times. The open architecture is a great asset for mixers designed to avoid air induction. As flow rates have increased the basic designs of pumpers and auxiliary mixers have required higher tip speeds and excessive power requirements. The additional PN will cause the fluids to look violent. As a result, the surface begins to splash, and vortices appear behind baffles. Placement of lids on the surface diminishes air induction, but the mixers are still imparting excessive power to the volume. The lids have only solved part of the problem, and they have made access to the dispersion and the mixers more difficult. The high PN is still creating a wide variation in droplet sizes, including those very small droplets (haze or fines) responsible for entrainment losses. Finally, simple designs are more economic by definition. The plant design in well known, trusted, and has a much lower capital cost due to the simplicity. Using the high efficiency impellers in a standard design produces the same result as if an alternate design is chosen. As such, the conventional design is an overall good investment. Optimizing the Design of Pumpers and Auxiliary Mixers Conventional pumpers are still in use today. They are called the General Mills or the Holmes & Narver pumper. Attached to the underside of a disk, are six straight blades that extend from the edge of the disk to the center. The inner edge is tapered toward the disk. The height of the blade is always 1/8 of the impeller diameter. The corresponding orifice is without exception between 0.33 and 0.37 times the impeller diameter. In the late 1970s Davy introduced the Davy BB. This impeller is still in use today and has found wide acceptance. It is placed on top of a draft tube, so that it can be placed in the natural media of which is the desired continuum. It has an upper and lower disk with eight curved blades between them. The lower disk is required to achieve head. The height of the blades is always 1/7 of the impeller diameter. The draft tube opening is always % of the impeller diameter. Since the relative dimensions of these pumpers are fixed, increasing the speed (N), increases flow (Q), tip speed (TS=nND), power (P) and power/volume (PN) proportionately. A scale-up criterion was used for these pumpers that was N3D2. It can be shown that, for a given impeller type, N3D2is proportional to PN. The critical value of N3D2is dependent on pumper type, power number and tank dimensions. Tip speed limits were imposed on pumpers based on operating experience without understanding the underlying mechanism of droplet breakup. Today, we know that tip speed is only part of the whole design criterion and PN plays a much bigger role in pumpmixer selection than previously thought.

1721

Characterizing pumper design The primary requirement of a pumper is to develop enough head to achieve the desired flow rate of an SX plant. Each pumper has a set of characteristic head-flow and power-flow curves (Giralico Post, and Preston, 1995). Since a 5050 mixture of PLS and Organic has a mixed SG of approximately 1.O, these curves have been developed using only water. Using Bernoulli’s law, the hydraulic power (PHyd) required to pump a fluid of average density (p) at a given flow rate (Q) over a given head (H) is:

This is the minimum power needed to generate just the head and flow, without any turbulence or shear. Using the power-flow curve, the power (P) needed by the pumper can be calculated. On existing plants, it can be measured. The hydraulic efficiency (E) of a pumper system can be determined by: &=- P H y d

P

(3)

The excess power (P,) is defined as the power above Phyd.

P, is the power responsible for the droplet breakup mechanism, along with the shear gradients emitting from the blade tips. This power results in turbulence. Most of this power is emitted in close proximity of the pumper’s blades, so that once the droplets are formed they usually do not decrease any further in size unless they are recirculated back to within this high shearing zone. Obviously, pumpers with high values of hydraulic efficiencies are directing their power more toward developing head and flow instead of creating very small droplets. LIGHTNM’s research has shown hydraulic efficiencies achievable up to 67% with a normal operating range of 3545% with the new R320 family of impellers. Among the conventional pumpers, the HolmestkNarver and other straight bladed turbines are typically in the 1520% range, while the Davy BB lies between 20-30%. Generally, entrainment losses are indirectly proportional to hydraulic efficiency. Common among all types of pumpers is that hydraulic efficiency decreases as scale decreases. In pilot plants, the range of hydraulic efficiencies is very small, although they compare in the same relative order. Straight bladed pumpers may have efficiencies as low as 5% and the best practical curved bladed pumpers may obtain a maximum of 20%. Much of this is due to poor box geometry design and higher residence times than on full scale. Another reason is that pilot plants do not need to create as much head as their full-scale counterparts. In large enough pilot plants, this can be defeated with novel mixer designs as discussed above. On the other hand, as scale increases, the range of hydraulic efficiencies increases dramatically as stated above. A mere 6% increase in hydraulic efficiency decreased overall organic-in-aqueous and aqueous-in-organic entrainment by one half over a three-month trial period’. Comparison was made at a full-scale copper SX-plant operating two different impeller designs in parallel. P N is not the answer to everything. Shear gradients are also important. Determination of shear gradients (y=&/ET) requires the knowledge of local velocity (v) profiles in the radial (r) direction. A laser operating in a back scattering mode (Giralico, Post, and Preston, 1995). conveniently determines this. Two types of impeller shear gradients are important for scale-up (Weetman and Oldshue 1988) the average shear gradient yavg and the maximum shear gradient ymap Ymax is responsible for the smallest droplets. If a mixer is designed with a uniform exit velocity profile, the droplet distribution will also be uniform. For scale-up, the following rules are typical for

1722

radial turbines (Oldshue 1984).

Ymax=51.N*D and

Ymg= & . N

(5)

and 6 2 are constants, which are impeller dependent. Equation 5 shows that ymaxis proportional to tip speed. has a lower value for long curved bladed impellers than straight bladed impellers’. Upon scale-up, the average and maximum shear gradients obviously diverge. Whereas tip speed is often constant, impeller speed always decreases with scale-up. Thus, the drop size distribution is always wider for larger scale equipment unless both gradients are similar, or if P N is lower in the larger scale. 51

These concepts show that scale-up is not simple, and that entrainment losses determined in pilot plants do not necessarily predict large-scale entrainment losses. Taking E, P N , y and residence times into account is important. Finding higher entrainment losses upon scale-up is not unusual. A conventional system that develops the desired head and flow with curved bladed pumpers having the highest possible hydraulic efficiency and still maintaining phase stability (minimum P N ) without air induction (maximum PiV) at tip speeds below 1000 Wmin results in the best possible SX-system. Fulfilling all these requirements is possible, provided the mix stage geometries are optimized with the pumper selection.

Optimized Pumper Designs No single pumper design is optimized for all operating conditions. Some generalities can be made, though, to help select the best pumper for each SX plant. To increase the flow rates of conventional pumpers, only two variables can be changed, either speed or pumper diameter. This is because the geometrical settings of the pumpers are futed.

-

Q= Nq N * D3

(6)

Upon scale-up N decreases. Therefore higher flow rates are only possible by increasing D (since Nq (the dimensionlessflow number) is fairly constant for geometricallysimilar impellers). When the impeller to tank diameter ratio, DTT, > 0.7 the radial flow becomes throttled resulting in higher than expected power draws and lower hydraulic efficiencies. At the same time, the power increases dramatically,by the fifth power of D.

Since Np, the dimensionless power number, is also relatively constant for a geometrically similar pumper, increasing D to increase flow rates also increases the power. Other methods are available to increase flow. For example, within certain limits, increasing the blade width can increase the flow rate. This increases both Nq and Np, but because the diameter can be decreased, tip speed and P (and P N ) can be decreased. This invariably increases hydraulic efficiency. Curving the blades does not necessarily increase the developed head over a straight bladed pumper, but Np can be halved. The net effect is higher hydraulic efficiencies at the desired head and flow. The blade length of curved bladed impellers affects the developed head and flow, but has relatively little effect on power. Obviously the curvature of the blades is crucial. Constant radius pumpers are easier to build and can operate over a wider range of flow rates. Pumpers with variable radii are more complicated to build but can achieve the very high hydraulic efficiencies in a narrow range of flow rates. A big factor in improving the process can be achieved by modifying the other very important part

of the pump, the orifice. The conventional straight bladed pumper design is typically accompanied by an orifice to pumper diameter ratio (DO/D) of 0.35. By increasing this ratio, more flow and head can be generated without increasing the power draw (provided the pumper is not throttled by design). This happens up to a ratio of approx. 0.5. Further flow increases require larger DO/D’s and higher power but, by increasing the orifice diameter, the systems head-flow

1723

performance increases and hydraulic efficiency stays in the 36-45% range. Dimensionless Numbers Dimensionlessnumbers are essential for the purpose of scale-up. The two more common dimensionless numbers have already been described. Rearranging equations 6 and 7 give:

Nq=- Q N-D3

and

Np=

P p-N3-D5

During small-scale experiments, where D is constant, the major influence on fluid dynamics is the impeller speed, N. By changing N and measuring the flow rate Q, generated by the system, Nq is determined. If the small-scale unit is attached to a torque cell, the measured torque and the speed yield the impeller power. From this, NP is determined. Plotting NP = f(Nq) is convenient (Giralico, Post, and Preston, 1995). This results in a curve characteristicto the pumper design. If the pumper and its corresponding orifice are built geometrically similar, the small-scale curve can be used to design the full-scale system with equations 6' and 7'. Equation 2 shows how the hydraulic power is related to the head, but does not show how to determine it. LIGHTNIN developed a dimensionless head number for this purpose. It is basically the hydraulic power divided by the tip speed squared.

During the small scale testing, the liquid will rise directly with the impeller speed. The difference in liquid level from N=O to N is the head, H. Plotting Nh=f(Nq) is also characteristic of the pumper design. Problems with Over-Efficient Pumpers If a pumper is designed merely based on optimizing hydraulic efficiency, the SX-plant may run into problems. This is particularly the case when retrofitting an existing unit, because all pump stage dimensions are given, and very difficult to change. Therefore, at a given flow rate, the residence time in a pump stage is also given. The minimum power required to maintain a stable dispersion is dependent on pumper type, location, organic solution properties, tank dimensions and mostly on residence time. Generally, as residence time increases, more power is required to hinder coalescence. Once this minimum power level is determined, the hydraulic efficiency cannot go above a maximum level. Otherwise, the dispersion will already begin to separate in the primary stage, leading to reduced mass transfer. As mentioned earlier, efficiencies of E = 67% are achievable. However, this is often too efficient and can result in phase separation. To maximize hydraulic efficiency, the residence times in the pump stages must be lower than one minute, sometimes as low as 20 seconds. The rest of the residence time must be carried out in the auxiliary mix stages. These design considerationsare being carried out on new installations where the SX extractant kinetics is taken full advantage. For existing units, the hydraulic efficiency may need to be artificially reduced so that enough power is available for phase stability. This can be done in several ways; i.e., higher off-bottom clearances, narrower blades, or the addition of radial vanes to the top disk of the impeller. Pumper Design for less than Minimum P N Operations If the existing unit has a high liquid depth to tank diameter ratio (Z/T>0.7) the best retrofit can actually exceed the maximum hydraulic efficiency required to maintain a stable dispersion. Tall tanks can be equipped with one or more A510 auxiliary up-pumping mixers in the upper portion of the primary stage. This patented feature (Post, Howk and Preston 1996) has been installed successfully in several full-scale installations, which had suffered from acute phase separation

1724

Next Page

problems. It can maintain the dispersion because the A510 auxiliary mixer requires about 1/10 of the power of a typical pumper. It can keep the surface active, without affecting the head and flow characteristics of the pumper. Thus, the pumper is designed just to deliver the desired head and flow at the lowest possible power consumption. The upper, built-in, auxiliary impeller maintains the dispersion in the otherwise quiet zones away fiom the pumper’s active zone. Retrofits have been so successful, that current design practice is to make the primary stage taller. Table 1: Some installations with the dual impeller concept.

Optimizing Auxiliary Mixer Design The optimized auxiliary mixer must achieve a high internal flow rate with the least amount of power. 100% efficient impellers would be ideal, but don’t exist. Excess P, is not required, because the droplet distribution has already been formed in the pump stage. Hydrofoiled impellers are the most efficient flow generating impellers available today. Composite impellers like the A6000 are the most efficient hydrofoils since they have true hydrofoil profiles including variable blade cross-section thickness and optimum twist and camber. They can generate the same flow rate as its metal counterpart, the A5 10, at about half the power. The composites are also ideal for solutions high in chlorides.

All other types of auxiliary mixers require too much power to achieve high internal flow rates. In fact some radial devices can cause an otherwise aqueous continuous operation to flip over to organic continuous, even at O/A ratios less than one. Too much power will also further decrease droplet size and increase the settling time. This is not necessary, since more than 90% of the equilibrium was achieved in the primary stage. Because of the low power consumption of hydrofoils, tip vortices are greatly reduced, and instead of further decreasing the droplet size, they can actually achieve some coalescence of the smallest droplets. Pointed in the upward direction, the internal flow generated by the A510 or A6000 goes upward near the shaft and down at the walls. This flow pattern supports the flow coming fiom the pump stage through the underflow weir or downcomer, reducing the head requirements of the pumper. The hydrofoil cannot be too small, because the power of the thrust is then focused at the center of the surface. The resulting flow pattern would look like an upward surging spring, which could induce unwanted air. The concentrated local power means that other regions may be underpowered with resulting organic puddles. The hydrofoil cannot be too large either. If the impeller diameter is too large, there will be no room for the downward flow near the walls. The spreading of the upward flow collides with the downward recirculation resulting in a thrashing and violent surface motion with unwanted air induction. This collision of the flows causes an increase in power demand and thus unnecessary turbulence. This chaotic flow pattern is also the result of any sized pitched bladed turbine, because the discharge angle of the flow is not axial like a hydrofoil, but off to an angle. It is not possible to get a top-to-bottom flow pattern with a pitched bladed impeller.

1725

BULLION PRODUCTION AND REFINING Charles 0. Galel and Todd A. Weldon2

ABSTRACT Bullion production and refining for the present day precious metal mine is generally limited to the melt refining of gravity separated concentrates, electrowinning sludges, and the product from zinc precipitation processing. Guidelines, given as basic rules-of-thumb, are presented for equipment selection and sizing. Some industrial compositions of electrowinning sludges and zinc precipitates are given, as are typical flux compositions used for melt refining these precious metal concentrates. The art of ultra-refining dore' is not included in this paper. Retorting mercury, cadmium, and selenium is discussed. EXTRACTION OVERVIEW Gold and silver extraction and concentration from mining ores and/or process solutions has been traditionally accomplished through mercury amalgamation, gravity concentration and separation, and sodium cyanide leaching coupled with carbon adsorption or zinc precipitation, independently or in combination. The precious metals recovered from these initial concentrating processes are upgraded by refining. For this paper, the term refining refers to both hydrometallurgical and pyrometallurgicalprocessing. For a brief review of the extraction and Concentrating methods: Mercury amalgamation is onc of the oldest, yet least used today, methods to recover fine particle size free gold and silver by alloying it with mercury. Gravity concentration and separation is also one the oldest methods of free gold and silver recovery. Gravity concentration recovers fine to nugget size particles. Sodium cyanide leaching of gold and silver is relatively new as an extraction medium, and leaching has allowed microscopic particles of gold and silver to be recovered. Sodium cyanide leaching is often coupled with carbon adsorption and electrowinning. The carbon is contacted with precious metal laden leach solutions where the metals load onto the carbon. The carbon is then stripped with a hot, sodium cyanide solution, which becomes highly concentrated with the precious metals. The precious metals are then plated onto the cathodes, either loosely or intimately, of electrolytic cells and recovered through downstream processing. Sodium cyanide leaching is less often coupled with carbon adsorption and zinc precipitation. The strip solution as in (4) above is contacted with fine particle size zinc dust wherein the precious metals are precipitated and collected in a filter. Sodium cyanide leaching may be coupled with direct electrowinning if the precious metal tenor is sufficiently high and deleterious impurities are not present. Sodium cyanide leaching is often coupled directly with zinc precipitation when the solution tenor is sufficiently high. Zinc precipitation as a concentration method dominates in the silver mining industry.

' Summit Valley Equipment & Engineering, Inc., West Bountiful, Utah, USA. Summit Valley Equipment & Engineering, Inc., West Bountiful, Utah, USA.

1747

REFINING In addition to electrowinning sludge and zinc precipitation concentrates, precious metal dusts, bag house and scrubber residues, floor sweepings, melt crucible residues, and furnace liners are also upgraded by refining. Refining is usually performed in two stages (Marsden and House 1993): (1) Treatment at the point of production to produce crude bullion that is easily handled and accurately sampled. (2) Refining of crude bullion to produce high purity gold and silver for market, Since the metallurgical practice of high purity refining of gold and silver is treated in detail in many fine publications, this paper primarily focuses on refining the precious metal concentrates to the crude bullion state. The authors’ guidance for equipment selection and sizing to produce crude bullion is given.

REFINERY The gold mine “Refinery” frequently contains ore or solution treating equipment for the production of the first precious metal concentrate, equipment to intermediately process the concentrate, and equipment for processing the concentrate into crude bullion. Thus, a refinery can house electrolytic cells, zinc precipitate filters, gravity concentration and separation tables, drying ovens, retorts, flux mixing equipment, melting furnaces and molds. The refinery layout varies greatly with the type of concentrating equipment selected. For electrowinning, corresponding valves and manifolds to the cell are located in the refinery area. For zinc precipitation, corresponding valves and manifolds are commonly located outside the refinery area to facilitate the changing of filter presses without entering the secure refinery. When a precipitation filter press is full, the flow stream is redirected to an empty filter press from outside the refinery area. Typical floor plans for a si!ver/gold mine with a zinc precipitation circuit and a gold mine with a carbodelectrowinning circuit are illustrated on the following page. ELECTROLYTIC RECOVERY In today’s gold extraction plants, electrowon metals are often produced as a loose cathode deposit or sludge. The sludge is washed from the cathode or may simply fall from the cathode to the bottom of the electrowinning cell. This sludge is flushed from the cell and pump into suitable filters where it is collected for further processing. The sludge is comparatively low in volume and can be relatively pure, 70 to 90 percent precious metals. Impurities often encountered are the base metals, mercury, cadmium, selenium, A1,0,, SO,, MgO, and CaO. Once filtered, a typical bulk density of the sludge is 1.5 grams/cm3 (drybasis) (Hall, 2001). By judiciously controlling the anode to cathode voltage potential at the cell, solution impurities such as copper can be minimized in the sludge product. The electrowinning cell can be considered as the first refining step. For electrowinning sludge filter sizing, the authors suggest using a conservative filtered sludge density estimate of 1,100 kg/m3 (dry). Note that downstream processing equipment is volumetrically controlled by the filtered sludge volume. ZINC PRECIPITATION Large silver producing mines, and gold mines with a silver to gold ratio of +5:1 or greater, utilize the zinc precipitation process. By current standards, a clarified and deaerated pregnant solution should have less than 5 ppm total suspended solids and a dissolved oxygen concentration of less than 0.5 ppm. Zinc powder is added in relation to precious metal content and oxygen and impurity concentrations. At the zinc addition point, diatomaceous earth is often added to the process stream and is used as a filtering aid. The filter aid provides body to the precipitate and decreases the rate of filter press blinding for some process solutions.

1748

Figure 2 Typical gold mine carbonlelectrowinningcircuit refinery

1749

Zinc precipitate quality can vary from 30 to near 85 percent precious metals. Start-ups and new operations often have low precious metal concentration precipitates due to process upsets and unrefined process control. Suspended solids not removed in the clarification filters, excess zinc (often added due to high oxygen concentrations), precipitating impurities, and body feed all lower the precipitate grade. By economy, the precious metal production from silver operations must be greater than gold operations. The silver metal concentrate is by production more voluminous as compared to gold. Silver refining equipment is necessarily larger in volumetric capacity versus gold plants. Precipitate handling equipment must be sized conservatively. The authors use a precipitate bulk density of 1,100 kg/m3(dry). This density is often considered too conservative, but has actually been measured at more than one location. Precipitate moisture after filtering will be 30 to 40 percent. When sizing drying ovens, the wet weight of the precipitate must be considered since most oven racks are weight limited. For silver mining zinc precipitates, retorts are larger and are generally designed to be loaded with fork trucks or mechanized pan-handling lifts. A large mercury retorting operation in a silver refinery can handle more than seven cubic meters of zinc precipitate per day varying in mercury concentration from one to twenty percent (dry basis).

MERCURY, CADMIUM AND SELENIUM REMOVAL, Retorting is universally accepted as the refining step for the removal of mercury, and may be used for cadmium and selenium. Vapor pressure curves for Hg, Zn, ZnO, Cd, CdO, Se, SeO, Te, and Sb are shown on the following pages. The vapor pressure unit of measurement for the curves is Torr, or mm Hg absolute pressure. Distillation or sublimation occurs when the retort atmosphere is of lower absolute pressure than the specie vapor pressure at a given temperature. Mercury The fundamental physics of mercury retorting have been known for a long time, but due to ever more stringent plant hygiene requirements, the equipment has been driven from the simple to the robust. To retort mercury, electrowinning sludge or precipitate is heated to 425"C/550°C in a vacuum environment ranging from 50 mm vacuum (50 mm below atmospheric pressure of 760 mm at sea level) to the low absolute pressures of 150 ton (150 mm pressure absolute). Tine mercury is removed from the sludge and/or precipitate as a vapor, transported from the heating chamber, and is condensed in water cooled heat exchangers. Mercury boils very easily (70.65 KcaVKg) (127.4Btu/lb), but also condenses very easily. Mercury will often condense in some very inconvenient places within the retort equipment. Care must be taken to maintain retort system surface temperatures in contact with mercury vapor above the mercury dew point. Mercury vapor is very heavy and can be difficult to transport from the retort chamber to the condenser. A controlled sweep gas is often used to assist in transporting the mercury to the condenser. The sweep gas reduces the partial pressure of mercury vapor, and thus, aids the distillation process. With a sweep gas, the condenser efficiency is reduced and must be sized accordingly. Equipment designed and operated at low absolute pressures (150 to 300 torr) is generally more successful in meeting the process requirements in a hygienically acceptable manner. To operate at the low absolute pressures, the equipment must be maintained nearly leak free. In these systems, the mercury vapor has less opportunity to condense outside of the condenser. In low pressure systems, fewer non-condensables will enter the condenser and the theoretical condenser efficiency will be higher. Mercury and water are the major condensable vapors. Air is the major non-condensable.

1750

Mercury Vapor Pressure loo00

lo00

100

10

1

Figure 3 Mercury Vapor Curve (Data from Weast 1983-1984)

1751

Zinc, Cadmium, Cadmium Wde, Selenium& Selenium Dioxide VaporPresSure

I

I

I

I

I

I

I

I

lo00

m I

E E

f3 v) v)

100

2

P

bp.

E

10

1

0

200

400

600

800

lo00

1200

Temperature, degrees C Figure 4 Vapor Pressure Curves (Data from Wenst 1983-1984)

1752

1400

1600

1800

Zinc Oxide Vapor Pressure

Figure 5 Zinc Oxide Vapor Pressure (Data from Weast 1983-1984)

1753

loo0

100

10

1

Figure 6 Vapor Pressure Curves (Data from Weast 1983-1984)

1754

Mercury containing sludges or precipitates should not be heated rapidly in a vacuum. Rapid and/or violent boiling of water contained in the sludge or precipitate will cause precious metal dusts to transport out of the retort towards the condensing system. A drying step, slow and controlled heat-up, is generally incorporated in mercury retorting systems to prevent excessive dusting. The sludges or precipitates must be placed in pans of 100 rnm to 150 mm maximum depth for retorting to be successful. Cadmium Cadmium metal removal via distillation i s thermodynamically possible, as is cadmium oxide distillation. Equipment constraints restrict the distillation of cadmium to the metal specie only. The cadmium oxide vapor pressure is too low to complete an economically viable distillation. During the distillation of cadmium metal, cadmium oxidation must be prevented. Cadmium distillation equipment has been placed into operation, but no operating data is available to the authors. The equipment is designed to heat the product to 700°C under a vacuum of 5 mm in a nitrogen atmosphere. At 535"C,cadmium will begin to distill. The cadmium is collected as a sublimate, which is manually removed from the system. Selenium Selenium can be successfully removed via distillation as documented by FMC Paradise Peak (Armburst 1989). Selenium is more cormonly associated with zinc precipitated high silver products. Selenium must first be oxidized to selenium dioxide in order to distill. The process kinetics are slow. Lengthy holding periods at temperature are needed to allow oxygen to diffuse and react, and the resulting oxide gas to migrate to the condensers (sublimer).

MELT REFINING The impurities commonly removed during melt refining are iron, zinc, calcium, magnesium, sodium, alumina, and silica by reaction with the flux and/or atmosphere. Mercury, selenium, zinc, and cadmium may volatilize during melting and escape into the furnace exhaust system. The exhaust system and scrubbing equipment must account for the collection of these metal fumes. Silver and gold also may volatize during malting, but if the melt temperature is maintained below 125OoC,the volatilization rate is generally low for gold. Silver has a higher vapor pressure than gold and its loss can be greater. Precious metal volatilization losses increase as the volatile impurity concentrations increase, notably tellurium, antimony, and mercury (Marsden 1993). Table 1 contains selected refinery feedstock from electrowinning and zinc precipitation processing facilities. Melt Refining Electrowinning Sludge If mercury, cadmium, and/or selenium are not present in the electrowinning sludge, the retorting step is eliminated. A process drying cycle of the sludge may not be required if the sludge cake has a low percent moisture. At one facility, fi!tered electrowinning gold sludge at 18 to 20 percent moisture is sometimes charged to the melting furnace without drying (Hall 2001). Flux and gold sludge are mixed (frequently by hand) and charged to an empty melting furnace. Fluxes for processing electrowinning cell sludge are generally kept to a minimum. Flux composition can be as simple as 100% borax (Na,B,O,), which supplies a cover during melting, to the more common mixtures of borax, niter (NaNO,), silica (SiO,), fluorspar (CaF,), and sometimes soda ash (Na,CO,). Generally, iiiter is used if base metals are to be oxidized. If alkaline and alkaline earth metals are the only impurities, niter is often eliminated. As a rule of thumb, a designer should allow for a flux volume equal to the electrolytic sludge volume being melted. If soda ash is used, a boiling slag often results during melting, thus sufficient crucible volume must be available to contain this boil. Table 2 contains some selected flux compositions from industry for electrowinning sludges.

1755

Table 1 Selected refinery feedstock from electrowinning and zinc precipitation processing (Kahl2002; McMillen 2001; Hall 2002; Haldane 2002; Frentress 2002, Anonymous)

Table 2 Selected flux compositions from industry (Kahl 2002; McMillen 2001; Hall 2002; ildane 2002; Frentress 2002. Anonvmous) Zinc Precipitate I Electrowinning Constituent Niter 48%

85%

Silica

17%

10%

Fluorspar

14%

Borax Soda Ash

Basis for Total

0.24~ Base Metal Wt

sludge weight

weight

Substantial base metals, Sludgehlag is 2:l

40% of sludge

Furnace Refining Zinc Precipitate After drying andor retorting, the precipitate is mixed with fluxes and melted. The fluxes generally contain substantial borax for fluidity, enough niter to oxidize all of the zinc and any other base metals, fluorspar (CaF,) for fluidity, and possibly soda ash to flux any excess silica and to lower the slag melting point. Table 2 contains some selected flux compositions from the industry.

1756

Flux mixing in large volume silver plants is generally mechanized. The mixing equipment varies from .75 m3mortar mixers to 4 m3 cement mixers to specialized “paddle mixers” of 3 to 6 m3 capacity. The mixers are sized to match the batch-melting requirement. As a rule of thumb, when sizing zinc precipitate mixing and melting equipment, the flux charge volume should range from 2 to 3 times the precipitate volume. Furnaces may be charged, melted and poured in one batch. In some operations the furnaces are alternately charged and melted, charged and melted, until sufficiently full of refined metal and slag to pour. To trap slag, cascade-casting arrangements are often used. Simple carousels with bar molds are also used, as are cone shaped slag pots.

Acid Digestion Acid digestion of zinc precipitates can be advantageous to remove excess zinc and limey precipitates prior to melting. The amount of slag will be reduced which reduces the cost of flux, goldsilver losses in slag, and the energy required to melt. Sulfuric acid is generally used for precipitate digestion. The digestion will produce small amounts of hazardous hydrogen cyanide (HCN), and voluminous amounts of flammable hydrogen (H2). If arsenic or antimony are present, arsine (ASH,) and stibine (SbH,) can be produced and are highly toxic. At times, nitric acid is added to reduce the production of arsine gas (Marsden 1993). To assure effective zinc removal, the acid digestion is finished with about one percent excess free acid concentration. The ventilation system must be efficiently designed to capture and dilute all the evolved gases. The acid digested precipitate must be filtered and washed, then dried and calcined or retorted prior to melting. Residual sulphates from acid washing promote equipment corrosion during retorting, and the residual sulphates can produce a “matte” during melting. SLAG TREATMENT In many operations, slag is ground and returned to the leach circuit. Some operations recycle small amounts of slag into the next melt, some sell the slag to a slag processor, and very few seriously process the slag in the refinery to recover precious metal fines. On-site slag processing can consist of crushing, grinding and screening (the oversize is re-melted), followed by gravity separation. FURNACE SELECTION Melt furnaces can vary extensively, as nored in the paper “Precious Metal Refinery Process Selection - An Overview” by G. Warren (Warren 2001). The choice between gas-fired, oil-fired or induction heated furnaces is made on the basis of fuel source economics, productivity, environmental considerations and refmery hygiene. In general, when highly concentrated gold sludge is being processed, crucible style melt furnaces are preferred. Fuel fired crucible furnaces are considerably less expensive than induction heated furnaces and the power consumption of induction furnaces can be quite high. The refinery will be cooler and require less ventilation with induction furnaces and the off-gas system can be much smaller. Induction furnaces can melt faster and crucible life should be longer than fuel fired furnaces. Induction stirring also assists metal to slag contact and reactions. In large silver refmeries, the fuel-fired, refractory lined reverberatory furnace appears to be the most common type of melting furnace. A few plants have large fuel fired crucible furnaces. The melting area ventilation systems are sized to handle the combustion exhaust at maximum firing rates. VENTILATION Industrial hygiene is very important in precious metal refineries. Heavy metals fumes and toxic gases can be produced during the refining processes. Mercury fumes are likely the most common problem, but other heavy metal fumes and dusts such as zinc, cadmium, silver, lead and selenium can be present.

1757

Extraction products such as zinc precipitates and electrowinning cell sludges that have any measurable mercury should be stored in covered containers and processed in encapsulated or well ventilated equipment. Slag mixing equipment should be ventilated. Melt furnaces may be partially enclosed and hooded. Retort systems should be essentially leak free with no vapor losses to the environment. Acid digestion equipment must be hooded and/or encapsulated.

Electrowinning Cells Electrowinning Cells must be ventilated to prevent hazardous vapors and byproduct gases from entering the refinery. Hydrogen is produced at the cathode and ammonia is produced by the oxidation of cyanide near anode. Both gases can react explosively under the right conditions and should never be allowed to collect in unventilated equipment or work areas. Mercury vapor is found in some electrolytic cell vent gases. To prevent hazardous gas and vapor buildup, a cell ventilation system should sweep fresh air uniformly across the entire electrolytic compartment. Preferably fresh air should ingress from the operating service-side and exhaust from the opposite side. The cell exhaust should be routed to a suitable scrubbing system. Ventilation flow rates as high as 1.7 m 3 h (1 cfm) per designed buss ampere are used. A 2,000 ampere cell can have a ventilation flow of 3,400 m 3 h (2,000 cfm). Retorts Mercury retorts should have exhaust hoods located near the charge doors. The hoods must be designed to capture heavy mercury vapors that fall as they exit the retort door, i.e. the hood is more effective at the bottom of the door on a mercury retort. If retort doors leak, mercury droplets will collect on the door near the leak as a tell tale sign. The hoods should be designed to capture and contain mercury droplets. Refinery Ventilation Systems Refinery ventilation systems should be designed to affect a floor ventilation sweep. Ideally, supply air should sweep down the walls on one side of the refinery, travel across the room at floor level to the other side of the refinery, and exit via a duct or curtain wall on or near the floor. Low lying areas, such a recesses or pits, should be adequately exhausted by adding ventilation system pickup ducts is these areas. Metal fumes are very dense, and transport with difficulty.

SUMMARY In the mining industry, the precious metal concentrating processes used throughout the world are gravity separation, electrowinning, and zinc precipitation. Products from these unit processes are often upgraded via retorting and acid digestion, albeit careful control of the metal concentrating processes can limit contamination. Oxidizing base metal contaminants in the concentrates via calcining is not commonly practiced and wiis not discussed. Melt refining is used as the final dore' bullion production unit process at most mines. Ultra-refining of bullion is usually completed by others after the bullion has been sold. Concentrate processing equipment should be sued conservatively to allow for new project startup process complications. To safeguard worker hygiene, careful attention must be paid to the design and operation of ventilation systems. REFERENCES Anonymous. Private Communication, Coeur Rochester. Armburst, W., D. Jensen, C.H. Sheerin, and R.A. Smith. 1989. Removal of Selenium From the Dore ' at FMC Gold Company's Paradise Peak Mine. Presented at the 2ndAnnual Elko Operators Convention. Frentress, R. 2002. Private Communication. Round Mountain Gold. Haldane, T. 2002. Private Communication. Glamis Gold Ltd. Hall, D. 2001. Private Communication.Placer Dome: Musselwhite Mine.

1758

Kahl, T. 2002, Private Communication. Barrick Goldstrike Mines. Marsden, J., I. House. 1993. The Chemistry of Gold Extraction. New York: Ellis Honvood Limited. McMillen, G. 2001. Private Communication. Florida Canyon Mining Co. Warren, G., 2001. Precious Metal Refinery Process Selection - An Overview. Proceedings-33rd Annual Meeting of the Canadian Mineral Processors. 69-80. Weast, R.C., Editor-in-Chief. 1983-1984. The Handbook of Chemistry and Physics. Florida: CRC Press, Inc.

1759

PLATINUM GROUP METAL BULLION PRODUCTION AND REFINING Corby G. Anderson’, Lance C. Newman=, and Greg K. Roset

ABSTRACT In recent years, the production and refining of platinum group metals (i.e. PGM’s) has become increasingly important. The variety of processes used in the recovery and refining of platinum group metals are classified into two major process categories. These process categories are based on the differences in the raw material sources: one, on the recovery and refining of precious metals from the concentrates obtained from mined platinum bearing copper and nickel sulfide ores, and the other from secondary raw material sources such as recycled industrial products. Examples of the latter include spent catalysts, electronic scrap, spent electrolytes, and jewelry scrap. This paper focuses on the primary production and refining of platinum group metals. However, the basic principles utilized are analogous for the recovery and refining of secondary and recycled PGM’s. INTRODUCTION There are two primary sources for the platinum group metals: those found in deposits in such as South Africa’s Bushveld complex, at North American Palladium in Canada or at Stillwater Mining in the USA; and those derived as a by-product of primary copper andlor nickel electrorefining as from anode slimes such as at INCO in Canada or in the former Soviet Union near Noril’sk and other sites (Flett, 1984, Hunt and Lever, 1969, Kirk-Othmer, 1968, Dayton and Burger, 1981, Dayton and Burger 1982, Loebenstein, 1983, Minty, 1999). Alluvial deposits of native platinum or alloys such as osmiridium are so scarce that they are not included. The composition and grade of raw materials differ, and depend on producing mine locations from which they are obtained, as do the methods of pretreatment and of preprocessing. Platinum, Pt, and palladium, Pd, are considered the major, primary platinum group metals. Rhodium, Rh, iridium, Ir, ruthenium, Ru, and osmium, Os, are considered the minor, secondary platinum group metals. The following annotated material excerpts (Benner, 1991) summarize primary PGM recovery and refining. PGM’S FOUND WITH NICKEL AND COPPER SULFIDE CONCENTRATES Platinum group metals can concentrate in sulfide matte that is produced during the smelting of nickel and copper ore concentrates. Historically, in the INCO Orford process, sodium sulfide is added to the copper-nickel sulfide matte, thereby forming two separate copper and nickel matte phases. The respective metals are recovered from those mattes by a converting process that yields a crude metal product suitable for electrorefining. The anode slimes obtained by electrorefining crude nickel contains the major portion of the platinum group metals. The slime concentrates are air roasted. Copper and nickel are leached out with sulfuric acid, to yield PGM-enriched residues from which the platinum group metals are further concentrated by melting them with charcoal and litharge to form a lead button. The resulting lead alloy is nitric acid leached. The acid insoluble residue becomes a raw material for further refining (Hougan and Zacharisen, 1975). The

Center for Advanced Mineral and Metallurgical Processing, Montana Tech of The University of Montana, Butte, Montana Stillwater Mining Company, Columbus, Montana Stillwater Mining Company, Columbus, Montana

1760

pressurized carbonyl process practiced by INCO Canada strips nickel as gaseous nickel carbonyl, and leaves residues that contain platinum group metals (Head, et al, 1976)

PGM’sfrom the J-M Reef at Stillwater Mining in the United States of America The platinum, palladium and other platinum group metals are associated with nickel-copper-iron sulfides in the J-M Reef within the Stillwater complex. Stillwater is the only primary source of PGM’s in the United States (Sharratt, 1994). The ore is mined and milled to produce a bulk nickel-copper sulfide concentrate. The concentrate is transported about 65 km to the Stillwater smelter, located in the town of Columbus, Montana, where it is smelted to produce a nickelcopper matte (Hodges, Roset, Matousek and Marcantonio, 1991). At the start-up of the smelter in 1990, the nickel-copper matte, containing approximately 42% Ni, 27% Cu, 22.5% S and 2.1% PGM, was shipped to a custom refiner in Europe. Later, Stillwater, in conjunction with Sherritt, developed process flowsheet for refining of the Stillwater matte to produce a high-grade PGM bullion concentrate, Sherritt’s two-stage acid pressure leaching technology (Weir, Kerfoot and Kofluk, 1986). As this is a fundamental example of primary PGM production, pertinent process details follow and are illustrated in Figures 1-4. At this time, Stillwater operates two PGM ore milling facilities. One is at the East Boulder mine while the other is at the mine at Nye. The operation of each is similar and the older Nye facility will be focused on as an operating example. The Stillwater Nye concentrator has been expanded from 450 to 2700 tonnes per day of ore capacity since it’s startup in 1987 (Turk, 2000). The feed material comes fiom a jaw crusher system designed to crush the mined ore to minus 150 mm. Further comminution is performed via a closed circuit SAG mill, pebble mill and ball milling circuit. This produces a cyclone sized ore product of p80 145 microns that reports to rougher flotation and flash flotation based on size. The Nye concentrator uses four major flotation reagents. These are potassium amyl xanthate promoter, Cytec 3477 promoter, CMC for talc suppression and sulfuric acid for pH adjustment. In essence the operating philosophy is focused on maximum recovery of all sulfides that bear the PGM’s. The rougher concentrate is subjected to two stages of cleaning followed by final column cell flotation. This produces a bulk copper and nickel sulfide concentrate containing the PGM’s. The tails from the rougher circuit are reground and subjected to middling and scavenger flotation. The final mill tails report to the mine paste backfill plant or the tailings pond. The Nye concentrator flowsheet is shown in Figure 1. Concentrate is transferred in bins from the East Boulder and Nye mills to the smelter as a filter cake. The concentrate is approximately 10% moisture and contains 0.2% of platinum and palladium per tonne in an approximate ratio of 3 to 1 Pd to Pt. The bins of filter cake are sampled, dried, and pneumatically conveyed to the concentrate bin for smelting in the electric furnace (Bushman and Roset, 2002, Hodges, Roset, Matousek and Marcantonio, 1991). Recycled PGM bearing automotive catalysts are also fed through this pneumatic system to the electric furnace. The electric furnace feed consists of dry concentrate, limestone, and dust and is fed through an airslide. Converter slag and crushed reverts are fed through two bins located above the furnace. The electric furnace is rectangular with 3 in-line prebaked electrodes. The furnace lining consists of magnesiachrome refractory bricks, with copper coolers installed in the slag zone. The furnace produces two products, slag and matte. Furnace slag contains oxide materials (mostly Si02 and FeO) and is removed several times a day into a pit for air-cooling. The slag exits the furnace at 1500-1550°C (2700-2800’F) and when cool is transferred back to the ore milling facilities for recovery of any metals left in the slag. Furnace matte contains about 0.8% per ton of platinum and palladium as well as copper, nickel, sulfur and iron. It is tapped from the furnace approximately every 8 hours into ladles and granulated in preparation for converting. Granulated furnace matte is converted in Top Blown Rotary Converters (TBRC’s) using oxygen. Converter slag is poured into ladles, granulated, dried and returned to the electric furnace. Converter matte is poured into ladles, granulated, dried and transported to the Base Metals Refinery (BMR). The converter matte now

1761

Figure 1. The Stillwater Mining Company Nye Operation PGM Concentrator flowsheet.

1762

4

Lss

Figure 2. The Stillwater Mining Company PGM Smelter Flowsheet.

1763

Figure 3. The Stillwater Mining Company Base Metals Refinery Atmospheric Leach Section Flowsheet. 1764

Figure 4. The Stillwater Mining Company Base Metals Refinery Pressure Leach Section Flowsheet.

1765

contains approximately 2.0 % per tonne of platinum and palladium. All off-gases first pass through baghouses for particulate capture before being cleaned of sulfur dioxide (SO2) in scrubbers. Over 99.7% of the sulfur dioxide is captured from the off-gases. A sodium regeneration circuit provides caustic for the scrubbing system. Gypsum generated in the regeneration process is pressure filtered into low moisture cake and is trucked away for resale to the cement or agricultural industries. The Stillwater smelter flowsheet is illustrated in Figure 2. The Stillwater Base Metals Refinery (i.e. BMR) process consists of matte grinding, atmospheric leaching, pressure leaching, PGM concentrate separation and iron precipitation. (Newman and Makwana, 1997). Figure 3 illustrates the atmospheric leach process flowsheet. First, the converter matte from the smelter is ground batch-wise in a tower mill to yield a 70% solids slurry in water, with 80% of the solids passing 74 pm. The ground matte is leached with the recycled acidic pressure leach solution and oxygen in a series of five cascading agitated tanks. Some of the nickel and iron are extracted from the matte, while some of the copper present in the pressure leach solution is precipitated according to the following typical reactions: Nio + H2S04+ 0.5 O2 +NiS04 Ni3S2+ H2S04 + 0.5 O2 Ni3S2+ 2 CuS04

+H20

NiS04 + 2 NiS + H20

+ Cu2S+ 2 NiS04 + NiS

Feo + H2S04+ 0.5 O2

+ FeS04 + H 2 0

(1) (2) (3)

(4)

The dissolved iron remains essentially in the ferrous state under the prevailing low pH (2.0 to 2.2) condition of the atmospheric leach process. Any PGM present in the feed solution are also co-precipitated with the copper. The unleached residue from the atmospheric leach process is separated by thickening and then leached further under elevated temperature and oxygen pressure. Figure 4 illustrates the pressure leach process flowsheet. The atmospheric leach residue consists essentially of millerite (NiS), digenite (Cu1.8S),djurleite (Cu1.96S)and iron in the form of magnetite and hydrated ferric oxide. The principal reactions in the pressure leach process are of the type shown below. NiS + 2 O2 Cu2S + H2SO4 + 2.5 02

+ NiS04

(5)

+ 2 C U S O+~ HzO

(6)

Magnetite, if present, does not dissolve at the relatively low acid concentrations (20 to 25 g L ) and temperature (130 to 135OC) prevailing in the pressure leach process. The atmospheric leach thickener overflow solution is polish filtered and then treated to precipitate most of the iron as ammonium jarosite, which is filtered and returned to the smelter. The iron precipitation process is necessary to meet the requirements of the nickel-copper sulphate solution customer. The iron precipitation process also acts as a backstop to precipitate or catch any soluble and insoluble PGM that may be present in the atmospheric leach solution,. The iron is precipitated as ammonium jarosite according to the following reactions.

+

Fe2(S04)3+ 3 H20 2 FeS04 + 0.5 O2 + H2S04 3 Fe2(S04)3+2 NH3+ 12 H 2 0 2 NH4Fe3(OH)6(S04)2+ 5 H2S04

1766

(8)

(9)

The matte grinding, pressure leaching, PGM concentrate separation and iron precipitation circuits are designed to operate batch-wise, while the atmospheric leach circuit is designed to operate on a continuous basis during each shift. The plant produces a final readily refined bullion concentrate of about 60%PGM content. As well, by-product nickel sulfate and copper metal are now produced from this zero discharge hydrometallurgical facility. PGM’s from the Merensky Reef in South Africa The platinum-bearing iron-chromite reef layers within the basic pyroxene deposit of the Bushveld igneous rock complex are accompanied by small particles of the sulfides of iron, copper and nickel. The platinum, partly in the form of native metal, is invariably found as a ferro-platinum alloy and/or as the sulfide, and arsenide of iron, copper or nickel. The platinum group metal content of the crude ore that ranges from 4 to 6 grams per ton, consists of 50 to 60 percent platinum and 20 to 25 percent of palladium. The crude ores are crushed and pulverized. Sulfide concentrates containing these metals are separated and recovered by flotation processing (Flett, 1984, Hunt and Lever, 1969, Kirk-Other, 1968, Dayton and Burger, 1981, Dayton and Burger 1982). Native platinum and their ferro-alloys are separated by gravity separation techniques. The sulfide concentrates are smelted in an electric furnace and produce a 25 percent copper-nickel matte. This matte is further desulfurized by oxidizing in a converting process, to produce high grade 75 to 80 percent copper-nickel matte which contains as much as 2 kg of platinum group metals per ton of matte. This high-grade matte is water granulated and finely crushed, then the base metals copper and nickel are leached from it by acid solution to leave a residue that contains most of the platinum group metals. These then become the raw materials for further recovery and refining of, the platinum group metals. This process is analogous to the previous in depth descriptions of the processes in use at Stillwater and firther details will not be elucidated in this paper. INCO CANADA CONVENTIONAL PGM BULLION CONCENTRATE REFINING METHOD A process flow chart employed at INCO Canada (Shin Kinzoku, 1977, Kirk-Other, 1968), shown in Figure 5 , is illustrative of the conventional methods based on dissolution and precipitation processing. Other conventional processes differ only in detail from the one shown in this figure. The chemistry differences are related to raw material characteristics. By following the process flow chart, the respective unit chemical process steps are outlined. The raw material is attacked with aqua regia to dissolve gold, platinum and palladium. The aqua regia solution is treated with ferrous sulfate. Gold is precipitated and separated for its recovery and further purification. Saturated ammonium chloride solution is added to the gold free solution to precipitate ammonium hexachloroplatinate. This precipitate is separated and calcined to form platinum metal sponge of 98 percent purity. This sponge is further refmed by repeated dissolution and reprecipitation. The filtrate from the ammonium hexachloroplatinate is treated with ammonium hydroxide solution to which is later added hydrochloric acid. Diammine palladium dichloride is precipitated, separated, and calcined in a hydrogen atmosphere. The crude palladium metal obtained is firther purified by repeated dissolution and reprecipitation. The original insoluble residue from the aqua regia attack is fused with fluxes composed of litharge, soda ash, borax and carbon. The lead alloy formed collects all the precious metals including the secondary platinum group metals. The lead alloy is leached in hot nitric acid solution to dissolve silver and lead. The lead in the leach liquor is precipitated as lead sulfate, and the silver, as its chloride. The nitric acid insoluble residue is fused with sodium bisulfate, forming a water-soluble rhodium sulfate salt. A water solution of this salt is treated with ammonium nitrite solution which precipitates ammonium rhodium nitrite, (NH4)3Rh(N02)6. The nitrite salt is later treated with hydrochloric acid to form ammonium rhodium chloride, (NH4)3RhC16,which is firther refined before conversion to rhodium metal. The water-insoluble residue from the sodium bisulfate fision is next fised with a mixture of sodium hydroxide and potassium nitrate. Osmium and ruthenium form their respective water-soluble

1767

S~withwOudC

1

Lud Alloy

1 Fusion with Na2SO4

D h b t i o n with Atpa Re&

-

Precipitationwith NH4Q

bIr

Figure 5. INCO Canada Conventional PGM Bullion Concentrate Refinery Flowsheet.

1768

potassium salts, (potassium osmate and potassium ruthenate). Iridium forms an oxide that is leachable with aqua regia solution. After the potassium salts of osmium and ruthenium are dissolved away in water, the solution is treated with hydrochloric acid and chlorine gas. Volatile tetroxides, Os04 and Ru04, are evolved and they are absorbed into a mixture of dilute hydrochloric acid and methanol. When the hydrochloric acidmethanol solution is heated, only osmium tetroxide is evolved. This is absorbed into a sodium hydroxide/methanol solution, €tom which ammonium osmium chloride is precipitated by the addition of ammonium chloride. This chloride salt is then calcined in a hydrogen atmosphere to produce crude osmium. This is further refined by repeated distillation and precipitation separation. Ruthenium remains as ruthenium oxychloride in the residue of the previous evaporation step. This is reduced by heating in a hydrogen atmosphere to obtain crude ruthenium. This is further treated for purification. The iridium oxide that is formed during the previous fusion treatment is dissolved in aqua regia and precipitated with ammonium chloride solution as ammonium iridium chloride. Crude iridium is obtained by calcining the ammonium iridium chloride compound in a hydrogen atmosphere. Further purification of the ammonium iridium chloride requires repeated dissolution and reprecipitation. This summary of a typical conventional process flow chart for platinum group metals describes the separation and production of the respective crude metals. A complete separation of the respective metals cannot be made in a simple dissolution-precipitation step. Co-precipitation occurs and other metal elements remain as impurities. To obtain pure metals of the respective elements, the crude metal requires further refining. Generally, refining processes involve many repeated steps of dissolution, conditioning and precipitation operations. PGM BULLION CONCENTRATE SOLVENT EXTRACTION REFINING TECHNOLOGY The degree of separation attained with the conventional and traditional separations used in refining platinum group metals cannot be considered to be efficient in terms of the yields, the complexity of the operation and the labor expended. Since the early 1970’s, considerable research and development efforts have been directed toward replacing traditional processes with new solvent extraction technology (Warshawsky, 1983, West, 1984, Edwards, 1984) . Flett 1983, Flett 1982, Dhara, 1984). Commercial refineries have incorporated this technology into their operations, employing various solvent extraction schemes. Major advantages of these solvent extraction processes are. Lower inventory due to the reduced overall processing time; higher separation efficiency; higher product purity; improved yields; flexibility; versatility and capability of continuous operation with process control. The separation and refining processes that employ solvent extraction technology are based on physico-chemical characteristics of precious metal solution chemistry such as the nature of the complex ionic species and their redox potentials. Extraction schemes vary with the kind of extractant used, extraction rates, distribution ratios, separation efficiency, etc. For the commercial and industrial operation of solvent extraction processes to be successful, various operational conditions and considerations have to be established with respect to efficiency and rate of back extraction (stripping), degradation and loss of extractants, flexibility and versatility of process, etc. Depending on differences in the forms of raw material, contents and grade of precious metals in them, and pretreatments, the respective commercial refineries have developed and applied different solvent extraction technologies and processes. Currently, there are three large refineries using solvent extraction technology, and future large-scale refineries will probably employ solvent extraction technology. The aqueous chemistry of precious metals as their chlorides in hydrochloric acid is complex. Thermodynamic and kinetic behaviors influence the solvent extraction scheme selected. The following are brief summaries of the respective element chlorocomplex behavior that have been

1769

studied and understood for application to solvent extraction. These include oxidation and coordination numbers, number of d-electrons and rates of ligand exchange reactions by amine extractants. The detailed information is contained in the cited literature. Gold, platinum and palladium dissolve relatively easily in hydrochloric acidchlorine mixtures. Osmium, ruthenium, rhodium and iridium and their oxides have slow or near zero dissolution rates in such mixtures. Alkaline fusion is often required before these elements will dissolve in hydrochloric acidchlorine solutions. Various basic amine extractants are capable of forming ionpair compounds in the organic phase. The degree of ion-pair formation depends on the size of anionic chlorocomplex ions and the ionic charge. These increase from MC14'-, MC162', to MC142' to MC163-. The basicity of amine extractants increases from primary, secondary, tertiary to quarternary. The more basic the amine extractant, the stronger is the ion-pair formation. This will result in good extraction capability but it will be difficult to strip. Where neutral organic solvation extractants involve the extraction of chlorocomplexes, the extractability increases from TPP, TBP to TOPO, following the basicity-increase of the esters. (TPP = Triphenyl phosphite, TBP = tributyl phosphite, TOPO = trioctylphosphine oxide.) Ligand d4 exchange rates depend on d-electron confirguations of the octahedral complexes. The order of the exchange rate is d5 > d4 > d3 > d6. With d4 square planar complexes, the order is palladium(II)> gold(II1) > platinum(I1). Redox reaction rates also depend on d-electron configurations. It is well known that the reduction of iridium(1V) to iridium(II1) is much faster than that of platinum(1V) to platinum(I1). Chlorocomplexes of gold, platinum and palladium formed in hydrochloric acidchlorine solution are generally those of higher oxidation states. Those of iridium, rhodium, osmium and ruthenium are complex. The latter two elements particularly present many kinds of mixed aquochlorocomplexes. The forms of chlorocomplexes or aquocomplexes depend on solution potentials, chloride ion concentration and acid concentrations. This summary is just a short abstract of the aqueous chemistry in solvent extraction technology. The process flow charts of three commercial refineries are described in the next section. INCO ACTON BULLION CONCENTRATE REFINERY Concurrent with the implementation of its new pressurized carbonyl process, employed to replace conventional nickel production from crude nickel metal, INCO examined its traditional lengthy and less efficient precious metals separation, recovery and refining process in light of solvent extraction. INCO developed and now operates a solvent extraction process. for precious metals recovery refining (Barnes and Edwards, 1982, Rimmer, 1974). The process flow chart is outlined Figure 6.

The raw feed material is copper electrorefining anode slimes from primary copper production. These, in turn, are derived from nickel carbonyl gasification residues that were reprocessed in a copper converter. The process is as follows: The raw material is attacked at 90-98'C with hydrochloric acidchlorine gas. The undissolved residues are attacked with nitric acid for silver and lead removal, and then fused with sodium hydroxide at 500 to 600'C. The alkaline fusion salt is then dissolved in hydrochlorickhlorine solution for the following treatments. After removal of excess chlorine from the solution, it is neutralized with sodium hydroxide. Sodium bromate, NaBr03, solution is added to it and the vapors of tetroxides of ruthenium and osmium are distilled away and absorbed in weak hydrochloric acid solution. Hydrolysis with sodium hydroxide follows and base metals such as copper are filtered out as their solid hydroxides. The solution is adjusted to contain a hydrochloric acid concentration of 3 to 4 mole/liter. Gold is extracted from the solution into dibutyl carbitol (DBC). The organic phase is scrubbed with dilute, 1-2Mhydrochloric acid to remove base metals. Gold is recovered from the scrubbed organic phase by direct reduction with aqueous oxalic acid. Palladium is extracted into an organic phase with

1770

stripping . with hot o&c aad

HCI

1

1

Solvent Extrdon withTBP

Figure 6. INCO Acton PGM Bullion Concentrate Refinery Flowsheet.

1771

Au

di-n-octyl sulfide, (DOS, R2S), which leaves a palladium free aqueous raffinate phase. This process utilizes the lability of the PdC1; ion with respect to its ligand exchange rate. The extraction exchange is described as follows:

The organic phase is scrubbed with dilute hydrochloric acid, then palladium is stripped from the organic phase with an aqueous ammonium hydroxide solution to form [Pd(NH3)F]aq.This ammoniacal solution is neutralized with hydrochloric acid which precipitates Pd(NH4)2C12.The stripping reaction is as follows: +4 [pdc12(~2~)21org

+

~ ~ 3 1 a q~

[ R ~+ s[pdo\r~3),2'1aq I ~ ~ ~ +2[c~aq

(1 1)

Platinum is recovered from the palladium free aqueous raffinate phase. The hydrochloric acid concentration in it is adjusted to 5 to 6 mole per liter. Iridium (IV) is reduced to iridium (111) by sparging sulfur dioxide gas into the solution. Platinum is then extracted into tri-n-butyl phosphate (TBP) by the following reaction:

After scrubbing the TBP organic phase with hydrochloric acid solution, it is stripped with water. Platinum is recovered as (NH&PtCl, with ammonium chloride. Iridium in the platinum free aqueous raffinate phase is then oxidized from the (111) state back to its (IV) state. It is extracted from this aqueous phase with TBP into an organic TBP phase. Iridium is then recovered from this TBP phase in a process similar to that described above for platinum. The separation of rhodium, is still under development according to the cited literature. MATTHEY RUSTENBURG PGM BULLION CONCENTRATE REFINERY The Matthey Rustenburg Refinery (Cleare, Charlesworth and Bryson, 1979, Reavell and Charlesworth, 1980 and Charlesworth, 1981) having similar reasons to those of INCO, to overcome the demerits of traditional separation and refining process, have started a solvent extraction process that uses oxime/amine extractants. This technology is based on the "straight chain process". The process flow chart is shown in Figure 7. The precious metal raw materials are dissolved in hydrochloric acid/chlorine solution. Silver forms insoluble AgCl which is separated and recovered here. Gold, in solution in the form of AuCl;, is extracted by either of the solution extraction schemes with TBP or with methyl isobutyl ketone (MIBK), into the organic phase. Its extraction reaction may be described as:

Other impurities, Fe, Te, etc., are extracted as well so that the organic phase is next scrubbed with hydrochloric acid for impurity removal. Gold is reduced and recovered from the organic phase with iron powder. Palladium is extracted by a beta-hydroxyoxime, making use of the ligand exchange reaction as follows: [PdC1:-laq

+ 2[RH]0r, +[PdR2]org + 2[H+]aq + 4[C1-Iaq

(14)

Since this extraction rate, i.e., the ligand exchange rate, is small, an organic amine reagent is added to accelerate the extraction. After the organic phase has been scrubbed with weak hydrochloric acid to remove base metal impurities, palladium stripping is done with aqueous ammonium

1772

Total Dissolution with H C 4 Q

1 Flltr8Oa

1

Solvent Extraction with MIBK

Treatment

Scrubbin0 with dil H a

with Fe powder

_____+

1

.

Scrubbing with dil Ha

Conditioningwith NaOH, etc.

1 Dlstillrrtlon

AbsorptlonwithHCI

1 Raluctloo with So2 sparging

4 Solvent Extraction with tri-n-oaylamirre

~

Skipping

with conc HCI

+

-.+

withconcHC1

Au

pd

I

1 Base Metals

1

Ag

_L*

.

Ba& Metals

Solvent Extraction with p-hydroxyoxim

-

AgClResidue

-

.-.

m3

Redistillation

-

Precipitation with N H 3

0s

RU

.

Pt

Oxidation Conditioning with HCI

4 Solvent Extraction with tri-n-octylamine

4

- Scrubbing

Stripping

Ir *Rh

Ion Exchange or Chemical Refining

Figure 7. Matthey Rustenberg PGM Bullion Concentrate Refinery Flowsheet.

1773

hydroxide solution. Palladium is precipitated from it as (NH4)2PdCl6 with hydrochloric acid. Osmium and ruthenium are separated as their volatile tetroxides after the solution has been neutralized with alkaline hydroxide. The tetroxides are absorbed into dilute hydrochloric acid solution and this is later redistilled for separation of osmium from ruthenium. After reducing the iridium (IV) to its (111) state, platinum is extracted by the tertiary amine, tri-n-octyl amine into the organic phase, as:

+

[~tc16~1aq+ 2 [ ~ ~ + 1 o r g E ( R H ) ~ P I~org c~~

(15)

Platinum is stripped from the organic phase with 10-12M hydrochloric acid. The platinum is precipitated from it as (NH4)2PtCIs with ammonium chloride. After oxidizing iridium (111) back to its (IV) state, and adjusting the acid concentration to about 4 mole per liter, iridium is then extracted into the organic phase of the tri-n-octyl amine phase in the same way as the platinum. The organic phase is scrubbed with dilute acid. The iridium is stripped into the aqueous phase after reducing it to its (111) valency state and recovered. Finally, rhodium is separated and recovered from the solution containing other element impurities. The detailed information on the operation and the chemistry rhodium recovery is not available at present. LONRHO PGM BULLION CONCENTRATE REFINERY The Lonrho Refinery solvent extraction process flowsheet is show in Figure 8 (Berry, 1979, Edwards, 1979). This process is a modified solvent extraction technology based on a process scheme developed at the National Institute of Metallurgy, South Africa. The Impala Platinum Refinery is said to have installed a similar process at Springs, South Africa. The raw material at Lonrho has a relatively high content of secondary platinum group metals, i.e. metals other than platinum and palladium, when compared with raw material feeds at other refineries. The raw material is first treated for base metal removal by leaching with acid solution. Then it is reduced with carbon and aluminum to produce an aluminum precious metal alloy. This alloy, upon dissolution in hydrochloric acidkhlorine solution solubilizes rhodium, iridium, ruthenium and osmium in a shorter period of time. The alloy is dissolved in hydrochloric acid. The remaining insolubles are dissolved in hydrochloric acidkhlorine solution. Silver is precipitated from the solution by dilution with water. Gold is reduced with a sulfur dioxide sparge. The solution is adjusted to a 0.5 - 1.O M acid concentration. The secondary platinum group metal valence states are adjusted to three. Platinum and palladium are extracted from the conditioned solution by the acetic acid derivative of a secondary amine, i.e. R2NCH2COOH.Both metals are stripped from the organic phase with hydrochloric acid. The principle of the separation of platinum from palladium is based on the fact that PdC1:is much more labile in its ligand exchange reactions than is PtCl;-. Only palladium is extracted from the raffinate into an organic sulfide phase, a ligand exchange reaction to form PdC12(RSR)2.This is stripped from its organic phase with aqueous ammonia as follows:

Palladium in the form of Pd(NH3)4C12is precipitated by the addition of hydrochloric acid to the aqueous ammonia strip. The acidified palladium free raffinate is made alkaline and osmium tetroxide vapor is distilled from the solution and absorbed in dilute hydrochloric acid. Ruthenium is separated for recovery as an ionic ruthenium nitrosyl complex, formed by the addition of nitric acid to the filtrate. The nitrosyl complex is transferred to an organic phase by a tertiary amine extractant. The ruthenium is stripped with 10 percent sodium hydroxide and precipitated as hydroxide. This is recovered and amine extractant is washed with hydrochloric acid and returned to the extraction step. Iridium complex ions in the solution are separated by the adsorption on to strongly basic resins. After desorption by a solution saturated with sulfur dioxide, followed by oxidation of the iridium to valence (IV),

1774

Carbon Reduction Ahrmtnum Reduction

IiotHclLuching

i

-

c c c

Reduction with S% spsrging SolventExtraction with *tic acid derivative of secondaryamine

c

Conditioning with HN%

4 Solvent Extraction with Ilimyl tatiary amine

4

1-

- -

Solvent Extraction withdi-N-he~yl + Pt

Stripping with HCl

Sulfide

.

4

Absorption with HQ

OsO4vapor

conditionblg

Ion Exchange with S t r o n g base nsin

__*

-

Stripping with NaOH

Dilution

Soveat Extraction

withqwusso2‘f

WithTBP

4u-

(32

Conditioning with NaQ, Na2S%

Figure 8. The Lonrho PGM Bullion Concentrate Refinery Flowsheet.

1775

-*

0s

b

Ru

f”

NH4C’

its chlorocomplex ions are extracted by TBP. The TBP phase is stripped with water and iridium in the water phase is precipitated as (NH4)21rC16by addition of ammonium chloride solution. Rhodium is precipitated out as its sodium salt by addition of sodium chloride and sodium sulfite solution and separated. This precipitate is then dissolved in acid solution and by the addition of ammonium chloride solution. recovered in the form of (NH4)~RhC16 SUMMARY In recent years, the production and refining of platinum group metals has become increasingly important. This paper focused on the primary production and refining of PGM’s. However, the basic principles utilized are analogous for the recovery and refining of secondary and recycled PGM’s. ACKNOWLEDGMENTS The authors gratefully acknowledge the expert assistance of Ms. Tami J. Cashell as well as thank Mr. G. Andrew Hadden for his literature search and carefil review of the manuscript. REFERENCES Barnes, J.E., and J.D. Edwards. 1982, “Solvent Extraction at INCO’s Acton Precious Metal Refinery”, Chemistry and Industry, March 6,15 1. Benner, L.S., T. Suzuki, K. Meguro, and S. Tanaka. 1991, Precious Metals Science and Technology, International Precious Metals Institute, Allentown Pennsylvania. Berry, R.I. 1979, “Refining Precious Metals”, Chemical Engineering June 18,90. Bushman, B., and G. Roset. 2002, Stillwater Mining Company, Personal communication, March. Charlesworth, P. 1981, “Separating the Platinum Group Metals by Liquid-Liquid Extraction”, Platinum Metals Review, 25(3), 106. Cleare, M.J., P. Charlesworth, and D.J. Bryson. 1979, “Solvent Extraction in Platinum Metal Processing”, Journal Chem. Tech. Biotech., 29,2 10. Dayton, S.H. and J.R. Burger. 1981, “Resouces, Rare Formations in Abundance”, Engineering & Mining Journal, 182(1 l), 67. Dayton, S.H., and J.R. Burger. 1982,”Platinum Mining Profiles”, Engineering & Mining Journal, 4, 182(1 l), 64. Dhara, S.C. 1984, Solvent Extraction of Precious Metals with Organic Amines, “Precious Metals”, p. 199, The Metallurgical Society, AIME. Edwards, R.I. 1976, “Refining of the Platinum Group Metals”, Journal of Metals, 28(8), 4. Edwards, R.I. 1979, “Selective Solvent Extraction for the Refining of Platinum Metals”, Proceedings of International Solvent Extraction Conference, Canada. Flett D.S. 1984, Visit Minute to Western Platinum Mines Ltd., Private Communication, Akio Fuwa. Flett, D.S. 1982, Solvent Extraction in Hydrometallurgy, Proceedings of International Seminar, EMU, London., 9,39. Flett, D.S. 1982, Solvent Extraction in Precious Metals Refining, Proceedings of International Seminar, EMU, London. Head, M.P. 1976, “Nickel Refining by the TBRC Smelting and Pressure Carbonyl Route”, Paper Presented at the AIME Meeting, NV. Hodges, G.J.,. G.K. Roset, J.W. Matousek , and P.J. Marcantonio. 1991, “Stillwater Mining Co.3 Precious Metals Smelter: From Pilot to Production,” Mining Engineering, Vol. 43, No. 7, 724-727. Hougan, L.R. and. H. Zacharisen. 1975, “Recovery ofNickel, Copper and Precious Metal Concentrate from High Grade Precious Metal Mattes”, Journal of Metals, 27(5), 6 . Hunt L.B. and F.M. Lever. 1969, “Availability of the Platinum Metals”, Platinum Metals Review, 19(4), 126.

1776

Kirk-Other. 1968, “Encyclopedia of Chemical Technology”, 2nd Ed., Vol. 15, p. 832, JohnWiley and Sons Inc. Loebenstein, J.R. 1983, “Platinum Group Metals in the USSR”, Platinum Group Metals 1983, IPMI, Williamsburg, Virginia. Minty, K.C. 1999, “North American Palladium Ltd., Lac Lles Mines Ltd., -An Update”; PRECIOUS METALS 1999, Proceedings of the 23rd Annual Conference, IPMI,Acapulco, Mexico. Newman, L. and M. Makwana. 1997, “Commissioning of the Stillwater Mining Company Base Metals Refmery”, Hydrometallurw and Refining of Nickel and Cobalt, Proceedings of the 27th Annual Hydrometallurgical Meeting, The Metallurgical Society of CIM, Sudbury, Ontario, August Reavell, L.R.P., and P.C. Charlesworth. 1980, “The Application of Solvent Extraction to Platinum Group Metals Refming”, Proceedings of International Solvent Extraction Conference, Belgium. R i m e r , B.F. 1974, “Refining of Gold from Precious Metal Concentrates by Liquid-Liquid Extraction”, bid, January 19,63. Sharratt, J.M., 1994, “StillwaterMining Company”, International Precious Metals Institute. Vol. 18, No. 1, Jan-Feb 3 - 6. Shin Kinzoku Handbook, 1977, “New Metals Data Handbook“, p. 282, Agne Publishing Company, Japan. Turk, D.J., 2000 Stillwater Mining Company Nye Concentrator Operation, SME Annual Meeting, Preprint 00-10, Salt Lake City, Utah. Warshawsky, A. 1983, Integrated Ion Exchange and Liquid-Liquid Extraction Process for the Separation of Platinum Group Metals, Hydrometallurgy Research, Development, Plant Practice”, p. 5 17, The Metallurgical Society, AIME. West, R.C. Ed. (1984),”CRC Handbook of Chemistry and Physics“, p. D-156, CRC Press Inc. Weir, D.R., D.G. Kerfoot, and R.P. Kofluk R.P. 1986, “Recovery of Platinum Group Metals from Nickelcopper-Iron Matte”, United States Patent. No. 4,571,262, February 18. I*

1777

FUNDAMENTALS OF THE ANALYSIS OF GOLD, SILVER AND PLATINUM GROUP METALS Corby G. Anderson’

ABSTRACT The analysis of silver and gold content in ores, concentrates and other materials is generally determined by classical lead based fire assay. As well, platinum group metals (PGM’s) such as platinum, palladium, rhodium, iridium, ruthenium and osmium are often separated or concentrated by fire assay. This paper outlines the basic concepts involved in this proven methodology. lncluded are general procedures for selecting appropriate tire assay charges based on approximate chemical compositions of the samples. Also included are descriptions of modem instrumental methods for the determination of the noble metals with a focus on the analysis of PGM’s. INTRODUCTION The determination of gold, silver, platinum, palladium, rhodium, ruthenium, osmium and iridium metals is often an important aspect of the discovery, development, design and operation of a precious metals mine, as well as the associated concentrator or metal production plant. As well, an abundance of financial issues in the mineral processing industry rely on the accurate and precise determination of the precious metals. The basic procedures and technologies involved in this discipline are discussed in this paper.

FIRE ASSAY FUNDAMENTALS Fire assaying is an established form of quantitative chemical analysis by which metals are separated and determined in ores and metallurgical products with the aid of heat and dry reagents. Traditionally, the object is to form a melt of at least two phases - a complex liquid borosilicate slag and a liquid lead precious metal collection phase of a controlled size. For PGM’s, collectors other than lead such as copper or nickel sulfide are often utilized. The high degree of solubility of the noble metals such as Ag and Au in molten metallic lead plus the great difference in specific gravity between the lead and slag permit the separation of the noble metals from the slag as lead alloys. The subsequent removal of the lead as lead oxide in a porous vessel known as a cupel by a carefully controlled oxidizing fusion separates the lead from the noble metals. The remaining metallic bead is then analyzed for the noble metals. By analogy, fire assay may be looked upon as ‘pyrometallurgical solvent extraction’ whereby a precious metal is collected and concentrated in a selective ‘solvent’. Then the solvent is stripped from the precious metal by cupeling.

The mixture of chemicals and ore used for the pot fusion provides a complex system. By analogy with simple systems such as a metal oxide with borax or silica, one may make reasonable guesses concerning some of the reactions, but a complete explanation of the reaction for even one ore composition must await an extensive examination of these multicomponent systems. Thus, the technique of a fire assay collection of the noble metals is largely an empirical process assisted to a degree by some fundamental principles. Initially the ore to be assayed must be in an exceedingly fine state of division and thoroughly mixed with the flux constituents. These conditions are necessary to ensure the intimate contact of each ore particle with particles of the melting flux. Ideally this contact must be maintained during the early stage of the fusing process. This is necessary in order to ensure in-situ a sufficiently complete reaction between ore and flux and the simultaneous production of the fine

’ Center for Advanced Mineral and Metallurgical Processing, Montana Tech, Butte, MT, USA. 1778

globules of lead by the reduction of litharge (PbO). To bring about this condition, the composition of the flux, the temperature, and its rate of increase must be carefully arranged. The optimum viscosity is somewhat dependent upon the proportion of borax and the character of the borates and silicates that are formed at or near incipient fusion. In the absence of this juxtaposition of lead and noble metal, one must depend upon the ability of the high-density noble metal to settle to the bottom of the pot during the subsequent fusion process. This settling process is facilitated by the increase in fluidity of the mixture at the elevated temperatures, and for platinum, palladium, and gold at least, it is also facilitated by the alloying tendency with the lead finally collected in the bottom of the crucible. Thus, for certain of the noble metals, a reasonably acceptable assay may be achieved even under unfavorable conditions. This opinion is supported by the fact that the classical lead fire assay does not accomplish the direct recovery of iridium through the formation of a homogeneous lead alloy. It is not unlikely that practically all of the iridium, and iridosmine, when the latter is present, are recovered by the fall of these high density metals through the low viscosity liquid in the later stages of the fusion. However, there is ample evidence that in general, all six of the platinum metals and silver and gold can be quantitatively recovered by careful use and manipulation of lead collection.

The beginnings of fire assaying can be traced to the finds in Troy I1 (about 2600 B.C.) and in the Cappadocian Tablets (2250 -1950 B.C.). These finds prove that very pure silver was made in the twenty-fifth century B.C. From this evidence we must conclude that the cupellation process, and therefore fire assaying, was invented in Asia Minor in the first half of the third millennium B.C. shortly after the discovery of the manufacture of lead from galena (Forbes, 1950). The first convincing evidence of the production of silver from lead ores is the cupel buttons found at Mahmatlar in the late third millenium B.C. and that are now in the Hittite Museum in Ankara, Turkey. Now, most methods combine fire assay precious metals collection with subsequent instrumental determination. The advantage of fire assay techniques for the determinationof noble metals is the ability to use a relatively large ore sample from which to concentrate these metals, in addition to eliminating virtually all the associated gangue minerals. The chemical reactions which take place in an assay fusion are very complex and beyond the scope of this paper. Detailed treatment of the theory involved can be found in textbooks on fire assaying (Bugbee, 1940; Shepard and Dietrich, 1940). However, one must initially determine whether the ore is neutral, reducing, or oxidizing in character. Neutral ores have no reducing or oxidizing power and usually are the siliceous, oxide, and carbonate ores. Reducing ores decompose litharge to form metallic lead and usually consist of sulfides and carbonaceous matter. Oxidizing ores contain ferric oxide and manganese dioxide which, when fused with fluxes, oxidize lead or reducing agents. Ores with considerable oxidizing power are comparatively rare. The fundamental overall chemical equation for the process of litharge reduction can be expressed as: 2 P b O + C + Pb+CO2 (1) The fundamentaloverall chemical equation for the process of lead oxidation is expressed as: 2 Pb + 02 -.) 2PbO

Fire assay is better facilitated if the chemical and mineralogical composition of the ore is known or can be obtained. For this reason it is advisable to make a semiquantitative analysis via a suitable instrumental method along with a coarse mineralogical examination of the sample either visually or with x-ray diffraction. From this information one can usually determine whether an ore is neutral, reducing, or oxidizing. All this information is necessary to prepare an optimum charge so that maximum recovery of the noble metals may be realized. If the description and an analysis of a sample or ore indicates the presence of sulfides, a preliminary fusion is recommended to establish the "reducing power" of the ore. Figure 1. illustratesthis (Ammen, 1984).

1779

f

Assay Crucible

Charge: 3g Ore, log Sodium Carbonate (NaCO,), 469 Litharge (PbO), 39 Silica (Si02),and Borax glass (Na,B,O,) r

Fusion: 20 Minutes at 20 Minutes at 1,OOO"C

4900"C;

/I I

Cast-Iron Mold

A Lead Prill Pill Weight

t

3 = R.P.

Figure 1 Establishing the reducing power of an ore The term "reducing power," as used in fire assaying, is defined as the amount of lead that 1 g of the ore will produce when fused with an excess of litharge. Conversely, oxidizing power is the amount of lead 1 g of ore will oxidize to lead oxide. When the need arises to establish the reducing power of an ore, a preliminary fusion is employed typically using the following charge constituted in a 10-g fireclay crucible: 3 g ore, 10 g NaXO3, 46 g PbO, 3 g Si02, and 1 g Na2B&I7. The fusion is performed at a temperature of 900°C to 1OOOOC for 40 minutes. The lead button produced h m this is weighed and the reducing power is then calculated. For example, if 3.00 g of ore is used and a lead button is obtained weighing 15.00 g,the reducing power of the ore is 15.00/3.00= 5.00. In some cases, a small amount of flour is added in the preliminary fusion to ensure that, in the case of neutral or oxidizing ores, a lead button is produced The approximate reducing power @P.) of some common minerals and reagents are listed in Table 1. Table 1. Approximate reducing power @P.) of some common minerals and reagents Mineral or reapent Flour FeSz Pyrite PbS, Galena CU~S, Chalcocite FeAsS, Arsenopyrite Sb2S3,Stibnite CuFeS2 Chalcopyrite ZnS, Sphalerite FeS, Pyrrhotite Fe, Metallic iron C, Carbon

R.P. 10-11 11 3-4 5

7 7 8 8 9

4-6 18-25

1780

Salt or Borax Cover

Closed Assay Furnace

3

-

Charge +

Preheat Crucible+

1,OOO”C ZtTFusion Is Complete

1 1

l-,a Remove Slag and Cube Button

Figure 2 Flow chart for crucible fusion

-D

-

‘

An oxidizing ore is one that will not reduce PbO. Flour or another reductant must be used to produce a lead button. Oxidizing ores seldom need more than a slight increase in the amount of flour used. Table 2. lists some oxidizing ores and reagents and their oxidizing power (O.P.):

T a b l e 2. Approximate oxidizing power (O.P.) of some common minerals and reagents Mineral or reagent Fe203,Hematite Mn0= Pyrolusite Fe304,Magnetite KN03 ,Niter Magnetite-ilmenite

O.P. 1.3 2.4

0.9 4.2 0.4-0.6

Having established the reducing power of the ore, the following calculations are used to determine the amount of oxidant in the form of nitre, requiredto obtain a desirable-sized button (28 -30 g) when 15 g of sample is used: Lead Total reducing effect of ore 15 x 5.00 = Lead button desired = Difference, ore equivalent that must be oxidized by niter = I g of niter oxidizes (from Table 2) Niter required = 4514.2 -

75.0 g 30.0 g 45.0 g 4.2 g 10.7 g

Dry reagents or flux components are added to the pulverized ore in a fireclay crucible to effect a fusion at an easily attained temperature.Each reagent serves a specificpurpose in the h i o n process, as follows:

Sodium carbonate (soda). - NaZC03, is a powerful basic flux and readily forms alkali sulfides during the crucible fusion. Some sulfates are formed in the presence of air, and for this reason Na2C03 can be considered a desulfurizing and oxidizing agent. It melts at 852" C; when heated to 950°C it undergoes a slight dissociation with the evolution of a small amount of COz and the liberation of about 0.4 percent of free alkali. Both the free alkali and sodium carbonate react to form silicates and aluminates Litharge.- PbO, is also a basic flux and acts as an oxidizing and desulfurizing agent. It melts at 883°C and on being reduced it provides the lead necessary for the collection of the noble metals. Litharge has such a strong affinity for silica that, if the crucible charge does not contain enough silica, the PbO will attack the crucible walls and, if left long enough, will eat a hole through the crucible. Silica - S O z , is a strong acidic reagent which combines with the metal oxides to form silicates, the foundation of almost all slags. It is added to the charge when the ore is deficient in silica to give a more fluid melt and to protect the crucibles from the corrosive action of 1itharge . Borax glass. - Na2B407,is extremely viscous when melted, but at a red heat it becomes fluid and a strong acid, dissolving and fluxing practically all the metallic oxides both acidic and basic. In addition, the fact that it fuses at a low temperature facilitates slagging of the ore and lowers the fusing point of all slags. For these reasons it is used in almost every crucible fusion. Calcium fluoride. - CaF2, is used especially when the aluminum content of the sample is 1 percent or more. It increases the fluidity of almost any charge.

1782

Flour. - Flour is a reducing agent because of the carbon that it contains and is commonly used in the crucible charge.

Potassium nitrate. - KN03, commonly known as niter, is a strong oxidizing agent. It melts at 339"C, but at a higher temperature it decomposes giving off oxygen which oxidizes sulfur and many of the metals. Potassium nitrate is used chiefly to oxidize sulfide-bearing ores. It is advisable to establish the reducing power of the sample because excessive amounts of nitre may cause boiling-over of the charge. The most important factor for a successful crucible fusion is the proper selection and amounts of the flux components. A good flux will produce a slag with the following characteristics: 1.

It must have a formation temperature within the temperature range of the assay furnace.

2.

It must remain sufficiently thick at or near its formation temperature to allow any precious metals present to be released from their chemical or mechanical bonds with the gangue before the flux allows the collector particles of lead to drop down and alloy with the precious metal values.

3.

It should become sufficiently thin when heated above its formation temperature to allow the reduced lead globules to settle through it easily.

4.

It should completely decompose the gangue to a fluid slag, and it should have a very low affinity for gold and silver.

5.

Its chemical composition should be such that it does not excessively attack or flux away the crucible.

6 . Its specific gravity should be low enough to give a good separation between the lead and the slag.

7.

When the slag is cold, it should be homogeneous and easily removed from the button.

8.

It should not retain higher oxides (oxides containing more than 2 0 atoms per molecule) of metal, and yet, at the same time, should contain all the impurities present in the gangue.

9.

It should be free of sulfides.

When the desired flux has been worked up based on the preliminary fusion to determine the reducing power of the ore and on the examination of the ore the fusion can proceed. Because the selection of flux components is based primarily upon a knowledge of the chemical or mineralogical composition of the sample, it is advisable to perform an analysis of the sample prior to constituting the flux. By comparing the results for elements (of 1 percent or more) obtained from the semi quantitative analysis with the approximate chemical composition of the various types of samples listed in Table 3 (Haffty, Riley and GOSS,1977), the proper charge can generally be constituted. In depth flux calculations are beyond the scope of this paper and are covered in other sources (Shepard and Dietrich, 1940) Experience has shown that for basic and ultra basic rocks, includingmineralized basic rocks, a flux high in borax should be used. This flux is also used for silicates where the Fe and Mg are each 5 -10 percent or more. A convenient flux to use for these types of samples consists of 30 g Na2C03,35 g Pbo, 4 g SO2,35 g Na2B407,1 g CaF2,and 3.2 g of flour for a 15-g sample. One component, Si02,may be varied 6om 4 to 8 g depending upon the silica content of the sample.

1783

Table 3. Charges for various types of samples 10. Greater than 10 percent to maximum for the compound. or greater than value shown. R.P.. Reducing power of the sample. Lpaders I . . . 1 indicate charge component was not usedl

Sample type and auxiliary application*

Typical analyses. in percent

Charge. in GamH Sample weight

Na,CO,

PhO

SiO,

Na,R,O,

Ca F*

Flour

KNO, ~~

Aluminum I oxide (reagent) Arsenopyri te R.P. = 2.82

Baaalt

Calcite

.......

A1

...........

..................

1

...............

(A1203)

Chromite concentrate (refractory) 6 times for 16 g.

Auxiliary flux 5 tjmrs for 15 g

30

70

15

10

8

3.0

0

G

15

25

90

8

10

1

0

3.5

15

30

35

4-8

35

1

3.2

0

15

20

50

14

12

0

3.0

0

15

30

70

15

10

8

3.0

0

3

30

35

10

30

1

3.5

0

30

...

5

25

...

...

G

Si A1

5

Mg

I -

gg

5 10

Si

... .{

15

Fe

.................. Ca

Ceramic'

G

2

(3)

...........................

.

.

I

a .

In

cv

:

0,

In

OW(300

-:

0

2 rl

*

v)

m

0

m

rl

In

1785

0

0

(

e

0

0

a

4

m

$ 5 :

o m

r

0

0

u)

m

0

m

d

In

0

2

0

0

0

m

z

W

In

rl

Lo

0

0

w

O *

ea

u)

rn

-

Table 3. Charges for various types of samples continued

Galena and Si sphalerite, composite R.P. = 3.16 ...........

Hematite (iron ore)

10

.............

Jasperoid ............... Si Kaolin rock'

............

Magnetite ore

Norite ..................

66

13

6

1

0

4.0

16

26

46

3

10

1

3.0

0

G

16

25

60

12-15

7

1

3.8

0

G

16

20

60

1

3

0

2.8

0

16

20

60

12

10

5

2.8

0

16

26

60

16

8

4

4.0

0

7 10 3 7 3 7

16

30

36

6 -12

36

1

3.2

0

G G G G G

16

30

35

4-4

36

1 -2

3.2

0

G

Si < 1 -

....

26

s i 7

.......... Fe

Manganese-rich ores

16

G(16.2) G(16.6)

A1 < I Fe 7 Mg < 1 Ca < 1 Na 1K <1Mn G 2

A1 Fe

Mg

Ca

373 .3-

G

v

0

I

rl

0

In

m

0

m

I 1 Icdrn

I

v

v+

0 0

2

0

I

In

*

m

rl

cv

0

0

2

rl

fu

In rl

0

2

0

0

0

m

0

0

0

m

m

*

0

0

In

OD

m

0

m

0

w

0

m

Ic

fx

In

0

m

0 (0

f u f u

rl

0

In m

4

4

0

0

u a m

c3

6 6

i i

Y

.-W

a

C3

0

0

4

In

0

m

In

0

d

d

In

0

k

1787

I

0

c.

IA

a

.-L0 (D

> L

0

-4

0

hl

Y

0

I

m

m

0

m

W

u!

OD

0

9

c-

0

7

r(

0

0

OD

0

0

m 4

0

5:

0

Y

m

I ?

0

m

:

d

In

r(

c-

cu

u3

m

tc

0

0

0

2

0)

4

2

0

m

0

N

g

A

c\1 d

10

0

0

0

0

m

hl

m

0

N

d

L3

0

0

ZLLLN

w a c

d

d

ln

0

1788

-

Table 3. Charges for various types of samples continued Sample typc and auxiliary applications

Tuff ....................

Typical analyses, in percent

(f

Charge. in grams

G

]

Sample

Na,CO,

PbO

SiO,

15

20

50

1

10 25 20

46 35 100

20 5

Auxiliary applications: Preliminary fusion to 3 determine R.P. of sample.. ................. !'Wash" for shot in slag ...................... ... ... Litharge for Ag or Au ........................ Determination of Ag and Au in inquarts (run 2 - 4 singly) ................ 1 inquart Determination of R.P. for flour .................................. ... 1

Also add 10 g of K&O, to the charge.

CaF,

Flour

KNO,

3

1

2.8-3.0

0

3

1

0 0

15

3 10

0

0 2.9 3.0

0

10

40

15

3

1

3.0

0

60

6

0

0

2.6

0

Na,B,O,

0

0

CaF2is another flux component which may vary widely (0 -15 g) depending upon the sample being fused. As mentioned previously, it increases the fluidity of most charges. When the elemental aluminum concentrationof the ore or sample is less than 1 percent, CaF2 is not added; if the aluminum concentration is 1-10 percent, 1 g of CaF2 is added to the charge; if it is 10-20 percent, 2 g is added. For high-grade aluminum bearing samples as much as 8 g may be used. Other samples not necessarily containing high aluminum but requiring an excessive amount (4 g or more) of CaF2 are black sands, magnetite, and calcium phosphate (bone ash). For difficult samples like chromite a high borax charge in addition to an auxiliary flux is used. The main charge contains 30 g Na2C03,35 g PbO, 10 g SO2,30 g Na2B407,1 g CaF2,and 3.2 g flour plus 15 g of sample and an added noble metal (Ag or Au) that serves to collect other noble metals during cupeIlation. These componentsare mixed well in a "30-g"!ireclaycrucible. The auxiliary flux consists of 10 g Na2C03,5 g S O 2 , and 25 g Na2B407,added to and mixed well in a "20-g" fieclay crucible. After fusion of the contents of both crucibles, the melt in the '*20-g"crucible is added to the main charge. When copper, nickel, or manganese is present in sulfides in appreciable amounts (2 -5 percent), all componentsof the flux are increased. This is to increase the volume of the melt so that the constituentsof the sample will dissolve more readily. For example, in the analysis of a sample described as a sulfidebearing fissure vein, the percentages of the following elements were determined: Fe > 10; Zn > 10; Mn >.2; Pb, 3; Cu, 1; Si, 2; and Al, 0.2. By a preliminary h i o n the "reducing power" of this sample was determined to be 5.37. The first flux, used for a sphalerite, was made to contain 20 g NazC03,60 g PbO, 8 g SOz, 6 g Na2B407,and 12 g KN03for a 15-g sample. After h i o n , this flux produced an undesixable stony slag. A repeat b i o n , where all components of the flux were increased with the exception of KN03, contained 30 g Na2C03,90 g PbO,12 g Si02, 10 g Na2B407,and 11 g KN03. This flux produced an acceptable slag and lead button. The preceding example shows that a flux used for a mined such as sphalerite (ZnS) would not necessarily produce an acceptable fusion when other elements or compounds are present in sufficient amounts. Siliceous samples or ores containing 60 percent silica or more and low in the ferromagnesian silicates usually give good fusions using the following flux: 20 g Na2C03,50 g PbO, 0-3 g Si02,3-5 g Na2B407,l g CaF2, and 2.8 g flour for 15 g of sample. Samples high in aluminum content are difficult to fuse and for this reason K2C03,in addition to the other flux components, is added to the charge. The mixture of K2C03and Na2C03,lowers the fusion temperature more than would either compound alone and, therefore, allows more time for reaction to take place. From the data given in Table 3 one can also adjust the charge if a mixture of minerals is contained in one sample. For example, Table 4 lists the charges for calcite and quartz, and an adjusted charge if the sample were to contain half calcite and half quartz; all quantities are in grams. Table 4. Fire Assay Fusion Flux Charge for Various Minerals and Mineral Mixtures. Calcite

Half Calcite,

Quartz half auartz

A seen, the adjusted charge is obtained by calculatingthe difference in the amount of each flux component and dividing by two. The result is then added to the smaller figure.

1790

Fire Assay Sample Preparation In preparing the sample for analysis one must constantly be on guard to prevent contamination as well as provide a representative sample. Many samples vary in form as well as composition. However, the basic procedure for pulverizing most samples is essentially the same. For example, if the samples are received in huge chunks, they are reduced to pieces of about 5 cm with a sledge hammer and steel plate. They are then passed through a large jaw crusher which reduces the size to about 1.9 cm. Following this, the samples are passed through a small jaw crusher which reduces them to pea size about 0.5 cm or smaller. When the amount of crushed sample is large, it is next passed over a Jones splitter one or more times to obtain enough representative sample to fill a 120-ml capacity container. The 120-ml sample is passed through a vertical ceramic pulverizer that grinds the sample to about 100% passing 100 mesh. In the fre assay, 15 g of sample is usually used. The concentration of each noble metal in the sample is reported in parts per million. In countries where the metric system is used, values are reported in grams per metric ton, which is the equivalent of parts per million. The latter can be easily converted (using appropriate conversion factor) to various units for expressing concentration whether it is North American, English, or other systems. In North America, it is customary to use a factor weight of sample such that each milligram of a noble metal in the sample is equivalent to one troy ounce in one avoirdupois ton (2000 lb) of ore. As one ton contains 29,167 troy ounces, the assay ton containing 29.167 g is the factor weight normally used. Milligrams per assay ton are reported as troy ounces per short ton of ore. In England and Australia the long ton of 2240 lb is used, and the factor weight becomes 32.667 g. These factor weights are termed assay tons. Theory of The Fire Assay Crucible Fusion Most ores are by themselves infusible, but if fmely pulverized and mixed in proper proportion with flux components in a fireclay crucible the mixture will fuse at an easily attained temperature. The ore and flux components are so intimately mixed that each particle of ore is in contact with particles of the flux components. As the temperature of the mass is gradually raised, part of the litharge (PbO) is reduced to lead commencingat about 550-600° C by the carbon in the flour, as previously illustratedin equation 1, or by the sulfides innate to sample. The mist of lead droplets produced collects or alloys with the noble metals released h m the sunnunding particles of decomposed sample. Part of the PbO forms slagmiscible compounds, such as the lead silicates, which are absorbed by the slag. The conditions should be such that the slag remains viscous until the ore particles are thoroughly decomposed and every particle of the noble metals has been taken up by the adjacent suspended droplets of lead. After this point has been reached, the temperature is raised to ensure that the slag is thoroughly fluid. This condition allows the lead droplets to accrete and fall like fine raindrops through the slag to form the lead button in which the noble metals are concentrated. Fire Assay Fusion Procedure The fitmace is brought to a temperature of 1OOO"C. Using heat protective equipment, the crucibles with their contents are now placed in the furnace in rows of four using a charging fork and starting at the rear of the furnace. A space of about 0.5 cm between crucibles is maintained in the event of boil over. After the furnace door is closed, the temperature is turned down to 900°C for about 20 minutes. In the meantime, the iron molds in which the fusions are to be poured are placed on the steel table with their aluminum covers. After 20 minutes the temperature is raised to 1OOO"C for an additional 20 minutes. At the end of this time, the h a c e door is opened, and the hsion crucibles are removed singly using the crucible-scorifiertongs. After each removal, the door of the b a c e is closed by tripping the switch with one hand, being careful to clench the fusion crucible firmly with the crucible-scorifier tongs in the other hand. This is illustrated in Figure 2 (Ammen, 1984). The bottom of the fusion crucible is tapped lightly on the steel table and the melt is swirled to ensure that it is liquid and of the right consistency. The melt is then poured into the iron mold, after which the crucible is xapidly rotated vertically so that the liquid does not run down the sides of the crucible. After every two hsions are poured, the cover of the mold is advanced to prevent the solidifytng slag from

1791

ejecting. After all fusions have been poured, the surface of the furnace is raked and smoothed with hot powdered bone ash stored at the rear of the furnace. After the crucibles and melts have cooled, the crucibles are examined for shot, and the slag of the cooled melt is broken with a steel rod and hammer. The small amount of slag remaining on the lead button is removed by tapping it with the steel rod. The button is then shaped into a cube with a hammer and anvil. The comers are rounded slightly for convenience of handling. If the button appears somewhat brittle, it should not be hammered more than necessary or it may flake and crack. The buttons are marked in pencil with the last three digits of the sample number and placed on a button tray, ready for cupellation. Fire Assay Refusions Repeat fusions of a sample are conducted (usually with a different flux) when an undesirable

fusion was obtained and sufficient sample is available. When the sample is insufficient, the slag from the first fusion can be reprocessed. Indications of an unsuccessful fusion are (a) the slag contains shot or globules of lead; (b) the melt is viscous and gives a sloppy pour; (c) the charge is too siliceous as is indicated by glassy streamers, or too basic which gives a muddy pour and a stony appearance. If conditions are too basic slag is high in litharge and an insufficient amount of lead was reduced. This is recognized by its higher-than average specific gravity and even a crystalline character. There are also unsuccessful fusions where the sample has not even decomposed. Unsuccessll fusions &lly fall into one or more of the categoriesjust listed based on their pour and the nature of their slag. Knowing which flux componentsto alter and the proper amounts to use to obtain a successful h i o n may require considerable thought and experimentation. In some instances generalities may be stated, but in others the problem is unique.

No specific set of conditions are known for the cause of shot in the slag. Consequently,no generalities can be stated to avoid the production of shot. However, some hypotheses or observationsmay be or have been suggested. For example, if the melt is viscous it may cause the globules of lead to main in suspension and prevent them fiom combining to form the lead button. Or, the globules of lead may become coated and thus prevented from combining. Making the melt more fluid in the former instance may aid in the combining and collection of the lead globules. In the latter instance, the causes for coating of the lead globules require more investigation However, the addition of a considerable amount of Na2B4@(30 -35 g), as in fusing chromite ores, seems to prevent coating of the globules in many cases. When a viscous pour is evident, steps must be taken to increase its fluidity by increasing one or more flux components,frequently Na2C03or PbO or Na2B407,but sometimes Si& or CaF2. If a highly siliceous pour is indicated, the usual treatment is to increase the NazC03by 5-10 g, and when the pour is too basic, to increase the Si& and Na2B407.Both pours need additionalPbO (about 20 g). If, during the pour and after cooling, material similar to the sample appears on the side of the fireclay crucible and in the slag, the sample obviously has not decomposed This condition suggests that the proper flux may not have been selected. Further study and experimentation to select the proper flux will usually solve this problem. However, in a few instances, such as when fusing chromite concentrates, it may be necessary to decrease the size of the sample from the routine 15 g to 3 g to obtain an acceptable fusion Fire Assay Scorification Scorification may sometimes be used as a substitute for a fusion assay. It is used extensively for certain types of silver and gold assays such as bullion or recycled highgrade precious metals. In these instances it involves mixing the ore sample with lead, and covering with borax. The reactions involved are similar to the pot fusion, but there is also a series of reactions between air and the constituents of the charge. The oxidized lead

1792

forms part of the fusion mixture, and the residual lead acts as a collector. The scorification assay is not, in general, recommended for accurate noble metal determination in minerals. The process is essential, however, for the reduction of button size or for cleaning a badly contaminated button. In this instance the button is transferred to a scorifier, and covered with a few grams of borax and silica. At the required temperature of 1050-1 100°C and in the presence of air the melted lead is oxidized, and together with the borax and silica forms a slag with the oxidized base metal contaminants. The slag moves progressively to the periphery of the scorifier, and the molten lead forms the center or eye of the fusion. The melt may then be poured as in the crucible fusion. This is illustrated in Figure 3 (Ammen, 1984). Fire Assay Cupellation The lead button obtained from the crucible fusion or scorificationis treated by a process called cupellation to separate the noble metals from the lead This consists of an oxidizing fusion in a porous vessel called a cupel. The lead oxidizes rapidly to molten PbO, 98.5 percent of which is absorbed by the cupel and 1.5 percent of which is volatilized. The cupel is a shallow cup typically made of compressed bone ash (calcium phosphate) with or without a binder added. A high-grade bone ash cupel will absorb its weight in litharge. When this process has been carried to completion, the noble metals are left on the cupel in the form of a bead. The cupel itself may be regarded as a membrane permeable to molten litharge and impermeableto lead and the noble metals. First, the size of the cupel is selected based on the weight of the lead button. For buttons weighing 32 g or less, 3.8cmdiameter cupels are used, and for buttons weighing 32-48 g, 4.4-cm-diameter cupels are used, Due to the intense heat of the h a c e and for convenience of handling, no more than 18-24 buttons should be cupelled at a time. The cupel should weigh approximately one-third more than the weight of the lead button being cupeled. Cupels can be made or purchased. There are several substitutes for bone ash-such as cement and magnesia. The end product of cupellation is a small dore bullion bead composed of the less-readily oxidized metals-generally, silver and gold, but also any metals of the platinum family that were present in the collection button. The cupellation procedure is simple, and if followed step by step, good results should result with few problems. However, without care, in excess of 5% of the silver values can be lost during cupellation. Thus, cupellation must be done methodically as follows: Step 1 : Heat the cupellation hrnace to 850°C. Step 2: Set out some cupels that are sound, dry, and free from dust. Purchased cupels are usually the proper density and are sound. The cupel has to be sufficiently permeable to allow the litharge to be easily absorbed, but at the same time not be so porous as to allow the bead or lead to sink in. The surface tension of the lead and of the resulting dore bead is such that neither will penetrate into the cupel. Place the cupels in the furnace with an extra row of empty cupels between them and the door. This row is to prevent the cupels in use from the thermal variances and shocks to which they would be subjected when the door of the furnace is opened. Heat the cupels for 15 to 20 minutes with the door closed and the temperature at 850°C. Step 3: Hammer the lead fusion buttons into rough cubes or "prills". Open the door, quickly place the prills in the hot cupels, and reclose the door. Use a peep hole in the furnace door occasionally to observe what is occurring inside. Each lead cube quickly melts and spreads out. The top of the lead pool in the cupel is covered with a black layer of lead oxide, slag, dirt, and other solids that have floated to the top. In a minute or two, this covering layer flows to the edge of the cupel and starts to soak into the cupel. The metallic lead, which had been covered, is then exposed. This is called "opening up" or "uncovering." At the opening-up stage, the remaining solids are swept aside and the exposed bare surface of the lead begins to bum. The lead oxide continues to drain off and soak into the cupel, although a small percentage of the lead vaporizes.

1793

0 fusion (Door Closed)

0 Oxidizing Roast (Door Open) Gangue Molten Lead

Q Eye Closed @ Eye Open (Door Open)

(Door Closed-Superheat)

Figure 3 Seorification (These lead fumes should be avoided at all costs as lead can be an extremely dangerous substance. Fire assayers should be tested regularly for lead contamination). A cupel that fails to "open" is called a "frozen cupel," and a frozen cupel usually results from a cupel's having too thick a layer of material over the lead to get rid of it all, or from a cupel's being too cool. Thus during the entire operation the door is kept closed until all the cupels have opened. Step 4: When the cupels have all opened and the lead is being oxidized and absorbed by the bone ash, the oxidizing reaction of the lead is exothermic, and the cupel acts as an insulator. This significantly raises the temperature of the lead being oxidized, and, unless the temperature of the furnace is lowered from 85OoC,the loss of silver will be promoted. In this case, the lead becomes considerably brighter than the furnace walls. This is called "driving the cupel" and is a poor practice. Once the cupels have opened, the hrnace temperature must be lowered to a point at which the forming litharge is maintained slightly above its melting temperature. At this point, the burning lead is red, instead of yellow or dazzling yellow-white. As long as the lead and litharge are kept at a temperature at which the lead continues to bum and at which the forming litharge remains molten long enough to soak into the bone ash cupel, satisfactory results will be obtained. If some of the litharge that vaporizes condenses as dendritic crystals on the cupel above the burning lead, it is a very good sign that any silver loss will be extremely minor. These dendritic crystals are called "feathers." Step 5 : Near the end point, when the lead is nearly gone and only a thin coating is left on the dore bead, raise the temperature of the furnace to keep the molten alloy of silver, gold, and other values in a liquid state and to drive off the last bit of lead. When the end point is close, there is a play of colors in the cupel, and the bead appears to spin about the cupel wildly. At the very end, the bead may become very bright for a split second; this action is called the "blick" or "wink." The cause of this is that the liquid metal super cools at the instant the last of the lead leaves it, and, when it goes from a liquid to a solid, the latent heat raises the temperature to produce the "wink" or "blick." If the dore bead is quite large and contains much silver, the cupel should be moved toward the front of the furnace and covered with a hot cupel so that the top surface will remain liquid (not crust over) as the bead solidifies. Silver, when molten, will release any dissolved oxygen released at the moment of solidification-the familiar "spit." In cupellation, if a solid crust forms around a liquid core, as the core then solidifies, the "spit" of released oxygen

1794

remains trapped by the crust, and builds up pressure until a miniature explosion occurs. The explosion fragments the crust (an occurrence known as "sprouting"), resulting in physical loss of some of the value. Step 6: Carefully clean and weigh the bead recovered from cupellation. Figure 4 illustrates cupellation (Ammen, 1984).

Dendritic Crystals (Feathers)

Figure 4 Cupellation With the platinum metals the final stage of cupellation is not usually clearly defined. The platinum metals, if present, will display a variety of effects on the surface of the cold silver bead. It may be emphasized here that the silver bead collection and its subsequent wet treatment, when properly applied, serves as an excellent method for the determination of gold, palladium, and platinum together with traces of rhodium, iridium, and ruthenium. It is quite unacceptable for osmium. For larger proportions of the more insoluble platinum metals, and where the amount of osmium is required, the direct wet treatment of the lead button is preferable. Osmium is largely volatilized during the cupellation step. Ruthenium, rhodium, and iridium have only a limited solubility in the silver and may be partially mechanically lost unless special precautions are taken.

The Fire Assay Silver and Gold Parting Process Parting is the process of separating silver from gold by dissolution in nitric acid. In the traditional fire assay process for gold and silver, when gold is alloyed with less than three times its weight in silver, it is difficult to accomplish a good clean parting because the gold protects the silver from attack by the nitric acid. When faced with such a bead, enough 99.9% pure silver must be added to make the silver in the sample weigh over three times as much as the gold. Usually this is done in a process known as inquartation by adding the pure silver at the initial stage of crucible fusion or first cupellation. The word inquartation means one quarter. Of course deduct the weight of inquarted or added silver must be deducted from the final totals. For the classical parting process, one part nitric acid mixed with seven parts distilled water is used. This is placed in a small beaker or porcelain dish, and brought to almost boiling. The goldsilver dore bead is then dropped into the acid solution. The acid will immediately start to dissolve or 'part' the silver from the dore bead. If the bead fails to start parting, full-strength nitric acid, drop by drop, is added until the bead starts to react with the acid solution. The nitric acid used for parting must be CP reagent grade. If the nitric acid contained any chlorine, it would in reality be a mild aqua regia, putting some of the gold from the dore bead into solution and coating the bead with a precipitate of insoluble silver chloride. If the gold sponge is allowed to

1795

disintegrate into small particles it becomes a problem. Therefore, the acid solution should be hot during the entire parting operation. If the bead to be parted weighs more than a few milligrams, it should be flattened on a polished anvil with a polished hammer and rolled or flattened to about ten-thousandths of an inch thick. A very large bead might have to be annealed several times to avoid being split or cracked during flattening. This is accomplished by heating the bead (while holding it in platinum tweezers) to a red heat and dropping it into cold water between flattenings on the anvil or rolls. The metal is then rolled into a spiral or comet for parting. The resultant gold sponge should be washed with distilled water. The sponge is then annealed by heating to a red heat, at which point it will assume its gold color. Alternative Fire Assay Collection Media Besides the use of classical lead collection, other fire assay collection media have been suggested for collecting small amounts of precious metals. Some involve the use of iron-nickel-copper alloys and tin ( Faye and Inman, 1961, Whitney 1982, De Neve 1986). Because of the lower fusion temperature required, copper alone has also been suggested as a collector of the precious metals. (De Neve, 1986, Banbury and Beamish, 1965, Agrawal and Beamish,1965, Banbury and Beamish, 1966, Diamantatos, 1987). The collection of the precious metals into nickel sulfide is also being used to an increasing extent. (Robert, van Wyk, and Palmer, 1971, Kruger and van Wyk, 1972, Kallmann and Maul, 1983). This approach has many advantages over the conventional lead-based fire assay system and will be described later in more detail. A new technique involving the collection of the precious metals into copper sulfide has recently been described and has several advantages over the nickel sulfide technique (Kallmann, 1986). First, copper sulfide collects gold quantitatively. Another advantage is the possibility of dissolving the copper sulfide in hydrobromic acid, thus allowing the separation of the Precious metals from large amounts of silver and lead. In both the nickel sulfide and copper sulfide collection schemes, the base metals are removed by acid treatments and the precious metals contained in the acidinsoluble residue are eventually determined instrumentally, or where justified, gravimetrically. The fire assay procedures previously mentioned provide separations from most matrix elements. Dissolution of the lead button in nitric or perchloric acids, dissolution of nickel sulfide in hydrochloric acid and of cuprous sulfide in hydrobromic acid allow the rapid removal of the collecting media.

If a lead button has been subjected to cupellation, the silver bead containing the precious metals can be treated with dilute nitric acid to dissolve the silver and most of the palladium. Platinum will also dissolve if the sample contains at least an equivalent amount of gold. (The silver bead naturally must be large enough to hold any rhodium, ruthenium, and iridium mechanically; minimum ratio of Ag to PGM is 1OOO:l). These three elements, together with gold, remain insoluble in dilute nitric acid and can be determined instrumentally after fusion of the insoluble residue with sodium peroxide and acidification of the leached melt with hydrochloric acid. In another recently introduced fire-assay technique, cupellation is interrupted when the weight of the lead button has been reduced to less than 1 gram. This lead bead is then analyzed for the precious metals by standard optical emission spectrometry techniques (ASTM, 1982). Platinum and palladium can be determined instrumentally after removal of the silver by precipitation as the chloride. If present in sufficient quantities, platinum and palladium can be determined gravimetrically after precipitation with ammonium chloride (for Pt) or dimethylgiyoxime (for Pd). Also of interest is the selective extraction of the palladium complex with chloroform ( Fraser, Beamish and McBride, 1954). For PGM’s a flux consisting, of a 2-to-I mixture of sodium carbonate and sodium tetraborate with nickel sulfide as the collector can be utilized. The presence of at least 10 g of silica is desirable for a good fusion. The silica is usually contributed by the sample, but, where it is not, it can be added as powdered silica. With a flux-to-sample ratio of I-to-I, 20 g of nickel oxide and I0 g of sulfur, a

1796

satisfactory button is obtained at 1000°C. For chromites a suitable fusion required a 340-1 flux-tosample mixture. For samples of concentrate an increase in the flux-to-sample ratio had little or no effect on the recovery of noble metals. The quantity of flux necessary for a consistently high recovery of all the noble metals lies between 90 to 105 g; below 90 the recovery is low and an excess above 105 the recoveries are erratic. The minimum weight of button is about 25 g. Buttons much larger do not increase the recovery of noble metals. Excess of sulfur produced a grayishyellow button which sometimes disintegrated on standing. The authors of the nickel sulfide collection investigated the efficiency of the method relative to the classical lead collection and a wet separation involving selective extraction and recoveries with tellurium. The results show that the collection of noble metals with nickel sulfide is, for all the noble metals except gold, superior or equal to collection by the lead method. However, the higher results obtained for platinum by the acid-extraction procedures indicate that its recovery is incomplete by both fusion procedures. Ruthenium, osmium, and iridium are not determined by the acid-extraction procedures. A further disadvantage of the integrated nickel sulfide method is that it is time consuming compared to the lead assay. However claims for its superiority for some sample types justify its inclusion here. By comparison, using nickel sulfide as the collector has several advantages over the leadcollection method. For instance, a smaller flux to sample ratio and a lower fusion temperature (1000°C as against 1200'C) are used. Furthermore, the method is applicable to all six platinum-group metals and can be applied to samples high in nickel and sulfur without the pretreatment that is required in the lead method. No change in flux composition is required for different types of samples, except for chromite ores, where the quantity of flux used must be higher. The advantages of the lead collection procedure are that it requires less time for analysis, particularly in determinations of total platinumgroup metals, and that the gold recovery is some 10 to 20% higher than with the nickel sulfide method. Apart from the possible incomplete collection of platinum and gold, the nickel sulfide procedure offers a precise and accurate method for the concentration and isolation of the noble metals in samples of ores, concentrates, and mattes. Its applicability to the different samples thus far encountered, together with the simplicity of the technique renders it an extremely useful procedure for the analysis of the noble metals. The method is applicable to the determination of all six of the platinum group metals in ores, concentrates, and mattes. The recovery of gold appears to be incomplete, and the lead collection method is therefore preferable when gold is to be determined. Samples containing sulfur need not be roasted before the fusion, although the sulfur content must be taken into account when the required amount of flux is calculated. If the sulfur content is unknown and the determination of osmium is not required, the sample can be roasted. If difficulty is encountered in the obtaining of a satisfactory fusion and button (as has been experienced with samples containing unusual amounts of zinc), the sample can be leached in concentrated hydrochloric acid before the fusion. (This procedure results in a satisfactory fusion and button from Merensky Reef samples.) For samples containing more than 10 ppm of total noble metals, platinum, palladium, rhodium, ruthenium, and iridium can be determined in one button. If the total concentration is less than 10 ppm, a separate button must be prepared for the determination of iridium. For all samples, a separate button is prepared for the determination of osmium.

DETERMINATION OF THE PRECIOUS METALS BY INSTRUMENTAL METHODS Instrumental methods introduced during the last 20 years have undoubtedly revolutionized the repertoire of the precious metals analytical chemist. Optical emission and x-ray fluorescence methods were already used in the precious metal industry to a limited extent. About 20 years ago AAS methods made their triumphant entry, followed in rapid succession by many other techniques

1797

designed to determine the composition and the properties of the precious metals, particularly those used to an increasing extent in the electronics industry: A brief, annotated, excerpted description of these follows (Benner, Suziki, Meguro, Tanaka, 1991, Furuya, and Kallmann, 1991). As well, the determination of PGM's with instrumental analysis will be illustrated. Optical Emission Spectroscopy (OES) This technique is particularly well suited for the determination of impurities in pure precious metals. In addition, when combined with preconcentration techniques based on chemical or fire-assay preconcentration techniques, it allows the determination of platinum and palladium concentrations in complex matrixes down to 0.03 ug or 0.001 odton. In this particular procedure, 20 mg of gold is used as a collector and the F't and Pd content of the gold is compared with that of gold standards containing known quantities of the two PGM's. With minor modifications, the method can be applied to the determination of trace amounts of rhodium and iridium. One attractive feature of OES with photographic recording is its capability of providing simultaneous qualitative andor quantitative information on many elements (typically, 2040 or more). The direct current spectral source. Although the high voltage ac and interrupted do are useful for some applications, they have largely been replaced by newer techniques, such as A A S and PES. Spark-Source Mass-Spectroscopy (SSMS) Spark-source mass-spectroscopy (SSMS) is a semi quantitative technique with ultrahigh sensitivity, which has detection limits in the low ug/g and ng/g ranges. When used instead of OES for final measurements, it allows determination of PGM content of complex matrices at the 1 ng/g level. In this technique, the sample is sparked in vacuum by a high-energy radio frequency spark to produce positive ions of the sample elements. A double-focusing spectrometer separates the ions according to their "mass-to charge" first in an electrostatic, than in a strong magnetic field. The ions thus separated are recorded photographically on an ion-sensitive photographic plate or are measured by means of photo multipliers. Accelerator mass spectrometry has recently been used to determine the osmium in terrestrial samples. X-Ray Fluorescence (XRF) XRF is based on measurements of the secondary x-rays emitted by the constituents of a sample excited by primary x-rays. Two different types of XRF instruments are available, energydispersive and wavelength dispersive. Considerable progress in the instrumentation, particularly, the precision of the detection and measuring devices, allows rapid determination of the Precious metals with excellent precision. Mufti-channel spectrometers facilitate the simultaneous determination of all the precious metals and many base metals. The main advantage of XRF over the various atomic emission or absorption techniques (OES, DCP, ICP, AAS) described separately, is that it is non-destructive, allowing recovery of the original sample after the determination. Its main disadvantage is that for quantitative work standards with the same chemical composition and physical characteristics as the sample must be available or prepared. If solid samples are to be analyzed, facilities are required to prepare a set or sets of PM-bearing alloys with highly polished surfaces. Thus, one laboratory routinely analyses platinum alloys containing 5-10% of palladium andor rhodium, another determines the PGM's in a tin button obtained fire assay. A solution technique used by another organization simplifies the preparation of standards, but at the expense of sensitivity. Microgram amounts of all precious metals have been determined by absorbent-pad-and cellulose pellet techniques. Computer programs have been designed to correct for the positive or negative effects of other Precious metals or those of base metals on the result of the PM being determined. As far as sensitivity is concerned, with wavelength-dispersive instruments and using a tungsten tube as the primary source of x-rays, the radiation of the K-lines of Ru, Rh, Pd, and Ag is 2-3 times more intense than that of the L-lines of Os, Ir, Pt and Au.

1798

Flame Atomic Absorption Spectrometry (FAAS) FAAS has largely replaced spectrophotometry as the work-horse in the precious-metals analytical laboratory. There are several reasons for this. (a) FAAS is virtually element specific. Thus a PM which cannot be determined spectrophotometrically at all in the presence of certain other Precious metals or base metals, can often be determined with comparative ease by FAAS. (b) In many instances, there is no interference by moderate concentrations of base metals, and even where this is, it can be dealt with by the standard-addition technique The limitations to the use of FAAS in precious-metals analysis mainly arise from the sensitivity which is poor for some PGM's, particularly iridium. The relative concentrations of PGM required to match the AAS response of a unit concentration of silver are: Pd 4, Rh 5 , Ru 9, 0 s 100, Ir 150, Pt 38, and Au 5 . A relative value of 40 or more would indicate that the element should not be handled by FAAS (Os, Ir) unless present in substantial quantities or isolated from the matrix. It must also be remembered that FAAS is a solution technique and therefore requires that the element(s) can be dissolved with relative ease and without introducing too much extraneous matter. This, of course, is difficult, if not impossible, in the case of many samples containing the Precious metals in a complex matrix. Electrothermal Atomic Absorption Spectrometry (ETAAS) This technique supplements FAAS, in as much as it offers greatly enhanced sensitivity. Unfortunately, the precision is much poorer because of the small volume of sample solution (5-50 itl) used and the difficulties encountered in reproducible sampling and control of the atomization conditions. In addition, since ETAAS also generally relies on the preparation of solutions of samples and standards, it is subject to the same dissolution limitations as FAAS. Attempts have been made to use solid samples, but with limited success. An interesting application of ETAAS for geochemical exploration work was recently described in which diantipyrylmethane was used for isolating the PGM's by extraction into chloroform. Plasma Emission Spectrometry (PES) Plasma emission spectrometry is a variant of atomic emission spectrometry (AES), based on the use of a plasma for excitation. Plasmas are increasingly used as a spectral excitation source for determining the precious metals. Generally, the sample is introduced in the form of a solution that is atomized by the carrier gas in various fashions. Two types of plasma are used: the direct current plasma (DCP) and the inductively-coupled plasma (ICP). Lasers are also used as an excitation source. Moderately priced instruments are readily available, based on the sequential principle (measurement of one precious metal at a time). There are also more complex instruments based on the use of multi-channel detectors (ICP) or cassettes (DCP) allowing the simultaneous determination of all the Precious metals and many base metals. There is even a fast sequential DCP instrument.(99-102) PES is decidedly superior to FAAS in regard to sensitivity and dynamic range while AAS has certain advantages over the arc and spark emission methods for PM analysis. PES methods tolerate the presence of moderate amounts of alkali metal salts. This will often permit fusion of a sample or a residue with alkaline fluxes such as sodium peroxide and sodium carbonate. The linear dynamic ranges (Itg/ml) of DCP for the eight precious metals at interferencefree wavelength are: Ru 0.5-50, Rh 0.1-30, Pd 0.1-30, Ag 0.4-60,0 s 0.5-100, Ir 0.5-5, Pt 0.3-75, Au 0.3-100. The corresponding ranges of ICP are similar, but depend to some extent on the instrumentation used and the availability of interference-free wavelengths. Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) This technique has recently been introduced, but not yet hlly examined as to its suitability for precious metal determinations. In this technique ICP is used instead of the spark-source, and ions can be extracted from the plasma (which is at atmospheric pressure) and introduced (at greatly reduced pressure) into a spectrometer for mass resolution and detection. ICP-MS may be of value

1799

to laboratories having no access to SSMS. Laser Mass Spectrometers have also been built but are still in the experimental stage. Inductively Coupled Plasma-Atomic Fluorescence Spectrometry (ICP-AFS) This technique has recently been advocated as an efficient tool for precious metal analysis. In atomic fluorescence, the plasma does not function as an excitation source but solely as an atomization cell to produce ground state or low energy excited state atoms. Excitation is mainly by resonance absorption of light from an external light source, and the fluorescence emitted by return to a lower energy state is viewed at an angle to the excitation beam. The sensitivity of the method is said to be comparable to that of FES techniques. The cost of the equipment compares favorably with that of AAS. Activation Analysis (NAA) Gamma rays, charged particles and particularly neutrons react with isotopes of the precious metals to produce radioactive nuclides. The characteristic radiation emitted by the nuclides produced can be used for the detection and determination of the Precious metals. In some instances, neutron activation is more sensitive than any other technique. Though instrumental NAA (INAA) can be applied effectively to gold and silver determinations, in the case of the PGM's some separations or concentration may be necessary, either before or after irradiation. Recently, particle induced x-ray emission spectrometry (PIXE) has been developed. It has been used in conjunction with preconcentration of the Precious metals by the nickel sulfide fine-assay technique. Controlled-Potential Coulometry (CPC) In controlled-potential coulometry, the substance is electrolyzed at a working electrode with the potential controlled or kept constant during the electrolysis by means of potentiostat. The current is integrated with an electronic integrator or coulometer. A reference electrode and a two-electrode electrolysis cell are employed. This technique has been shown to be very effective for the determination of all the Precious metals. Unfortunately, it has not yet received the attention from precious metal analysts that it richly deserves. Other instrumental techniques that may be of value to the precious metals analyst for specific applications include polarography (now rarely used), differential pulse polarography, anodic stripping voltametry, ion-selective electrode potentiometry and most recently, ion chromatography. The applicability of the latter technique to PGM analysis was recently discussed by Heberlin. The compositional change on the surface of a cold-worked silver-palladium alloy has been observed by x-ray photoelectron spectrometry. Many studies on the surface composition of homogeneous alloys of gold-silver and of silver-palladium have been reported by scanning electron microscopy-energy dispersive x-ray detection (SEM-EDX), Auger electron spectrometry (AES), SIMS and XPS. Studies on the surface absorption of PGM's on the surface of catalysts have been reported by a number of different techniques, such as UPS XPS, LEEDS and AES. Since such techniques go far beyond the scope of this short review, the appropriate literature should be consulted for more details. The Instrumental Analysis of Minor PGM Levels Instrumental techniques are the mainstay of analysis for low (ppm) levels of PGM's. These methods include inductively coupled plasma (ICP) spectroscopy, direct current plasma (DCP) spectroscopy and atomic absorption spectroscopy ( De Neve and Hofmans, 1989, Homeier and Smith, 1989, Shore, 1989, Skrabak and Demers, 1988, Blumberg, 1997, Hofmans and Adnaenssens, 1993). Recently combination techniques such as ICP-mass spectrometry have been used to reached sub-ppm levels of quantification. The "plasmayytechniques are performed on

1800

solutions derived from either direct acid dissolution of a sample or from the parting solution resulting from fire assay of a sample. Atomic absorption spectroscopy, best suited for aqueous solutions, is used for the analysis of parting solutions where silver is precipitated. Many elements, including Pt and Pd, are not precipitated with the silver. Often buffers are added to increase measurement sensitivity. Direct acid dissolution requires matrix matching for the major elements present in solutions. Analysis of parting solutions requires precipitating Ag as the chloride or taking suitable precautions if the solution is analyzed directly. The accuracy of these comparative techniques is based on the use of solution standards traceable to NIST primary standards. The ability of modem instruments to scan wavelength regions and correct for background, matrix and spectral interferences increases accuracy. The use of internal standards improves the precision of the analysis. Error sources include incomplete sample dissolution, dilution problems, cross contamination of glassware, improper selection of concentration ranges and unmatched enhancement effects of sample or reagent preparation matrices. The Instrumental Analysis of Major PGM Levels Ideally suited for low-level analysis, instrumental techniques lose their advantage when concentrations reach the point where dilution and instrumental errors become large. PGM concentration ranges above 6% necessitate a different methodology to obtain the required accuracy and precision. Gravimetric wet chemical methods have traditionally been used. These methods encompass the time-honored techniques of separation, pH adjustment, precipitation, filtration, evaporation, reduction, drylng and other laboratory manipulative skills. However, X-ray fluorescence spectroscopy (XRF)has been used successfully for the analysis of many PGM alloys (Wissmann and Nordheim, 1993). The method is quick and sample preparation is minimal requiring only surface milling. The main drawback is the need for exacting, well-characterized standards. This can be prohibitive when dealing with a multitude of unknown samples. However XRF using standardless techniques based on “fundamental parameters” can serve the important function of providing fast and accurate preliminary assays necessary for the optimization of the final analysis technique (Savolainen, 1999). With standardless techniques, XRF instruments can check for all elements from Na to U. Recently, an instrumental method with the accuracy and precision of a fire assay has been developed by combining some of the best features of instrumental and gravimetric methods. The procedure incorporates mass dilutions with the internal standardized, drift corrected, multirepetitive ICP technique developed at NIST. (Salit, 2000, Salit, 2001). Precision is improved by: 1) making all dilutions by mass; 2) measuring analyte and internal standard simultaneously; 3) analyzing four separately prepared sample weighing and standards ten times with each analysis consisting of five integrations; and 4) drift correcting all results using six-figure fitted polynomials. (Salit, and Turk, 1998). Internal standards and their approximate mass fractions have been determined for all the PGM elements. High accuracy requires the use of NIST certified solutions and matrix-matched standards. The only error to be cognizant of is incomplete dissolution of the sample and the analyte. The method is robust in that there are numerous checkpoints to determine if a problem occurred. For example, if precision levels from repetitive analyses are out of range, this indicates the presence of unacceptable instrumental error. The Instrumental Analysis of High Purity PGM’s High purity PGM’s, 99% and above, are analyzed by determining the concentration levels of impurity elements. There are several instrumental techniques available including ICP or DCP, spark emission and mass spectroscopy. (Arniaud and Liabeuf, 1992). For the plasma methods, samples must be dissolved with an acid that requires proper cleaning techniques and acid blank matching. Standards are easily made from certified solution standards. Spark emission techniques require initial calibration for each of the elements and this, in turn, may require certified standards containing 35 to 40 elements to cover all likely impurities. Once an instrument is calibrated, only

1801

a check standard and recalibration sample is necessary. The analysis then becomes very fast since sample preparation consists simply of surface milling. Mass spectroscopic analysis of solid samples, although very sensitive (ppb levels) and not requiring specific standard development, is not suited for daily routine analyses because of the time involved in instrument setup. SUMMARY The analysis of silver and gold content in ores, concentrates and other materials is generally determined by classical lead based fire assay. As well, platinum group metals (PGM’s) such as platinum, palladium, rhodium, iridium, ruthenium and osmium are often separated or concentrated by fire assay. This paper outlined the basic concepts involved in this proven methodology. Included were general procedures for selecting appropriate fire assay charges based on approximate chemical compositions of the samples. Also included were descriptions of modern instrumental methods for the determination of the noble metals with a focus on the analysis of PGM’s. ACKNOWLEDGEMENTS The author is grateful for the assistanceand expertise provided by Ms. Tami Cashell in the preparation of this paper. As well, Dr. Dave Kinneberg and Mr. Arnold Savolainen are acknowledged for their technical input and review. REFERENCES Agrawal, K.C, and F.E. Beamish. 1965, Z. Anal. Chem., 21 1,265. Agricola, Georgius. 1556, De re metallica, translated by H. C. Hoover and L. H. Hoover, 1950 [reprinted from The Mining Magazine, London, 19121: New York, Dover Publications, 638 p. Amiaud, D. and N. Liabeuf 1992, “Analysis of Impurities in Palladium and Platinum Matrices by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES), RhodiudSampling and Analysis. R.C. Kaltenbach and L. Manziek, eds., International Precious Metals Institute, Allentown, PA pp. 221- 234. Ammen, C. W. 1984, Recovery and Refining of Precious Metals, Van Nostrand Reinhold, New

York. Annegam, H.J., C.C. Erasmus, J.P.F. Sellschop, and M. Tredous. 1983, “Sensitivity Amplification by Sample Preconcentmtion in Ion Beam Analysis”, Nucl. Instr. Methods; Phys. Res., 218,33. ASTM, 1982, Annual Book of ASTM Standards, Part 42- E 400,53 I . Banbury L.M., and F.E. Beamish. 1965, Z. Anal. Chem., 21 1,178 Banbury, L.M., and F.E. Beamish. 1966, Z. Anal. Chem ,218,263 Bamett, P. R., W.P. Huleatt, L.F. Rader, and A.T. Myers, A. T. 1955, ‘3pectrographicdetermination of contamination of rock samples after grinding with alumina ceramic.”, Am. Jour. Sci., v. 253, no. 2, p. 121-124. Beamish, F.E., and J.C. van Loon. 1977,Analvsis of Noble Metals, Academic Press, NY Beamish, F.E., 1966, The Analvtical Chemistrv ofNoble Metals, Pergamon Press, Oxford. Beamish, F.E., and J.C. van Loon. 1972, Recent Advances in the Analvtical Chemistrv of the Noble Metals, Pergamon Press, Oxford. Benner, L.S., Suzuki, T., M e w , K., and Tanaka, S., 1991, Precious Metals Science and Technology, International Precious Metals Institute, Allentown Pennsylvania. Blumberg, P. 1992, “Quantitative Analysis of Rhodium in Material Containing Iridium,” RhodiudSampling and Analysis, R.C. Kaltenbach and L. Manziek, eds., International Precious Metals Institute, Allentown, PA, pp. 91-96. Blumberg, P. et al., . 1997, “ICP Evaluation of Precious Metals,” Precious Metal RecoverylRefining Seminar,j. Vogt, ed., International Precious Metals Institute, Allentown, PA, pp. 101-1 12. Bright, J. translator, 1965, Jeremiah, v. 21 of The Anchor Bible: Garden City, N. Y., Doubleday & co.

1802

Bugbee, E. E., 1940, A Textbook of Fire Assaying [3rd ed.]: New York, John Wiley & Sons, 3 14 p. Clarke, F. W., and W.F. Hillebmd. 1897, Analyses of rocks, with a chapter on analytical methods, laboratory of the United States Geological Survey, 1880 to 1896: U.S. Geol. Survey Bull. 148, p. 9. Covell, D. F., 1959, “Determination of gamma-ray abundance directly from the total absorption peak.”, Anal. Chemistry, v. 3 1,p. 1785-1790. Dahood, Mitchell, S. J., translator, 1966, Psalms I: 1-50, v. 16 of The Anchor Bible: Garden City, N.Y., Doubleday & Co. De Neve, R. and H. Hofinans. 1989, “Determination of Precious Metals by Emission or Absorption Spectrometric Techniques at MHO,” Precious Metals Sampling and Analysis, S. Kallmann, ed., International Precious Metals Institute, Allentown, PA, pp. 41-52. De Neve, R., 1986, “Analysis of Precious Metals Combining Classical Collection Procedures with Modem Instrumentation.”, paper presented at 10th Annual IPMI Meeting, Lake Tahoe, NV. Diamantatos, A., 1986, Analyst, 111,2 13. Diamantatos,A., 1987,Talanta, 8,34. Egan, A. 1986, Atomic Energy of Canada Limited, private communication. Emmons, S. F., 1886, Geology and mining industry of Leadville, Colorado: US.Geol. Survey Mon. 12, 770 p. in neutron activation analytical determinations by uranium fission: Jour. RadioanalyticalChemistry, v. 10, p. 137-138. Faye and Inman, 1961,Anal. Chem., 33,278 Forbes, R. J., 1950,“Metallurgy in antiquity.”, Leiden, Netherlands, Brill, EJ., 1964, “Studies in ancient technology.”, v. VIII, 2d revised ed.: Leiden, Netherlands Fmya, K. and S. Kallmann. 1991, Chapter 5 - Determination of the Precious Metals, Precious Metals Science and Technology, L.S. Benner, et al., ed., International Precious Metals Institute, Allentown, PA, pp. 223-246. Fraser, J.G., F.E. Beamish, and W.A.E. McBride. 1954, Anal. Chem. Greenland, L. P., J.J. Rowe, and J.I. Dinnin. 1971, Application of triple coincidence counting and of tire-assay separation to the neutron-activation determination of iridium: U.S. Geol. Survey Prof Paper 750-Byp. 13175-13179. Grimaldi, F. S., and M.M. Schnepfe. 1970, Determination of iridium in mafic rocks by atomic absorption: Talanta, v. 17, p. 6 17 -62 1. H a m , Joseph, and L.B. Riley. 1968. Determinationofpalladium, platinum, and rhodium in geologic materials by fire assay and emission spectrography: Talanta, v. 15, p. 111-117. H a m , Joseph, and L.B. Riley. 1971, Suggested method for spectrochemical analysis of geologic materials by the fire-assay preconcentration-intermittent d c arc technique, in Methods for emission spectrochemicalanalysis, 6th ed.: Am. SOC.Testing and Materials, p. 1027-1031. H a m , J., L.B. Riley and W.D. Goss. 1977, A Manual on Fire Assaying and Determination of the Noble Metals in Geological Materials, US Geological Survey Bulletin 1445, Washington D.C.. Harrar, J.E. and M.C. Waggoner. 1981, Pap. Surf, 68,41. Herberlin S. ,1987, Paper presented at ZPMI Western Regional Analytical Symposium, San Jose, CA. Hillebmd, W.F., and E.T. Allen. 1905, Comparison of a wet and crucible-firemethods for the assay of gold telluride ores: U.S. Geol. Survey Bull. 253,30 p. Hofmans, D. and E. Adriaenssens. 1993, “Determination of Precious metals by ICP at UM: Advantagges and Quality Assurance,” Precious Metals 1993, R.K. Mishra, ed., International Precious Metals Institute, Allentown, Pa., pp.235-246. Homeier, E.H. and D.W. Smith. 1989, ‘‘Determination of Platinum in Catalyst Residues by Inductively Coupled Plasma Atomic Emission Spectroscopy,” Precious Metals Sampling and Analysis, S. Kallmann, ed., International Precious Metals Institute, Allentown, PA, pp. 67-75. Huffman, Claude, Jr., J.D. Mensik, and L.F. Fbder. 1966, “Determination of silver in mineralized rocks by atomic-absorption spectrophotometrf‘, U.S. Geol. Survey Prof Paper 550-B, p. 13189-13 191.

1803

Huffman, Claude, Jr., J.D. Mensik, and L.B. Riley. 1967, “Determinationof gold in geologic materials by solvent extraction and atomic-absoxption spectrometry”, U.S. Geol. Survey Circ. 544,6 p.Koda, Y., 1970, Determination of radioruthenium using a polyethylene film: Jour. Radioanalytical Chemistry, v. 6, p. 345-357. Kallmann, S. 1986, “Sampling and Analysis of Spent Automotive Catalyst,” Platinum Group Metals Seminar 1985, E.D. Zysk, ed., International Precious Metals Institute, Allentown, PA, pp. 233240. Kallmann, S. 1983,” Interdependence of Instrumentaland Classical Chemical Methods of Analysis of Precious Metals”, paper presented at Eastern Analytical Symposium, New York, NY. Kallmann, S. and C. Maul. 1983, Talanta, 20,2. Kallmann, S., 1986, Talanta, 22,75. Kruger, M.M. and E. van Wyk. 1972, Nat. Inst. Met.Rep. S. Afi. Rept., 1432. Kudo, M., Y. Nihei, T. Machiyama, K. Furuya, and H. Kamada. 1977, “Some Problems on Quantitative Surface Analysis of Copper-Nickel and Palladium-Silver Alloys by Means of XRay Photoelectron Spectroscopy(XPS-ESCA)”, Bunseki Kagaku, 26,173. Lancione, R. 1983, ‘”recious Metal Analysis by Inductively Coupled Plasma-Atomic Fluorescence Spectrometry,paper presented at Eastern Analytical Symposium, NY. Lenahan, W.C. and R. de L. Murry-Smith, 2000, Chapter XVIII - The Determinationof the Platinum Group Metals, Assay and Analytical h c t i c e in the South African Mining Industry, Chamber of Mines of South Africa, Johannesburg,pp. 43 1-505. McLaughlin, R E., 1992, “Modifications of Tradituional Methods in the Analysis of Precious Metals,” Rhodium/Samplingand Analysis, R.C. Kaltenbach and L. Manziek, eds., International Precious Metals Institute, Allentown, PA, pp. 235-244. Millard, H. T., Jr., and A.J. Bartel. 1971, “A neutron activation analysis procedure for the determination of the noble metals in geological samples.”, in A. 0. Brunfelt and E. 0. Steinnes, eds., Activation analysis in geochemistry and cosmochemistry, NATO Advanced Study Inst., I1 jeller, Norway, Sept. 7-12, 1970, Proc.: OsloBergen-Tromso, Norway, Universitetsforlaget, p. 353-,358. Moreland, John, and A.T. Myers. 1973, Notes on use and maintenance of vertical pulverizers for geologic materials: U.S. Geol. Survey open-file report, 6 p. Myers, A. T., and RG. Havens. 1970, ‘“spectmchemisfryapplied to geology and geochemistry by the US. Geological Survey in the Rocky Mountain region.” in Proceedings of the second seminar on geochemical prospecting methods and techniques, Ceylon, 1970: U.N. ECAFE Mined Resources Devel. Ser. 38, p. 286-291 Robert, R.V.D., E. van Wyk, andR. Palmer. 1971, Nat. Inst. Met. Rep. S. Afr. Rept., 1371. Rowe, J. J. 1973, “Determination of gold in phosphates by activation analysis using epithermal neutrons”, U.S. Geol. Survey, Jour. Research, v. 1, no. 1, p. 79-80. Rowe, J. J., and F.O. Simon. 1968, “The determination of gold in geologic materials by neutronactivation analysis using fire assay for the radiochemical separations.”, US. Geol. Survey Circ. 599,4 p. Salit, M.L., et al. 2001, “Single-Element Solution Comparisonswith a High-Performance Inductively Coupled Plasma Optical Emission Spectrometric Method,” Analytical Chemistry, Vol. 73, No. 20, pp. 482 14829. Salit, M.L., et al. 2000, “An ICP-OES Method with 0.2% Expanded Uncertainties for the Characterizationof LiAl02,” Analytical Chemistry, Vol. 72, No. 15, pp. 3504-35 1 1. Salit, M.L. and G.C. Turk. 1998, “A drift Corntion Procedure,” Analytical Chemistry, Vol. 70, No. 15, pp.3 184-3 190. Savolainen, A.M., et al. 1999, “Preliminary Analysis and Method Optimization Based on X-ray Fluorescence - Application to Precious Metals,” Analytical Technologies in the Mineral Industries, Cabri, L.J., et al., eds., The Mineds, Metals and Materials Society, pp 55-71.

1804

Shepard and Dietrich. 1940, Fire Assaying, McGraw-Hill Book Company, Inc. New York. Shore, L. 1992, ‘The ICP Determination of Impurities in Rhodium,” Rhodiudsampling and Analysis, R.C. Kaltenbach and L. Manziek, eds., International Precious Metals I h t u t e , Allentown, PA, pp. 97-120. Shore, L. 1989, “ICP vs. DCP More Than Just a Coin Toss,” Precious Metals Sampling and Analysis, S. Kallmann, ed., International Precious Metals Institute, Allentown, PA, pp- 115-143 Skrabak, J.W. and D.R. Demers. 1988 ,“Applications of ICP-Atomic Fluorescence Spectrometry to the Analysis of Precious Metals,” Precious Metals1988, R.M. Nadkarni, ed., International Precious Metals Institute, Allentown, PA, pp.559-570. Smith, E.A. 1987, The Sampling and Assay of the Precious Metals, 2nd ed. (reprint), Met-Chem Research Inc., Boulder, COYpp. 408424. Wissmann, F. and U. Nordheim. 1993, ‘Trecise Quantification of Precious Metals by ICP - A comparison of SophisticatedICP Methods Versus Classical Wet Chemical Analysis with regard to Accuracy, Ease of Use and Cost-Effectiveness.”, Precious Metals 1993, R.K. Mishra, ed., International Precious Metals Institute, Allentown, PA, pp. 117-130. Whitney, J. 1982, “X-Ray Fluorescence Analysis of Precious Metal Concentrates Melted with Tin”, paper presented at Eastern Analytical Symposium,NY.

1805

Management of Tailings Disposal on Land Bruce S. Brown. PhD., P.E. Principal, Knight Piesold Consulting

INTRODUCTION The management of the disposal of mill tailings is of critical importance to the success of any mining project. Failures of tailings facilities have resulted in loss of life, devastating environmental damage, the closure of mining operations, dramatic declines in share value and, in some countries, the personal liability of the mine management. The issues surrounding the development of tailings facilities are therefore often the main focus of regulatory scrutiny during the permitting process for a mine development. Successful management of tailings disposal requires a good understanding of the complete life cycle of a tailings facility. Only in this context can decisions be made that will result in overall minimization of risk for all stakeholders. This paper is designed to give people who are responsible for the management of the development ,operation and/or closure of tailings facilities an understanding of what is required for each step in the lifecycle of a tailings facility. Life Cycle of a Tailings Facility The typical life cycle of a tailings facility starts with the recognition of the need for a tailings facility for a new or existing mining operation. The steps in the cycle will include most or all of the following: Identification of all regulatory requirements and laws governing the design, operation and closure of a tailings facility. Definition of the quantity and physical and chemical characteristicsof the tailings to be stored. Siting study. Environmental baseline studies. Scoping level design for prefeasibility. Preliminary design for feasibility. Environmentalimpact assessment and permitting. Detailed design, construction drawings and specifications. Regulatory review. Construction of Stage 1 of the facility. Startup and commissioning. Operation and monitoring. Ongoing staged construction. Safety reviews and risk management. Closure and decommissioning. Reclamation. Post closure monitoring. Each step in the life cycle is closely related to the others and decisions made in the early stages of the life cycle will have a profound effecd on the options available in the later stages. The term “designing for closure’’relates to this whereby the options for closure are fully considered in the initial stages of development.

1809

Regulatory Framework The development, operation and closure of tailings facilities are, to some extent, regulated in most areas of the world. In the USA, regulation is generally carried out at the state level and in some instances at the county level. Regulation is typically prescriptive whereby specific design features are mandatory for given waste types. In Canada, the regulatory process has both provincial and and federal components. In countries that have little or no enforceable regulations covering the design and construction of tailings facilities, World Bank standards are commonly used as a basis for design. The first step in the development of tailings facility is to determine all regulatory requirements and laws that are applicable. In some countries, there are no such regulations or the regulations that exist are not enforced. In these cases it is often the guidelines and standards of the lending agencies that finance the project that are applied. Notwithstanding the existence or lack of regulations in a given jurisdiction, responsible mining companies apply appropriate measures to international standards. Some developed countries have laws that require companies to apply as a minimum, the standards and regulations of the home country when mining in other countries.

DEVELOPMENT Site Selection The selection of the most appropriate site for a tailings facility requires the evaluation of many and often conflicting criteria. In addition to determining the “best’ site, there is often a regulatory requirement that it be demonstrated that all potential sites have been fully considered. A formal site selection process is often the best way to do this. A desk study of regional topographic mapping is the first step where areas that can reasonably contain the required volume of tailings are identified. An outline for a facility is laid out at each site to the level that the major dimensions and bulk quantities of major works can be determined. The outline should also delineate the impacted areas, the hydrological catchment areas, and show the relationship of the site to the mining project as a whole. The selection of a preferred site or combination of sites for one or more mine facilities is often accomplished using various decision-making tools and presentation techniques. After the initial identification of alternative sites, a fatal flaw or exclusion criteria assessment is often conducted to eliminate those alternatives with clear feasibility problems, such as a major fault located through the facility footprint, a shortage of available water, or an alternative that cannot be permitted (e.g. submarine tailing disposal). The remaining feasible alternatives may then be compared in physical terms, including, for instance, impacts on the surrounding environment, required construction materials, and pumping and hauling distances and elevations. At this stage, appropriate decision criteria may be selected and categorized in major accounts such as ‘environmental’, ‘engineering’ or ‘economic’ and subsequently weighted to allow the numerical ranking of the feasible alternatives in a decision matrix. From the selection of preferred alternatives, combinations of preferred mine components are identified, resulting in a limited number of feasible, overall mine development plans. Overall mine development plans, comprising, for example, preferred combinations of open pits, waste rock storage areas and tailings storage facilities may then.be compared using a larger scoped multiple accounts evaluation method. In addition to the physical considerations discussed above, a multiple accounts evaluation will consider non-physical accounts, sub-accounts and indicators, such as socioeconomic impacts, aboriginal issues, financial or economic analyses and the comparison of project risks to arrive at a defensible and balanced project development plan for permitting and construction. Site Characterization Climate and Hydrology. The climate and hydrology of a region are strongly linked, so it is important to understand both topics in order to fully appreciate the water management challenges facing any particular mine development. In this context, climate generally refers to patterns of temperature and atmospheric moisture, while hydrology refers to patterns of surface

1810

runoff. The total amount and timing of the precipitation runoff at a site dictates many aspects of a mine’s tailings disposal system, ranging from environmental concerns such as what level of impact the tailings disposal might have on background water quality, to engineering concerns such as how much live storage is required in a tailings pond or whether or not diversions or a spillway are needed, and if so, how large they should it be. Typically, little or no site-specific climatic and hydrologic data is available for most mine development sites, particularly at the time of initial design. In most instances, a data collection program is initiated during project feasibility and environmental impact studies, but to a large degree, these data sets are very short term and of limited use at the time of project design. As a result, the hydrometeorologic characterization of a project area is generally conducted on the basis of regional data sets, which are extrapolated to the project site according to known or suspected weather patterns, similarities of watershed characteristics, and an understanding of the fundamentals of hydrometeorologic systems, including lapse rates, orographic effects and rainfallhnoff mechanisms. The hydrometeorologic characterization of a site typically involves estimates of temperature, evaporation, precipitation (rain and snow) and runoff, on both average monthly and annual bases, including some measure of variation. Also required are values of extreme precipitation and flow, which are usually presented in terms of likelihood of occurrence. In addition, in colder climates, patterns of snow accumulation and melt are needed.

Regional Geology. The regional geology identifies the bedrock geology and structural setting in the vicinity and under the tailings facility site. The findings from this element of site selection will classify large scale features. Specific site investigations will be carried out based on the regional features to accurately map the features on a more concentrated scale. The location of these large features is a very important component of site characterization. The type of bedrock or presence of regional faulting greatly affects the location and type of tailings facility to be built. Regional geology maps can usually be obtained from various government agencies. Generally, large maps in the order of one to a million scale are easily obtained. The availability of smaller scale maps, such as one to two hundred and fifty thousand or one to one hundred thousand, is much more variable due to the economical, political and geological characteristics of the area in question. Many industrialized countries have government geology departments at the federal and provincial or state level, while many developing countries only have limited geological data at the federal level. In addition, the size and population of the country, state or province may control the map scale. For example, Canada is a large, industrialized country; however, the scale and accuracy of regional mapping is large and incomplete in many areas because these areas are vast and the populations are small. Another source that is becoming useful for gathering regional information is the internet. Many countries have maps in digital form for viewing and printing. Once regional maps have been obtained, it is recommended that a qualified person carry out field reconnaissance to verify the mapping. Smaller scale mapping should be carried out if these maps do not exist or there is a complicated geological or structural setting at the proposed site. It is very important to identify weak bedrock geology units or active regional faults at this stage for site characterization. These identifications will aid designers and form the framework for detailed site investigations. Terrain Analysis. A terrain analysis is carried out to assist in characterizing a tailings facility site. This analysis is designed to identify terrain units and natural hazards around the site that could impact the tailings facility or the tailings deposit and to identify potential materials for construction. The terrain analysis will not necessarily accurately determine the characteristics in the footprint of the facility but is a useful tool for defining where detailed site investigations are required. There are two levels of terrain analysis. The f ist level comprises air photo interpretation and the second level involves ground reconnaissance to verify the observations made by air photo interpretation. The detail of each of these two levels of study is based on the following 14 main terrain attributes: bedrock geology, quaternary geology, geomorphology, weathering, erosion and

1811

deposition, climate, vegetation and pedology, hydrogeology, geotechnics, volcanic activity, neotectonics and seismicity, natural dams, human activity and land use. This list is exhaustive and many of these attributes will not be present at a given tailings facility site. However, it is prudent to check for all these attributes initially to determine potential natural hazards. In areas of steep, wet terrain, it is recommended that 1:15,000 scale air photographs be used for the first level of reconnaissance. Steep terrain is generally characterized with significant portions of the area having slopes greater than 50% or 27". If drier, flatter terrain is present in the vicinity of the site, 1:40,000 scale air photographs are adequate. If reconnaissance flights and air photo production can be acquired, this exercise can be completed in conjunction with topographic surveys. The second level comprised ground proofing of the air photo interpretation. The requirements for ground reconnaissance verification of the air photo interpretation is based on the terrain attributes identified in the fust stage. Generally, areas where the slopes are greater than 50% require foot traverses, while areas of flatter terrain is completed by ground checks supported by helicopter or vehicle. Unstable or potentially unstable areas identified during the air photo interpretation must be checked in detail by foot traverses and 1:5000 scale mapping. The person who carried out the air photo interpretation should complete the field checking. The results of the terrain assessment should be made into a terrain hazard map, outlining areas of potential natural hazards such as landslides, debris flows and snow avalanches. This information can be used for risklfatal flaw analyses associated with the tailings facility and to quantify safety concerns during construction and operation. In addition, the information from the analysis can be used to delineate possible local borrow materials for embankment construction.

Hazard Classification. The hazard classification of a tailings facility is required to characterize potential hazards and the consequences of failure of the tailings facility. This enables appropriate design parameters to be selected, including design earthquake events and design flood events. For new facilities, a hazard classification should be carried out during feasibility design studies and confirmed for final design. In the United States there is not just one set of guidelines in use. The hazard classification guidelines adopted for a project will depend on the project location and maybe also the owner. Typically, each state has an agency that regulates most of the dams (water-retaining and tailings dams) within its jurisdiction, but various other agencies oversee a significant number of projects. These include the US Army Corps of Engineers and the US Bureau of Reclamation. In Canada, the hazard classification for a tailings dam is typically carried out using the Dam Safety Guidelines (1999) of the Canadian Dam Association. Internationally, the International Commission on Large Dams provide similar guidelines (ICOLD 1989). A hazard classification is carried out by assessing the hazard potential and consequences of failure of a dam. Several factors are considered, and typically include the size (storage capacity) of the tailings facility, height of dam, the potential economic loss, environmental impact and potential loss of life that would result from failure. Once a hazard rating has been assigned to a tailings dam, appropriate design earthquakes and flood events are selected. A high hazard classification requires more stringent design criteria. For seismic design of a tailings facility, two levels of design earthquake are generally considered: the Operating Basis Earthquake (OBE) and the Maximum Design Earthquake (MDE). The OBE is typically determined using probabilistic seismic hazard methods, and represents the level of earthquake shaking at the dam site for which only minor damage is acceptable. A tailings dam is expected to function normally after an OBE. The MDE represents the maximum ground shakmg for which the dam is designed. For large tailings dams classified with a high hazard, the MDE is often characterized as an event corresponding to the Maximum Credible Earthquake. Damage to a tailings dam is acceptable from the MDE, provided the integrity and stability of the dam is maintained and the release of impounded tailings is prevented (ICOLD, 1995). A similar approach is used to determine the design flood for a tailings facility. The resulting design flood will range from 1 in 50 year recurrence period to the Probable Maximum Flood depending on hazard rating and size of dam.

1812

Seismicity. The seismic stability of a tailings dam is an important component in the design and operation of a tailings facility. In regions of high or even moderate seismicity, it is often the seismic loading that controls dam stability. Consequently, a seismicity review of the region where the project is located should be carried out during the initial stages of project development. Initial studies at the conceptual or prefeasibility level may only include a review of existing information regarding the regional seismicity, and preliminary seismic design parameters may be obtained from seismic hazard maps, if available, for the region. However, for the feasibility and final design stages of a tailings facility, more sophisticated methods of analysis are required. These typically include both deterministic and probabilistic methods of seismic risk analysis. A probabilistic analysis is carried out to define a unique probability of occurrence for each possible level of ground acceleration experienced at a site. The methodology used for the probabilistic analysis is based on that presented by Cornell (1968). The likelihood of occurrence of earthquakes within defined seismic source zones is determined by examining seismicity data. Using historical earthquake records for the region, magnitude-frequency recurrence relationships are established for each potential earthquake source or fault zone. The seismic design parameters selected for the design of a tailings facility are dependent on the level of seismicity in the region and the geologic and tectonic conditions at and in the vicinity of the site. Unlike the probabilistic analysis, the deterministic method does not account for the likelihood of occurrence of a predicted ground acceleration. Seismic source zones or fault systems in the region are defined and maximum earthquake magnitudes assigned to each source. The resulting deterministic acceleration at the study site for each source is considered to be the maximum acceleration that can occur, on the basis of geologic and tectonic information. The maximum acceleration produced by this procedure is referred to as the maximum credible acceleration and the corresponding earthquake as the Maximum Credible Earthquake (MCE). The MCE is defined as “the largest reasonably conceivable earthquake that appears possible along a recognized fault or within a geographically defined tectonic province, under the presently known or presumed tectonic framework” (ICOLD, 1989). Site Investigations. Geotechnical site investigations for a tailings facility need to be carefully planned with an initial design concept in mind, and should be conducted to an increasing level of detail as the facility moves from scoping level to preliminary and on to detailed design. Further, site investigations should be planned to investigate the geologic interpretation of the site developed during the site characterization phase. The key objectives of site investigations for tailings facilities are: Confirm any potential natural hazards identified during the terrain analysis phase. Characterize the foundation materials through sampling and laboratory index tests such as particle size distribution and plasticity. Characterize the existing groundwater conditions through dnlling investigations including in-situ hydro-geologic properties of the foundation soils such as permeability. Determine the geotechnical properties of the foundation soils, such as shear strength and compressibility. Confirm the availability and characteristics of the earth or rock fill materials required to construct the facility to the proposed design concept, including mine waste materials from open pit development. Typically test pits, drill holes and seismic refraction surveys are sufficient tools for site investigations for detailed design of a facility. Test pits are used to investigate potential construction material borrow sources and the shallow foundation conditions for the tailings dam and the basin. Hydraulic excavators with minimum 5 m (16.5 ft) depth of excavation are typically used to excavate test pits. Test pits alone may be sufficient for scoping level designs.

1813

Drilling investigations of the foundation are required for design of the tailings embankment. A properly planned and executed site investigation of the foundation soil or rock includes careful selection of the drilling method@) to ensure the necessary data is collected. A geotechnical engineer needs to supervise all drilling activities, to ensure sampling and testing are completed at appropriate locations. Electronic piezo-cone penetration tests (CPTU) are very efficient at estimating in-situ soil properties in sands, silts and clays, but cannot be pushed into gravelly material. CPTU's can extend the data from more traditional drilling, sampling and Standard Penetration Testing (SPT). For gravelly materials, the Becker Drill is a common method that provides samples and penetration data correlated to the SPT. Seismic refraction surveys include seismic lines that provide vertical profiles of depth to harder foundation soils or bedrock. Seismic refraction with both compression and shear wave velocity measurements are particularly valuable for design in seismically active areas. Samples collected from the investigations are tested in a soils laboratory for index, strength and compressibility characteristics, and undisturbed samples of foundation materials are recommended for the latter two types of tests. For the materials from potential borrow areas, index tests must include compaction testing to confrm their suitability for placement in the tailings dam, basin liner or drainage systems. Strength and compressibility tests are also required for the fill materials, and should be performed on remolded samples compacted to the density criteria that will be required by the construction specifications. Waste Characterization. Characterization of both the solid and solution portion of the mill tailings is an essential step in the development of the tailings management plan. The acid generating potential and metal leaching characteristics of the solid tailings mass and the chemistry of the liquid effluent will effect the design of the tailings containment facility and be important considerations in solution management and reclamation planning. Acid rock drainage (ARD), metal leaching, and contaminantlmetal release from mine tailing facilities are recognized environmental concerns. To ensure that natural aquatic systems are not significantly degraded, it is important to fully understand the tailings material by conducting a thorough waste characterizationprogram. A phased approach to waste characterization is often a prudent way to proceed. The relatively inexpensive static tests can be performed on a large number of tailings solids samples, with the results of the initial testing used to plan for additional, more detailed testing requirements. This stage of the characterization program will include testing to determine the trace element content of the waste, acid base accounting (ABA) or equivalent to determine the relative balance of potentially acid generating and potentially acid consuming minerals, and leach extraction testing to measure the soluble components of the samples. Whole rock analysis and mineralogical descriptions may also be conducted. The trace element testing will usually consist of a full suite of metal analyses (ICP-MS/ICPES). The trace element concentrations will indicate which metals are naturally high in the waste and may be a concern for future leaching. Acid base accounting, or the determination of the relative amounts of acid generating and acid neutralizing minerals in a sample, can be accomplished through a number of test procedures, with the Sobek Acid Base Account Test and the Modified Acid Base Account Test being the most common. These tests measure the acid potential (AP),also called the maximum potential acidity (MPA), of the sample based on its sulfur or sulfide content. The samples neutralizing potential (NP) is determined by titrating a pulverized sample of the material with an acid, with the resulting NP representing the acid neutralizing capacity of the sample. The samples net neutralizing potential (NNP = NP - AP),ratio of NP to AP, and paste pH are also determined. The results of the ABA testing are compared to general guidelines to assess whether the samples are likely to generate acid. Guidelines may vary for different areas but are generally based upon the net neutralization potential ("P) and the ratio between neutralization potential and acid potential (NP/AP).

1814

Leach extraction testing measures the soluble component of the sample. Some commonly used tests include the B.C. Special Waste Extraction Procedure (SWEP), the U S . EPA 1312 procedure, and the U S . EPA TCLP leach test. Leach extraction test results will determine the leaching potential of the waste and highlight those metals that are likely to leach from the tailings under neutral pH conditions. Static testing provides valuable insight into the acid generating and metal leaching potential of the waste. Static testing does not, however, provide information about the rate of acid generation or neutralization and should not be used to predict drainage water quality in the field. The next step in the characterization process for the tailings solids is kinetic testing. If static testing results show high variability between samples a relatively high number of samples should be considered for kinetic testing, while if samples were largely consistent, fewer samples are needed for the kinetic testing phase. Kinetic testing is used to c o n f m the samples’ acid generating or neutralizing characteristics while determining the rates of reaction for acid generation and neutralization. Kinetic tests are most often conducted in a laboratory, where chemical weathering is simulated over time in cells or columns. The composition of the drainage collected from the test containers can be used to predict drainage water quality in the field if concentrations are corrected for the effects of temperature, flushing, and particle size. The test container drainage is analyzed for total and dissolved metals (ICP-MS), conductivity, total sulfur, sulfate, sulfide, and pH. Kinetic testing procedures include humidity cell tests, humidity column tests, column leach tests, Soxhlet extraction test, and field plot tests. It is equally important to characterize the tailings solution or the aqueous portion of the tailings slurry. A sample of the tailings solution from the pilot test work should be analyzed for a range of parameters including total and dissolved metals (ICP-MS), nutrients, and reagents and reagent by-products used in the process. For example, if cyanide is used in the process, the effluent should be tested for the full range of cyanide species (total cyanide, free cyanide, WAD cyanide, strong acid dissociable cyanide, cyanate, thiocyanate). The speciation of any metals of concern should be determined since the species of metal present has a significant effect on the metal’s availability and toxicity. Turbidity and suspended solids are also measured. Effluent concentrations are compared to effluent guidelines such as Canada’s Metal Mining Liquid Effluent Regulations (MMLER) or the U.S. EPA Effluent Guidelines. A toxicity test of the effluent is also often required to ensure that the effluent is not acutely toxic to aquatic life. The toxicity tests are commonly performed on Duphniu or rainbow trout. Effluent concentrations, coupled with the site hydrology, can also be used to perform water quality modeling exercises if discharge to nearby waterways is considered. Predicted water quality concentrations are then compared to regional water quality guidelines or criteria for the protection of aquatic life. Design Options Conventional Tailings. Tailings produced by most conventional milling processes comprise a slurry of solids and solution with a solids content ranging from -30-50% depending on the process. The conventional process of depositing the tailings is to transport the slurry by pipeline to the tailings facility and to discharge it into the tailings facility . The slurry is deposited on beaches or into a pond where the solids settle leaving supernatant process water that collects in a supernatant pond. The supernatant water, along with rainfall runoff that collects in the pond is usually recycled back to the mill for reuse in the process. Thickened Tailings. The discharge of thickened tailings is a concept that has been in use for over 20 years in the mining industry. Thickened tailings disposal requires dewatering of the tailings slurry to about 50 to 60 percent solids, at which the tailings behave more like a highly viscous fluid, compared to a liquid slurry. The main advantage of this disposal method is that impoundment embankments may be significantly reduced in size or even eliminated, due to the reduced requirements to handle supernatant water.

1815

Paste Tailings. The term Paste Tailings is generally used to describe a tailings slurry that does not segregate during mixing, transportation or placement and has a working consistency similar to wet concrete. Paste tailings can be transported through a pipeline but, unlike a slurry, the paste has no critical flow velocity below which solids settle out in the pipe. In recent years there have been significant advances in dewatering technology and a number of processes are now available to produce paste tailings. These include conventional thickenedfilter methods and the Paste Production Storage Mechanism (PPSM). The amount of water required to maintain a paste consistency depends on the type of tailings. Generally, as the fines content of the solids fraction increases, the amount of water retained by the paste also increases. Paste tailings typically consist of approximately 60 percent solids for a fine tailings, up to about 80 percent for a coarse tailings material. The properties of a tailings paste can be modified as required, by adding materials such as Portland cement and fly ash to alter chemistry and strength characteristics. For above ground disposal this may be useful for stabilization of a paste tailings berm, erosion protection or reclamation of the tailings surface by providing improved trafficability. Dewatered tailings have been used as backfill in underground mining for well over a century. The main uses of dewatered tailings in underground mining have been and still are for stope support, to provide a working floor and for waste disposal. The design of paste backfill using dewatered tailings requires an understanding of the physical-chemical as well as the mechanical properties of the tailings materials. Backfill design requires the determination of the fill composition and water requirements for fill preparation to produce an acceptable, cost-effective mix. Compressive strength and permeability are recognized as the most important mechanical properties. Backfill strength is commonly enhanced for high early strength requirements by additives such as Portland cement, ground slag or fly ash. Chemical additives to improve permeability include flocculants, accelerators and retarders. The most commonly used forms of backfill are high density slurries and paste fills which can be tailored for high early strength with reduced cement consumption. The design of a successful backfill operation, based on dewatered tailings, requires the assessment of backfill preparation, storage and transport, bulkheads, quality control, fill material preparation, placement methods, maintenance, labour and void preparation. Backfill pilot plant testing is commonly initiated for trials prior to the implementation of full-scale operations. Full production systems can often be designed utilizing most of the components of an ordinary run-ofmill tailings disposal system. Paste tailings have been used as underground mine backfill for a number of years. The surface disposal of paste tailings is, however, a relatively new concept which only recently has been considered as a viable alternative to conventional tailings disposal. Dewatering of tailings reduces or eliminates the initial consolidation phase of a slurry. In certain cases, dewatering prior to surface disposal may offer advantages in terms of material handling, placement strategy and environmental impact when compared to conventional tailings disposal. These advantages could-includethe following: Decreased time required for the tailings mass to achieve its ultimate density and volume due to the higher initial solids content. Reduced storage requirements due to more rapid consolidation and reduced volumes of supernatant water release. Reduced seepage water from ongoing tailings consolidation. Reduced dam construction requirements (smaller embankments and reduced fill volumes). Reduced geotechnical hazards associated with containment structures due to improved material stability and strength and reduced excess pore pressures. Reduced short and long-term environmental liability due to reduced seepage. Increased operational and design flexibility for tailings facilities due to improved handling and storage characteristics.

1816

Dry tailings. For “dry” tailings the moisture content of the tailings is typically reduced by filtration methods to about 20 to 30 percent of the dry solids weight. At these low moisture contents, the tailings no longer exhibit the flow characteristics of a slurry or paste and form a “dry cake” material. This can usually be handled using conventional earth moving equipment and transported by truck or conveyor system. New technology is being developed that may allow the transportation of the “dry” tailings by froth transport in pipelines. Current filtering technology allows large tonnages of tailings to be dewatered more efficiently and economically.

Co-disposal of tailings with other mine wastes. In some cases other mine waste products such as waste rock can be combined with dewatered tailings for co-disposal within the same tailings facility. For paste and dry tailings materials a composite material can be created with the waste rock for transport to the facility and placement. Potential advantages of the co-disposal of dewatered tailings and waste rock include: stability of the composite material when compared to the tailings alone Decreased storage volume with the tailings filling the voids in the waste rock. Decreased ARD potential when waste rock is encapsulated in paste tailings. Reduced disturbance area for waste storage. Materials for the construction of containment structures. There are many types of structures built to contain the tailings solids and solutions and to protect the surrounding environment from contamination. The structures or embankments are classified by the materials used to construct them and by the shape of the embankment cross section. The use of waste products from the mining process including waste rock from the mine and the tailings from the milling process in the construction of the embankment can be economically and environmentally beneficial. For example, waste rock can be directly hauled from the mine to the embankment for the cost of the overhaul. This can often be done at a relatively low cost. Tailings can be processed using hydro cyclones or controlled spigotting to classify the tailings into a course and fine fraction. The course fraction can be used for embankment construction, again at a relatively low cost. The use of waste products for embankment construction can reduce the need for excavating natural construction materials and thereby reduce the overall area of land disturbance. It is necessary to fully characterize the waste materials before using them in embankment construction. Mineralized waste rock may be subject to oxidation and release of metals that could cause environmental impacts. Oxidation can also lead to mechanical breakdown of rock particles that can result in a loss of strength. The course fraction of tailings can also be subject to oxidation and acid drainage and can be subject to water and wind erosion. The alternative to using waste materials is the use of borrowed natural materials. The availability of different types of materials such as low permeability soils, sand and gravel, durable rock, etc., dictates the design of the containment structures. Excavation of these materials from within the storage basin of the facility can have the added advantage of increasing the storage capacity of the facility. Embankment types. There are four basic types of embankments shown on Figure 1. These are as follows: Downstream Upstream Centerline Modified Centerline The downstream embankment is constructed in stages so that the centerline of the embankment crest moves downstream with each stage. This embankment type uses the most construction materials and is therefore the most expensive option for embankment design. The downstream embankment is completely independent of the physical properties of the tailings

1817

deposit. There are some jurisdictions such as California where this is the only option available for embankment design due to dam safety regulations. The upstream embankment is constructed in stages where each stage is constructed on the tailings beach immediately upstream of the previous stage. The centerline of the embankment crest therefore moves upstream with the construction of each stage. This embankment type uses the least amount of construction materials and is therefore the least cost option for embankment design. In the past, a large proportion of tailings facilities have been constructed using the upstream method combined with spigotting of tailings to produce a course fraction for embankment construction. There have been some major failures of embankments constructed using these methods. The failures have in the most part been due to a lack of drainage in the embankment section resulting in high water levels. Earthquakes have caused liquefaction of the saturated tailings and resulting loss of strength has caused flow slides or large scale deformations and loss of freeboard. These failures have led to the complete ban of upstream construction methods in some jurisdictions, most notably, ChilC.

It is interesting to note that no upstream dam has failed that has been rigorously designed using modern engineering principles to ensure that the embankment is adequately drained and the phreatic surface is controlled. The centerline embankment is constructed in stages so that the location of the centerline of the embankment crest does not change with each stage. The upstream toe of each embankment stage is constructed over the tailings beach but the majority of each new stage is founded on the previous embankment stage. This method relies on some strength and structural support for the upstream slope but does not rely on the tailings characteristics for overall stability. Liquefaction of the tailings as a result of earthquake loading could result in, at worst, some localized instability of the upstream slope of the embankment stage. This would not result in significant damage. The centerline method is a compromise between the higher cost downstream embankment and the higher risk upstream embankment. A variation of the centerline embankment is the modified centerline embankment. This method allows the embankment crest centerline to move slightly upstream optimizing the quantity of construction materials required in the downstream shell zone of the embankment. Modem analytical techniques are used to determine how far the embankment crest centerline can be moved upstream with staged construction and still be independent of the physical characteristics of the tailings materials for overall stability. These techniques have resulted in significant cost savings

1818

without compromising embankment stability. Modified centerline embankments have been designed, permitted and constructed in USA, Canada and South America.

Geotechnical Design The goal of the geotechnical design of a tailings facility is to provide a sound, stable facility at a minimum construction cost. As the design moves from scoping level to detailed design, an increasing level of design detail accompanies an increasing level of confidence in the construction cost estimates. Levels of contingency to the estimated cost of a tailings facility can be decreased as the design level increases. With due consideration of local and national regulatory requirements, the geotechnical designer must address the following key issues: Structural stability of the tailings dam and appurtenant structures under the key loading conditions: construction, static long-term, seismic loading (if applicable) and under loads resulting from rapid draw-down of the water level if that is possible. Stability against seepage flow through core zones, basin liners, drains and filters that prevent the migration of tailings solids or fill materials through the dam or into the basin drains . Control of seepage rates below allowable limits through the life of the facility. Seepage flow rates to the surface pond and to the drainage systems resulting from tailings consolidation need to be included in addressing this issue. Stability of the facility against surface erosion by water or wind.

For scoping level designs, these stability issues may be addressed using experience and relatively simple calculations. For preliminary design, computer analyses for structural and seepage stability are typically employed. Limit equilibrium stability analyses and finite element seepage analyses are easy to perform these days and are normally provided in a preliminary design. For detailed design, additional analyses such as numerical modeling of structural stability are often employed, particularly where the assessed risk for the facility is moderate or high. For seismic structural stability of tailings dams, it is most efficient to stage the stability analyses such that complexity of the analyses increases at each design level. The ICOLD bulletin on tailings dams and seismicity (1995) states: “Seismic stability assessment of tailings dams.... is often carried out in a staged approach, which involves starting with a simpler analysis and progressing to more complex analyses as required by the specific case. In ascending order of cost and complexity, these are: Static Limit-equilibrium Stability Analysis using Steady-State Strength; Simplified Seismic Stability Analysis; and Finite Element Seismic Stability Analysis”

The geotechnical characteristics of the tailings can only truly be determined after deposition takes place for a period of 1 or 2 years and in-situ testing techniques are used to measure design parameters. Hence, design of Stage 1 of a tailings facility should incorporate reasonably conservative assumptions on the tailings characteristics that will be confirmed by site investigations after construction of the first stage or two of the facility. Also, finite element analysis is usually not justified until after the in-situ investigations of the tailings

Basin Liner Systems. The requirements for lining the tailings facility basin depend on the waste characterisation of the tailings to be stored, the hydrogeologic characteristics of the basin and regulatory requirements. Regulatory requirements can be either performance or prescriptive in nature. Performance regulations usually set limits to leakage rates or levels of impact on receiving waters. Prescriptive regulations mandate a specific liner design on the basis of the waste

1819

characterization. In either case, the waste characterization of the potential seepage water from the tailings facility is the driving criteria in liner design. It is possible to find sites that are naturally impermeable and would therefore not require constructed liners. The difficulty is in conclusively demonstrating during the site investigation phase of the design that no discrete leakage pathways exist. Constructed liners range in cost and complexity from simple single layer plastic membranes to multi-layer compound liners with drainage layers, friction layers and leak detection layers. A range of liner types is shown schematically on Figure 2. Low permeability soil liners have some distinct advantages. Firstly there is a significant time for the travel of the initial seepage front through the liner. Secondly, there may be a reduction in the concentration of waste constituents in the seepage water as a result of dispersion, diffusion and adsorption by the soil. Thirdly, the liner will consolidate under loading by the tailings deposit and this will result in reductions in permeability. A disadvantage of soil liners is that seepage takes place over the whole area of the liner. For large facilities, even if the permeability of the soil liner is low, the total seepage rate can still be significant. Other disadvantages are that soil liners are erodable, are subject to desiccation cracking, frost heave, and osmotic consolidation. Synthetic liners comprise thin plastic membranes. Typical plastics used for these liners include HDPE, LDPE, PVC, etc. The liners are placed in sheets and the seams are welded in the field to create a continuous membrane. Although the plastic materials are effectively impermeable, the liners invariably have a finite permeability due to leakage through pin holes and seaming defects. Typical leakage rates for a well installed membrane liners are 30 - 300 gallons per day per acre of liner. (0.3 to 3 cubic meters per day per hectare). The leakage through a soil liner comprises a small unit leakage over a large area while leakage through a membrane liner comprises high unit leakage over very small areas (pin holes).

Figure 2 - Liner Systems ,-Underdrain/Erosion

Protection

,-Membrane

Soil Liner with Erosion Protection/Droinage

Compound Liner

Layer

Underdrain

f

Leak Detection Drain

-Membrane

Underdrain /4embrane

(Leak

Detection Drain

. ..

.”

_.A,&

‘.-.L

f

//////I////

Low Permeability Soil //////////

/4

//////////

/

Double Compound Liner with Leak Detection

Double Liner with Leak Detection

A compound liner comprises a membrane liner placed in direct contact with a soil liner. This combines the best of both liner types (a small unit leakage rate over a very small area) and results in leakage rates substantially lower than either individual liners. The main factor controlling the leakage rate through a liner is the hydraulic head on the liner. For a single liner the head can be significant. The head on the primary liner can be controlled by the installation of a secondary or inner liner separated from the primary liner by a drainage layer. The drainage layer is designed to have sufficient capacity to remove any leakage through the inner liner without significant build up of hydraulic head. This restricts the head acting on the primary or outer liner and thereby limits leakage out of the facility.

1820

Tbe drainage layer between liners is often used as a leak detection system to measure leakage through the inner her. Many regulatory agencies have limitations on the leakage flow from this drain. In order to control this leakage rate through the inaer liner, a compound h e r is sometimes required. A more meaniagful performance specification is to limit the head buildup in the drain as this directly controls the leakage rate through the outer liner and from the facility. An important consideration in the design of liner systems is the interface shear strength between adjacent liner components. Interface strength between the membrane liner and soil liner, membrane liner and drainage media, etc, can be very low. This can result in instabilities even on relatively gentle slopes. It important to identify the most critical interface shear strength in the proposed liner system for inclusion into stability analyses. Specialized laboratory testing of interface shear strengths is generally required. With a liner in place, the tailings deposit will need an underdrainagesystem if the deposit is to become fully consolidated. An under consolidated deposit has a lower density and therefore requires a larger storage capacity for a given tonnage. The under consolidated deposit also has a low strength that can cause problems during construction of centerline or upstream embankment raises and when placing cover systems for reclamation. An underdrainage system provides a pathway for the tailings deposit to drain and reduces the hydraulic head at the base of the deposit. The effect of this is to halve the drainage path for consolidation. This reduces the time required for the deposit to consolidate by a factor of four. It reduces the hydraulic head in the liner system ant therefore will reduce the leakage from the facility. CoastroetiOn Considerations Detailed designs for the construction of tailings facilities comprise technical specificationsfor the work and detailed drawings issued for construction. These documents can be included in a contract for the construction of the facility by a third party contractor or be used by the mine to carry out the construction using their own equipment and staff. The design of tailings facilities, l h that of most large civil structures are based on the available topographic mapping and information generated by the site investigation. It is inevitable that there will be conditions encountered in the field during construction that differ to some degree from those assumed during the design. It is essential that a representative of the design engineer be onsite to provide technical direction for construction. This will ensure that significant variations in site conditions are recognized and that the design is modified as required to meet the design intent. This service is often integrated with the quality assmudquality control (QNQC)role. QNQC is required during the construction of a tailings facility so that conformance of the work with the technical specifications and drawings is measured and recorded. This continuity from the design through construction is also important from a liability perspective. A designer that has no involvement with the construction can take the position that any shortcoming is a result of the construction or due to changed conditions that were only evident during construction. The constructor conversely cao take the position that there was shortcomings in the design. Continuity of design engineer over the life of a facility can also be important as, by it’s nature, the mining industry generally has a high staff turnover at a given mine site. Often the design engineer provides the only continuous presence over the mine life. An important and often overlooked aspect of construction of a tailings facility is the preparation of detailed as built drawings of the completed facility. The preparation of as built drawings takes place at the end of the project when budgets are exhausted and enthusiasm is low. These drawings however are invaluable as the basis for ongoing staged construction and if there are any problems with the performance of the facility.

1821

OPERATIONS Tailings Deposition The method of placement of tailings into the facility depends on the form of the tailings and the desired tailings and water pond configuration. All tailings except for dry or dewatered tailings are generally conveyed to the tailings facility by pipeline. Paste tailings are included in this group even though the paste viscosity is very high, and the friction losses are significant, paste tailings can be conveyed in a pipeline using high head, positive displacement pumping systems. Conventional and thickened tailings are usually spigotted onto a tailings beach. The solids settle onto the beach and supernatant water is released. The supernatant water flows down the beach to a surface pond. The spigotting of tailings generally result in segregation of the tailings solids. The courser fraction of the tailings will settle out first resulting in a steeper beach. The finer fraction is conveyed further from the discharge point with the finest fraction settling out beneath the surface pond. The degree of segregation can be controlled by the energy of the tailings discharge stream and the solids content of the slurry. A single discharge point can result in a turbulent, high energy tailings stream that can carry the course fraction of the slurry further from the discharge point. This reduces segregation and flattens the tailings beach. Spreading the discharge over multiple spigots or the use of spray bars results in a low energy, lamina flow of tailings onto the beach. The settlement of the courser fraction is faster and the beach slope is increased. The solids content affects segregation as slurries with higher pulp densities are more viscous and the settling rate of the larger particles is impeded. Thickened tailings, as described above, minimizes segregation but the increased viscosity results in a steeper beach with slopes up to 2%. Paste tailings require special deposition techniques due to the very high viscosity of the tailings. The paste is non segregating and does not produce significant supernatant water. Tailings deposit slopes of up to 10% or greater are possible and therefore the discharge point for the tailings must be moved often to place the tailings into the facility. Two methods of placement of paste tailings are the top-down and the bottom-up methods. The top-down method is shown schematically on Figure 3.

Figure 3 - The Top-Down Method of Paste Placement Ultimate Facility

L

Pigot

\

Point

Lpaste Layers (Angle of Repose)

Native Ground

This method deposits the paste tailings downslope from a point upslope from the final toe of the deposit and the discharge is progressively moved towards the final toe.

1822

Figure 4 - The Bottom-Up Method of Paste Placement Structural Fill Zone (Compacted Paste)

Paste Layers (Angle of Repose)

Native Surface

The bottom up method is shown on Figure 4. This method deposits the paste tailings upslope from the final toe of the deposit. A structural fill zone is required to contain the deposit and provide a platform for spigotting. The main advantage of the top-down method is in water management. Runoff from the deposit reports to the downstream toe where it can be collected and managed. In the bottom-up method, the low point where runoff accumulates is always moving upslope making collection and management more problematic. The main advantage of the bottom-up method is that the downstream extent of the paste deposit is well defined. Any erosion or sloughing of the paste deposit is upstream and will always be confined. With the top-down method the toe buttress is the only containment to prevent the migration of tailings by erosion or sloughing outside the limits of the tailings facility. If this migration is excessive then it may be necessary to raise the buttress. When hydocyclones are used for segregating course tailings for embankment construction, the underflow has many of the characteristics of paste tailings. The course underflow can be directly placed in the embankment section, conveyed to the embankment by pipeline or launder or stockpiled and placed using earthmoving machinery. The cyclone overflow comprises the fine tailings fraction and the majority of the water from the feed slurry. The overflow can be deposited by spigotting onto beaches in a conventional manner. The deposition of dewatered tailings or dry stacking requires earthmoving methods and can be achieved by truck haul dump and spreading or by conveyor stacking. The main issue with dry stacking is trafficability of the surface of the deposit. This can be problematic in areas of high rainfall and snowfall. In these cases, temporary storage facilities may be required to store tailings until such time as trafficability can be established after a rainfall event or until snow can be cleared from the area of tailings placement.

Water Management A key part of the operation of a tailings facility is the management of water in the facility. Water enters the tailings facility as process water in the tailings slurry, direct precipitation, and runoff from surrounding undiverted catchments. Water is stored permanently in the tailings deposit as pore water. Water is also lost to evaporation and seepage. In almost all cases, water is recycled to the mill for use in the process. Generally, if the total quantity of the water lost to permanent storage, evaporation and seepage is greater than the water into the facility from precipitation and runoff, the facility is in water deficit and makeup water is required to sustain the milling operation. If this is not the case then the facility is in water surplus and the excess water will need to be treated and released or adequate storage will be required to contain the water. Precipitation and evaporation usually vary seasonally. This results in some months producing water surpluses and some months requiring makeup water. Water can be stored in the wetter months and used as makeup water in the dryer months. A detailed water balance is required for a tailings facility that accounts for the water inputs and outputs on a month by month basis. The water balances should consider climatic variation to ensure that there is sufficient makeup capacity for dry periods and sufficient treatment or storage capacity for wet periods. The water balance is used to predict the variation in size of the supernatant pond on a seasonal basis as the facility is filled.

1823

A further aspect of water management is to ensure that the tailings facility has adequate capacity at all times to store, route, or otherwise deal with the runoff from extreme precipitation events. Overtopping as a result of storm runoff is one of the most common causes of tailings facility failure. It is imperative that adequate storage is provided to contain the supernatant pond resulting from the design wet season plus the runoff from the design storm plus required freeboard for wave run up. It is often prudent to include an emergency spillway as a last line of defense. This will enable water to spill from the facility if the design storms are exceeded and prevent overtopping and potential failure of the facility.

Staged Construction In almost all cases, the construction of tailings facility is staged over its service life. The facility is staged so that each stage provides the minimum volume for storage of tailings, the supernatant pond and extreme storm runoff for the period up to the construction of the next stage of the facility. This defers capital expenditure for tailings facility construction thereby minimizing initial capital investment. The other advantage of staging the construction of a tailings facility is that an observational approach can be taken to the ongoing design of the facility. When the initial design is done, many assumptions are made regarding the physical characteristics of the tailings including the expected deposit density, strength characteristics, liquefaction potential etc. These assumptions are necessarily conservative. After several years of operations, the performance of the tailings deposit can be investigated producing hard data on these characteristics. Staged construction allows for the design to be modified and optimized to incorporate these data often resulting in reduction in construction and operating costs for tailings management. Monitoring and Surveillance Designs for tailings facilities are based on basic assumptions as to the behaviour of the embankments, the tailings deposit, the hydrogeology and the water management. It is essential that adequate instrumentation is installed and that a comprehensive monitoring and surveillance program is implemented. The monitoring program should include a formal inspection carried out at least annually by a senior engineer with full knowledge of the design and operating criteria for the facility. The results from the program should be reviewed on a regular basis, preferably by the design engineer to c o n f m that the design requirements are being met. A regular review of monitoring data can provide early warning of developing problems. This allows remedial action to be taken before the problem develops into a major problem or worse, a failure. Many regulatory jurisdictions require that a regular report be prepared that fully describes the operation of the tailings facility over the reporting period and records the volumes and types of wastes that have been deposited, the findings of the formal inspections and the resulting recommendations, the data from instrumentation and interpretationsthereof. Safety Review and Risk Management Detailed safety reviews, should be carried out for tailings facilities at the initial design stage, at a stage during operations and at closure. These reviews and risk assessments are best carried out by third party experts unrelated to the owner or the design engineer. The review should cover the adequacy of the design, an audit of the QNQC program, and a review of the construction documentation. Similarly, a formal risk analysis should be carried out at least once in the early stages of the operation of a tailings facility. In an engineered risk assessment, the risk associated with each failure mode and for each considered structure is considered to arise as a result of two sufficient and necessary components: likelihood and consequences. Mathematically, risk is simply the product of likelihood and consequences expressed in units of probability and dollars respectively. A life, resource or operation may only be considered to be at risk if both the potential (likelihood) for a failure and an impact (consequence) of that failure exist. Risk may be increased due either to an increased likelihood of occurrence andor increased consequences of occurrence.

1824

A quantitative risk assessment is typically considered to be more robust and thorough due to the finite expected values that are determined for each failure mode associated with the structure being considered. In other words, if it is determined that there is a 10 percent chance (likelihood) of a hydrometeorological event causing one million dollars of erosion damage (consequence) over the life of the mine, then the expected value (risk) of that failure mode or hazard is $100,000. The expected value of this hazard may then be compared directly to other identified failure modes in order to assist with decision making during operations or design stages of a project. The obvious challenge is the estimation of probabilities or likelihoods and the quantification of consequences in ? What value is placed on a terms of dollars. Is the likelihood of an event 1 x lo4 or 1 x pristine lake? In contrast, a qualitative risk assessment relies on a subjective rating of perceived likelihoods and consequences that are combined according to descriptive codes and compared with one another. Although apparently simplistic, a qualitative risk assessment may provide valuable insight to the vulnerability of various project components where insufficient or potentially misleading quantitative data is available. Regardless of the method used, the results provide information on which risk management decisions can be made. The overall goal of risk management is to ensure that efforts are focused to minimize the over all risk of operating the facility and that priorities are set accordingly..

CLOSURE Planning for Closure The main requirements for closure of a tailings facility is to provide long term secure and stable storage for the tailings materials. The closure works must ensure that the tailings solids are not transported from the facility by wind or water erosion and that the contaminants in the tailings deposit are not dispersed into the environment by seepage, vegetation uptake etc. The design for closure of a tailings facility commences at the earliest stages of conceptual design during initial development. A conceptual remediation and closure plan is developed at this stage and the objective of this plan is to provide a systematic approach to decommissioning the facility and to return the disturbed lands to a habitat capability similar to pre-mining conditions, or to an acceptable alternative. To achieve this, a closure plan must be flexible enough to allow for future changes in the mine plan and to take advantage of information obtained from ongoing reclamation research. An alternatives analysis is often carried out to select the best option to suit both current conditions and potential changes. As well, measures to progressively reclaim the site during operations are often included to reduce the environmental and safety risks at the site. The closure plan is often the basis for estimating the cost of closure and setting the level of bonding required by regulatory agencies from mining companies to ensure that the reclamation is funded. Ongoing development and refinement of the closure plan along with progressive reclamation during the operation of the tailings facility can be the basis for reduction of the reclamation bond. The land use prior to development provides a guide to the criteria for post closure land use for the site. Tailings facilities located in a desert location will have different criteria for closure than facilities sited on productive farm or range lands. The first step in designing a tailings facility is therefore to determine how the facility is going to look after closure and reclamation. This includes how the final site drainage will be established, how the facility will be covered, revegetated, stabilized to prevent erosion etc. Once these concepts are finalized, the design of the facility including the tailings deposition, location of the supernatant ponds liners and drainage systems can be carried out to complement closure concepts. When planning closure, another consideration is if the mine closes earlier than planned due to economic or other circumstances. This can leave the tailings surface a long way below the planned final spillway location and can result in costly works to establish a workable closure configuration. To meet the requirements for closure, the tailings surface generally requires stabilization. Stabilization can range from direct revegetation of the tailings surface, if possible to multiple

1825

layered cover systems. Direct revegetation is an option where the available nutrients and climate make establishment of sustainable vegetation possible. This will prevent wind erosion but will generally do little to prevent infiltration of precipitation that will eventually seep from the deposit. If there are any contaminants such as heavy metals, the plants may take these up and make them accessible to grazing animals. Simple revegetation is therefore only applicable where the tailings are benign. Cover systems for tailings deposits are designed to carry out a number of functions. These can include providing a capillary break to prevent the migration of contaminants out of the deposit, a barrier to limit the infiltration of precipitation into the tailings deposit, an erosion resistant layer to prevent wind or water erosion and growth media for the establishment of vegetation. Cover systems can cover very large areas and can require significant resources for construction. It is important to identify the source of materials at the earliest stages and to ensure that the environmental impact of accessing these materials is not excessive. It is very important to conserve potential materials such as topsoil during the construction of the facility so that they are available for reclamation works. Flooded tailings surfaces have been designed and constructed for tailings where acid drainage is an issue. Flooding of the tailings surface maintains the tailings in a state of saturation and prevents the oxidation of sulfide minerals. This method of closure is very effective in preventing oxidation but has significant associated issues. The flooded deposit is a constant source of seepage and can result in ongoing transport of contaminants. The flooded deposit is a perpetual liability that will require ongoing inspection and maintenance. Mine tailings that are unconsolidated or only partially consolidated at closure will continue to consolidate after closure, resulting in on-going settlement of the tailings surface and the generation of consolidation seepage. Estimating the degree of consolidation of a tailings deposit at closure is therefore a key consideration for reclamation, particularly when surface covers are to be constructed or estimates of post closure consolidation seepage rates are required. Predictions of the consolidation rate, the magnitude of surface settlement and seepage flows after closure can be determined using data from in-situ field testing, laboratory testing of tailings samples and computer modeling of the tailings consolidation process. If the ongoing consolidation of the tailings and the time frame for this are problematic, then measures can be designed to significantly reduce the time for consolidation. Some under consolidated tailings deposits will take decades or even centuries to fully consolidate and this may postpone final reclamation indefinitely. The installation of vertical drains into the tailings deposit can reduce the time for consolidation of the tailings deposit to a few years. Closure planning requires a water management plan. Once tailings deposition and recycle of process water stops, precipitation can cause water to accumulate in the facility. In addition, underdrainage from the tailings deposit, seepage recovery, etc. continues after the mill has stopped. Methods of dealing with the water depends on the level of contamination. Treatment can be as simple as sediment control and release or require complex biological treatment to remove a variety of contaminants. It is sometimes possible to dispose of excess water using controlled land application. Water treatment plants are usually costly to build and require significant ongoing operating costs. It is very important for closure planning to predict the level of water treatment and the length of time that the treatment will required.

Post Closure Maintenance and monitoring All tailings facilities require a degree of maintenance and monitoring after reclamation is complete. The level of effort required depends on the complexity of the reclaimed structure and the requirement for ongoing treatment of effluents from seepage and surface runoff. As previously stated, flooded closures require water retaining structures that require frequent inspections and maintenance to ensure that the facility continues to be safe. Diversions, drainage ditches, spillways, cover systems etc. are all subject to deterioration. Root systems from vegetation can penetrate liners, animals can burrow through embankments and covers, beavers can dam diversion ditches and spillways. A post closure maintenance program should anticipate the requirements to deal with these issues and provide the funds necessary to carry these out.

1826

CONCLUSIONS This paper provides a glimpse of the scope and complexity of work required to successfully manage a tailings facility through it’s complete life cycle. Tailings facilities are non revenue generating and money spent on development, operations and closure comes directly from the bottom line. The successful operation of a tailings facility however is of fundamental importance as potential failure can represent the greatest risk to the environment and a mining company’s reputation, viability and share value. ACKNOWLEDGEMENTS This paper was prepared with input from a number of engineering and environmental specialists in the Knight Piesold Consulting group whose contributions are gratefully acknowledged. REFERENCES

Canadian Dam Association (CDA), (1999), “Dam Safety Guidelines”. Cornell, C.A. (1968) “Engineering Seismic Risk Analysis” Bulletin of the Seismological Society of America, Vol. 58, p.1583-1606. ICOLD (1989) “Selecting Seismic Parameters for Large Dams”, International Commission on Large Dams, Bulletin 72. ICOLD (1995) ‘Tailings Dams and Seismicity”, International Commission on Large Dams, Bulletin 98.

1827

Design of Tailings Dams and Impoundments Peter C. Lighthall, Michael P. Davies, Steve Rice and Todd E. Martin

ABSTRACT The state of practice for tailings dam and impoundment design is summarized. The design process, which embraces construction, operational and closure issues together with requisite technical aspects, has evolved over the past several decades though the engineering principles have remained the same. The design process has evolved to meet the demands of a regulatory environment that has become increasingly stringent together with more challenging mining economic conditions, as well as alternative tailings management technologies have been developed; including filtered dewatered, stacked deposits; thickened dewatered systems; frozen tailings deposits; and paste disposal. In response to a number of highly publicized tailings impoundment failures, tailings management systems, dam safety programs, and risk assessment techniques have been created and are now part of standard practice and appropriate dadimpoundment stewardship. The concept of environmental sustainability is now an integral component of the tailings dadtailings impoundment design process and appropriate project conceptualization followed by stewardship of the design are essential to this sustainability. INTRODUCTION Environmental factors have increasingly driven tailings dam designs in recent years, but not necessarily to the benefit of engineering safety of tailings dam structures. For while design tools have improved and designs may have become more rigorous with technological improvements, the safety record of tailings dams has not markedly improved. Highly publicised failures continue to occur, resulting in a negative image for the mining industry, and particularly for Canadian mining companies, which have been involved in several of the more prominent failures. Unfortunately, the failure statistics do not tell the real story, which is that design, construction, operation and management of tailings facilities has advanced tremendously over the past thirty years. The security of tailings facilities is now a recognized priority at a corporate level in most large mining companies and the concept of sustainable mining, which clearly involves appropriate mine waste stewardship, is an accepted part of the modern industry. Tailings, and waste rock, are the waste products of the mining industry. Their disposal adds to the cost of production, and consequently, it is desirable to accomplish their disposal as economically as possible. This requirement for low cost led to development the upstream method of tailings dam construction, which was the standard for tailings disposal up to the mid-l900’s, irrespective of site conditions. With the advent of sound engineering practice, it became recognised that there are significant weaknesses and risks in the upstream method of construction under many site conditions. To augment the upstream method, embankment designs were developed using downstream and centreline construction methods. Sound civil engineering designs for embankment slopes, transition zones and filters were applied to tailings dams on an industry wide basis for the first time, beginning in the 1960’s. Once designed and constructed based on an empirical and experience-based approach, tailings dam design has since evolved into a formal specialist engineering discipline. Over the past 30 years, the most significant change in tailings disposal technology has been the recognition of the long-term geochemical risks, particularly the potential for acid drainage and metal leaching. The tailings dam designer of 30 years ago was not tasked with understanding and addressing geochemical issues. Today, geochemical characterisation is one of the most important aspects of tailings disposal planning, and designs for operation and closure are focussed on geochemical issues. These issues often govern not only the type of tailings dam, but also may govern tailings dam site selection. Contaminant loading analysis is now required on practically all

1828

tailings dam design projects. Closure strategies to prevent long-term geochemical impact to receiving environments are often the driving factors in design of tailings impoundments. The other significant trend over the last 30 years has been the considerable number of highly publicised failures of tailings dams that have continued to occur with alarming frequency over the past three decades. In response to these large numbers of failures, mining corporations, financial institutions, environmental groups, government regulators and even the general public have instituted much more rigorous scrutiny of tailings management systems. This paper attempts to describe how the state of practice of tailing impoundment design, construction and operation has changed, and to outline some of the technologies that have been developed in recent years.

MINE TAILINGS Tailings are the finely ground barren minerals left following ore extraction processes. They are typically a product of milling, although in several industries, e.g. the oil sands mining activity in Northern Alberta, large volumes of tailings can be produced without mechanical crushing. In milling, the process begins with crushing mine-run ore to particle sizes generally in the range of millimetres to centimetres. Crushed ore is then further reduced by grinding mills to sizes less than 1 mm in ball mills, rod mills, and semi-autogenous (SAG) mills. Water is added to the ground ore, and the material remains in slurry form throughout the remainder of the extraction process. The grain size distribution of tailings depends upon the characteristics of the ore and the mill processes used to concentrate and extract the metal values. A wide range of tailings gradation curves exist for various mining operations and consequently, tailings may range from sand to claysized particles. For most base metal mines 40% to 70% of the tailings will pass a No. 200 sieve (74 pm). However, some milling processes such as gold extraction may grind the ore so that 90% or more of the tailings pass the No. 200 sieve. Tailings may also include metal precipitates from neutralisation sludges or residues from pressure leaching processes. Such materials may exhibit long-term chemical stability concerns and need to be disposed in secure, lined facilities. BACKGROUND AND HISTORY OF TAILINGS DAMS Mining has been carried out in some form for at least 5 000 years. In forms more similar to modern mining, crude millstone crushing and grinding of ore were initially practised in the New World in the 1500’s, and continued through the mid-1800’s. The largest change over those centuries was the introduction of steam power, which greatly increased the capacity of grinding mills and hence, the amount of barren by-product (tailings) produced. Minerals of economic interest were initially separated from crushed rock according to differences in specific gravity. The remaining tailings were traditionally routed to some convenient location. The location of greatest convenience was often the nearest stream or river where the tailings were then removed from the deposition area by flow and storage concerns were largely eliminated. Later in the 1800s, two significant developments, which changed mining dramatically, were the development of froth flotation and the introduction of cyanide for gold extraction. Flotation and cyanidation greatly increased the world’s ability to mine low-grade ore bodies, and resulted in the production of still larger quantities of tailings with even finer gradation (i.e., more minus 74 pm material). However, tailings disposal practices remained largely unchanged and, as a result, more tailings were being placed and transported over greater distances into receiving streams, lakes and oceans. Around 1900, remote-mining districts began to develop, and attract supporting industries and community development. Conflicts developed over land and water use, particularly with agricultural interests. Accumulated tailings regularly plugged irrigation ditches and “contaminated” downstream growing areas. Farmers began to notice lesser crop yields from tailings-impacted lands. Issues with land and water use that led to the initial conflicts then led to litigation in both North America and Europe. Legal precedents gradually brought an end to uncontrolled disposal of tailings in most of the western world, with a complete cessation of such practices occurring by about 1930.

1829

To retain the ability to mine, industry fostered construction of some of the first dams to retain tailings. Early dams were often built across a stream channel with only limited provisions for passing statistically infrequent floods. Consequently, as larger rainfalls or freshet periods occurred, few of these early in-stream dams survived. Very little, if any, engineering or regulatory input was involved in the construction or operation of early dams. Mechanized earth-moving equipment was not available to the early dam builders. As a result, a hand-labour construction procedure (the initial upstream method) was developed. A low, dyked impoundment was initially filled with hydraulically-deposited tailings, and then incrementally raised by constructing low berms above and behind the dyke of the previous level. This construction procedure, now almost always mechanized, remains in use at many mines today. The first departure from traditional upstream dam construction likely followed the failure of the Barahona tailings dam in Chile. During a large earthquake in 1928, the Barahona upstreamconstructed dam failed, killing more than 50 people in the ensuing, catastrophic flowslide. The Barahona dam was replaced by a more stable downstream dam, which used cyclones to procure coarser-sized material for dam construction from the overall tailings stream. By the 1940's, the availability of high-capacity earthmoving equipment, especially at open-pit mines, made it possible to construct tailings dams of compacted earthfill in a manner similar to conventional water dam construction practice (and with a corresponding higher degree of safety). The development of tailings dam technology proceeded on an empirical basis, geared largely to the construction practices and equipment available at the time. This development was largely without the benefit of engineering design in the contemporary sense. Nonetheless, by the 1950's many fundamental dam engineering principles were understood and applied to tailings dams at a number of mines in North America. It was not until the 1960's, however, that geotechnical engineering and related disciplines adopted, refined, and widely applied these empirical design rules. The 1965 earthquake-induced failures of several tailings dams in Chile received considerable attention and proved to be a key factor in early research into the phenomenon of liquefaction. Earthquake-induced liquefaction remains a key design consideration in tailings dam design. Issues related to the environmental impacts from tailings dams were first seriously introduced in the 1970's in relation to uranium tailings. However, environmental issues related to mining had received attention for centuries. Public concerns about the effects of acid rock drainage (ARD) have existed for roughly 1,000years in Norway. Public concerns were similarly expressed hundreds of years ago in Spain and in Greece. In the early 1970's, most of the tailings dam structural technical issues (e.g. static and earthquake induced liquefaction of tailings, seepage phenomena and foundation stability) were fairly well understood and handled in designs. Probably the only significant geotechnical issue not recognised by most designers was the static load-induced liquefaction (e.g. the reason for many previously "unexplained" sudden failures). However, issues related to geochemical stability were not as well recognised, and tailings impoundments were rarely designed with reclamation and closure in mind. Over the past 30 years, environmental issues have grown in importance, as attention has largely turned from mine economics and physical stability of tailings dam to their potential chemical effects and contaminant transport mechanisms. Physical stability have remained at the forefront, as recent tailings dam failures have drawn unfortunate publicity to the mining industry, with severe financial implications in many cases. In response, a great many mining companies, at a corporate level, have identified safe tailings management as a priority, and have made resources available to address that priority. A significant tailings impoundment failure will almost certainly have a direct cost in the tens of millions of dollars and indirect costs, including devaluation of share equity, often many times the direct costs. In all of the tailings dam failure cases, a few examples of which are noted later in this paper, relatively simple, well-understood structural failure mechanisms were found to be at fault in causing the incidents.

1830

THE ROLE OF GOVERNMENT REGULATIONS AND WORLDWIDE STANDARDS As a new century unfolds, regulators and non-government organisations worldwide are becoming increasingly educated about tailings dam design and stewardship requirements. Lending agencies have also dramatically increased their technical requirements prior to funding new or expanded projects. This trend in education is a welcome development as candid discussions on risk levels for any given technical issue can be carried out with good understanding from all stakeholders. However, an unwelcome development has been the significant amount of non-technical and often misguided opposition and it is this latter situation that causes the most grief for the mining industry. No longer can a mine development proponent simply agree to meet the criteria of the senior governing authority and provide evidence of credible design. Often, several levels of government and non-government organisations must be satisfied with the proposed mining development. The tailings impoundment is often the most critical component of a mine development in the eyes of regulators and third party interest groups. Many developing countries, where the international mining industry is focussing considerable attention, have only recently enacted regulations pertaining to tailings disposal. These regulations typically are based on those in place in more developed jurisdictions. Regulators in developing countries, however, all too often lack the resources and the expertise to fully implement these regulations. As a result, regulations in developing countries are often technically prescriptive. Such technically prescriptive regulations provide a false sense of security, since failures are so often the result of a combination of design flaws and improper stewardship. Many failures have occurred at facilities that conformed with all regulations, except for the most important of all (the dam failed). The problem of limited resources for regulators is by no means unique to developing jurisdictions. Government budget cutbacks in jurisdictions such as Canada and North America mean that the mining industry is striving to a condition of “co-regulation”, in partnership with regulators. This is a welcome trend, placing the initiative for continuous improvement and safety of tailings disposal facilities squarely with the mining industry. Although visible exceptions continue to arise, as noted later in this paper, the modern tailings facility is typified by a well-designed and constructed facility that has met several levels of regulatory and non-regulatory scrutiny and has received corporate attention to the highest level. Most of the world’s mining companies have multi-national operations, and to continue operation, and to attract share capital, must be seen to have exemplary environmental and safety records. Regulators, mining companies and international environmental organisations have developed numerous programs to ensure a high level of security of tailings disposal systems. Examples of some of these programs include: The Mining Association of Canada (MAC), has recently published a document entitled “A Guide to the Management of Tailings Facilities” (MAC, 1998); The Canadian Dam Association (CDA) recently updated its dam safety guidelines (CDA, 1999). The update focussed in large part on incorporating elements specific to the safety of tailings dams; The International Committee on Large Dams (ICOLD), and related organizations, have published numerous materials with regards to tailings dams; The United Nations Environment Programme, Industry and Environment (UNEP), and the International Council on Metals and the Environment (ICME) have been active in recent years in sponsorship of seminars, and publication of case studies (UNEP-ICME, 1997 & 1998), related to tailings management; Several major Canadian-based mining companies have established corporate policies and procedures to ensure that all personnel involved in stewardship of tailings facilities, from the corporate level to the operators, clearly understand their roles and responsibilities (e.g., Siwik, 1997); Numerous mining companies, including Syncrude, Kennecott Utah Copper, and Inco, retain a board of eminent geotechnical consultants to provide independent review and

1831

advice in terms of the design, operation, and management of their tailings facilities. These programs are described in McKenna (1998), Dunne (1997) and (McCann, 1998); and Many mining companies have regular third party risk assessment programs for their tailings facilities, in which experienced consultants, usually teamed with the owner’s personnel, carry out audits of tailings facilities. Of all of the above measures, the authors consider the last two to be of most importance.

SITE CHARACTERIZATION A major factor in managing tailings impoundment failure risk is to carry out adequate site characterisation for siting and designing tailings impoundments. Experience in tailings impoundment design and construction has served to emphasise the critical importance of a proper understanding of the geology of impoundment sites, and of an appreciation of how geology will affect the design, construction, and performance of the tailings facility. Many of the catastrophic structural failures have occurred as a direct result of inadequate site characterization. Site characterization has always relied on carrying out thoughtful site investigations to develop a thorough understanding of site geology. The judgement of experienced engineering geologists should play a major role in site characterization. Traditional tools of geological mapping, air photo interpretation, test pitting, geotechnical drilling continue to form the basics of site investigation. However, significant changes and technological improvements have been made, so that tailings designers have a much wider range of tools from which to choose. As well, the needs of site characterization have added new demands. Some significant advances/changers in recent years include: Generally improved technology in site investigation techniques; Increased emphasis on water management and environmental characterization of the site, particularly with regards to hydrogeology; Greater importance being stressed on recognising geochemical issues, such as acid rock drainage (ARD), and in terms of attenuation of groundwater contaminant transport; Greater emphasis on closure and reclamation considerations and the site investigation required to support that design; and Development of site investigation methodologies for characterization of liquefaction potential that have evolved significantly.

Technological Advances The technology available for use in execution of site investigation programs has advanced greatly in recent decades. There are more and better-equipped geotechnical drilling contractors, with more powerful and efficient drilling equipment. Examination of aerial photographs represents an essential method of site assessment, and the quality of aerial photography has improved considerably. Remote sensing, and satellite imagery techniques are now available and prove invaluable on many projects. Geographic Information Systems (GIs) have been developed and represent a major advance in the collation and ultimate usefulness of site data. The world earthquake database, and the means for using it in probabilistic characterisation of site seismicity, a significant consideration for tailings impoundment design, has also advanced. Geophysical methods have also become more useful as techniques have become more reliable and analysing the data within complex solutions more readily available with the dramatic increase in computer processing power. Perhaps the most significant advance has been the development and widespread application of electronic piezocone technology for geotechnical and environmental characterization of clay, silt, and sand soils. The piezocone provides end bearing, friction, and porewater pressure data on a near-continuous basis as the probe is advanced. From these data, such geotechnical information as soil type, shear strength, in situ state, sensitivity, relative density/consistency, and liquefaction

1832

susceptibility can be determined using semi-quantitative relationships. Resistivity measurements are also possible, enabling contaminant plumes in groundwater to be delineated. The piezocone can be used effectively as a piezometer to characterize porewater pressure gradients, and for in situ estimation of hydraulic conductivity and consolidation parameters. Piezocone technology is particularly well suited to geotechnical and environmental profiling of mine tailings deposits.

Emphasis on Environmental Site Characterisation The increasing emphasis on environmental protection in siting, design, construction, operation, and closure of tailings impoundments has placed increased emphasis on environmental aspects of site investigations. In terms of siting and design of a tailings impoundment, groundwater quality protection is perhaps the most significant environmental protection aspect requiring investigation. To evaluate potential groundwater quality impacts, the following are essential: Establish baseline (pre-development) groundwater quality conditions, by collecting surface water and groundwater samples (monitoring wells). Identify principal hydrogeologic units (overburden and bedrock), and develop a hydrogeologic model of the site. In most cases, a larger, regional hydrogeologic model is also required. Model tailings impoundment development and estimate contaminant loadings. This may require characterization of attenuation capacity in the hydrogeologic units. Based on that model, and on parametric (sensitivity) analyses, determine compliance versus noncompliance at appropriate locations. Seepage modelling, and contaminant transport modelling, can now be carried out using very powerful yet simple to use finite element and finite difference computer models. These models include both saturated and unsaturated flow regimes, which allow better prediction of behaviour at the important interface between the tailings and the atmosphere, where much of the geochemical activity is occurring. The development of these tools has to some degree driven the need to obtain the data that allows their effective use. The foundation of these models is a reasonable hydrogeologic model, and the foundation of a hydrogeologic model is an understanding of local and regional geology.

Geochemical Characterization The role of geochemical issues in tailings impoundment design, and particularly closure, is equally as important as geotechnical issues driven primarily by the critical issue of acid rock drainage (ARD). Tailings that are potentially acid generating require closure strategies, and therefore impoundment designs, that will preventkontrol acid generation. This issue is a major component of initial mine studies involve addressing the acid generation potential of tailings and/or waste rock in a comprehensive manner. Susceptibility of dam fill and foundation materials to structural change due to the effects of ARD generation in the tailings deposit must also be considered. For example, if an impervious core dam is being considered, then the mineralogy of the core material should be checked, as dissolution of carbonates within the material could greatly increase the permeability of the core. Similarly, geochemical effects on materials being considered as a clay liner for the impoundment must also be considered. As another example of the importance of this issue, there have been many documented case histories of ARD resulting in clogging of internal drainage zones within tailings dams, requiring in many cases extensive remedial measures.

NEW DEVELOPMENTS IN TAILINGS DISPOSAL A number of improvements have been made in tailings disposal technology and tailings dam design, both to improve on the weaknesses of previous practices and also to take advantage of tailings processing technologies. Geochemical aspects now largely drive the siting, of a tailings

1833

impoundment, the design of retention structures, and tailings disposal technology. There have been technologies put forward as panaceas for tailings disposal problems, which have turned out to be flawed in practice. These improvements can be categorized as changes in basic management practices and changes in tailings characteristics through pre-discharge dewatering.

Designing for Geochemical Issues Geochemical issues have become highly prominent as severe acid generation problems became apparent at a number of mature mines around the world. Some of these mines, which had been operated by smaller mining companies, became orphan sites, leaving significant legacies for future generations. The majority of the acid drainage mine sites have become very expensive legacies for the major mining companies that owned them. It has been necessary to develop and operate acid drainage collection and treatment systems for continued operation and closure of numerous mines. Capital costs for ARD collection and treatment systems have been in the several tens of millions of dollars, with ongoing operating costs up to several millions of dollars annually. As a result, companies developing new mines have focussed on methods to predict and prevent or reduce acid generation from tailings. Considerable research, for example CANMET’S Mine Effluent Neutral Drainage (MEND) program, was carried out in the 1980’s and 1990’s, to assess viable methods of acid drainage control. The most significant conclusion of the past 20 years is that it is far easier (economic) to prevent ARD in the first place than to control it. From a number of existing sites where tailings had been placed in lakes in northern Canada, it was concluded that long-term submergence of acidic wastes was probably the most effective means of ARD control. Considerable work has also been done on placement of impervious closure covers over tailings to prevent ingress of air and water. Sophisticated designs of multiple-layer covers, incorporating impervious zones, pervious capillary barriers and topsoil for vegetation growth, have been developed. Covers have been found to present the risk of long term cracking or erosion, and to be ineffective in excluding air, so are less favoured solutions than submergence from the geochemical standpoint. Some of the main technologies for reduction of ARD potential from sulphide bearing tailings are the following: Design for submergence by flooding the tailings at closure. This is a solution, which is being increasingly encouraged and accepted by regulators. However, the authors are concerned that flooded impoundments may create a risky legacy. The more traditional closure configuration for tailings impoundments has been to draw down water ponds as completely as possible, to reduce the potential for dam failure by overtopping or erosion. To raise water levels in impoundments formed by high dams could present considerable long-term risk. One of the reasons that closed tailings impoundments have traditionally proven to be generally more safe, from the physical stability perspective, than operating impoundments is the relatively more “drained” condition of closed impoundments that do not include a large water pond. The flooded closure scenario represents an “undrained” condition that does not allow this improvement in physical stability to develop, so the risk does not decrease with time. 2. Treatment of tailings to create non-acid generating covers. To avoid the necessity of flooding impoundments, non-reactive covers of tailings can be placed on the top of the impoundment on the last few years of operation. It has been shown in several mining operations, for example at the Inco Ontario Division central milling operation in Copper Cliff, Ontario, that by the relatively inexpensive installation of some additional flotation capacity, pyrite can be removed to the level that the tailings can be made non-acid generating. The upper non-acid generating tailings placed on top can be left as a wide beach for dam safety, while the underlying mass of potentially acid generating tailings remains saturated below the long-term water table in the impoundment. Normally, the small amount of pyrite removed by flotation can be disposed as a separate tailings stream, placed in the deepest part of the impoundment where it can be left flooded. 1.

1 a34

3.

Lake or ocean subaqueous disposal. The surest, safest and most cost-effective solution to prevent ARD is sub-aqueous disposal in a lake or the ocean. Tailings will remain permanently submerged and have shown to be non-reactive under water and to have few permanent environmental impacts. The challenge for this solution is that regulators have become reluctant to permit lake or ocean disposal, and there are not always appropriate sites available. In addition, the public often reacts emotionally and negatively to the concept of such disposal, despite the considerable benefits of these approaches. The authors are aware of at least two examples where public pressure incited regulators to demand that existing operations switch from ocean and lake disposal to on-land impoundments, with the result that environmental problems actually increased. The authors do note a slight trend to re-acceptance of subaqueous disposal, particularly in the marine environment, as the true environmental impact of the technique can be demonstrated to be almost negligible in certain instances. Moreover, the corporate risks and environmental liabilities associated with surface tailings storage on many projects grows to the point where project viability is threatened without looking to environmentally acceptable alternatives including subaqueous disposal.

Improved Basic Design Concepts Improved Upstream Construction. Considerable attention has been given to improving traditional upstream dam construction to make the technique not only economical but also stable under both static and dynamic conditions. Numerous failures of upstream constructed dams have occurred. The failures have been the results of earthquakes, high saturation levels, steep slopes, poor water control in the pond, poor construction techniques incorporating fines in the dam shell, static liquefaction, and failures of embedded decant structures. Most failures have involved some combination of the above weaknesses. Based on the above experiences, and through the use of improved analytical tools (computer programs for stability, seepage, and deformation under both static and seismic conditions), safe, optimised designs have been developed. Some of the key design features that have been added include: Underdrainage, either as finger drains or blanket drains, to lower the phreatic level in the dam shell; Beaches compacted to some minimum width to provide a stable dam shell. Beaches are compacted by tracking with bulldozers, which are also used for pushing up berms for support of spigot lines; Slopes designed to a lower angle than was used for many failed tailings dams. Slopes are generally set at 3 horizontal to 1 vertical or flatter, depending on the otherI measures incorporated into the designs. Steeper slopes, without an adequate dr ined and/or compacted beach, create the potential for spontaneous static liquefaction - a phenomenon not widely recognized in 1972 but one responsible for a number of majorI tailings dam failures.

a

1835

Figure 1 below shows a typical section of an improved upstream design.

1

FINGER DRAIN

1

Figure 1 Typical section of improved upstream tailings dam design Lined Tailings Impoundments. With the advent of larger gold mining operations, and the almost universal use of sodium cyanide as an essential part of gold extraction, the need came about to develop impervious impoundments to contain cyanide solutions. Although cyanide is in most forms an unstable compound that naturally breaks down on exposure to air, it can be very persistent and migrate long distances in groundwater. As well as cyanided gold tailings, other types of tailings may also be considered potentially contaminating. For protection of aquifers, where tailings impoundments are not sited over impervious soils or bedrock and embankment cutoffs are not sufficient to reduce seepage, it is often necessary to design and construct a liner over the base of a tailings impoundment. Great progress has been made in liner design and construction practise. Liners may be as simple as selective placement of impervious soil to cover outcrops of pervious bedrock or granular soils, or may need to be a composite liner system constructed over the entire impoundment. Where geomembrane liners are used, it is normal practise to incorporate a drainage layer above the geomembrane, to reduce the pressure head on the liner and minimise leakage through imperfections in the liner. Another benefit of such under-drainage is that a low pore pressure condition is achieved in the tailings, giving them a higher strength than would exist without such under-drainage. The drainage layer typically consists of at least 300 mm of granular material, with perforated pipes at intervals within the drainage layer. The pipes are laid to drain water extracted from the base of the tailings deposit and to discharge to a seepage recovery pond. Figure 2 below shows two typical configurations of lined impoundments. Figure 2a shows a liner extending up the face of the embankment, requiring special detailing of drainage pipe penetrations through the liner. In Figure 2b, the liner extends beneath the embankment. In the latter case, care must be taken to design for lower foundation shear strength for the downstream slope of the embankment, as the liner may form a plane of weakness.

1836

LSINGLE GEOMEMBRANE

PIPE

a) LINED IMPOUNOMENTS/LINED PERVIOUS ROCKFILL DAM

LINER

EMBANKMENT

TAILINGS7

GEOMEMERANE LINER f3TENOS BENEATH EMBANKMENT

LINED SEEPAGE COLLECTION POND

\DOUSE

b) LINED IMPOUNDMENTS/UNLINED EMBANKMENT

Figure 2 Conceptual sections of lined impoundments with underdrains Dewatering Technologies. As shown on Figure 3 below, the basic segregating slurry is part of a continuum of water contents available to the tailings designer in 2000. Although tailings dewatering was previously practised for other purposes in the mining process, until recently the only form of tailings for most tailings facilities was a segregating, pumpable slurry with geotechnical water contents of well over 100%.

Tailings Slurry (Typically segregating)

Thickened Tailings

(Dewatered, >100%saturated - ideally non-segregating)

Paste (additive (s) to thickened tailings)

Most efficient water conservation Negligible seepage losses from stack

“Wet” Cake (At or near 100%saturation)

Progressive covering & reclamation Stable tailings mass Minimal containment requirements

“Dry” Cake

(Unsaturated - typically 70 to 85% saturation)

Simple water management

High operating Cost

Figure 3 Classification of Tailings by Degree of Dewatering (after Davies and Rice, 2001)

1837

There are several candidate scenarios where dewatered tailings systems would be of advantage to the mining operation. However, dewatered tailings systems have less application for larger operations for which tailings ponds must serve dual roles as water storage reservoirs, particularly where water balances must be managed to store annual snowmelt runoff to provide water for year round operation.

“Dry” Cake filtered tailings disposal. Development of large capacity, vacuum and pressure belt filter technology has presented the opportunity for disposing tailings in a dewatered state, rather than as a conventional slurry. Tailings can be dewatered to less than 20% moisture content (using soil mechanics convention, in which moisture content is defined as weight of water divided by the dry weight of solids). At these moisture contents, the material can be transported by conveyor or truck, and placed, spread and compacted to form an unsaturated, dense and stable tailings stack (often termed a “dry stack”) requiring no dam for retention. While the technology is currently considerably more expensive per tonne of tailings stored than conventional slurry systems, and would be prohibitively expensive for very large tonnage applications, it has particular advantages in the following applications: 0

0

0

In very arid regions, where water conservation is an important issue. The prime example of such system is at the La Coipa silvedgold operation in the Atacama region of Chile. A daily tailings production of 18,000t is dewatered by belt filters, conveyed to the tailings site and stacked with a radial, mobile conveyor system. The vacuum filter system was selected for this site because of the need to recover dissolved gold from solution, but is also advantageous for water conservation and also for stability of the tailings deposit in this high seismicity location; and In very cold regions, where water handling is very difficult in winter. A dewatered tailings system, using truck transport, is in operation at Falconbridge’s Raglan nickel operation in the arctic region of northern Quebec. The system is also intended to provide a solution for potential acid generation, as the tailings stack will become permanently frozen. A dry stack tailings system is also being planned for a new gold project in central Alaska. Relatively low tonnage operations. A separate tailings impoundment can be avoided all together by having a tailingdwaste rock co-disposal facility. Regions where a “dry landscape” upon closure is required. The tailings area can be developed and managed more like a waste dump and therefore avoids many of the operation and closure challenges of a conventional impoundment.

Moreover, filtered tailings stacks have regulatory attraction, require a smaller footprint for tailings storage (much lower bulking factor), are easier to reclaim and close, and have much lower long-term liability in terms of structural integrity and potential environmental impact. Figure 4 below shows a photograph of a large dry-stack tailings system. Davies and Rice (2001) present a state-of-practice overview of dry stack filtered tailings facilities.

1838

Figure 4 Example Dry-Stack Tailings Facility ThickeneNpaste technologies. It is critical before basing mining operations on new technologies to carry out adequate engineering studies to demonstrate feasibility. Several tailings disposal technologies have been introduced to the mining industry that, over the past 30 years, have not proven out to be as effective as may have been hoped. While all have contained good ideas, they have often been wholly or partially unsuccessful, or have not found extensive application to date. However, two developments will likely see renewed emphasis over the coming years. 1. Paste disposal. The development of improved thickener technology has led to tremendous advances in paste tailings for underground backfill operations. Paste tailings are essentially the whole tailings stream, thickened to a dense slurry (previously only the coarse fraction of tailings was separated from tailings for use as backfill). Cement is added to the paste and the material is pumped underground to use as ground support in mined out stopes. Advocates of paste technology have promoted its use for surface tailings disposal, claiming that it can be placed in stable configurations with the cement providing adequate strength. However, the technology has not been shown to be economically pragmatic at a large scale for many surface applications. 2. Thickened disposal. Thickened tailings are paste without the additives. Thickened disposal is a technique that has been proposed for over 25 years and has been implemented in a few operations. The main premise of thickened disposal is that tailings may be thickened to a degree that they may be discharged from one or several discharge points to form a non-segregating tailings mass with little or no water pond. In the most classical connotation, thickened tailings is assumed to form a conical mass with the tailings surface sloping downwards from the centre of the cone. A thickened tailings system, if successful, should require lower retaining dykes, as storage is gained by raising the centre of the impoundment. It had been proposed that, at the start of operations, the tailings could be thickened to a lesser degree, when flatter slopes on the impoundment would suffice, then later thickened to a higher degree as it became necessary to raise the central point of the cone. In most instances where thickened tailings was implemented, thickening technology was not capable of producing a consistent non-segregating

1839

material, so fines would form a very flat slope and require additional dyking at the toe. As well, flatter than projected slopes were experienced, and it was not possible to steepen these slopes to avoid extensive land use impact. From the above experiences, the thickened disposal did not become accepted. It has, however, been very successful in very arid regions, such as the gold mining districts of Australia. In recent years, high density thickening technology has been developed which make it useful to re-examine thickened disposal. The authors are aware of several major mining projects considering thickened tailings as an alternative management practice.

INSTRUMENTATION AND MONITORING General The reasons for instrumentation and monitoring of tailings dams are as set out by Klohn (1972). Reworded, these reasons are as follows:

To check that the environmental performance of the facility is meeting design intent, with no downstream out-of-compliance impacts on surface water quality and groundwater quality. 2. To confirm that the dam and impoundment are safe from the physical standpoint, from construction through operation and closure. 3. To provide the data required for confirmation and/or optimisation of design and construction through successive stages of impoundment construction and development. 1.

The key consideration to be accounted for in instrumentation and monitoring of tailings impoundments is that they are dynamic structures, changing nearly continuously, constructed over periods of several to many years, often under the stewardship of changing personnel. An important consideration that has come into focus in recent years is the realisation that monitoring must continue into the closure phase, and that only in rare and favourable circumstances can a truly “walk-away’’ closure scenario, requiring no ongoing monitoring or maintenance, be achieved.

Technology and Communication Improvements Technology improvements in instrumentation and monitoring, and in the ability to communicate the results, have been remarkable over the last 30 years. Instrumentation has become more advanced, robust, accurate, and reliable. It can be read more quickly and efficiently due to improved data logging equipment. It is now possible to monitor instrumentation remotely, using automatic data loggers and telemetry to transmit data to the office. Personal computers make plotting of the data in graphical form simple and rapid. Results can be emailed to the design engineer’s office for quick review. Threshold (alert) levels can be included on these plots to visually define where the data plots in relation to safe versus unsafe conditions. Photographs can likewise be emailed for quick review by other mine personnel and/or the designer. Video surveillance cameras may even have some application at very remote sites, particularly for monitoring during the closure phase. Despite these favourable trends, there is a significant caveat: technology advances are not yet at the point where they can replace visual inspection of the structure by a qualified, experienced engineer. Nor can technology serve as a substitute for application of imagination and judgement to the interpretation to monitoring data. Training of Operations Personnel and Documentation of Monitoring Program The international mining industry, and the mining industry in Canada in particular, has become increasingly focused on tailings dam safety issues. Instrumentation and monitoring represent important tools in tailings dam safety programs. It is becoming increasingly common for mining companies to require that their employees responsible for tailings dam operation receive training in dam monitoring.

1840

The instrumentation and monitoring program for a tailings dam is normally included in a comprehensive Operations Manual (a document required by legislation in an increasing number of jurisdictions). Having an Operations Manual in place is now considered to be state of practice for tailings facilities management, and provides the following benefits: 1. It provides a concise, practical document that can be used by site operating personnel for guidelines on operation and monitoring of the tailings facilities. 2. It serves as a useful training document for new personnel involved in tailings management and operations. 3. Its existence provides reassurance to senior level management, and to regulators, that formalised practices are in place for the safe operation of the facility. 4. It demonstrates due diligence on the part of the owner.

EnvironmentalPerformance Requirements for environmental performance have become more stringent, paralleling environmental regulations. Environmental performance monitoring can include the following: Surface water quality, downstream and upstream of the tailings impoundment. Groundwater quality, downstream and upstream of the tailings impoundment. Tailings pond supernatant water quality. Acid Base Accounting (ABA) testing of waste rock and tailings to determine susceptibility to acid rock drainage generation. Air quality (dust). Fisheries resources, and maintenance of minimum flows to fish bearing watercourses. Progress of revegetation where test plots and/or progressive reclamation are underway. Most mining operations have on staff an Environmental Superintendent or Coordinator, who typically takes a keen interest in the operation and monitoring of the tailings impoundment.

Dam Safety Monitoring The increasing focus on tailings dam safety brings with it an increasing awareness of the importance of a good monitoring and instrumentation program to confirm that the tailings dam is in a safe condition. Guidelines in terms of monitoring of tailings dams are provided by ICOLD (1994) and the Canadian Dam Association (1999), among others. Monitoring to confirm tailings dam safety involves the following components: Periodic, detailed visual inspections of the tailings dam and its associated appurtenant structures (spillway, decants, diversion ditches). These inspections are carried out and documented by mine personnel who have received training in potential modes of dam failure and their warning signs. Any unusual conditions or concerns must be immediately reported. Definition of “green light” (safe), versus “yellow light” (caution) and “red light” (unsafe) conditions, and a pre-determined course of action (who to contact, what to do, increased frequency of monitoring, etc.) if yellow or red light conditions are noted. Reading of instrumentation (piezometers, survey monuments, inclinometers, seepage weirs, etc.) according to a set schedule, presentation of the results in graphical form, and interpretation and reporting of the results to the appropriate personnel. The scope and frequency of dam safety monitoring will change through the life of a tailings impoundment, depending on the phase (construction versus operation versus closure) and on the consistency of monitoring results.

1841

Confirmation and Optimisation of Design and Construction Tailings dam engineering practice places considerable reliance on monitoring of the structure performance to confirm satisfactory performance, and to confirm design assumptions. Tailings dam construction, because it happens on a near-continuous basis, provides the opportunity to optimise design and construction over the life of the facility. This is the basic tenet of the observational method (Peck, 1969), a risk management method accepted and used in geotechnical engineering to avoid initial designs that may be overly conservative and overly expensive. As tailings disposal represents a cost rather than a profit centre to mining operations, the advantages of this method to optimise design and construction are obvious. The elements of the observational method are illustrated schematically on Figure 5. Monitoring and instrumentation represent an integral component of the method. The observational method has a number of limitations, as follows: The method is not suitable for failure modes that can develop very quickly, with little or no warning, examples being static liquefaction of loose tailings, or a brittle, overconsolidated clay deposit that undergoes significant loss in strength with minimal straining. Such failure modes can only be properly addressed by good design. 0 Once unfavourable conditions are noted, there must be sufficient time, and resources available, to react, putting in place measures that are pre-determined, an essential requirement of the method. The method cannot compensate for an inadequate site investigation program. The monitoring program in support of the observational method must be properly designed, not just to confirm anticipated conditions, but, more importantly, to detect unanticipated, unfavourable conditions. Failure to recognise these limitations represents an abuse of the process, which then goes from being the observational approach to the "hope for the best" approach.

Design based on most probable conditions

:onditions and the m s t unfavorable :onceivable deviations from these :onditions

Investigation sufficient to determine the general nature, panem and propelties of the deposits

+

ldenlify potentlal fallure modes, and "green llghf" vs. "red light"

r

I

\

Evaluale results oN-of /,dam , surveillanceprogram

-

Selection of parameters to be observed as construction proceeds, and wediction of values based on design analyses and assumptions

necessary based on the observed field conditions Reevaluate surveillance design program and

Design dam surveillanceprogram, quantlfy "green light" condHlons

\/

I

YES

I I

+I

Design must Include a dam surveillance program & operatlonal guldance

Field measurement of parameters and evaluation of actual conditions relative to conservative conditions assumed for +design

I

lmplement dam surveillance oroaram , -

Selection in advance of a course of action or design modificationfor every forseeable significant deviation d the Observed findings from those predicted 4+-- in design

I 1I

Response to Unusual Conditions Emergency Acllon Plan Emergency Response Plan

1

I

Calculation of values of the same parameters that would correspondto the most unfavorable conditions +--conceivable

I IL

igure 5 Elements of the Observational Method (adapted from Peck, 1969)

1842

Establish and quanllfy "yellow light" and "red light" condltlons

Operational Monitoring Tailings and water management plan must be formulated for any tailings impoundment, and these must be monitored regularly. Elements of such plans include:

0

Tailings deposition schedule; Storage versus elevation relationship for the impoundment; Operation of diversion structures; A mass balance model; and Pond filling and dam raising schedule.

Tailings and water management plans are projections, and require updating and calibration against actual conditions on a regular basis, usually no less frequently than annually. Operational monitoring data required for this purpose are as follows: 1. Measured precipitation, evaporation, runoff and snowpack data (mass balance models

2.

3. 4.

5. 6.

typically assume average annual conditions, broken down to a monthly basis). Runoff data is particularly useful in confirmation and adjustment of assumed runoff coefficients. Regular tailings beach surveys and soundings to determine above and below water tailings slopes, and to allow the elevation versus storage volume curve for the impoundment to be updated. Recording of tailings discharge points, elevations, and tonnages of tailings discharged from each point. Pond level measurements, no less frequently than monthly. Operation of reclaim barge, decants, spillways, etc. Other water inflows to the impoundment (e.g. mine water) and outflows (e.g. water discharged directly, or discharged following treatment).

The Operations Manual should describe the data required, the frequency with which it is to be collected, and the manner in which it is to be collated and reported.

CONSTRUCTIONAND OPERATION PROBLEMS Davies et a]. (2000) note that if one becomes a student of tailings impoundment case histories, an interesting conclusion arises. Tailings dam failures, each and every one, are entirely explainable in hindsight. These failures cannot be described as unpredictable accidents. There are no unknown loading causes, no mysterious soil mechanics, no "substantially different material behaviour" and definitely no acceptable failures. There is lack of design ability, poor stewardship (construction, operating or closure) or a combination of the two, in each and every case history. Tailings impoundment operational "upsets" or more catastrophic failures are a result of design and/or constructionloperationalmanagement flaws - not "acts of god". Should an severe upset or breach failure occur, several ramifications can be expected including, but not limited to: 0

0

Extended production interruption; Possible injury and, in extreme cases, loss of life (there are more than 1100 documented fatalities attributable to tailings impoundment failures, Davies and Martin, 2000); Environmental damage; Damage to company and industry image; Financial impact to mine, corporate body and shareholders; and Legal responsibility for company officers

For perspective, a few recent failures are presented with their summary characteristics and impacts.

1843

Stava, Italy, 1985 Static liquefaction collapse of two dykes, with release of 190,000 m3 mud wave, travelling 10 km down a valley and killing 268. Legal ramifications for mine owners. Bafokeng, 1974 and Merriespruit, 1994, South Africa Tailings paddocks overtopped as a result of high rainfalls on ponds with excess water storage. Release of 3 million m3 and 600,000 m3 respectively, with 12 and 17 deaths and immense property damage. Omai Gold Mine, 1995, Guyana Dam failure due to inadequate internal filter desigdconstruction led to release of cyanide solution to river. Although minimal short-term environmental impact was indicated, extended production loss, loss of share value and worldwide outrage resulted. Los Frailes, 1998, Spain Shallow foundation failure led to translation of dam shell and subsequent breach of one "cell" of tailings impoundment resulting in the release of 4 to 5 million m30f water and tailings slurry. Slurry inundated some significant areas of agricultural land. DESIGN FOR CLOSURE AND RECLAMATION OF TAILINGS IMPOUNDMENTS General Many old mining operations, and the environmental problems associated with them, are now in the care of governments. Regulators, the public, and lending agencies are no longer willing to accept mining operations resulting in long term environmental legacies for governments. Therefore, closure and reclamation considerations have become perhaps the most important driving factors in siting, design, and permitting of mining projects. Successful closure and reclamation of tailings impoundments to a secure condition are an obvious necessity. Keeping closure in mind from day one is very much in the best interest of a mine operator, as it facilitates the most effective closure and reclamation of the site, and also avoids high costs at the end of the mine life. Regulators increasingly demand that mining companies post reclamation bonds for existing or proposed new mining developments. Impounded tailings so often represent the most significant closure risk, and so it follows that these issues are best dealt with when a project is in its conceptual stages. Design Criteria for Closure The design criteria (flood, earthquake, static stability) to which tailings dams must be designed depend on the time of exposure to the hazard. Because the exposure period of closed impoundments is perpetuity, the tailings dam must in most cases be designed to endure the most extreme events, such as the Maximum Credible Earthquake (MCE) and the Probable Maximum Flood (PMF). Modern tailings dam design must account for these requirements at the outset, so that expensive retrofit measures are not necessary at the end of the mine life. Design must also account for the cumulative effects of major floods (e.g. damage to riprap and/or spillways) and multiple earthquakes in high seismic zones. Other design aspects that must be considered for closure are, for example, the erosion resistance of the dam, and the durabilility (weathering resistance) of the materials of which the dam is constructed. "Dry" Closure Where closure does not involve submergence of tailings below a water cover, modem practice requires either direct revegetation of the tailings surface, and/or covering of the tailings surface with a material that is more erosion resistant than tailings. Mining companies now typically carry

1844

out extensive reclamation studies (appropriate species, topsoil availability and requirements, etc.) during the feasibility study phase of projects. The "dry" closure scenario can be used for closure of tailings with potential for generation of ARD, in concert with collection and treatment of seepage and (if necessary) surface runoff from the impoundment. Eventually, the sulphides would be depleted, and/or the contaminant generation rate would become sufficiently low as to eliminate the need for continuing collection and treatment. The "short term" disadvantage of having to collect and treat must be weighed against the longer term advantage of improved dam safety and reduced failure consequences associated with now "inert" tailings.

Submerged Closure of ARD Tailings Impoundments Permanent submergence of tailings behind a water retaining dam to prevent ARD resolves the key geochemical issue, but places more stringent requirements on the safety of the tailings dam. The safety of a dam will not increase over time because the "drained" closure condition is not achieved. The closure spillway(s) now represents a particularly critical component of the closed impoundment, and, like the dam, must be inspected and maintained on a regular basis. The dam safety program implemented through the closure phase should be no less comprehensive than one that would be implemented for a conventional water retaining dam with similar failure consequences. In fact, the failure consequence of a tailings dam impounding sulphide tailings would be higher than that for a conventional water retaining dam. Failure of the tailings dam would, besides uncontrolled release of water, also represent an environmental failure in the release of sulphide tailings to the environment. Submerged closure of ARD tailings impoundments, therefore, is not a "walk-away'' closure scenario. Other ARD Control Alternatives Other ARD control alternatives include: Permanent neutralization of the ARD source by addingenhancing neutralization potential within the tailings mass, such as could be achieved by mixing in finely crushed limestone with the tailings; or Sulphide removal and separate storage towards the latter stages of the mine life, such that the portion of the tailings containing sulphides remains saturated and/or under very low oxygen flux. As both understanding of the ARD problem and the technologies available for mitigation of the problem continue to advance (e.g. improving water treatment technologies), it is likely that closure options less adverse to dam safety than permanent submergence will continue to be developed and become increasingly cost effective.

CONCLUSIONS A review of the state-of-practice in tailings dadimpoundment design and construction shows that great technical progress has been made. Better investigation and design tools are available. New technologies in thickening and filtration of tailings have provided the opportunities for alternatives to disposing of tailings as conventional slurries. New concepts include "dry-stack" tailings, thickened tailings and paste tailings. Geomembrane liners are commonly used where tailings may present a risk of groundwater contamination, and design and construction methods for lined impoundments have been developed. Improvements have been made to the traditional upstream construction method to reduce stability risks. Environmental considerations have become increasingly more important in tailings dam design and permitting. Closure planning has become an integral part of initial design and permitting. Designs must address ARD and include measures for long term control and/or prevention of ARD.

1845

In spite of the improvements in tailings disposal practices, a number of highly publicised failures have overshadowed the advances. These failures have led to increased scrutiny of mining projects by regulators, environmental groups, and financial institutions. Numerous guidelines and risk assessment programs have been developed by the industry to reduce tailings dam failure risks.

REFERENCES Canadian Dam Association (CDA) 1999. “Dam Safety Guidelines”. Davies, M.P. and S. Rice (2001). An alternative to conventional tailings management - “dry stack” filtered tailings. . “Tailings and Mine Waste 2001, ” Fort Collins, Colorado, pp. 41 1420. Davies, M.P., T.E. Martin and P.C. Lighthall (2000). Tailings dam stability: essential ingredients for success. Chapter 40, Slope Stability in Surface Mining, SME, pp. 365-377. Dunne, B. 1997. Managing design and construction of tailings dams. “Proceedings of the International Workshop on Managing the Risks of Tailings Disposal ”. ICME-UNEP, Stockholm, pp. 77-88. International Committee on Large Dams (ICOLD) 1994. “Tailings Dams - Design of Drainage Review and Recommendations”, Bulletin 97. International Council on Metals and the Environment (ICME) and United Nations Environment Programme (UNEP) 1997. “Proceedings of the International Workshop on Managing the Risks of Tailings Disposal”, Stockholm. International Council on Metals and the Environment (ICME) and United Nations Environment Programme (UNEP) 1998. “Case Studies on Tailings Management”. Martin, T.E. and M.P. Davies (1999). Trends in the stewardship of tailings dams. “Tailings and Mine Waste 2000, Fort Collins, Colorado. McCann, M. 1998. Sustaining the corporate memory at Inco’s Copper Cliff operations. “Case Studies on Tailings Management”, UNEP-ICME, pp. 55-56. McKenna, G. 1998. “Celebrating 25 years: Syncrude’s geotechnical review board”. Geotechnical News, Vol. 16(3), September, pp. 34-41. Mining Association of Canada 1998. “A Guide to the Management of Tailings Facilities”. Peck, R.B. 1980. “Where has all the judgement gone? The 5” Laurits Bjerrum Memorial Lecture”, Canadian Geotechnical Journal, I7{4),pp. 584-590. Siwik, 1997. Tailings management: roles and responsibilities. “Proceedings of the Znternational Workshop on Managing the Risks of Tailings Disposal ”. ICME-UNEP, Stockholm, pp. 143-158. “

1846

Hazardous Constituent Removal from Waste and Process Water Luny Twidwell’, J q McCloskey and Michelle Gale-Lee’

ABSTRACT A review of technologies appropriate for removing contaminant constituents from wastewater and plant effluents is presented. Emphasis of this presentation is placed on the removal of oxyanions of arsenic and selenium with additional consideration given to the removal of heavy metals such as copper, lead, zinc, and thallium. Current technologies and potentially new technologies appropriate for achieving present day environmental requirements are summarized and discussed.

INTRODUCTION It is indeed a formidable undertaking to summarize appropriate technologies capable of removing oxyanions and heavy metals from waste and process waters to the pg/liter concentration level in a relatively brief presentation. Therefore, the emphasis of this paper has been placed on the removal of oxyanions of arsenic and selenium. Literature review publications will be noted for the removal of heavy metal contaminants.

Environmental Standards Maximum concentration level (MCL), secondary drinking water standards and aquatic life standards are presented in Table 1. Table 1. Environmental considerations for toxic element constituents

SUMMARY OF TECHNOLOGIES Extensive literature reviews have been published that identify potential technologies for removing arsenic, selenium, and heavy metals from waste waters, e.g., arsenic (Nriagu 1994, MWTP 1994, ~

1

Montana Tech of the University of Montana, Butte MT [email protected] MSE-Technology Applications, Butte MT [email protected]

1847

Twidwell et al. 1999, Harris 2000, Welham et al. 2000, Riveros et al. 2001), selenium (MWTP 1999, Twidwell et al. 1999% 2000), heavy metals, especially copper, lead, zinc (Young 2000, SenGupta 2002), and thallium (Grzetic and Zemann 1993, Nriagu 1998, MWTP 1999% WilliamsBeam and Twidwell 2001). The identified treatments include technologies now being utilized and technologies that show potential for being implemented in the near future, examples include: precipitatiodadsorption, treated activated carbon, tailored ion exchange, and reduction processes. Detailed discussions for each of these technologies are presented elsewhere (as referenced above). An abbreviated annotation of the literature is presented in the following sections.

Arsenic Precipitation. The classic work of Chukhlantsev (1956) on arsenate solubility as a hnction of pH is often referenced in the literature. Plots of his data as log solubility versus pH have been extrapolated to predict extremely low solubilities for individual metal arsenates. Some of these extrapolations have been shown by to be greatly in error, e.g., calcium, iron, lead and barium (Robins 1981, 83, 88; Nishimura and Tozawa 1978, 85; and Essington 1988). The extrapolations are in error at higher pH values because of the relative thermodynamic stability of hydroxides and carbonates (compared to arsenates) and the extrapolations predict arsenic solubilities that are much too low in outdoor storage environments. Great care should be taken when using metal arsenate extrapolated solubility values. Calcium Arsenate-There have been a number of published investigations (Laguitton 1976, Nishimura et al. 1978, 85, 88: Plessas 1993; Robins 1984; Bothe and Brown 1999; and many others) that have demonstrated that arsenic can be effectively removed by lime precipitation, i.e., near drinking water standards are achievable if the CdAs mole ratio >3 at pH's >9 (Nishimura and Itoh 1985). Lime precipitation of arsenic with subsequent outdoor storage was the accepted industry disposal method until the middle 1980's. Robins (1981, 84) demonstrated in the early 1980's that calcium arsenate is not stable in an outdoor storage environment in the presence of air at pH levels above 8.3. Carbon dioxide in air converts the calcium arsenate to calcium carbonate with the subsequent release of arsenic back into the aqueous phase. Therefore, lime precipitation of arsenic with subsequent outdoor storage is not presently viewed as an appropriate disposal technology. However, Valenzuela et al. (2000) and Castro et al. (2000) report that lime precipitation (with subsequent calcination to crystallize the calcium arsenate compounds) is practiced at several Chilean copper smelters. Ferric Arsenate-The use of autoclave conditions allows for the formation of ferric arsenate (scorodite). The solubility of crystalline scorodite at ambient temperatures has been reported by Krause and Ettel(l989) to be <50 clgn (at pH -4). The long term stability of pond stored products have been demonstrated by INCO at their CRED facility in Sudbury, Ontario, e.g., ferric arsenate is precipitated under autoclave conditions, producing a product that is 0.5-2% As. The final arsenic bearing filter cake is transported to a tailings storage area. Krause (1992) reports that the outfall from the storage area has shown arsenic concentrations, in general, to be <20 pg/L. The Campbell Mine at Red Lake Ontario also uses autoclave formation of ferric arsenate salts. This product is stored (since 1991) along with flotation tailings. The final outfall arsenic concentration is approximately 100 pg/L (Riveros et al. 2001). Several recent studies on the Campbell tailings have shown that anaerobic reduction of ferric arsenate by sulfidic materials and bacterial activity results in increases in the pore water concentration of arsenic (Riveros et al. 2001). Detailed reviews of the stability of ferric arsenate are presented by Riveros et al. (2001) and Welham et al. (2000). Mineral-Like Arsenates-Several studies have been conducted investigating the removal of arsenic from solutions by formation of mineral-like arsenate compounds (Porubaev et al. 1972, Nikolaev et al. 1974,76, Comba 1988, Bothe and Brown 1999,99a). Porubaev et al. and Nikolaev et al. demonstrated effective removal of arsenate by lime addition in the presence of phosphate. Comba (1988) investigated the formation of mimetite, (Pb5(As0&C1). The arsenic solubility was minimal at pH 5-6, e.g., 4 pg/L,. The compound was shown to be stable in the presence of carbon

1848

dioxide by thermodynamic calculation and by experimental verification. The author suggests that mimetite may be a method of removing arsenic from process solutions because of the effectiveness of the precipitation and the ease of filtration (because of the morphology of the precipitates). However, mimetite will not pass the EPA Toxicity Characteristic Leach Procedure (TCLP) for lead because of the complexing of lead by acetate. Therefore, it cannot be considered an outdoor storable waste form. Bothe and Brown (1999, 99a) demonstrated that calcium hydroxyarsenate compounds form at ambient temperature. Arsenic solubilities at near neutral conditions were 10 pg/L and 500 pg/L for C~(0H)2(As04)2.4H20 and ca~(Aso,,)@H, respectively. Khoe et al. (1991) have reported the formation of ferrous arsenate. This compound shows a minimum solubility (at pH 7.5) of -8 p a . The authors suggest that ferrous arsenate would be environmentally stable if stored in sub-surface anoxic conditions (where ferrous iron would not be oxidized).

Adsorption. A large number of investigations have focused on adsorption of arsenic on ferric oxyhydroxide (referred to here as ferrihydrite) and on alumina/aluminum hydroxide surfaces. A relatively small portion of the literature describing the adsorption of arsenic onto these surfaces is annotated below. A detailed literature review is published for arsenic by the EPA Mine Waste Treatment Program (MWTP 1994). FmShy&tk4n important detailed review of conditions for formation and the stability of ferrihydrite is presented by Jambor and Dutrizac (1998). Some of the characteristic features of amorphous ferrihydrite formation and its conversion to more crystalline forms (goethite and hematite) include the following. The approximate formula for ferrihydrite is generally considered to be 5Fes03.9H20(Schwertmann et al. 1982,83,91; and Eggleton 1987,88). The surface area of freshly precipitated ferrihydrite is 180-300 m2/g (Schwertmann et al. 1983, 91). The rate of crystallization of ferrihydrite to hematite or goethite at 25°C is a fhction of time and p€I, e.g., conversion was half complete in 280 days at pH 4. Conversion of ferrihydrite to goethite results in a relatively large change in surface area, i.e., freshly prepared ferrihydrite showed a surface area of about 150 m2/g; when converted to goethite at 25"C the area was reduced to 92 m2/g; when converted to goethite at 90°C the area was reduced to 9 m2/g (Schwertmann et al. 1983). The fact that conversion of ferrihydrite occurs reasonably rapidly and that the conversion results in a significant decrease in surface area may hold important negative consequences for long term outdoor storage stability for surface adsorbed arsenic. However, the ferrihydrite conversion rate may be mitigated (changed from days to years) by the presence of other species and solution conditions during precipitation. Factors that decrease the rate of conversion of ferrihydrite to more crystalline forms include: lower pH (Schwertmann and Thalmann 1976, Schwertmann and Murad 1983), lower temperatures (Cornell and Schwertmann 1996), presence of silicate (Schwertmann and Thalmann 1976, Schwertmann and Fechter 1982, Cornell et al. 1987a), aluminum (Schwertmann 1984, Schulze and Schwertmann 1984), manganese pavies 1984, Cornell and Giovanoli 1987, Scott 1991) and organics (Schwertmann 1966, Cornell and Schwertmann 1979, Cornell 1985). FerrihydritdArsenic-Studiesinvestigating the adsorption of anions on ferrihydrite surfaces are not new. Municipal water treatment systems have utilized ferric, ferrous and aluminum hydroxide precipitation for many years to cleanse heavy metals and phosphates from solution as a final polishing stage. In fact, doping manuals are available: ferric chloride (AWWA 1988), aluminum sulfate (AWWA 1988a). The US EPA has chosen ferric precipitation as the Best Demonstrated Available Technology (BDAT) for arsenic bearing solutions. EPA notes, however, that arsenic bearing ferrihydrite still leaches and may require hrther treatment, such as vitrification. EPA still has concerns about the stability of arsenic bearing ferrihydrite precipitates under long term storage conditions (Rosengrant and Fargo 1990). Several recent and detailed reviews of arsenic adsorption and stability considerations are presented by Twidwell et al. (19991, Harris (2000), Welham et al. (ZOOO), Riveros et al. (2001), and SepGuta (2002a).

1849

Pierce and Moore (1980) have reported that ferrihydrite has a tremendous adsorption capacity for aqueous As(V), i.e., greater than 1.0 mole As(V)/mole Fe (maximum at pH 4). Puls and Powell (1991) also reported significant As(V) adsorption on ferrihydrite, e.g.. up to one percent arsenic. Fuller, et al (1993) have also confirmed the significant capacity of ferrihydrite for arsenate adsorption, e.g., 0.7 mole As(V)/mole Fe was achieved when iron was hydrolyzed and precipitated in the presence of As(V) in the pH range 7.5-9.0. Their work also showed that adsorption of A m on previously precipitated ferrihydrite was very effective but much less effective than when the iron was precipitated in-situ with the As(V), e.g., 0.25 mole Aslmole Fe was achieved by solid slurry adsorption. The authors also reported that the rate of As(V) removal was much faster for insitu iron hydrolysis and precipitation (<5 minutes) than for the slurry adsorption (which showed a rapid pick-up, then a continued slow adsorption for days). One must be aware that the test work demonstrated that there was a release of a portion of the adsorbed arsenic with time, e.g., desorption at pH 8 resulted in an increase in the Fe/As mole ratio in the solid fiom 5.3 to 8.3 over a 750 hour test period. This represents a significant decrease in the solids arsenic content. Also, Belzile and Tessier (1990) add to the concern that adsorbed As(V) on ferrihydrite may not be a long term stable storable waste form. They have compared the data (over the pH range 4-8) of Pierce and Moore (1980) for adsorption on ferrihydrite with the data of Hingston (1968) for adsorption on crystallinegoethite. The adsorption equilibrium constants for the two materials are orders of magnitude apart, i.e., adsorption is much more favorable on ferrihydrite than on goethite. Their conclusion was that “goethite cannot efficiently compete with amorphous Fe oxyhydroxides for adsorbing arsenaie in natural waters.” An important unknown at this time is whether the product &om ferrihydrite adsorption of arsenic will be stable if storage conditions become anaerobic. Brannon et al. (1987) have demonstrated that anaerobic lake sediments convert As(V) to As(IlI) (pH 5-8.0). However, when the anaerobic conditions were shifted by aerobic leaching the previously reduced As(III) was reconverted to more immobile As(V) which was associated with aluminum and iron oxyhydroxides. Masschelegn (1991) found that arsenic concentration in moderately reducing soils (0-100 mv) was controlled by the dissolution of oxyhydroxides and that the arsenic concentration in solution increased by a factor of 25 times when the redox potential was reducing, e.g., -200 mv (pH 5). Also unknown is the effect of bacteria in buried storage systems.

Removal of arsenic fiom solution by ferric precipitation has been and is presently practiced at numerous facilities, e.g., the Noranda Home Smelter, the Giant Mine, the Con Mine, and the Teck-Corona mine (Riveros et al. 2001); the Kennecott Utah Smelter, Placer Dome Lonetree and Getchell mines (on a periodic basis), and Barrick’s gold mining operations in Nevada (McCloskey 1999). An example at a mine water site (pilot scale demonstrations) is the Susie Mine, Rimini, MT (MWTP 1997). A case study is presented by Banerjee (2002) to illustrate the optimization and utilization of the fmihydrite adsorption technology for the removal of arsenic fiom a groundwater at a northeastern US industrial facility. The water contained approximately 30 m g L arsenic as As(III), As(V), and organic arsenic. A hll scale treatment plant is presently in operation and has demonstrated consistent arsenic removal from 18-48 m g k to <5 p&. The plant treats 550,000-750,000 literdday groundwater using a FdAs weight ratio of 10, hydrogen peroxidae weight ratio of 0.5, and pH of 7.5-8.0. Other examples for the treatment of wastewater and/or effluent plant water by femhydrite precipitation and adsorption are presented by the US EPA (EPA 2000,Ol). Sengupta and Greenleaf(2002) present a case study based on the work of Chwirka et al. (2000) for arsenic removal from Albuquerque, NM drinking water. Chwirka et al. investigated the use of Ion Exchange (I-, Activated Alumina (AA), Nanofiltration (NF), and Ferric

1850

CoagulatiodMicrofiltration( C M ) for lowering the As(V) concentration from 100 to lop@. They concluded that Ch4F was the least expensive option for the treatment of 8,700 m3/d and that option has been selected for implementation in Albuquerque. The cost estimate for each process was: capital cost (1999 US$); M $5.2 million; AA $4.6 million; NF $3.9 million; C/MF $4.1 million and annualized capital plus O&M yearly cost; IX $447,000;AA $444,000; NF $390,000; C/MF $273,000. Chwirka and Narasimhan (2002) present a spreadsheet that is available to the reader to compare the capital and operating cost of the four technologies presented above. The details for using the spreadsheet (which can be downloaded) are presented in the reference. Other example publications illustrating the application of femhydrite adsorption in drinking water treatment plants are presented by EPA (2000), Hering et al. (1996), McNeill and Edwards (1999, and Cadena and Kirk (1995). Fm’hydrita/AArsenic Summary-Two femc precipitation arsenic removal technologies are presently practiced by industry, i.e. ambient temperature fzrrihydrite precipitation and elevated temperature autoclave precipitation of ferric arsenate. The adsorption technology is relatively simple, can be performed at ambient temperatures, and the presence of commonly associated metals (copper, lead, zinc) and gypsum have a stabilizing effect on the long term stability of the product. The disadvantages of the adsorption technology are the formation of voluminous waste material that is difficult to filter, the requirement that the arsenic be present in the fblly oxidized state (arsenate), and the question as to long term stability of the product in the presence of reducing substances. The autoclave technology produces less waste product, the product is easily filtered into a dense, compact filter cake, and less iron is required to sequester the arsenic. The disadvantages of the femc arsenate precipitation are that the treatment process is more capital intensive, the compound may dissolve incongruently if the pH is >4, and it may not be stable under reducing or anaerobic bacterial conditions (Riveros et al. 2001, Rochette et al. 1998). Riveros et al. (2001) concluded fiom their extensive review of the literature that “for practical purposes, arsenical femhydrite can be considered stable provided the FdAs molar ratio is greater than 3, the pH is slightly acidic and that it does not come in contact with reducing substances such as reactive sulphides or reducing conditions such as deep water, bacteria or algae.” Swash, et al (2000) have concluded that fiom the point of view of safe disposal of arsenic, that there is no clear experimental evidence yet favoring the low temperature precipitates over the high temperature precipitates or vise versa (Riveros et al. 2001). A detailed review of the stability of scododite, ferrihydrite and femhydrite/arsenate adsorption is presented by Welham et al. (2000). Their conclusions include “there are significant problems with the use of jarosite and scorodite as phases for the disposal of iron and/or arsenic from metallurgical systems. Neither phase is stable under typical atmospheric weathering conditions with transformation to goethite predicted to occur. The currently permitted discharge level of arsenic is only achieved due to the slow kinetics of the transfonnation releasing arsenic over time. Crystalline scorodite is two orders of magnitude less soluble than amorphous iron (HI) arsenate precipitate often formed in low temperature systems.” AlrcmidAluminum Hydrcucide-Alumina and aluminum hydroxide have been investigated for arsenic removal from solution by adsorption, example publications include: Hingston et al. (1970, 72), Gulledge and O’Connor (1973), Anderson (1974, 76), Leckie et al. (1980), Diamadopoulosand Benedek (1984), Ghosh and Teoh (1985), Goldberg (1986), and Lake (1990). Activated alumina is widely utilized as an adsorbent in treatment of drinking waters. Example publications covering this subject include: Ghosh and Teoh (1985), Rozelle (1986), Fox and Sorg (1987), Fox (1989), and Lake (1990). Lake (1990) concluded that activated alumina is competitive in effectiveness for As(V) adsorption with other drinking water purification adsorbents and is the most cost effective adsorbent for drinking water treatment. Stewart and Kessler (1991) conducted a pilot study for treatment of six million liters of water by activated alumina. Adsorption reduced the As(V) concentration to analytical detection limits; filters lasted for over a hundred days and operating costs were estimated to be $0.20/1000 liters.

1851

The use of activated alumina has not been extensively studied for application to dilute solution effluent wastewater or contaminated groundwater. Some studies have been conducted that demonstrate it may be applicable to these waters. Sisk et al. (1990)evaluated the use of activated alumina on a pilot scale to treat As(V) bearing groundwater. Alumina and aluminum hydroxide were shown to be ineffective for As(LU) species except at pH> 9.The authors found that activated alumina was capable of extracting As(V) (at pH 5) to <50pgL. The adsorption process is greatly affected by the presence of competing anionic species, especially sulfate. As(V) is much more effectively adsorbed by amorphous aluminum hydroxide than by alumina but ferrihydrite adsorption is much more effective than either of the two. Activcrted Gubon-Activated carbon adsorption (ACA) has been used extensively for treating drinking waters but only when the waters contained concentrations near the drinking water standard and when the arsenic is present as A$V) (Fox and Sorg 1987). ACA treaiment of arsenic bearing multicomponent wastewater has been shown to be relatively ineffective (Fu 1983,Huang and Fu 1984,Sisk et al. 1990,Wolff and Rudasill 1990, Rajakovic 1992,and many others). Rajakovic (1992)reported on the use of metal impregnated AC. He found that copper was the most effective additive for enhancing @H 4-9)the adsorption of arsenic. The arsenic adsorption was increased from essentially no adsorption to 48 mg Adg C. Huang et al. (1984,89) have also investigated carbon adsorption and metal ion-doping as a means of enhancing arsenic extraction, e.g., these investigators doped activated carbons with various metals; barium, copper, ferrous and ferric iron. Ferrous iron was found to be greatly superior to all the other metals tested. Enhancement in As(V) adsorption was ten-fold (for pH range 4-5)over acid washed AC. Sulfuric acid was used to strip the adsorbed arsenic and the carbon was then regenerated by soaking in a ferrous salt solution. Essentially complete arsenic extraction was achieved fiom a 15 mg/L solution by the stripped and regenerated carbon. Ion Exchange (IX). Ion exchange has been evaluated by a number of investigators for polishing drinking waters and dilute solution groundwater (Ghosh 1985, Rozelle 1986, 87, Fox 1989, Lankford 1990,Rajakovic 1992,SenGupta et al. 2002,SenGupta and Greenleaf 2002,Zappi 1990, and many others). The evaluations have often involved comparative removals with other drinking water clean up technologies, e.g., RO, AA, IX,NF, Femc C W , and ACA. In general ion exchange has been shown to be competitive in effectiveness for removing arsenic from drinking waters. However, other treatment processes have been shown to be more economical, especially femc C/MF (Chwirka and Narasimhan 2002).

There have only been a few investigations concerned with M treatment of As(V) bearing multicomponent waste water. Chanda et al. (1988) investigated the use of ferric form resin exchangers @ow XFS-4195,Chelex 100) and found that the resins were several fold more effective than ferrihydrite for As(V) adsorption. Arsenate was adsorbed to below drinking water standards, was effectively stripped fiom the resin, and the resin could be reactivated and repeatedly reused. The saturation loading capacity of the resins varied between 45-70mg/g of wet resin. This study was conducted on synthetic arsenic bearing solutions and did not contain competing anions (such as sulfate, phosphate, humic). Hauck et al. (1990)also investigated femc form ion exchangers and found complete arsenic removal from wastewater. Clifford et al. (1990) investigated strong-base chloride anion-exchange resins. They demonstrated essentially complete extraction of As(V) in the presence of As(II1). Sisk et al. (1990) investigated the removal of arsenic from groundwater (on a pilot scale basis) by two ion exchange resins (Amberlite IRA 402 and Ionac A-641).Both ion exchangers reduced the arsenic content to below the drinking water standard. Reverse Osmosis (RO)/Membrane Separation. RO is the most universally used treatment technology for point-of-udpointf-entry treatment of drinking waters (Huxstep and Sorg 1981,

1852

88, Rozelle 1987, Fox 1989, Clifford 1990, Gandilhon et al. 1992, and many others). Rogers (1989) studied point-of-use RO using polyamide membranes at 78 sites. Arsenic (0.059 mfl), iron, manganese, chloride and total dissolved solids were lowered to below the drinking water standard over a twenty month test period. Fox and Sorg (1987) evaluated five different membranes (in a small continuously operating pilot scale test system) on natural Florida groundwater containing 15 elements. They found all the membranes removed >95% of all dissolved species. Fox and Sorg (1987) report that five techniques are in use for point-of-use treatment: reverse osmosis, activated alumina, ion exchange, granular activated carbon, and distillation. They state that EPA has approved the first three of these for removal of inorganic contaminants as Best Available Technologies (BAT). Fox (1989) found that low pressure RO was not completely effective at high arsenic concentrations (i.e., at 1.08 mg Adliter). However, high pressure units handled all concentrations from 0.1-1.0 mg Miter). There are numerous other references illustrating the application of RO to drinking waters that are not quoted here.

Reduction Processes. Removal of arsenic by reduction with iron has been reported by several investigators (Santana 1996, McCloskey 1999, Dahlgren 2000, Nikolaidis 2002, Cockhill 2002, and Hadden 2002). Arsenic has been stripped to less than analytical detection limits from laboratory samples utilizing iron reduction technology (Santana 1996, Dahlgren 2000, Hadden 2002). MSE has utilized a proprietary catalyzed cementation process to strip arsenic from a large variety of process and mine waters to less than analytical detection limits. The process has been demonstrated on a pilot scale (1-5 gallondmin) at an industrial site (McCloskey 1999).There are a number of studies investigating the use of iron bearing reactive permeable barrier walls to treat groundwater plumes. Most of the reported work has dealt with the destruction of organic compounds, reduction of Cr(VI) to C r o , or radionuclides (O’Hannesin and Gillham1992, Kaplan et al. 1994, Cantrell et al. 1995, Powell et al. 1995, Appleton 1996). The use of zero valence iron appears to hold promise for future applications for arsenic, selenium and heavy metal removal. The advantages of the technology include: the process is independent of the arsenic species valence state; it is not influenced by the present of sulfate and other anions; and heavy metals more noble than iron are coextracted.

SeIenium Precipitation. Precipitation of selenate and selenite compounds as a water treatment technology to lower selenium to
1853

Parida et al. (1997) investigated the adsorption of Se(lv) onto various surfaces. They found that ferrihydrite adsorbed significantly more Se(1V) than the other phases. The order of adsorption was ferrihydrite (225 m2/g) > GFeOOH (1 17 m2/g) > yFeOOH (61.1 m2/g) > PFeOOH (77.8 m2/g) > aFeOOH (70.8 m2/g), e.g., at 20 mg Se(IV)/L, ferrihydrite was loaded to 48 mglg while ctFeOOH was only loaded to 10 mg/g. Approximately ninety percent removal of Se(Iv) was achieved fiom a 20 mg Se(IV)/L. solution by using 1.2 g femhydritefl. at a pH of 3. Whereas, aFeOOH removed only about 25% of the Se(1v) under the same conditions. Adsorption was a relatively strong knction of solution pH, i.e., adsorption fell rapidly as the pH was increased fiom 3.5 to 9.5. There was essentially no adsorption at or above pH 9.5.Note that even at pH 3 removal to the pg& was not achieved. MSE piloted the ferrihydrite adsorption process for the EPA MWTP on a smelter wastewater that contained two mg/L Se(vI) at a pH of 5-8. Removal to <50 pg/L required extremely high additions of ferric ions (MWTP 2001). Balistrieri and Chao (1990) investigated the role of the presence of other anions on the adsorption of selenium by ferrihydrite. The order of adsorption at pH 7 was phosphate > silicate = arsenate > bicarbonate/carbonate> citrate = selenite > molybdate > oxalate > fluoride = selenate > sulfate. Therefore, the adsorption of selenium is an anion that is in competition with other anions. Hayes et al. (1987) studied the adsorption of selenium oxyanions on cLFeOOH surfaces. Their conclusion was that Se(IV) adsorbs in a bidentate manner (inner-sphere adsorption which is much stronger than outer-sphere adsorption). Se(VI) adsorbs as an outer-sphere hydrated complex, which is why it can be easily replaced by other solution anions such as sulfate. Merrill et al. (1986, 87) and EPRI (1980, 85) found that the optimum pH for Se(IV) removal was 6.5 and optimum iron dosage was 14 mgL for a water initially containing 40-60 pg& SHIV). The effluent selenium concentration resulting fiom the treatment of 115 literdminute in a continuous pilot facility was < 10 pgL. WSPA (1995) reported similar results for the treatment of a petroleum biotreater effluent (the solutions contained cyanide as well as selenium and sulfate), e.g., the optimal pH was 5-7; optimal ferric chloride dosage was 14-28 mg iron/L.. Selenium adsorption occurs on precipitated ferrihydrite. However, the use of ferrihydrite adsorption is not utilized industrially because: even though Se(IV) is effectively removed at pH<8 rarely can pg/L concentrations be achieved; S e ( V I ) is poorly adsorbed at any pH (which mean that reduction of the Se(VI) prior to adsorption is required); the presence of other aqueous species in the solution influences the removal of both Se(1v) and Se(VI); and long term stability of this product in outdoor storage is questionable. AlumindAluminum Hydroxide -Activated alumina has been studied for selenium adsorption by several investigators: Trussell et al (1980, 91), Yuan et al. (1983), Hornung et al. (1983), Altman and Hegerle (1993). Trussell et al. (1980) investigated the adsorption of Se(1v) and Se(VI) on activated alumina. Se(1V) was effectively adsorbed to analytical detection limits over the pH range 3-7 (for 100-200 pg SeL, one hour exposure). The loading capacity was 90 mg S e L of alumina. Se(VI) adsorption was much less effective (loading capacity was 7 mg Se/L of alumina). The order of selectivity for anions over the pH range 3-7 was hydroxide > arsenate > selenite > sulfate > selenate > arsenite. Sulfate and bicarbonate had little effect on Se(IV) but greatly affected Se(vI) adsorption. Trussell et al. (1991) demonstrated that selenium adsorbed on alumina could be effectively stripped using 0.5% NaOH. A cost estimate was presented for treating 3.8 million literdday containing 100 pg Se/L. The cost estimate was 6dlOOO liters for Se(1V) and 21d1000 liters for Se(vI). The above studies were conducted on groundwater not on mine waters. Investigators Altman and Hegerle (1993) and Batista and Young (1994, 97) have applied alumina adsorption to an actual mine effluent (FMC Gold Paradise Peak Mine in Nevada). They demonstrated that mine waters

1854

that have appreciable silica present cannot be effectively treated by alumina adsorption. The alumina adsorption capacities for selenium and silica were determined to be 0.2 mg S e / L alumina, 3.6 mg S f i alumina, respectively. In the absence of silica the alumina adsorption capacity for selenium was 0.9 mg/L alumina. Activated Cnrbon-Activated carbon adsorption of either Se(1V) or Se(VI) has been shown to be ineffective, e.g., SHIV) or S e O at concentrations from 30-100 pg/L showed < 4’Yo removal using dosages of activated carbon up to 100 mg/L (Sorg 1978).

Ion Exchange. IX is used widely for treating drinking waters, dilute metal bearing solutions, wastewater, and groundwater. However, its application for removing selenium has some successes and some failures (Maneval et al. 1985, Boegel and Clifford 1990); see the MWTP (1999) for a more detailed discussion. The Western States Petroleum Association (WSPA 1995) investigated the use of IX (strong base anion resins) for treating refinery wastewater (sourwater and biotreater effluent, containing 114,870 pg SeL). Their conclusions included: IX is very inefficient for selenium because other anionic species may load in preference to the selenium, especially sulfate. SHIV) loaded well but after 200-600 bed volumes sulfate displaced the selenium. IX is not able to routinely produce effluentswith <50 pg/L total selenium. Efforts to identify selectiveresins were unsuccesshl. Tailored chelating polymer resins have been investigated (Ramana and Sengupta 1992) for extracting selenium in the presence of high concentrations of sulfate. The resins investigated included Dow 2N (loaded with copper, 1.7 meq/g) and IRA-900 (no copper loading). The resins were evaluated for extraction in a solution containing 250 mg sulfate/L. The selectivity sequence at pH 9.5 was selenate > sulfate > selenite > nitrate > chloride. Resin loading capacities and cost evaluationswere not presented by the authors. Virnig and Weerts (1993) considered the use of liquid ion exchange as a way to treat spent gold heap leach effluent. The ion exchange reagent was CyanoMet I$ a proprietary reagent. The reagent extracted anion complexes including gold, silver, nickel, copper, zinc, and selenium. The extracting reagent contained 30 weight percent CyanoMet R. There were three stages of extraction followed by two stages of strip (four percent NaOH). The selenium removal from the influent feed (1 1 mg Se/L) was very good, e.g., the final raflinate solution contained 70 pg Se/L. Pilot studies were conducted at a Nevada gold mine. The feed rate was 1 gallon/minute. The selenium removal fiom the influent feed (1.7 mg S&) was very good, e.g., the final rafinate solution contained 36 pg S e a . This technology was not implemented at the mine site.

Membrane Separations. Reverse Osmosis (R0)-RO is listed by EPA as one of the Best Available Technologies (BAT) for selenium removal (Pontius 1995, Kapoor et al. 1995), i.e., the removal effectiveness is quoted as being >SO% ifregardless of valence state. Reverse osmosis/membrane ultrafiltration processes require that the solutions to be treated contain very dilute concentration of solids. Therefore, pretreatments are normally required in order not to foul the separating membrane. Whereas, RO is readily applicable to drinking waters it is doubtikl that this separation technique is applicable to acid mine waters except as a final polishing step. N@m$&&n (Nfl-NF appears to be a technology on the horizon for treating some low metal containing selenium bearing mine waters (Kharaka et al. 2000). NF has been used commercially for sulfate removal from seawater prior to injection into oil field reservoirs (Bilstad 1992); for sulfate removal fiom concentrated brines (Kharaka et al. 2000); for organic compound removal from paper plant effluents (Afonso et al. 1992); and for organic compound removal from groundwater (Fu et al. 1994). Publications demonstrating the application of NF to mine waters

1855

were not found but the technology has been applied to agricultural waters (high in Se(VI), sulfate and total dissolved solids) taken fiom San Joaquin Valley drainage (Kharaka et al. 2000). Nanofiltration is based the use of membranes constructed of a porous inert layer of polysulfone and a negatively charged hydrophobic rejection layer. These membranes reject multivalence anions, including sulfate. The technology is similar to RO but the NF system is operated at pressures that are about one-third of that required for RO. Kharaka (2000) demonstrated >95% removal of selenium for waters (pH 6.3-8.5) containing 24-308 pg Se(VI)/L; 2,080-26,100 pg sulfatd; and 780-38,800 pg TDS/L. Similar recoveries were demonstrated for Se(VI) concentrations up to -1000 pgL.

Reduction Processes. There are a number of proposed reduction processes in the literature. However, only metallic and biological reduction appear to be capable of lowering the selenium to
1856

Heavy Metals Precipitation. Banerjee (2002) has reviewed the literature on heavy metal removal by compound precipitation. His conclusion is that hydroxide precipitation of heavy metals (copper, lead, zinc), in a properly designed system, can achieve metal removal to about 500 p a and if sulfide precipitation is utilized removals to about 100 pg/L can be achieved. However, for achieving the removal of heavy metal concentrations to
SUMMARY Various technologieshave been evaluated for removing arsenic from aqueous solutions. Several of the technologies appear appropriate for treating drinking water where the initial arsenic concentration is already relatively low ( 4 0 0 p@) and where there are few competing anions present, e.g., ferrihydrite precipitatiodcoagulation; alumina adsorption, membrane, and ion exchange technologies are all capable of achieving the new US EPA MCL of 4 0 pg&. However, only a few technologies appear to be appropriate for treating multicomponent arsenic bearing waste and mine waters. The US EPA BDAT and the most commonly utilized technology is ferric arsenate precipitation andlor fmihydrite adsorption. Both of these technologies are capable of achieving arsenic concentrations that are <10 pg/L. It appears that the use of these technologies will contiixie to dominate into the near hture. However, because long term stability of the ferric

1857

products in outdoor storage environments is still a concern, other technologies must continue to be investigated. The other nonadsorption technologies discussed above must also be scrutinized with respect to final arsenic bearing product disposal or storage. The membrane and ion exchange processes produce a concentrated brine solution that still must be further treated. The metallic reduction process appears to be capable of lowering arsenic concentrations to analytical detection limit levels, however, the final product will be an arsenic bearing metallic product that must also be further treated or stored in an environmentally acceptable manner. The US EPA BDAT for removing selenium from wastewater is ferrihydrite adsorption. The technology is inappropriate for achieving selenium concentrations of <10 p a unless the selenium exists exclusively as Se(1V). Those technologiesthat appear to be potentially appropriate for achieving selenium concentrations of < 1OpgL include: membrane technologies, such as nanofiltration; tailored ion exchange; and reduction processes such as metallic iron reduction and biological reduction. The membrane and ion exchange processes produce a concentrated selenium bearing solution that must be firther treated. The reduction processes produce either a selenium bearing metallic sludge or a selenium bearing biological sludge material. Both these product contain elemental selenium at relatively high concentrations. The disposal options for handling these products must be further considered but one of the options includes the recovery of elemental selenium as a potentially marketable feedstock. The final solution to the problem of how to safely dispose of arsenic and/or selenium bearing waste forms has yet to be determined.

ACKNOWLEDGMENTS Literature reviews for arsenic and selenium removal technologies were conducted for the US EPA, Office of Research and Development, Mine Waste Technology Program. The work was performed under an Interagency Agreement between the EPA National Risk Management Research Laboratory and the US Department of Energy National Energy Technology Laboratory (NETL). Montana Tech of the University of Montana is a subcontractor to MSE Technology Applications, Inc. which manages the MWTP through the NETL at the Western Environmental Technology Ofice under contract No. DE-AC22-96EW96405. REFERENCES Adams, J., T. Pickett, J. Hogge. 2000. Biotreatability for Se removal process, In: Minor Elements ’ 2000. (Ed) C. Young, SME, Littleton, CO. 53-66. Afonso, M. et al. 1992. Nanofiltration removal of chIorinated organic compounds from alkaline bleaching effluent in a pulp and paper plant. Wat. Res. 26, 1639-43. Altringer, P., D. Larsen, K. Gardner. 1989. Bench-scale process development of selenium removal from wastewaste using facultative bacteria. In: Biohydromet. I989, CANMET, 643-57 Altringer, P., R Lien, K. Gardner. 1991. Biological and chemical selenium removal from precious metals solutions, In: S p p . Em. Mmgemenffor the 1990’s. SME-AIME, Littleton, CO. 135-42. Altman, K., K. Hegerle. 1993. Se removal processes considered for the heap leach detoxification project paradise peak mine, FMC Gold. In: Gearingfor a New Era in Gold, Beaver Creek, CO. Randol Gold, Denver, CO. 349-53. Anderson, M. 1974. Arsenate adsorption on amorphous Al hydroxide. MS Thesis (74). PhD Thesis (76) John Hopkins Univ.; and, 1976. Anderson, M., J. Ferguson, J. Gavis. Arsenate adsorption on amorphous Al hydroxide. J ColloidandInterjace Sci. 54, (3), 391-99. Appleton, E. 1996. A nickel-iron wall against contaminated groundwater, E.S.&T. 30 (12) 537-39. AWWA . 1988. AWWA standard for Fe chloride--liquid, ground, or lump. AWWA,VIII, B407-88, 1Sp; and, 1988a. AWWA standard for Al sulfate--liquid, ground, or lump. AWWA,IX, B403-88, 17 p.

1858

Balistrieri, L., T. Chao. 1987. Selenium adsorption by geothite. Soil Science Soc. Am. J. 51, (5); and, 1990. Adsorption of Se by amorphous iron oxyhydroxides and manganese dioxide. Geochim. Cornwhim- Acta, 54,739-51. Banerjee, K. 2002. Case studies for immobilizing toxic metals with iron coprecipitation and adsorption. In Env. Separation of Heavy Metals. (Ed) SenGupta, Lewis Publ., NY, NY. 18 1-204. Batista, J., J. Young. 1994. The influence of aqueous silica on the adsorption of Se by activated alumina. AAWA Water Research, 167-81; and, 1997. Removal of Se from gold heap leachate by activated alumina adsorption. M S , Warrendale, PA 29-36. Belzile, N., A. Tessier. 1990. Interactions between As and iron oxyhydroxides in lacustrine sediments, Gemhim. et Cosmochm. Acta. 54, 103-10. Benjamin, M., N. Bloom. 1981. Interactions of strongly binding cations and anions on amorphous iron oxyhydroxide. In: Ahorptionfiom Aqueous Solutions7(Ed) P. Tewari, N.Y., N.Y. Plenum. Benjamin, M., K. Hayes, J. Leckie. 1982. Removal of toxic metals from power-generation waste streams by adsorption and coprecipitation.JWPCF. 54, (1 l), 1472-8 1. Bilstad, T. 1992. Sulfate separation from seawater by nanofiltration, In: Produced Water. (Eds) J. Ray, F. Engel, Plenum Press. NY,NY.503-09. Boegel, J., D. Clifford. 1986. Se oxidation and removal by ion exchange, EPA/600/2-86/031, Washington, DC. Bothe, J., P.W. Brown. 1999. As immobilization by calcium arsenate formation, EmSci.Technol0g. 33(21). 3806-1 1; and, 1999a. The stabilities of calcium arsenates at 23f1°c, J: H a . Materials. 69(2). 197-207. Brannon, J., W. Patrick. 1987. Fixation, transformation and mobilization of arsenic in sediments, Em.Sci. Tech.2 1,450-59. Cadena, F., T. Kirk. 1995. Arsenate precipitation using ferric iron in acidic conditions. WRRI Rep. 293. New Mexico WaterResources Research Institute. Las Cruces, NM. Cantrell, K., D. Kaplan, T.Wietsma. 1995. Zero valence iron for the in-situ remediation of selected metals in groundwater.J: Haz. Materials. 42,201-12. Castro, S., P. Munoz. 2000. Removal of As(lII) in solution by co-precipitation of calcium arsenite and calcium hydroxide. Fzph Int. Con$ on Clean Technologiesfor the Mining Industries. (Eds) M. Sanchez, et al. Univ. Concepcion, Chile. 121-29. Chandra, M., et al. 1988. Ligand exchange sorption of arsenate and arsenite by chelating resins in Ion Exch., Sorbents 8, ferric ion form: iminodiacetic chelating resin Chelex lOO.-React. Pol’., (l), 85-95; and, 1988a Ligand exchange sorption of arsenate and arsenite by chelating resins in ferric ion form: weak-base chelating resin XFS-4195. 7, (2-3), 251-61. Chukhlantsev, V. 1956. The solubility product of metal arsenates, Z k .Neorg. Khim.,I, 1975-82, II, 529-35; Unrr. Anal. Khim.,11,529-35;J Anal. Chem., 11,565-71. Chwirka, J., B. Thomson. 2000. Removing As from groundwater. JAWWA 92 (3), 79-88. Chwirka, J., R Narasimhan. 2002. AWWA arsenic treatment cost estimating tool. Web Site: www.awwa.ordgovtafflAWWAArsenicCostsRev.xls. Clifford, D., 1990. Ion ExchanEe and Inorganic Adsorption, Water Quality and Treatment, McGraw-Hill Inc. NY, NY. 561-640. Cockhill, L. 2002. Se removal by column cementation,MS Thesis.Montana Tech, Butte, MT. Comba, P., L. Twidwell. 1988. Removal of As from process and wastewater. Proc. AsMetaZZurgy Fundamentals andApplications, (Eds) R Reddy, et al. TMS, Warrendale, PA. 305-19. Corapcioglu, O., C. Huang. 1987. The adsorption of heavy metals onto hydrous activated carbon. WaterResearch, 21, 1031-44. Cornell, R 1985. ERkt of simple sugars on the alkaline transformation of ferrihydrite into goethite and hematite. Clays Clay Minerals, 33, 219-27; and, Cornell, R., R. Giovanoli. 1987. Effect of Mn on the transformation of femhydrite into goethite and jacobsite in alkaline media. 35, (I), 11-20; and, Cornell, R., R. Giovanoli, P. Schnidler. 1987a. Effect of silicate species on the transformation of ferrihydrite into goethite and hematite in alkaline media. 35, (l), 21-8. Cornell, R., U. schwertmann. 1979. Influences of organic anions on the crystallization of femhydrite. Clays C l q Minerals. 27,402-1 0. Cornell, R., U. Schwertmann. 1996. The Iron Oxides. VCH. NY, NY. 574 p.

1859

Davies, S. 1984. Mn(II) oxidation in the presence of metal oxides. PhD Zhesis, Calif Inst. Tech., Pasadena, CA, 170 p. Dahlgren, E. 2000. Sonic enhancement of Se removal by cementation fiom wastewater, MS Thesis.Montana Tech, Butte, MT, 169. Diamadopoulos, E., A. Benedek. 1984. Al hydrolysis effects on P removal from wastewater. JWPCF, 56, (1 I), 1165-72. Downing, et al. 1988. Removing Se from water. US Pat. 4,725,357. Dzomabak, D., F. Morel. 1990. Surface complexation modeling: hvdrous ferric oxide. John Wiley and Sons, NY,NY. Eggleton, R 1987. Noncrystalline Fe-Si-Al-oxyhydroxides. C&y, C&yMiner&. 35 (l), 29-37; and, Eggleton, R., W. Fitzpatrick. 1988. New data and a revised structural model for ferrihydrite, 36 (2), 113-18. EPA. 2000. Technologies and costs for removal of As from drinking water. EPA-R-00-028. EPA Office of Solid Waste and Emergency Response, Washington, DC. EPA. 2001. Treatment technologies for site cleanup: annual status report. EPA-542-R-01-004. EPA Officeof Solid Waste and Emergency Response, Washington, DC. EPRI. 1980. AdsorptiodCoprecipitationof trace elements from water with iron oxyhydroxide. Electric Power Res. Znst. WRI CF-1513, Palo Alto, CA. Ergas, S. et al. 1990. Microbial process for removal of Se from agricultural drainage water, Univ. Calif At Davis, Report to USBOR, Contract 9-FC-20-07720. Essington, M. 1988. Solubility of barium arsenate. SoiISci. Soc.Am.. 52. 1566-70. Fox, K. 1989. Field experience point-of-use systems for As removal. J A W A . 81, (2), 94-101; and, Fox, K., T. Sorg. 1987. Controlling As, F, and U by point-of-use treatment. 79, (lo), 81-8. Franzreb, M., J. Watson. 2002. Elimination of heavy metals from wastewater by magnetic separations. In: Em. Separation of Heavy Metals: Engineering Processes. Lewis Publishers, NY,NY.97-135. Fu, P. 1983. Treatmnt of As(V) containing water by activated carbon process. MS nesis, Univ. Del., Newark, Del. Fu, P. et al. 1994. Selecting membranes for removing precursors, JAWKA, 86, 55-72. Fuller, C. et al. 1993. Surface chemistry of ferrihydrite: ii. kinetics of arsenate adsorption and coprecipitation. Geochim. et Cosmochim.Actu. 57,2271-82. Gandilhon, A, N. Tambo, 1992. Experimental study on a membrane ultrafiltration process for drinking-water production.Aqua. 41 (4), 203-08. Ghosh, M., R Teoh. 1985. Adsorption of As on hydrous aluminum oxide. In: Toxic and Haz. Wastes:SeventhMid-Athtic Znd Waste Con$ Technomic Publ. Co. Lancaster, PA, 139-55. Goldberg, S. 1986. Chemical modeling of arsenate adsorption on Al and Fe oxide minerals. Soil Sci. Soc.Am. J. 50, 1154-60. Grzetic, I., J. Zemann. (Eds). 1993. Neues jahrbuch f i r mineralogie, thallium chemistry, In: Proc. Int. Symp. On TI Chemishy, Geochemistry,Mineralogy, Ores and Environmental. 166. (I), E. Schweizerbart'scheVerlagsbuchhandlung, Stuttgart, Germany. 43-52. Gulledge, J., J. O'Connor. 1973. Removal of As(V) fiom water by adsorption on Al and Fe hydroxides. JA WWA, 548-52. Hadden, A. 2002. Galvanically enhanced removal of Se from wastewater. MS Thesis. Montana Tech, Butte, MT. Harris, B. 2000. The removal and stabilization of As from aqueous process solutions: past, present, fiture. In: Minor Elements 2000, Processing and Em. AspecB of As, Sb, Se, Te, and Bi. (Ed) C. Young, SME. Littleton, CO. 3-21. Hayes, K., et al. 1987. In Situ x-ray absorption study of surface complexes: Se oxyanions on alpha FeOOH. Science, 238, 783-86. Hering, J., et al. 1996. Arsenic removal by ferric chloride. JAWWAI, 88 (4), 155-67. Hingston, F. 1970. Specific adsorption of anions on goethite and gibbsite. PhD Xhesis, Univ. W. Australia; and, 1972. Anion adsorption by goethite and gibbsite. J. SoilScience. 23,2. Hingston, F., et al. 1968. Adsorption of selenite by goethite, In: Ahorption from Aqueous solution. ACS, Advanced Chemistry Series. NY, NY. 79,82-90.

1860

Hornung, S., J. Yuan,M. Ghosh. 1983. Se removal in fixed bed activated alumina adsorbers. In: Create a New Excellence. AWWA Annual Conference. 299-3 18. Howard, J. 1972. Control of geochemical behavior of Se in natural waters by adsorption on hydrous ferric oxides. Trace Substances in Em. Health ,V, 485-95; and, 1977. Geochemistry of Se: formation of ferroselite and Se behavior in the vicinity of oxidizing sulfide and uranium deposits. Gemhim. Commhim. Acta, 41, 1665-78. Huang, C., P. Fu. 1984. Treatment of As(V) containing water by the activated carbon. JwpcF, 56, (3), 80-88.

Huang, C., L. Vane. 1989. Enhancing As(V) removal by a Fe(I1)-treated activated carbon. J W C F , 61, (9), 1596-1603. Huxstep, M. 1981. Project summary:inorganic contaminant removal from drinking water by reverse osmosis, NTIS PB 81-224 420, Government Printing Ofice, Washington, DC. 2; and, Huxstep, M., T. Sorg. 1988. Reverse osmosis treatment to remove inorganic contaminants from drinking water, EPAl600l2-87l109, N17S PB88-147780. 50. Jambor, J., 3. Dutrizac. 1998. Occurrence and constitution of natural and synthetic ferrihydrite, a widespread iron oxyhydroxide. Chemical Review, 98,2549-85. Kaplan, D., K. Cnatrell, T. Wietsma. 1994. Formation of a barrier to groundwater by injecting metallioiron colloids: effect of influent colloid concentration. In: In Situ Remediation: Scientific Basis for Current and Future Technologies, Battelle Press, Columbus, OH. Kapoor, A., T. Tanjore. 1995. Removal of Se fiom water and wastewater, Em. Sci.Tech. 49, (2), 137-47

Karim, Z. 1984. Characteristics of ferrihydrite formed by oxidation of FeClz solutions containing different amounts of silica. Clays Clqy Minerals. 32, 18 1-4. Kauffinan, J., W. Laughlin, R. Baldwin. 1986. Microbiological treatment of uranium mine waters, Env. Sci. Tech.20 (3), 243-48. Kharaka, Y., J. Thorsden, R. Schroeder. 2000. Nanofiitration membranes used to remove Se and other minor elements from wastewater, Proc. Minor EIements'2000. (Ed) C. Young, SME, Littleton, CO. 371-80. Khoe, G., J.C.-Y Huang, R. Robins. 1991. Precipitation chemistry of the aqueous ferrous-arsenate system. EPD Congress. (Ed) D.R Gaskell, TMS, Warrendale, PA 103-16. Koren, D., W. Gould, L.Lortie. 1992. Se removal fiom waste water, In: Proc. Waste Processing and Recycling inMin. &Met. I&stries, CIM, Edmonton, Alberta, Can. 171-82. Krause, E. 1992. Arsenic removal at INCOs Cred plant and disposal practice, Proc. Arsenic Workshop,EPA Office of Solid Waste and Emergency Response, Washington, DC. Krause, E., V. Ettel. 1989. Solubilities and stabilities of ferric arsenate compounds, Hydromet., 22, (3), 311-37.

Laguitton, D. 1976. As removal from gold-mine waste waters: basic chemistry of lime addition method. CIMBull., 69, 105-9. Lake, L. 1990. Alumina for POUIPOE removal of F and As water treatment. Poznt-ofiUse/Entry Treatment of Drinking Water,Noyes Publ., Park Ridge, New Jersey, 88-9. Leckie, J., D. Merrill, W. Chow. 1985. Trace element removal from power plant waste streams by adsorptiodcoprecipitation with amorphous iron hydroxide. Separation of Heavy Metals and Other Trace Contaminants.(Eds) R. Peters, B. Mo Kim, AICHE, 28-42. Leckie, M., et al. 1980. Adsorptionlcoprecipitation of trace elements fiom water with iron hydroxide. EPRI Document CS-1513, Electric Power Res. Inst., Palo Alto, CA.; and, M. Benjamin, K. Hayes, J. k k i e . 1982. Removal of toxic metals fiom power generation waste streams by adsorption. JKPCF, 54, (1 1). 1472-81. ManeVal, et al. 1985, Se removal from drinking water by ion exchange, EpA/600/2-85/074. US Government Printing Office, Washington, DC. Masscheleyn, P., R. Delaune, W. Patrick. 1991. As and Se chemistry as affected by sediment redox potential and pH. I. Env. Qual. 20, 522-27; and, Effect of redox potential and ph on arsenic speciation and solubility in a contaminated soil. Env. Sci. Tech. 25, 1414-19. McCloskey, 1999. Private communicationsfor one of the authors, MSE-TA, Butte, MT. McGrew, K., J. Murphy, D. Williams. 1996. Se reduction via conventional water treatment, Randol GoldForum '96,Randol Gold, Denver, CO. 129-41.

1861

McNeil, L., M. Edwards. 1995. Soluble As removal at water treatment plants. JAWWA, 105-13. Merrill, D., et al. 1986. Field Evaluation of As and Se removal by iron coprecipitation. JWPCF, 58, (1); and, 1987. Environ. Progress, 6, (2), 82-8. Memll, D., P. Maroney, D. Parker. 1985. Trace element removal by coprecipitation with amorphous iron oxyhydroxide: engineering evaluation. Electric Power Research Institute Coal CombustionSystems Division, Report EPRI, CS 4087. Palo Alto, CA. 264. MWTP. 1994. Issues identification and technology prioritization report: As. EPA Mine Wmte Treatment Program. Activity I, Vol5. MWTP-41. MSE-TA, Butte, MT., 169. MWTP. 1997. Arsenic oxidation demonstration project-final Report. EPA Mine Waste Treatment Program. Activity III.Project 7. MWTP-84. MSE-TA, Butte MT. MWTP. 1999. Issues identification and technology prioritization report: Se. P A Mine Waste TechnologyProgram, Activity I, Vol.W, MWTP-106. MSE-TA, Butte MT., 118. WIT. 1999a. Issues identification and technology prioritization report: TI. EPA Mine Waste Treatment Program. Activity I, Vol Vm.MWTP-143. MSE-TA, Butte MT., 76. MWTP. 2001. Final report-Se treatmenthemoval alternatives demonstration project. EPA Mine Waste Treatment Program. Activity 1. MWTP-143. MSE-TA, Butte MT., 69. Nikolaev, A,, et al. 1974, 76. As(1II) oxidation in aqueous solutions by air oxygen in the presence of a catalyst. Izv.Sib. Otd- Akad. Nauk SSSR, Ser. Khim, Nauk..5. 36-40; and, 1976. Oxidation of As(II1) in aqueous solutions in the presence of native pyrolusite. 3, 48-51; and 6,3640. Nikolaidis, N., J. Lackovic, G. Dobbs. 2002. As remediation technlogy-AsRT. US Put. Appl. 60/050,250. 9. Nishimura, T., I Itoh. 1985. The calcium-arsenic-water-air system. In: Proc. Impurity Control Disposal, 15* CIM Hydromet. Meeting. Vancouver, BC. 2.1-2.19. Nishimura, T., K. Tozawa. 1978. On the solubility product of ferric, calcium and magnesium arsenates Bull. Znst. Min. Dress. andkfet.34,202-6. Nishimura, T., K. Tozawa. 1985. Removal of As from wastewater by addition of calcium hydroxide and stabilization of As precipitates by calcination. In: Proc. Impurity Control and DispsaZ, 15th Hydromet. Sym. CIM, Vancouver, CAN. 3.1-3.18. Nishimura, T., K. Tozawa, R.G. Robins. 1983. The calcium-arsenic-water system. W / A u s W J o i n t Symp., Met. Session JD-2-1, Sendai, Japan. Nishimura, T., K. Tozawa, R. Robins. 1988. Stabilities and solubilities of metal arsenites and arsenates in water and effect of sulfate and carbonate ions on their solubilities. In: Proc. As Met. Fundamentals and Applications, (Eds) R Reddy, et al. TMS. Warrendale, PA., 77-98. Nriagu, J.O. (Ed.). 1994. Arsenic in the environment: Part 1 cvcling and characterization, Advances in Env. Sci. and Tech, Wiley and Sons, NY, NY. 448. Nriagu, J.O. (Ed.). 1998. Thallium in the environment, vol. 29, Advances in Env. Sci. and Tech., Wiley and Sons, NY,NY. 284. O’Hannesin, S., R. Githam 1992. A permeable reaction wall for insitu dehalogenating organic compounds”, Proc. 5 Can.Geotechnica1Soc. Conf., Toronto, Canada. Oremland, R. 1991, 93. Selenate removal from waste water. US Patent 5,009,786;5,271,831. Parida K., et al. 1997. Studies on ferric oxide hydroxides. iii. adsorption of selenite on different forms of iron oxyhydrides. J. ColloidZnterJace Science. 185, (2), 355-62. Pengchu, A., D. Sparks. 1990. Kinetics of selenate and selenite adsorptioddesorption at the goethite/waterinterface. Env. Sci. Tech. 24, (12), 1848-56. Pierce, M., C. Moore. 1980. Adsorption of arsenite on amorphous iron hydroxide from dilute aqueous solution. Em. Sci. Tech 14, (2), 214-16. Plessas, K. 1992. Removal of As from wastewater. MSc i’??esis,Montana Tech, Butte, MT. 118. Pontius, F. 1990. An update of the federal drinking water regs. JAWWA 87 (2), 48-58. Porubaev, V., et al. 1972. Use of phosphates for the purification of waste waters to remove nonferrous metal ions and arsenic, Rud Tsvet. Metal. 10, 1 10-18. Powell, R., R. Puls, C. Paul. 1994. Innovative solutions for contaminated sit management, Water Env. Fed. Miami, FL,.485-96.

1862

Powell, R., R. Puls, S. Hightower. 1995. Coupled iron corrosion and chromate reduction: mechanisms for substanceremediation. ES&T, 29 (8), 1913-22. Puls, R., R. Powell. 1991. Colloidal ferric oxide transport studies in laboratory model systems using shallow aquifer material. EPA/600/D-91/098. US Printing Oflce, Washington, DC. 5 p. Rai, D. et al.1984. Chemical attentuation rates, coefficients and constants in leachate migration, vol. 1, a Critical review, EPRIRep. EA-3356. Ele. Power Res. Inst., Palo Alto, CA. Rajakovic, L. 1992. Sorption of As onto activated carbon impregnated with metallic silver and copper. separation Sci. Tech. 27, (1 l), 1423-33; and, As removal from water by chemisorption filters. Env. Pollution. 75, (3), 279-87. Ramana, A, A. Sengupta. 1992. Removing SeOv) and A s 0 Oxyanions with Tailored Chelating Polymers, J. Em. Engn. ASCE, 118 (9,755-75. Reed, B. 2002. Removal of heavy metals by activated carbon. In &. Separation of Heavy Metals. (Ed) A. SenGupta, Lewis Publishers, NY,181-204. Riveros, P., J. Dutrizac, P. Spencer. 2001. As disposal practices in the metallurgical industry. Can. Met. Quart. 40, (4), 395420. Robins, RG. 1981. The solubility of metal arsenates. Met. Trans. 12B. TMS. Warrendale, PA. 103-9. Robins, R.G. 1983. The stabilities of As(V) and As(1II) compounds in aqueous metal extraction systems. In: Proc. Hyc-iometallurgyResearch, Development and Plant Practice, (Ed) K . OsseoAsare, J.D. Miller. TMS. Warrendale, PA. 291-310. Robins, R.G. 1984. The stability of As in gold mine processing wastes. PreciousMetals, (Eds) V. Kydryk, D. Corrigan, W. Liang. TMS. Warrendale, PA, 241-49. Robins, R.G. 1988. Arsenic hydrometallurgy. Proc. A s Metallurgy FuAmentals and Applications. (Ed) R. Reddy, et al. TMS. Warrendale, PA. 215-48. Rogers, K. 1989. Point-of-Use treatment of drinking water in San Ysidro, NM. EpA/600/2891050, NTIS PB90-108838, US Printing OfJice, Washington, DC. 65 p. Rochette, E., G. Li, S. Fendorf. 1998. Stability of arsenate minerals in soils under biotically generated reducing conditions. SoilSci. Soc.Am. J.62, 1530-37. Rosengrant, L., L. Fargo. 1990. Final best demonstrated available technology (BDAT) background document for K031, K084, K101, K102, As wastes @004), Se wastes (Dolo), and P and U wastes containing As and Se listing constituents. EPA/53O/SW-90/059A. EPA, Wash., DC. 124. Rozelle, L. 1987. Point-of-Use water treatment for removal of chemical contaminants from drinking water. In: A WWA Sem. on Water Quality Concerns in the Distribution System. AWWA, Denver, CO. 121-38. Runnells, R., R. Skoda. 1990. Redox modeling of As in the presence of Fe: applications to equilibrium computer modeling. In: Proc. Em. Res. C o d on Groundwater Quality and Water Disposal,EPRI, Palo Alto, CA. 22-1-1 1. Sanjay, H., A. F a a h , D. Walia. 2002. Humasorb: a coal-derived humic acid-based heavy metal sorbent.In: Env. Separation of Heavy Metals: Engr. Processes. Lewis Publishers, NY,347-74. Santina, P. 1996. SulfUr-Modified iron (SMI) process for As removal. US Pat. 5,575,919. Schulze, D., U. Schwertmann. 1984. The influence of Al on iron oxides: x. properties of Al-substituted goethites. Clay Minerals, 19, 521-39. Schwertmann, U. 1966. Inhibitory effect of soil organic matter on the crystallization of amorphous femc hydroxide. Nature, 212, 64546. Schwertmann, U. 1984. The influence of Al on iron oxides: ix. dissolution of Al-goethite in 6 M HCI. ClayMznerals, 19, 9-19. Schwertmann, U., R Cornell. 1991. Iron oxides in the laboratory: preuaration and characterization. VCH Verlagsgesellscha$ mbH, Weinheim, Germany, 137 p. Schwertmann, U., H. Fechter. 1982. The point of zero charge of natural and synthetic ferrihydrite and its relation to silicate. CIayMinerals, 17,471-6. Schwertmann, U., E. Murad. 1983, 85. Effect of pH on the formation of goethite and hematite from ferrihydrite. Clays, Clay Minerals. 31 (4), 277-84; and, 1985. Properties of goethites of varying crystallinity. 33, (5), 369-78.

1863

Schwertmann, U., H. Thalmann 1976. The influence of Fe(II), Si, and pH on the formation of lepidocrocite and ferrihydrite during oxidation of aqueous FeClz solutions. Crcly Minerals. 11, 189-200. Scott,M. 1991. Kinetics of adsorption and redox processes on Fe and Mn oxides: reactions of As(IIII) and SefIV) at goethite and birnessite surfaces. Ph.D. Thesis, CIT, 266 p. SenGupta, A. 2002. Principles of heavy metal separation: an introduction. In: Em. Separation of Heavy Metals: EngineeringProcesses. Lewis Publishers, NY, 1-14. SenGupta, A (Ed). 2002a. Environmental separation of heaw metals: engineering processes. Lewis Publishers, N.Y. 380 p. SenGupta, A., J. Greenleaf. 2002. Arsenic in subsurface water: its chemistry and removal by engineered processes. In: Env. Separation of Heavy Metals: Engr. Processes. Lewis Publishers, NY.265-306. Sengupta, S., A. SenGupta. 2002. Trace heavy metal separation by chelating ion exchangers. In: Em. Separation of Heavy Metals: Engineering Processes. Lewis Publishers, NY, NY. 45-91. Sisk W., et al. 1990. As contaminated groundwater treatment pilot study. In: Superfind ’90. 1 lth Natl. Cod. Washington, DC.901-6. Smith, R, E. Jenne. 1991. Recalculation, evaluation, and prediction of surface complexation constants for metal adsorption on iron and manganese oxides. Env. Sci Tech 25, (3), 525-30. Sorg, T., G. Logsdon. 1978. Treatment technology to meet interim primary drinking water regulations for inorganics: part 2. J A W A . 379-93. Sparkman, J., et al. 1990. Adsorption of oxyanions by spent western oil shale: selenite. Em. Geol. Water Science, 15, (2), 93-9. Stewart, H., K. Kessler. 1991. Evaluation of As removal by activated alumina filtration at a small community public water supply. J. New England Water WorhAssoc: 105, (3), 179-99. Stiksma J., et al. 1996. Iron addition for impurity control at Shenitt’s nickel refinery. In: Proc. Iron control and dzpsal._lEds)J. Dutrizac, G. Harris. CIM. Montreal, Quebec, 287-98. Swash, P., A. Monhemius, J. Schaekers. 2000. Solubilities of process residues fiom biological oxidation pretreatments of refractory gold ores. Minor Elements 2000. (Ed) C. Young. SME, Littleton, CO. 115-22. Trussell, R., A. Trussell, P. Krafi. 1980, 91. Se removal from groundwater using activated alumina. EPA-600/12-80-153. Washington, DC; and, 1991. Se removal with activated alumina. A W A Res. Found. Water Quality Res. News. 19,4-5. Twidwell, L., et al. 1993. Removal of As from wastewater and stabilization of As bearing waste solids: summary of experimental results. J Haz. Mat. 36,69-80. Twidwell, L., et al. 1999. Technologies and potential technologies for removing As from process and wastewater, In: Proc. REWAS’99, Global Sjmp. on Recycling, Waste Treat. and Clean Tech. (Eds) I. Gaballah, J. Hager, R Solozaral, San Sabastian, Spain, TMS Warrendale, PA 1715-26. Twidwell, L. et al. 1999a. Technologies and potential technologies for removing Se fiom process on Recycling, Waste Treat. and Clean and wastewater, In: Proc. REWAS’99, Global S’p. Tech. (Eds) I. Gaballah, J. Hager, R. Solozaral, San Sabastian, Spain, T M S Warrendale, PA, 1645-56; and, 2000. Technologies and potential technologies for removing Se from process and waste water: update. In: Proc. Minor Elements 2000, (Ed) C. Young, SME, Littleton, CO, 5366. Valemela, A., K. Fytas, M. Sanchez. 2000. As management in pyrometallurgical processes. part II. recovery and disposal. Int. Con$ on Clean Technologiesfor the Min Indusm-es. (Eds) M. Sanchez, F. Vergara and S. Castro. Univ. Concepcion, Chile. 107-21. V i i g , M., K. Weerts. 1993. CyanoMet R-A process for the extraction and concentration of cyanide species from alkaline liquors, Randol Gold Forum ‘93,Beaver Creek, CO. 333-36. Watson, J. 1999. SeDaration methods for waste and environmental applications. NY. Marcel Dekker. 600 p. Weir, D., I. Masters. 1980, 82. Removing As fiom aqueous solutions. Can. Pat.358,966, (1980); US. Pat. 4,366,128,(1982); Fr. Pat. 2,488,869, (1982).

1864

Welham, N., K. Malati, S. Vukcevic. 2000. The stability of iron phases presently used for disposal from metallurgical systems-a review. Minerals Engineering, 13 (8-9). 91 1-3 1. Williams-Beam, C., L. Twidwell. 2001. Potential technologies for removing TI from mine and process wastewater: an abbreviated annotation of the literature. .EM!P, March, 2002 Wolff, C., C. Rudasill. 1990. Baird and McGuire Superfund Site: investigation of As and Pb removal from groundwater. Su-mrhnd '90,. 1lth Natl. Cod. EPA. Washington, DC. 371-75. WSPA. 1995. Se removal technology study final report. Western States Petroleum Association, Concord, C A Young, C. (Ed) 2000. Minor Elements 2000, Processing and Environmental Aspects of As, Sb, Se, Te,and Bi. In: Proceedngs MAnnualMeeting, Salt Lake City, UT.SME, Littleton, CO. 408. Yuan, J. et al. 1983. Adsorption of As and Se on activated alumina. In: Proc. A X E , Em. Engr. Division Specialty Con$ (Eds) A Medine, M. Anderson, Boulder, CO, July 6-8. 433-41.

-

1865

Treatment Of Solutions And Slurries For Cyanide Removal

'

Michael M. Botz, P.E. and Terry I. Mudder, Ph.D.

ABSTRACT

A variety of proven and reliable chemical, physical and biological treatment processes have been developed for the removal and recovery of cyanide from mill tailings and process solution. The purpose of this chapter is to discuss various cyanide treatment processes, their common areas of application and treatment performances that can be reliably achieved at full-scale. Emphasis is placed upon treatment processes with proven field success, as well as those processes exhibiting significant potential for specific application at mine sites. WATER AND CYANIDE MANAGEMENT An integral and key component of water management systems at precious metals mining sites is the approach adopted to manage cyanide-containing solutions and slurries. Excluding the bulk storage of cyanide reagents such as sodium cyanide, most cyanide present at mining sites will be present in water solutions. Therefore, to a great extent the management of water and the management of cyanide can be considered as one and the same and should be simultaneously considered when developing a water management plan. All mining sites that utilize cyanide for metals recovery should have a comprehensive and well-maintained cyanide management plan. A good cyanide management plan will include descriptions of how cyanide-containing solutions and slurries are to be handled, stored, contained and monitored, and in many cases the plan will also include a description of treatment plants used to remove cyanide from solutions or slurries. At sites where natural cyanide attenuation is important, the cyanide management plan should address the specifics of predicting and monitoring the effectiveness of the attenuation processes. Despite the critical importance of having a formal written cyanide management plan, many mining operations have not developed such a plan. The lack of a cyanide management plan, in some cases, has contributed to adverse environmental incidents involving cyanide (Mudder and Botz 2001a). Attempts to remedy this situation have been recently made by the United Nations Environment Programme (UNEP), which is developing an international code for the management of cyanide (www.cyantists.com/cyanide.htm). Implementation and adherence to this code, augmented by experienced scientific and engineering judgment, will help reduce both the number and severity of environmental incidents involving cyanide.

1 Elbow Creek Engineering, Inc., Joliet, Montana. 2 TIMES Ltd., Sheridan, Wyoming.

1866

BACKGROUND In the mining industry, cyanide is primarily used for extracting silver and gold from ores, but cyanide is also used in low concentrations as a flotation reagent for the recovery of base metals such as copper, lead and zinc. At these operations, cyanide treatment systems may be required to address potential toxicity issues in regard to wildlife, waterfowl and/or aquatic life. This may include the removal of cyanide from one or more of the following:

Slurry tailings from milling operations Bleed solution from Merrill-Crowe operations Excess solution from heap or vat leaching operations Supernatant solution from tailings impoundments Seepage collected from ponds or tailings impoundments Cyanide treatment is classified as either a destruction-based process or a recovery-based process. In a destruction process, either chemical or biological reactions are utilized to convert cyanide into another less toxic compound, usually cyanate. Recovery processes are a recycling approach in which cyanide is removed from the solution or slurry and then re-used in a metallurgical circuit. Selection of an appropriate cyanide treatment process involves the consideration of many factors, but generally the number of candidate processes for a particular application can be narrowed following an inspection of the untreated solutiodslurry chemistry and the desired level of treatment. The common applications for cyanide treatment in the mining industry are the following: 1.

Tailings slurry treatment is employed when the cyanide level must be lowered prior to being discharged into a tailings storage facility. In this application, the initial tailings s l u q WAD cyanide level typically ranges from about 100 to 500 mg/L and treatment to less than 50 mg/L WAD cyanide is commonly established as the goal for wildlife and waterfowl protection.

2.

Solution treatment is employed when the cyanide level in a decant or process solution must be lowered prior to being discharged into the environment. Treatment of WAD cyanide to below 1.0 mg/l is occasionally required to ensure the protection of aquatic ecosystems. Treatment technologies for decant solution commonly employ chemical oxidation and polishing processes, which are applicable to relatively low concentrations of cyanide and generate a high quality effluent.

CYANIDE ANALYSIS The term “cyanide” generally refers to one of three classifications of cyanide, and it is critical to define the class of cyanide that is to be removed in a treatment plant. The three classes of cyanide are: (1) total cyanide; (2) weak acid dissociable (WAD) cyanide; and (3) free cyanide as shown in Figure 1. Each of these forms of cyanide has a specific analytical methodology for its measurement, and it is important that the relationship between these forms be understood when analyzing cyanide-containing solutions. As indicated in Figure 1, for a given solution the total cyanide level is always greater than or equal to the WAD cyanide level, and likewise, the WAD cyanide level is always greater than or equal to the free cyanide concentration.

1867

{

Strong Metal-Cyanide Complexes of Fe

ga:ide

WAD Cyanide

Weak and Moderately Strong Metal-Cyanide Complexes of Ag, Cd, Cu, Hg, Ni and Zn Free Cyanide

CNHCN

Figure 1 Classification of Cyanide Compounds

The appropriate approach to assessing the quality of untreated and untreated samples in most situations is to analyze for WAD cyanide since this includes the toxicologically or environmentally important forms of cyanide, including free cyanide and moderately and weakly complexed metal-cyanides. Total cyanide includes free cyanide, WAD cyanide plus the relatively non-toxic iron-cyanide complexes. Complete characterization of a cyanide solution generally includes analyses for pH, total cyanide, WAD cyanide, thiocyanate, cyanate, ammonia, nitrate and base metals such as copper, iron, nickel and zinc. CYANIDE DESTRUCTION Most cyanide destruction processes operate on the principle of converting cyanide into a less toxic compound through an oxidation reaction. There are several destruction processes that are well proven to produce treated solutions or slurries with low levels of cyanide as well as metals. In the following sections, several cyanide destruction processes are discussed along with their typical areas of application. With all of these processes, laboratory and/or pilot testing is required to confirm the level of treatment achievable and to evaluate the associated reagent consumptions.

The INCO Sulfur Dioxide and Air Process The sulfur dioxide (SO2) and air process was developed by INCO Limited in the 1980's and is currently in operation at over thirty mine sites worldwide. The process utilizes SO2 and air at a slightly alkaline pH in the presence of a soluble copper catalyst to oxidize cyanide to the less toxic compound cyanate (OCN'). SO2 + 0 2 + HzO + CN-

C U + ~Cuta/ysr

> OCN- + SOi2 + 2H'

The theoretical usage of SO2 in the process is 2.46 grams of SO2 per gram of CN- oxidized, but in practice the actual usage ranges from about 3.5 to 5.0 grams SO2 per gram of CN- oxidized. The SO2 required in the reaction can be supplied either as liquid sulfur dioxide or a sulfur salt such as sodium metabisulfite (Na2S205)or sodium sulfite (Na,SO,). Oxygen (02) is also required in the reaction and is generally supplied by sparging atmospheric air into the reaction vessels. The reaction is typically carried out at a pH of about 9.0 in one or more agitated tanks, and lime is added to neutralize the acid (H') formed in the reaction to maintain the pH in this range. Lime usage is generally on the order of about 3.0 to 5.0 grams per gram of CN' oxidized. As indicated, copper (CU'~) is required as a catalyst, which is usually added as a solution of copper sulfate (CuS04-5H20)to provide a copper concentration in the range of about 10 to 50 mg/L, depending upon the corresponding cyanide concentration. In solutions where sufficient copper is already present, supplemental addition of copper may not be required. A flowsheet for the sulfur dioxide and air process is shown in Figure 2.

1868

I

-l

1 I r

A@ated RcacmnTarJ,

Figure 2 The Sulfur Dioxide and Air Cyanide Destruction Process Upon completion of the cyanide oxidation reaction, metals previously complexed with cyanide, such as copper, nickel and zinc, are precipitated as metal-hydroxide compounds. Iron cyanide removal is affected through precipitation according to the following generalized reaction where ‘M’ represents copper, nickel or zinc: 2M+*+ Fe(CN)i4 + M2Fe(CN)6 (solid) The primary advantage of the sulfur dioxide and air process is with slurry tailings, but it is also effective for the treatment of solutions for the oxidation of free and WAD cyanides. Representative results for treatment of solution and slurry with the sulfur dioxide and air process are shown in Table 1 (Ingles and Scott 1987). Table 1 Treatment Results Using the INCO Sulfur Dioxide and Air Process Solution Tailings Slurry Parameter Untreated Treated Untreated Treated (mg/L) (mg/L) (mg/L) (mfm 450 0.1 to 2.0 115 0.1 to 1 .o Total Cyanide Copper 35 1 to 10 17 0.2 to 2.0 Iron 1.5 <0.5 0.7 0.02 to 0.3 Zinc 66 0.5 to 2.0 18 <0.01

As indicated in Table 1, the sulfur dioxide and air process is capable of achieving low and environmentally acceptable levels of both cyanide and metals. Generally, the best application of this process is with tailings slurries containing low to moderately high initial levels of cyanide and when treated cyanide levels of less than about 5 mg/L are required. In some cases, solutions treated with this process may be of suitable quality to permit their discharge. The process does not remove thiocyanate quantitatively, although a few percent of this cyanide related compound are typically removed during treatment.

1869

The Copper Catalyzed Hydrogen Peroxide Process The hydrogen peroxide treatment process chemistry is similar to that described for the INCO process, ’out hydrogen peroxide is utilized rather than sulfur dioxide and air. With this process, soluble copper is also required as a catalyst and the end product of the reaction is cyanate. H202 + CN-

cu+2Co/a/y,\r

> OCN‘ + H20

The primary application of the hydrogen peroxide process is with solutions rather than slurries due to the high consumption of hydrogen peroxide that occurs in slurry applications. The process is typically applied to treat relatively low levels of cyanide to achieve effluent quality that may be suitable for discharge. The hydrogen peroxide process is effective for the treatment of solutions for the oxidation of free and WAD cyanides, and iron cyanides are removed through precipitation of insoluble copper-iron-cyanide complexes. As indicated in the above reaction, hydrogen peroxide reacts with cyanide to form cyanate and water, a process which limits the build-up of dissolved solids in the solution being treated. The theoretical usage of H202in the process is 1.31 grams H202per gram of CN‘ oxidized, but in practice the actual usage ranges from about 2.0 to 8.0 grams H202 per gram of CNoxidized. The H202used in the process is typically provided as a concentrated liquid in 50% or 70% strength. Although the reaction can be carried out over a wide pH range, it is usually conducted at a pH of about 9.0 to 9.5 for optimal removal of residual metals such as copper, nickel and zinc initially complexed to cyanide. If iron cyanide must also be removed to low levels, then a lower pH is needed to maximize the precipitation of copper-iron-cyanides at the expense of lowering the removal efficiencies of copper, nickel and zinc. It is common in these instances to use a two-stage system with intermediate removal of the precipitated iron cyanide. As indicated, copper (Cu”) is required as a soluble catalyst, which is usually added as a solution of copper sulfate (CuS04-5H20) to provide a copper concentration in the range of about 10 to 50 mg/L, depending upon the initial cyanide and copper concentrations. Upon completion of the indicated reaction, metals previously complexed with cyanide, such as copper, nickel and zinc, are precipitated as metal-hydroxide compounds. Representative results for treatment of solution with the hydrogen peroxide process are shown in Table 2 (Mudder et al. 200 1b). Table 2 Treatment Results Using the Hydrogen Peroxide Process Solution Parameter Untreated Treated (mg/L) (mg/L) Total Cyanide 19 0.7 WAD Cyanide 19 0.7 20 0.4 Copper
As indicated in Table 2, the hydrogen peroxide process is capable of achieving low levels of both cyanide and metals. Generally, the best application of this process is with solutions containing relatively low initial levels of cyanide and when treated cyanide levels of less than about 1 mg/L are required. Oftentimes, solutions treated with this process may be of suitable quality to permit their discharge. As with the INCO process, this process does not remove thiocyanate quantitatively, but does remove a few percent of this cyanide related compound during treatment.

1870

The Caro’s Acid Process Peroxymonosulfuric acid (H2S0,), also known as Caro’s acid, is a reagent used in a recently developed cyanide treatment process that has found application at a few sites. Caro’s acid is produced from concentrated hydrogen peroxide and sulfuric acid in an exothermic reaction (Norcross 1996):

Due to its instability, Caro’s acid is produced on-site in-situ and used immediately for cyanide detoxification with only minimal intermediate storage. At room temperature, Caro’s acid is stable for several hours, however at elevated temperature it is stable only for several minutes, decomposing to liberate oxygen, water and sulfur trioxide (SO3). Production of Caro’s acid is typically conducted with 1.5 to 3.0 moles of H2S04 per mole of H 2 0 2to yield a product of up to 80% purity. Normally, 70% hydrogen peroxide solution and 93% sulfuric acid solution are used to generate Caro’s acid. The overall oxidation reaction of Caro’s acid with cyanide is shown below.

H~SO + CN~ +.OCN- + SOi2 + 2H’ The theoretical usage of H2S05 in the process is 4.39 grams H2S05 per gram of cyanide oxidized, but in practice 5.0 to 15.0 grams H2SO5 per gram of cyanide oxidized is required. Acid produced in the reaction (H’) is typically neutralized with lime, if necessary, and the reaction is normally carried out at a pH in the range of about 7.0 to 10.0. Caro’s acid is used in slurry treatment applications where the addition of a copper catalyst is not desirable, which is typically only in situations where the sulfur dioxide and air process is not suited. In solution applications, other destruction processes, such as the hydrogen peroxide process, are preferred to the Caro’s acid process. A flowsheet for the Caro’s acid process is shown in Figure 3.

Tailing Sliirry

-e _-_

Aetated Reaction Tank

Figure 3 The Caro’s Acid Cyanide Destruction Process

1871

~

Treated Slurry c

Representative results for treatment of slurry with Caro’s acid are shown in Table 3 (Mudder et al. 2001b). Table 3 Treatment Results Using Caro’s Acid Slurry WAD Cyanide Concentration Test Number Untreated Treated (mg/L) (mg/L) 1 44.5 8.5 2 37.5 4.2 3 46.0 14.0 4 39.8 4.0 5 115.0 27.1 6 113.1 16.3 7 101.5 18.7

As indicated in Table 3, the Caro’s acid process is capable of achieving levels of WAD cyanide below 50 mg/L, which are generally suitable for discharge into tailings impoundments. Generally, the best application of this process is with tailings slurries containing low to moderate initial levels of cyanide and when treated cyanide levels of less than about 10 to 50 mg/L are required. The Alkaline Chlorination Process Alkaline chlorination at one time was the most widely applied of the cyanide treatment processes, but it has gradually been replaced by other processes and is now only used occasionally. Alkaline chlorination is effective at treating cyanide to low levels, but the process can be relatively expensive to operate due to high reagent usages. The cyanide destruction reaction is two-step, the first step in which cyanide is converted to cyanogen chloride (CNCI) and the second step in which cyanogen chloride hydrolyzes to yield cyanate. Cl2 + CN- + CNCl

+ CI-

CNCl + H 2 0 + OCN- + CI- + 2H’

In the presence of a slight excess of chlorine, cyanate is further hydrolyzed to yield ammonia in a catalytic reaction. OCN-

+

H20

c/,Culu/y.sl > NH: + HCO; + O K

If sufficient excess chlorine is available, the reaction continues through “breakpoint chlorination” in which ammonia is fully oxidized to nitrogen gas (N2). 3CI2+ 2NH:

+ N2 + 6C1- + 8H’

In addition to reacting with cyanide, cyanate and ammonia, the alkaline chlorination process will preferentially oxidize thiocyanate, which in some cases can lead to excessively high consumptions of chlorine. It is the removal of thiocyanate that makes this cyanide treatment process unique when compared to other chemical oxidation processes. 4CI2 + SCN- + 5H20 + S O i 2 + OCN- + SCI- + 1OH’

1872

The primary application of the alkaline chlorination process is with solutions rather than slurries due to the high consumption of chlorine that occurs in slurry applications. The process is typically applied to treat low solutions flows initially containing low to high levels of cyanide to achieve cyanide levels that may be suitable for discharge. The process is effective for the treatment of solutions for the oxidation of free and WAD cyanides, but a lesser amount of iron cyanides are removed depending on the levels of other base metals in the solution being treated. As can be seen in the above reactions, a significant increase in the treated water dissolved solids concentration may result, particularly with chloride. The theoretical usage of Clz to oxidize cyanide to cyanate is 2.73 grams C12 per gram of CNoxidized, but in practice the actual usage ranges from about 3.0 to 8.0 grams CI2 per gram of CNoxidized. The Clz used in the process can be provided as a liquid Clz, as a 12.5% solution of sodium hypochlorite (NaOCI) or as a solution of calcium hypochlorite (Ca(OC1)2). Chlorine consumptions for the oxidation of ammonia and thiocyanate can be calculated from the above reactions. In addition, the above reactions generate varying amounts of acid (H’) which is typically neutralized by adding lime or sodium hydroxide to the reaction vessels. The reaction is carried out at a pH of greater than 10.5 to ensure potentially irritating cyanogen chloride is rapidly hydrolyzed to cyanate. An advantage of the process is that copper is not required as a catalyst as with the sulfur dioxideiair and hydrogen peroxide processes. Upon completion of the cyanide oxidation reaction, metals previously complexed with cyanide, such as copper, nickel and zinc, are precipitated as metal-hydroxide compounds. Representative results for treatment of solution via alkaline chlorination are shown in Table 4 (Ingles and Scott 1987). Table 4 Treatment Results Using the Alkaline Chlorination Process Solution Parameter Untreated Treated (mg/L) (mg/L) Total Cyanide 2,000 8.3 0.7 WAD Cyanide 1,900 Copper 290 5 .O Iron 2.4 2.8 Zinc 740 3.9

AS indicated in Table 4, the alkaline chlorination process is capable of achieving low levels of both cyanide and metals. Generally, the best application of this process is with low flows of solutions containing high to low initial levels of cyanide when treated cyanide levels of less than about 1 mg/L are required. Oftentimes, solutions treated with this process may be of suitable quality to permit their discharge. Iron-Cyanide Precipitation Free, WAD and total cyanides will all react with ferrous iron to yield a variety of soluble and insoluble compounds, primarily hexacyanoferrate (111) (Fe(CN)i3), Prussian blue (Fe4[Fe(CN)&) and other insoluble metal-iron-cyanide (MxFey(CN),) compounds such as those of copper or zinc (Adams 1992).

+ H+ -+ Fe(CN)i3 + %H20 Fe” + 6CN- + %02 4Fe” + 3Fe(CN)i3 + %02 + H+ + Fe4[Fe(CN),]3 + %H20 These reactions act to lower the free and WAD cyanide concentrations by converting them to stable iron cyanide compounds (soluble and insoluble), while the iron-cyanide concentration is lowered as a result of precipitation reactions.

1873

The iron-cyanide precipitation process is limited in its suitability to situations where the precipitation reactions can be controlled and the precipitated solids can be separated and properly disposed. Proper handling and disposal of the cyanide precipitates generated in the process is important and represents the major disadvantage of this non-destructive process. In the past, this process was widely used to convert free and WAD cyanides to less toxic iron-cyanide compounds, but its present utility is primarily as a polishing process to reduce total cyanide concentrations to less than about 1 to 5 mg/L. There are a number of environmental drawbacks to this process, including the generation of cyanide precipitates and the formation of stable and soluble ironcyanide compounds that will persist for many years and may require hrther treatment. The process is optimally carried out at a pH of about 5.0 to 6.0 and iron is added as ferrous sulfate (FeS04-7H20). Ferrous sulfate usage ranges from about 0.5 to 5.0 moles Fe per mole of CN- depending on the desired level of treatment (Adams 1992; Dzombak et al. 1996). As indicated in Table 5, the iron-cyanide precipitation process is capable of achieving relatively low levels of total cyanide at pH 7.0 and an Fe:CN molar ratio of about 4: 1 (Dzombak et al. 1996). Table 5 Treatment Results Using the Iron-Cyanide Precipitation Process Solution Parameter Untreated Treated (mg/L) (mg/L) Total Cyanide 8.8 0.89 Effluent Polishing with Activated Carbon Activated carbon has a high affinity for many metal-cyanide compounds, including the soluble cyanide compounds of copper, iron, nickel and zinc. Activated carbon is suitable for use as a polishing treatment process to remove cyanide to low levels, when the initial cyanide concentrations are already below about 2.0 mg/L. This is a simple and effective process, convenient for installation at sites where activated carbon is used in metallurgical processes for precious metals recovery. At these sites, newly purchased carbon can be used for water treatment, and then when the carbon breaks through and is no longer suitable for water treatment, it can be transferred to the metallurgical circuit for continued use. This has been done at a number of sites to produce high quality effluents without impacting gold recovery operations. At inactive sites, regeneration andlor disposal of loaded carbon must be considered. Representative results for treatment of solution using activated carbon adsorption are shown in Table 6 (Botz and Mudder 1997). Table 6 Treatment Results Using the Activated Carbon Adsorption Process Solution Parameter Untreated Treated (mg/L) (mg/L) Total Cyanide 0.98 0.20 Copper 0.02 <0.02 lron 0.22 0.02 0.15 0.15 Nickel Zinc 0.02 <0.02 Biological Treatment Biological treatment processes have become more widespread in the mining industry due to the success of the plant installed at Homestake Lead, USA in the 1980’s. In this plant, an aerobic attached growth biological treatment is used to remove cyanide, thiocyanate, cyanate, ammonia and metals from tailings impoundment decant solution prior to discharge into a trout fishery. The plant has been operating successfully for over fifteen years, producing high-quality effluent.

1874

A multiple stage suspended growth biological treatment plant was installed by Homestake Nickel Plate, Canada in the mid-1990's to treat tailings impoundment seepage. This plant is a suspended sludge system with both aerobic and anaerobic treatment sections to remove cyanide, thiocyanate, cyanate, ammonia, nitrate and metals. Another biological treatment process was developed for the Homestake Santa Fe, USA mine to treat draindown from the decommissioned heap leach operation. This process, known as the passive Biopass process, is suitable for solution flows of less than about 10 m3/hour for the removal of cyanide, thiocyanate, cyanate, ammonia, nitrate and metals. The applicability of biological processes for the treatment of cyanide solutions in the mining industry has been somewhat limited, but is growing again with several applications being developed throughout the world. Their applicability is primarily with continuous solution flows with temperatures above about 10°C. The key advantage to biological treatment is the ability to simultaneously remove several compounds in a single process, often at a much lower cost than would be encountered with other treatment processes. In situations where cyanide and one or more of its related compounds of cyanate, thiocyanate, ammonia and nitrate must be removed, biological treatment should be considered. Representative results from the above three biological treatment plants are presented in Table 7 (Given et al 2001; Mudder et al. 2001a; Mudder et al. 200 1c).

Table 7 Treatment Results Using Biological Processes Parameter

Total Cyanide WAD Cyanide Thiocyanate Ammonia Nitrate Copper Iron Nickel Zinc

Homestake Lead Untreated (mg/L) 3.39 2.34

Treated (mg/L) 0.37 0.03

--

--

5.3 1

0.27 21.9 0.04 0.27 0.03 0.0 1

-0.49 0.1 to 5.0 0.01 to 0.04 0.01 to 0.1

Homestake Nickel Plate Untreated (mg/L) 1.04 0.33 379 25.3 2.8 0.02 0.06

---

Treated (mg/L) 0.44 0.04 0.08 0.15 0.13 0.005 0.02

---

Homestake Santa Fe Untreated (mg/L)

Treated (mdL)

14

<0.2

--

--

--

-_

---

55.6 10.3

0.8 <0.5

----

--

__

--

Other Cyanide Treatment Processes There are a number of other treatment processes that have been applied at full scale to treat cyanide, but implementation of these processes has been limited for a number of reasons. Ion exchange and reverse osmosis are frequently considered for treatment of decant solution, but with both of these processes waste brine is generated as a by-product. Disposal or further treatment of this brine is difficult and expensive, and in some cases the brine may be hazardous and require special handling. The amount of waste brine generated in by ion exchange and reverse osmosis typically ranges from about 10% to 30% of the volume of water treated. Ion exchange and reverse osmosis processes are also relatively 'expensive and complex to construct, operate and maintain. Due to these drawbacks, ion exchange and reverse osmosis are limited to situations where waste brine can be easily disposed or treated, or where very high quality effluent is required. An advantage of these processes, particularly reverse osmosis, it that in some cases simultaneous removal of cyanide, cyanate, thiocyanate, ammonia, and nitrate can be affected. Ion exchange is occasionally used to target ammonia or nitrate removal from decant solution, and additional information in this regard is provided later in this chapter.

1875

Ozone is a strong oxidant and capable of oxidizing free and WAD cyanides to cyanate, ammonia and nitrate. However, ozone initially oxidizes thiocyanate to cyanate or cyanide depending on the solution pH (Botz et al. 2001). The reaction rate is rapid and generally only limited by the rate at which ozone can be absorbed into the solution. Low effluent cyanide concentrations can be achieved with ozone, but treatment may result in the formation of cyanate, ammonia and nitrate. Ozone is relatively expensive to produce and this has limited its use for cyanide destruction, particularly for large water flows, but may find application in small-volume polishing applications. Iron cyanides are also oxidized by ozone, but the reaction rate is too slow at ambient temperature for practical application. At elevated temperature and in the presence of ultraviolet radiation, iron cyanides are converted to cyanate by ozone. Ammonia can be oxidized to nitrate by ozone, but an alkalirie pH is required. CYANIDE RECOVERY A number of cyanide recovery processes have been investigated over the previous one hundred years, but only two have found widespread application, as described in the following sections. The need for development and implementation of processes for recovery of cyanide will be important to the success of current and future mining operations. The requirement stems from concerns over low metal prices and the realization that more stringent environmental regulations will be developed, thereby restricting the concentrations of cyanide that can be discharged into tailings impoundments. With increasing concern over groundwater and wildlife issues, there will be increasing pressure to regulate cyanide entering tailings impoundments more closely. The advantages of cyanide recovery include the lowering of cyanide and metals levels entering a tailings impoundment, the economic benefit of recycling cyanide and the potential reduction in downstream treatment requirements for cyanide and its related compounds of cyanate, thiocyanate, ammonia and nitrate. Stripping and Absorption The stripping and absorption approach to recovering cyanide, also known as the acidificationvolatilization-reneutralization (AVR) and Cyanisorb processes, remove cyanide from solution as hydrogen cyanide gas. At a pH of less than about 8.0, free cyanide and some WAD cyanide compounds are converted to hydrogen cyanide gas, which can then be air-stripped from solution. Once removed from solution as hydrogen cyanide gas, the hydrogen cyanide is easily absorbed into an alkaline solution of sodium hydroxide or lime. The three main reactions involved with the cyanide recovery process are as follows. 2CN- + H2S04+ 2HCN(,,,

+ 2S04-2

(acidification) (stripping)

HCN,,) + NaOH

+ NaCN + H 2 0

(absorption)

This basic process has been used at about ten sites worldwide to affect WAD cyanide recoveries ranging from about 70% to over 95% with both slurries and solutions. Its application is primarily with solutions or slurries with moderate to high concentrations of cyanide under both high and low-flow conditions. The presence of dissolved copper in untreated solution can cause difficulties with conventional cyanide recovery processes, and in some cases pre-treatment of the solution with sodium sulfide (Na2S) may be required to precipitate copper sulfide (Cu2S) prior to cyanide recovery.

1876

The flowsheet for the Golden Cross cyanide recovery plant that operated in New Zealand during the life of the mine is shown in Figure 4. This plant processed tailings slurry and reduced the overall cyanide consumption at the site by about one-half. A key benefit of operating this cyanide recovery plant was after closure of the mine when tailings impoundment seepage was directly discharged to a trout fishery. This was possible because of the low concentrations of cyanide and related compounds present in the impoundment during mine operation.

Clean Gas

Two Stripping Towers Sulfuric Acid

CIL Tailings Slurry

*

Air

c=x=>-

*

Treated Slurry c To Tailins Impoundment Neutralization Tank

Figure 4 The Golden Cross Cyanide Recovery Plant Flowsheet Table 8 lists representative treatment results from several cyanide recovery processes (Moura 2001 ; Goldstone and Mudder 2001; Omofoma and Hampton 1992). Table 8 Treatment Results Using the Stripping and Absorption Cyanide Recovery Process WAD Cyanide Percent Plant Untreated Treated (mg/L) (mg/L) Recovery Golden Cross, New Zealand -250 <30 85% to 90% (Slurry) DeLamar, USA -300 <2 5 >go% (Solution) Cerro Vanguardia, Argentina -300 <30 >90% (Solution)

1877

Cyanide Recovery through Water Recycle and Tailings Washing Cyanide can effectively be recovered and re-used by recycling cyanide-containing solutions within a metallurgical circuit. This is commonly conducted using tailings thickeners or tailings filters to separate solution from tailings solids, with the solution being recycled in the grinding andor leaching circuits. This approach to recovering cyanide should be evaluated for all operations utilizing cyanide, and its performance can be determined by a simple mass balance calculation. Cyanide recovery affected in this manner is purely a physical process, with the recovery of cyanide accompanying the recovery of water from slurry tailings. In some cases, thickeners or filters can be used in conjunction with a chemical-based cyanide recovery process (Botz and Mudder 200 1b) to increase the overall cyanide recovery percentage. Cyanide recovery in water recycle and wash circuits are capable of achieving 90% cyanide recovery, but their implementation must be preceded by a careful examination of the site water balance. NATURAL CYANIDE ATTENUATION It is well known that cyanide solutions placed in ponds or tailings impoundments undergo natural attenuation reactions which result in the lowering of the cyanide concentration. These attenuation reactions are dominated by natural volatilization of hydrogen cyanide, but other reactions such as biological degradation, oxidation, hydrolysis, photolysis and precipitation also occur. Natural cyanide attenuation occurs with all cyanide solutions exposed to the atmosphere, whether intended or not. At several sites, ponds or tailings impoundments are intentionally designed to maximize the rate of cyanide attenuation, and in some cases resultant solutions are suitable for discharge. Advantages of natural attenuation include lower capital and operating costs when compared to chemical oxidation processes. Two approaches have been developed to predict the rate of cyanide attenuation in ponds and tailings impoundments. The first method is empirical in nature and uses experimentally derived rate coefficients to estimate the rate of attenuation using a first order decay equation (Simovic et al. 1985). This approach is relatively simple to apply, but its applicability at a given site must be verified by conducting field testwork and the results may not be accurate under changing weather, pond/impoundment geometry or chemistry conditions. The second approach to modeling natural cyanide attenuation was developed by Botz and Mudder (2000). This approach utilizes detailed solution chemical equilibria and kinetic calculations to predict the rate of cyanide losses from ponds and impoundments through a variety of reactions. The reactions of cyanate, thiocyanate, ammonia and nitrate can also be modeled with this approach. This approach can be time-intensive to apply at a given site, but the results are accurate under a wide variety of weather, pond/impoundment geometry and chemistry conditions. Examples of natural cyanide attenuation in tailings impoundments are presented in Table 9 as observed at several mines in Australia (Minerals Council of Australia 1996). These data correspond to WAD cyanide reductions of ranging from about 55% to 99%, reflective of the varying tailings chemistries, climatic conditions and tailings impoundment geometries at these sites. An additional example of natural attenuation of both cyanide and its related compounds is given in Figure 5 (Schmidt et al. 1981).

1878

Table 9 Examples of Natural Cyanide Attenuation WAD Cyanide in WAD Cyanide in Tailings Impoundment Tailings Discharge Decant Solution (mg/L) (mg/L) 210 94 48 10 57 0.5 150 20 125 22 186 20 82 12 99 9

34 32

--

1

Percent WAD Cyanide Reduction 55% 79% 99% 87% 82% 89% 85% 91%

CNT CNS CNO 3m P Shallow pond

26 24 22

0

6

20

9

18

0

.

16

Figure 5 Example of the Natural Attenuation of Cyanide and Related Compounds TREATMENT OF CYANIDE RELATED COMPOUNDS The primary constituents of concern in cyanidation solutions include not only the various forms of cyanide, but the cyanide related compounds thiocyanate, cyanate, ammonia and nitrate. In many cases, these compounds are important from both a water quality and toxicity standpoint and low levels must be achieved in treated waters. The following sections provide general background information on the removal cyanide related compounds.

1879

Thiocyanate Treatment Thiocyanate is formed through the interaction of cyanide with sulfur-containing compounds, particularly sulfide minerals such as pyrite, chalcopyrite or arsenopyrite. As with cyanate, thiocyanate is not a cyanide compound but related to cyanide and usually only found in solutions that also contain cyanide. Thiocyanate is far less toxic than cyanide and exhibits unique chemical, analytical and treatment characteristics. Thiocyanate removal from decant solution is not routinely practiced in the mining industry, but there are full-scale treatment plants that remove thiocyanate. In the case of biological treatment of cyanide, thiocyanate is also removed as part of the process, and may be required if land application of the solution is contemplated. There are no documented cases of thiocyanate resulting in adverse environmental impacts to aquatic life in the mining industry. Removal of thiocyanate from decant solution can be accomplished with one of several available destruction methods (Mudder and Botz 2001 b). It is possible to chemically regenerate cyanide from thiocyanate, however this process has not been implemented on a full-scale basis (Botz et al. 2001). The chemical destruction methods utilize an oxidant such as chlorine to convert thiocyanate to cyanide at an alkaline pH, and then cyanide oxidation rapidly continues to yield cyanate, ammonia and nitrate. Several oxidants are capable of oxidizing thiocyanate, but only chlorine and ozone yield suitably rapid reaction kinetics. Alkaline chlorination and ozone efficiently destroy thiocyanate and are capable of lowering thiocyanate concentrations to a few mg/L. If the residual chloride or chlorine content in treated solution is of concern, then the ozone process may be preferable since ozone dissociates to oxygen or water. Both chlorine and ozone can be used to simultaneously oxidize cyanide, cyanate, and thiocyanate, and often the choice of the appropriate oxidant is based on considerations of cost, byproduct generation and process efficiency. A lower cost alternative in many cases for thiocyanate destruction is biological treatment. Microorganisms in an aerobic environment readily oxidize thiocyanate and the reactions are rapid at temperatures above about 10°C to 15°C. As with the chemical destruction methods, biological thiocyanate treatment processes can readily be configured to simultaneously remove cyanide, cyanate, and ammonia along with thiocyanate. At a North American site, a biological treatment plant is used to lower cyanide, cyanate, thiocyanate and ammonia from initial concentrations of about 0.4 mg/L, 300 mg/L, 500 mg/L and 40 mg/L, respectively, to final concentrations of about 0.08 mg/L, <5 mg/L, 0.3 mg/L and 0.5 mg/L, respectively (Given et al. 2001). The advantages of biological processes over chemical processes for thiocyanate removal are that capital and operating costs are relatively low and the concentration of reaction by-products is low. However, biological processes are kinetically slow at colder temperatures and do not respond well to rapid fluctuations in solution flow or chemistry. Cyanate Treatment The compound cyanate is related to cyanide and is often found in waters that contain cyanide. Cyanate originates from the oxidation of cyanide but exhibits different chemical, analytical, treatment and toxicity characteristics. Treatment of solutions for cyanate removal is uncommon because it is far less toxic than cyanide, is generally present in metallurgical solutions at low concentrations and does not persist in the environment for long periods of time. In some cases the cyanate concentration in decant solution may be sufficiently high as to warrant implementation of a cyanate removal process, and typically this is the result of cyanate produced in a cyanide destruction process. This would be limited to situations where decant solution were to be discharged to the environment and the concentration of cyanate in untreated solution would be toxic to aquatic organisms.

1880

The authors are not aware of any full-scale water treatment facilities operating in the mining industry that specifically target the removal of cyanate or contain a limit on the level of cyanate that can be discharged. However, there are fUll-scale water treatment plants that incidentally remove cyanate along with their intended purpose of removing cyanide or other related compounds. Most notable are several biological treatment plants where processes to remove cyanide, thiocyanate, and ammonia also result in the removal of cyanate. An example is a mine in North America where decant solution is biologically treated for cyanide, thiocyanate, and ammonia removal (Given et al. 2001). These species are removed to low levels in the aerobic treatment system, although the cyanate level is also reduced from about 300 mg/L to less than about 5 mg/L. The biological mechanism of cyanate removal is first the oxidation of cyanate to ammonia, and then ammonia removal proceeds through a biological process termed nitrification. Subsequent nitrate removal through a biological denitrification process may also be warranted depending upon the resultant nitrate concentration in solution. Information regarding the removal of ammonia and nitrate through chemical, biological, and physical means is presented later in this chapter. Cyanate may also be removed from solution using chemical oxidation or hydrolysis processes. Chemical oxidation with chlorine at a slightly alkaline pH will convert cyanate to ammonia, though the chlorination process can easily be configured to complete the oxidation of ammonia into nitrogen gas through a process termed “breakpoint chlorination”. Ozone at an alkaline pH is capable of converting cyanate directly to nitrate, thereby avoiding the intermediate formation of ammonia. The advantage of the ozone process is that the concentration of dissolved species in treated solution is not increased significantly since ozone dissociates into oxygen and water. Cyanate also can be hydrolyzed to ammonia at an acidic pH, though the reaction is relatively slow at low temperatures, which in some cases may require solution heating. With the oxidation or hydrolysis reactions, cyanate is converted either into ammonia or nitrate and subsequent removal of these compounds may be required depending upon their resultant concentrations.

Ammonia Treatment Ammonia is toxic to aquatic organisms, particularly fish, but is usually present in metallurgical streams at low concentrations. The two sources that are responsible for the majority of ammonia that may be present in decant solutions are the following: 1. A mixture of ammonium nitrate and fuel oil (ANFO) is often used as a blasting agent at mining operations. A small percentage of ANFO used in blasting will remain unreacted and report as ammonia and nitrate is slurry tailings discharged into a tailings storage facility. The concentration of ammonia originating from this source is generally low, though in some circumstances ammonia removal from decant solution is required as a direct result of ANFO usage. 2.

Ammonia is one of the breakdown products of thiocyanate and cyanide, and forms through the hydrolysis of cyanate in a tailings storage facility. If cyanate is present in decant solution at an elevated concentration, then often there will be a correspondingly elevated concentration of ammonia.

i881

Through a combination of these two sources, ammonia removal from decant solution is occasionally required at mining operations, particularly if treated solution must be discharged into the environment. Primarily the concern is with toxicity to aquatic organisms since ammonia is generally not present in decant solution at concentrations that would be toxic to wildlife or waterfowl. If natural attenuation in a tailings storage facility is not sufficient to limit the concentration of ammonia, then implementation of a decant solution treatment system may be required for water to be discharged to the environment. Treatment options considered typically include biological, chlorination and ion exchange processes, though in some cases air stripping may be considered. Ammonia is readily oxidized through biological nitrification by an aerobic biological treatment process. This process is practiced at many municipal wastewater treatment plants throughout the world. The reaction product from this process is first nitrite and then nitrate, which is less toxic than ammonia but may also require removal depending upon its concentration. It is not uncommon for biological treatment plants to reduce ammonia to concentrations below 1 mg/L, however applications are generally limited to situations where the solution flow and chemistry do not fluctuate rapidly. Biological conversion of ammonia to nitrate can be conducted at relatively low temperatures, but reactions rates are higher at temperatures above about 10°C. Ammonia removal through the breakpoint chlorination process converts ammonia directly into nitrogen gas, thereby avoiding nitrate formation as would be encountered with other ammonia destruction processes. The chlorination process is efficient at removing ammonia to low levels and can also be configured to affect the simultaneous removal of cyanide, cyanate and thiocyanate as well as ammonia. A disadvantage with chlorination however, is that as ammonia concentrations increase, larger quantities of chlorine may be required depending upon the solution flow rate. In addition, chlorine added to solution will ultimately convert to chloride and increase the dissolved solids concentration in treated solution. If the concentration of either chloride or total dissolved solids is of concern in the treated water, then alternatives to chlorination should be considered. Chlorine is also highly toxic to aquatic organisms and a dechlorination process must follow any chlorination process used to treat water for discharge to the environment. In some cases, ion exchange is used to reduce ammonia concentrations without causing a significant increase in the dissolved solids concentration. Ion exchange is also not as prone to process upsets as a result of flow and chemistry fluctuations in comparison to biological treatment processes. Consideration of ion exchange is appropriate when concentrations of interfering species such as sodium, calcium and magnesium are relatively low and when the solution pH is less than about 9.0. Ion exchange resins are only suitable for solutions and not slurries. Under these conditions, ion exchange resins can be selective towards ammonia removal and the process may be economical for full-scale implementation. A significant disadvantage of ion exchange is that resins must be periodically regenerated using concentrated solutions of sodium chloride or sulfuric acid. These solutions along with all ammonia removed from solution will be present in the waste regenerant solution, and disposal of this waste solution is often difficult and expensive. In addition, ion exchange resins can become fouled due to the presence of certain dissolved metals in solution, and resin fouling can lead to high costs for purchasing new resin and for disposing fouled resin. Because of these disadvantages, ion exchange has only been practiced to a limited extent at mining operations.

1882

Air stripping of ammonia from solution at a pH above about 11.O is effective at reducing its concentration and in select applications this process may be economical. For example, if the solution to be treated has a pH near 1 1.O, then stripping of ammonia can be conducted with little or no initial pH adjustment. The disadvantage arises when the initial pH is below 11.0 and must be adjusted to the alkaline region using lime or sodium hydroxide. This adds to the cost of the process and increases the concentration of dissolved solids in the treated solution. In addition, if the pH of stripped solution must be lowered to less than 9.0 before being discharged into the environment, then sulfuric acid addition may be required. This also adds both to the cost of the process and to the concentration of dissolved solids in treated solution. Scale formation in process equipment may also be problematic due to carbon dioxide absorption from atmospheric air that will occur at elevated solution pH values. Because of these drawbacks, ammonia stripping is limited to applications when other treatment approaches would not be feasible or economical. Nitrate Treatment Nitrate is a relatively non-toxic compound at the concentrations typically observed in decant solutions and usually is not of concern relative to wildlife, waterfowl or aquatic organism toxicity. The primary concern with nitrate generally is related to drinking waters where elevated nitrite and nitrate concentrations can be harmful to humans, particularly young children and infants. In addition, nitrate is a biological nutrient and in some cases can lead to accelerated algae growth in waters, thereby consuming dissolved oxygen and impairing the ability of fish to survive. Nitrate is a relatively stable compound in surface waters and because of this, its removal from waters discharged to the environment is often required. There are relatively few treatment technologies that can be implemented on a full-scale basis to reliably lower nitrate levels, though the few that are available are effective and economic in many cases. The most widely applied nitrate treatment technology is biological denitrification which proceeds under anoxic conditions. In this process, nitrate is converted to nitrogen gas, which is then vented to the atmosphere. Like most biological process, denitrification is best suited for situations where the solution flow and chemistry do not fluctuate rapidly and where the solution temperature is above about 10°C to 15OC. The process does not significantly increase the concentration of dissolved solids, but does require the addition of a supplemental food source such as methanol or molasses. A key advantage of biological denitrification is that it can be coupled with an aerobic biological process to affect the removal of cyanide, cyanate, thiocyanate, ammonia and nitrate. Under many circumstances, biological treatment systems are inexpensive to construct, operate and maintain and will provide high-quality effluent. Ion exchange can also be used to remove nitrate to low levels, but as described for ammonia removal, the disadvantages of waste brine disposal and resin fouling have limited its application in the mining industry. Under conditions where the concentration of interfering compounds such as chloride and sulfate are relatively low, ion exchange may be an economical approach.

SUMMARY As indicated in the previous discussion, there are several treatment processes that have been successfully used worldwide for cyanide removal at mining operations. The key to successhl implementation of these processes involves consideration of the following: Site water and cyanide balances under both average and extreme climate conditions The range of cyanide treatment processes available and their ability to be used individually or in combination to achieve treatment objectives Proper testing, design, construction, maintenance and monitoring of both water management and cyanide management facilities

1883

By carehlly considering these aspects of water and cyanide management before, during and after mine operation, operators can reduce the potential for environmental impacts associated with the use of cyanide. Another aspect of cyanide treatment to be considered is the potential environmental impact of the cyanide related compounds cyanate, thiocyanate, ammonia and nitrate. These compounds may be present in mining solutions to varying extents and may require treatment if water is to be discharged. Each of these cyanide related compounds is affected differently in the treatment processes discussed and this should be considered when evaluating cyanide treatment alternatives for a given site. Table 10 provides a simplified summary of the general applications of various treatment technologies for the removal of iron cyanide and WAD cyanide. This table represents a very simplified summary, but can be used as a conceptual screening tool when evaluation cyanide treatment processes.

Table 10 Preliminary Guide to Selecting Cyanide Treatment Processes Iron Cyanide WAD Cyanide Slurry Treatment Process Removal Removal Application J J J S02/Air Hydrogen Peroxide J J J J Caro's Acid Alkaline Chlorination J J J J J Iron Precipitation J J Activated Carbon J J Biological J J Cyanide Recovery J Natural Attenuation J J

Solution Application J

J J J J J J J

REFERENCES Adams, M.D. 1992. The Removal of Cyanide from Aqueous Solution by the Use of Ferrous Sulphate. Journal of the South African Institute Mining & Metallurgy. 92: 1. 17-25. Botz, M.M. and T.I. Mudder. 1997. Mine Water Treatment with Activated Carbon. Proceedings Randol Gold Forum. 207-210. Botz, M.M. and T.I. Mudder. 2000. Modeling of Natural Cyanide Attenuation in Tailings Impoundments. Minerals and Metallurgical Processing. 17:4. 228-233. Botz, M.M. and T.I. Mudder. 2001a. An Overview of Cyanide Treatment and Recovery Methods. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited. Botz, M.M. and T.I. Mudder. 2001 b. Cyanide Recovery Applications for CCD Circuits. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited. Botz, M.M., W. Phillips, T. Polglase and R. Jenny. 2001. Processes for the Regeneration of Cyanide from Thiocyanate. Minerals and Metallurgical Processing. 18:3. 126-1 32. Dzombak, D.A., C.L. Dobbs, C.J. Culleiton, J.R. Smith and D. Krause. 1996. Removal of Cyanide from Spent Potlining Leachate by Iron Cyanide Precipitation. Proceedings WEFTEC 159'~ Annual Conference & Exposition. Given, B., B. Dixon, G. Douglas, R. Mihoc and T. Mudder. 2001. Combined Aerobic and Anaerobic Biological Treatment of Tailings Solution at the Nickel Plate Mine. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited. Goldstone, A. and T.I. Mudder. 2001. Cyanisorb Cyanide Recovery Process Design, Commissioning and Early Performance. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited.

1884

Ingles, J. and J.S. Scott. 1981. State-of-the-Art of Processes for the Treatment of Gold Mill Effluents. Environment Canada. Minerals Council of Australia. 1996. Tailings Storage Facilities at Australian Gold Mines. Moura, W. 200 1. Private Communication. AngloGold South America. Mudder, T.I. and M.M. Botz. 2001a. A Global Perspective of Cyanide. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited. Mudder, T.I. and M.M. Botz. 2001b. An Overview of Water Treatment Methods for Thiocyanate Removal. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited. Mudder, T.I., F. Fox, J. Whitlock, T. Fero, G. Smith, R. Waterland and J. Vietl. 2001a. Biological Treatment of Cyanidation Wastewaters: Design, Startup, and Operation of a Full Scale Facility. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited. Mudder, T.I., M.M. Botz and A. Smith. 2001b. Chemistry and Treatment of Cyanidation Wastes, 2"d Edition, London: Mining Journal Books Limited. Mudder, T.I., S. Miller, A. Cox, D. McWharter and L. Russell. 2001c. The Biopass System: Phase I Laboratory Evaluation. In The Cyanide Monograph, Ed. T.I. Mudder and M.M. Botz. London: Mining Journal Books Limited. Norcross, R. 1996. New Developments in Caro's Acid Technology for Cyanide Destruction. Proceedings of RandoI Gold Forum. 115-117. Omofoma, M.A. and A.P. Hampton. 1992. Cyanide Recovery in a CCD Merrill-Crowe Circuit: Pilot Testwork of a Cyanisorb Process at the NERCO DeLamar Silver Mine. Proceedings Randol Gold Forum. 359-365. Schmidt, J.W., L. Simovic and E.E. Shannon. 1981. Development Studies for Suitable Technologies for the Removal of Cyanide and Heavy Metals from Gold Milling Effluent. Proceedings 36Ih Industrial Waste Conference, Purdue Universify. 83 1-849. Simovic, L., W. Snodgrass, K. Murphy and J . Schmidt. 1985. Development of a Model to Describe the Natural Attenuation of Cyanide in Gold Mill Effluents. In Cyanide and the Environment, Vol. II. Ed. D. Van Zyl. 41 3-432.

1885

Strategies for Minimization and Management of Acid Rock Drainage and Other Mining-Influenced Waters Ron L. Schmiermund’

ABSTRACT Long-term treatment of non-process, mining-influenced waters (MIWs) is a significant economic, legal, and public relations liability. Problematic waters may not resemble classic ARD. Onerous treatment requirements can often be traced to shortcomings in characterization andor handling of waste rock. A holistic and non-prescriptive approach to waste management and MIW is recommended, including consideration of genetic and weathering aspects of subject ore deposits, flexible mine plans, testing programs adapted to the deposit and its environment, and recognition that future water treatment costs must be justified by present-day mining practices. This chapter identifies aspects of the approach and provides a framework for decision-making. INTRODUCTION AND SCOPE The adverse effect of mining on natural waters has been documented for nearly 2000 years and was almost certainly widely recognized long before that. The same causes and, unfortunately, many of the same consequences remain today to greater or lesser extents, depending on the history of a given mining district and its location in the world. In fact, water-related problems are frequently cited as the most obvious environmental issues associated with mining (Evans, 1996). These problematic waters are collectively referred to here as MIWs (Schmiermund and Drozd, 1997). A discussion of this topic may seem out of place in this volume, but it can be argued that contamination of natural water by mining is rapidly becoming the most significant technical challenge to mining and milling. Without an environmentally acceptable mining and processing plan, the most sophisticated mechanical and chemical techniques for recovering resources are wasted because the mine will not be permitted in a growing percentage of the world’s countries. Without an economically and environmentally viable method for closing a mine, all operational profits may well be consumed in very long-term post-closure maintenance. Many readers may automatically equate acid-rock drainage (ARD), typically with high sulfate concentrations and often with elevated metals, with MIWs. Although ARD may be the most common and most visible MIW, the topic is considerably broader than A m . MIWs include nonacidic, even hyper-alkaline waters, with or without high sulfate or common toxic heavy metals. Nitrate, cyanide, chloride, fluoride, or suspended sediments may constitute the offending constituents. The complexity of ARD alone, let alone the full spectrum of MIWs, is extreme and has been extensively explored by numerous authors. It is beyond the scope of this chapter to address the technical aspects of ARD or other MIWs except where they have direct bearing on the discussion, but the reader will be directed to key publications. It is the intent of this chapter to provide a framework for anticipating water-related problems as early as during exploration, throughout development, and during mine planning. Various mine components and their potential for contaminating water are discussed along with strategies for timely intervention in the processes that affect water. Finally, geologic, laboratory and field

I

Knight Pitsold and Co., Denver, Colorado.

1886

methods for predicting impacts of mining-related materials on natural waters are discussed. Throughout, prevention is the preferred management approach. A similar scope of subject matter has been addressed in much more detail, but still in an abbreviated form, by Morin and Hutt (1997), and the reader is directed there for practical theory and case studies. Indeed, a significant problem with these subjects is their breadth and variety and the need to translate knowledge at the microscopic scale to practice at the huge scale of modem commercial mining.

Summary of Relevant Geochemical Processes A myriad of chemical reactions impact MIWs. The following list attempts to include the generally most influential ones. Inorganic oxidation of simple iron sulfide minerals (pyrite, marcasite, pyrrhotite) by oxygen. Reaction products include protons, sulfate, and ferrous iron. Bacterially-catalyzed oxidation of simple iron sulfide minerals by ferric iron. Reaction products as above. Further oxidation of pyrite oxidation products, specifically ferrous iron to ferric oxyhydroxide solids (e.g., goethite, limonite, “yellow-boy”). Inorganic or organically-catalyzed oxidation of other metal or semi-metal sulfide minerals. Reaction products include sulfate, f oxidized metal ions, f metal or semi-metal oxyanions (e.g., molybdate, arsenate, selenate), and f protons. Reaction of sulfide oxidation products, especially protons, with carbonate minerals and select silicates to neutralize acidity. Concentration of sulfide-mineral oxidation products by evaporation with subsequent precipitation of secondary sulfate minerals. Dissolution of secondary sulfate minerals derived from sulfide-mineral oxidation. Dissolution of primary evaporite minerals or other relatively high-solubility minerals associated with an ore deposit (e.g., gypsum, sylvite, trona). Attenuation processes that limit mobility and/or toxicity of dissolved constituents, especially metals, including sorption, precipitation, co-precipitation, plant uptake, complexation with volatilization, photooxidation, and bacterial metabolism. Mechanisms that enhance mobility, including complexation, sorption to suspended sediments and colloidal-facilitatedtransport. Incorporation of components of industrial process into natural waters by dissolution or mixing, including cyanides, nitrates, organic flotation reagents, lixiviants, etc. For further information about the mechanisms listed above, the reader is directed to several chapters appearing in Plumlee and Logsdon (1998) and Filipek and Plumlee (1998) but especially Nordstrom and Alpers (1999) and Alpers and Nordstrom (1999). In addition, Perkins, et al. (1995) and other reports published by Mine Environment Neutral Drainage, or MEND, provide extensive discussions and references. See Tremblay and Hogan (2001) for an overview and summary of the MEND program. U.S. Geological Survey websites and InfoMine (www.infomine.com) provide additional information and portals to numerous other information sources.

MINING SITUATIONS THAT MAY ADVERSELY AFFECT WATER AND POTENTIAL INTERVENTION STRATEGIES The following situations or operational components may be found in modem metal mines and with a few limitations can be applied to coal mining and other non-metal mines. With the exception of direct discharges or accidental leakage from process facilities such as ponds, tanks, or pipelines, the sources of most natural water contamination can be traced to these situations or components. In Situ Exposed Rock in Surface and Underground Mines Most ore minerals and many associated gangue minerals are inherently chemically unstable at the earth’s surface. Consequently, when deposits are exposed during mining either in surface or

1887

underground mines, the contained minerals begin to chemically re-equilibrate in response to the new environment. This is especially relevant to many hypogene metal sulfide deposits, and sulfidic over and underburden associated with coal, evaporates, and chemical precipitate deposits. The products of such re-equilibration may become contaminants of surface waters, including pit lakes, and groundwaters and areas of exposure in mines are sources of those contaminants. Chief among them are the decomposition products of ore and gangue sulfide minerals (acidity, sulfate, and metals). In addition, artifacts of mining activities may be present in waters originating from areas of exposed rock, including nitrogen compounds from blasting agents, salts used for dust suppression and deicing, and general suspended sediments. Important variations to the foregoing are found in deposits that have either already reequilibrated, or partially re-equilibrated, with the surface environment (i.e., they have already undergone weathering) or consist of minerals that react very slowly under surface conditions. These include oxidized portions of originally sulfidic rock masses, supergene-enrichment zones of ore deposits (typically originally sulfidic), primary non-sulfidic ore deposits (e.g., iron, titanium, chromium, aluminum, and tin oxides), silicate mineral deposits, and most placers. The single most important commonality among deposits in this group is the relative paucity of sulfide minerals and thus the decreased capacity to produce ARD and associated phenomena. However, deposits in this group should not automatically be considered inert, and care should be taken in evaluating data from characterization tests designed for fresh rock (see Prediction Techniques). Open-pit high walls are obvious locations of exposure and weathering of ore and associated mineralized waste rock. Underground openings of all descriptions, but especially rubbleized or block-caved areas, are potential sites for intense weathering activity due to the focusing of air and water flows and possibly higher temperatures and relative humidities. At various times, all portions of a an ore body being actively mined will be exposed to weathering, but mining usually proceeds rapidly enough so that the main concern lies with material that will be exposed in the final pit wall or underground openings and thus be subject to weathering for long post-mining periods. An especially problematic and symbiotic situation results when underground and surface mines come in contact via shaft and drift penetrations of pit high walls. Block models employed in ore reserve estimation are convenient methods for characterizing the final pit walls, but provisions must be made early for generating and including the necessary data in the model. Unfortunately, the kind of data required for environmental predictions is not commonly collected during mine planning because it usually relates to non-ore minerals (e.g., pyrite, pyrrhotite, arsenopyrite, etc.) or non-ore elements. It is not uncommon that the database for a very well characterized deposit is virtually devoid of any information useful for estimation future environmental impacts. Examples of information that would be useful for environmental purposes and might be collected from exploratioddevelopmental drilling programs are given below. An early evaluation of the deposit from an environmental standpoint can better define the important variables to be collected. Distribution and abundance of gangue sulfide minerals, especially pyrite, marcasite, or pyrrhotite. Pyrite morphology (e.g., disseminated or in veinlets, coarsely crystalline or sooty). Total sulfur assays and a correlation to sulfide sulfur and visual estimates of pyrite in cores. Assays of important potential contaminants (e.g., arsenic, selenium, thallium, noncommercial copper, zinc, or lead). Distribution and abundance of reactive non-sulfide gangue minerals, especially carbonate minerals with distinctions made between calcite, dolomite, and siderite. Distribution and identity of secondary sulfate and oxides, aided by use of infrared spectrophotometer, e.g., alunite,jarosite, and scorodite). Distribution of rock types that may be relevant to chemical and physical weathering (especially carbonates and reactive silicates). Distribution of hydrothermal and supergene alteration zones with clear documentation of the mineralogic characteristic minerals of each zone.

1888

Information on primary and secondary porosity, fracture density, and fracture characteristics. Any information on water depths encountered during drilling, artesian flow rates from drill holes. Chemical analyses of water flowing from drill holes. Retain complete cores or cuttings (including overburden) from representative holes, especially near peripheral areas of the pit (i.e., near the ultimate pit limit). Retain assay pulps for later environmental-related analyses. Efforts to quantify the processes associated with weathering of pit high walls have resulted in complex conceptual and numericaI models that consider reaction rates, surface areas, fracture densities, diffusion of gases and liquids, and moisture movement. Alpers and Nordstrom (1999) review the complex geochemical aspects and existing models that pertain to reaction in this environment. Physical aspects of the wall-rock environment are complicated by the fact that pit walls are extremely irregular, and the presence of broken rock and the effects of blasting are difficult to quantify (Morin and Hutt, 1997). Tunnels and shafts penetrating the pit wall may dominate the recharge of groundwater to the pit. Because of the inherent complexity, any model of pit wall or underground face weathering behavior may require considerable calibration prior to accepting the results. One such model is MINEWALL 2.0 developed under the MEND program to address a wide range of situations. It is available in extensive documentation and demonstrated applications (MEND, 1995a; MEND, 1995b; MEND, 199%). Implementation of MINEWALL requires empirical inputs that characterize in situ sulfide oxidation rates and information on the production of solutes from exposed rock over time. To supplement in-field measurements that can estimate these variables or to substitute for them if provisions were not made for their collection, laboratory data (especially kinetic test data [see paragraph entitled “’Kinetic’ Laboratory Methods for Anticipating Weathering Behavior”]) have been used in various ways. Extrapolation of smallscale, accelerated laboratory experiments will always carry risks of yielding misleading information. Morin and Hutt (1997 and 1999) argue the value of using observational data as the source of rate and yield data for models and provide examples of data generated from field exposure experiments (see paragraph entitled “Site-specific Field Methods”). However, considerable foresight and time are required to obtain such data. Commissioning of field experiments should occur as soon as bulk rock becomes available, but valuable information can also be gained from undisturbed natural outcrops. More theoretical approaches to obtaining oxidation rates and solute production rates have been developed based on the Davis-Ritchie (D-R) equations (Davis and Ritchie, 1986a and 1986b), which assume that oxygen diffusion is the rate-limiting step in sulfide oxidation. One of the models/codes incorporating the D-R conceptual equations is PYROX (Wunderly, et al., 1996). Other proprietary codes incorporating the equations exist, but it should be recalled that the D-R equations are mainly about oxygen diffusion. They were originally applied to heap leach piles and subsequently to dumps and tailings where oxygen diffusion through pore spaces can be shown to be rate-limiting. Care must be taken to correctly address inherent differences in diffusion characteristics at exposed rock faces. In addition, the D-R equations do not always consider water films on reactive grains that may be significant in some dumps but perhaps not on exposed rock faces in arid climates. Fennemore, et a]. (1998) argues that the D-R approach overestimates oxidation progress in arid climates. Intervention Strategies For surface mines: Collect data on distributions of potentially problematic minerals and elements in much the same way that ore grades data are collected. Evaluate the costs of removing as much reactive rock (including non-ore) as possible; compare handling costs to incremental increases in water treatment costs.

1889

0

0

Identify problematic rock masses and control blasting in those areas; consider installation of long-hole drains and grouting during mining while equipment and access are available. Consider the real costs of long-term water treatment relative to benefits of extending a mining operation into and exposing problem material. Remove rubble from benches and design for rapid draining and reduced infiltration, especially in areas of more reactive rock.

For underground mines:

0

0

Consider mine designs that allow for post-mining flooding and oxygen exclusion. Consider mine designs that dewater reactive rock passively. Modify surface recharge areas where possible to limit flows into mine. Position shafts to maximize post-mining flooding. Install critical permanent bulkheads while access and equipment are available. Evaluate backstoping with beneficial material capable of reversing undesirable chemistry.

Waste Rock Dumps and Ore Stockpiles Waste rock dumps in conjunction with ore stockpiles present the same issues for contamination of waters as do pit walls or tunnels because they can contain the full variety of rock types present in a given ore deposit. However, the reactivity of the rock dumps can be higher than the corresponding in situ lithologies because of increased surface area and the potential for enhanced exposure to air and water. In addition, waste rock dumps may be relegated to environmentally unfortunate locations, including natural drainages, for logistical reasons. Dumps generally do not receive the level of engineering study necessary to control their internal structure, and thus water and air permeability, because dumps generate no revenue. However, failure to properly design and operate waste rock dumps can consume the revenues generated by other mining units in the form of long-term treatment of contaminated seepage. Waste rock dumps should be thought of as very large and exceedingly complex (bio)chemical reactors beyond the scope of this discussion. The interiors are difficult to investigate and are typically poorly known both physically (Smith, et al., 1995) and chemically (Ritchie, 1994). In addition to the waste rock material itself, reactants include infiltrating water from precipitation, groundwater seepage or runoff, and atmospheric oxygen. Although the boundary conditions for these external reactants can be easily quantified (e.g., precipitation amounts and atmospheric partial pressure of oxygen), their fluxes into the dump and distributions within the dump are known to be complex functions of many variables. Ongoing products of the dump reactors (i.e., seepage) represent their risk to natural waters and are easily quantified but difficult to predict over the life of the dump. Although problematic effluents from dumps are most commonly acidic, sulfate- and metal-enriched (i.e., ARD), they can be representative of every type of MIW. MEND (1996) describes individual waste rock dumps and discusses methods of predicting their environmental behavior. From the operator’s perspective, it is most useful to understand the aspects of waste rock dumps that can be controlled to minimize production of contaminated seepage. Fortunately, preventative measures are reasonably well understood and technically feasible. Probably the greatest benefit can be realized by properly selecting and preparing the dump site, applying construction techniques that segregate and encapsulate the most reactive wastes in less reactive material, and creating an internal structure that limits circulation of reactants. The greatest obstacle appears to be recognition of the necessity for investing capital and operating funds before and during construction to minimize closure costs and to prevent perpetual water treatment costs. Operators should be keenly aware that prevention of MIW drainage problems, especially ARD generation, is the only demonstrated viable long-term option. Once initiated and established, the reaction sequences that create ARD are virtually unstoppable. They may be retarded by oxygen deprivation or application of biocides, but reversal is highly unlikely. Elimination of infiltration would eventually stop seepage but is essentially impossible to achieve andor maintain. Peak contaminant (@loading from waste rock dumps may not occur until after mine closure, so early favorable indications should not be assumed to preclude later adverse behavior.

1890

Intervention Strategies Characterize footprint of proposed waste rock dumps with respect to existing surface and subsurface water flows and potential post-dump flow paths (e.g., faults, karst features). Evaluate interceptioddiversion ditches and sub-dump culverts or drains to isolate waste rock from water as much as possible. Design outer surfaces to minimize infiltration. Different climates and precipitation characteristics will dictate different designs. Coordinate dump design with mine plan based on block model containing environmentally-relevant information. Objective should be to isolate acid-generating (or other problematic) waste from external reactants and to associate it with acid-consuming (or other beneficial) material. Careful planning is required to minimize excess handling. Block-model data can be enhanced through rapid analysis of blast hole cuttings for critical or indicative environmental parameters (e.g., total sulfur, leachable sulfate, whole-rock arsenic, etc.). Real-time GPS-based systems for directing haul trucks based on ore grade can be similarly used to direct waste rock dumping. Identify waste rock or other waste materials with best contaminant-attenuating or reaction-inhibiting characteristics for preferential placement in the dump. Waste rock with high ANP can be placed below or mixed with rock with high APP to neutralize acid that might be produced. Materials with capacity to sorb contaminant metals can be placed downgradient of metal-rich waste rock to act as a primary barrier to migration. Organic material such as wood chips can be mixed with sulfidic material to act as a sacrificial reductant or oxygen “getter.” Possible application of biocides or sulfide mineral passivators to especially reactive sulfide material to inhibit reactions. Evaluate techniques for placing and otherwise modifying dump material to limit infiltration and vertical percolation of water and circulation of air. An effective internal permeability structure may be engineered through creation of fine-grained layers and/or capillary breaks. Selective dumping, truck traffic pattern, and ripping and dozing may create the desired effects. Limit air infiltration by eliminating or covering coarse-grained toe deposits. Create accurate as-built records of dump’s internal construction in the event that drilling for sampling or injection of biocide is required. Consider constructing sampling access ports to demonstrate performance and/or provide early warning of problems. Tailings and Tailings Storage Facilities (TSFs) Modem mines manage tailings through storage in engineered facilities or controlled placement in natural structures with favorable characteristics. Many of mining’s worst and sustained environmental legacies are sourced in uncontrolled tailings disposal into rivers and coastal waters. Engineered TSFs are most commonly designed to drain to allow compaction and physical stabilization of the tailings and thus will be sources of waste water that must be managed longterm. Other facilities simply leak. Waters exiting TSFs evolve over time from pure process water with a wide range of characteristics, depending of the beneficiation process, to rain water that has equilibrated with the tailings. Stability of the impoundment structure itself has also proven to be of great concern in both metal- and coal-mining operations. Severe high-profile failures have resulted in very significant impacts to natural waters and extremely high costs for remediation. Tailings are obviously derived from the same materials that require consideration in the cases of in situ exposures and waste rock dumps, but they differ in many important ways that can influence their chemical stability and thus the risk that they pose to natural waters. Tailings represent concentrations (relative to the overall deposit) of gangue minerals, in particular non-ore sulfide minerals that may have potential for ARD production and may be hosts of toxic metals. Tailings have increased and frequently high surface areas resulting from size separation and/or comminution and as a result may experience increased reaction rates. .

1891

Tailings are typically placed in TSFs as a slurry and retain high moisture contents for long periods, thus potentially facilitating the initiation of water-rock reactions and a cycle of water contamination. In addition, the liquid fraction of the tailings slurry may contain artifacts of the beneficiation process such as cyanide, acids, caustics, dissolved metals, and organic compounds. Depending on the placement of the TSF relative to the surrounding topography and water table, tailings may be a source of recharge to groundwater or a point of discharge. Robertson (1994) discusses this and other aspects of water movement associated with TSFs and notes that specifics are highly site-dependent. Predicting the potential impacts of TSFs on surrounding surface and groundwaters must consider the processes that render solutes mobile within the tailings mass as well as how those solutes might be transferred to the environment. Solutes may exist in the tailings slurry liquid fraction when it arrives at the TSF or arise from further in situ reactions of the tailings solids with the process fluids or new reactions with the atmosphere andor with the pore fluids as they evolve over time. As is the case of waste rock dumps, the most important of these reactions is likely to be pyrite oxidation. Elberling, et al. (1994) discusses the fbndamentals of a model for pyrite oxidation in tailings that are incorporated into computer codes such as RATAP (Scharer, et al., 1994) and WATAIL (Scharer, et al., 1993) and integrated approaches as described by Nicholson, et al. (2000). Wunderly, et al. (1996) discusses the PYROX model that was incorporated with PLUME2D into MINTOX (Molson, et al., 1997), another model for TSF evolution (Alpers and Nordstrom, 1999) for a more complete history of the evolution of MINTOX. From the foregoing discussion, it is obvious that tailings and TSFs require careful design and maintenance to avoid realization of the various potentials for contamination of natural waters. However, there are several characteristics of tailings masses that can be taken advantage of and perhaps optimized to limit risks. The fine-grained natures of tailings and the associated retention of moisture serve to impede the diffusion of oxygen from the atmosphere, thus dramatically limiting the rate sulfide mineral oxidation. Further, when the sulfide minerals are abundant throughout the tailings mass, oxygen is rapidly consumed very near the surface. In dry climates, oxidation of sulfides near the surface of the tailings mass may result in precipitation of secondary sulfate minerals that further limit oxygen infiltration, tend to stabilize the tailings by forming a crust, and may be self-repairing after rainfall events. Intervention strategies include: Optimize siting of TSFs to contain seepage and constraidfocus its exit from the facility. Design tailings placement to maximize efficiency of drainage. Selectively place problem materials Separate facility for waste streams containing problem materials (as opposed to COmingling prior to disposal). Physically stabilize tailings surface to prevent erosion and exposure of fresh surfaces. Choose material for final layer that optimizes weathering retardation. Heap, Valley-Fill, and Dump Leach Facilities Both heap and dump leaching can pose risks to ground and surface water for the following reasons:

1. Excursions of pregnant solutions occur even in cases of well-constructed pads with adequately designed liners. Dump leaching, especially when initiated as an afterthought, may pose higher risks because liners are not present and formal pregnant solution collection systems were never installed. 2. Handling of large volumes of pregnant and barren solutions, especially where large retention ponds are involved, incurs risk of leakage. 3. Rinsing, as may be required and appropriate for closure, generates volumes of solution that must be disposed of andor detoxified. Treatment of rinsates may result in sludges requiring safe disposal to protect natural waters. 4. Post-closure infiltration of precipitation can result in seepage from spent leach piles.

1892

5.

Leaching of sulfidic material is chemically equivalent to creating ARD and, as such, can be a very difficult process to stop.

Intervention Strategies Engineered pads offer distinct environmental advantages; consequently, intervention scenarios can be more proactive. Because a pad that functions as designed is likely to have the lowest operating and the best environmental performance, many of the intervention scenarios below will have multiple benefits. Fully characterize the material’s leaching characteristics, including rinse-down, prior to initiating leaching operations. Use sufficiently large apparatus to assure meaningful results. Anticipate disposal of rinse-down solutions. Engineer leak-detection systems that provide the earliest advice of a leak and information on the location of the leak. Design segmented pads that would allow for partial shutdown in the event of a localized leak. Require and preserve accurate as-built drawings of leach pads and pre-dump topography. Consider chemical reactions that might occur within the pad and block drainage systems, resulting in unacceptable head pressures on the liner and leakage. This may be especially important where a separate oxidant phase, in the form of air, is injected at the base of the pad and moves counter-current to solution flow paths. Install sampling access to during pad loading. This might consist of a vertical array of lysimeters to allow sampling of pore waters and gases during and operation and rinsing. Design sampling access into the leachate collection systems upstream of mixing boxes and manifolds to allow sampling of individual cells

In Situ Leaching/Solution Mining Examples of the application of in situ mining methods include: Surface application of acidic solutions to disseminated sulfide and oxide deposits with solution recovery via pumping from abandoned underground workings Subsurface injection of acidic solutions in copper sulfide deposits with solution recovery via pumping wells Injection of alkaline solutions into sandstone-hosted uranium oxide deposits with recovery via pumping wells In situ solubilization of trona and other evaporites left in pillars by injection of undersaturated process solutions into old underground workings and pumping from resultant mine pools The introduction into an aquifer of any solution that differs chemically from native groundwater risks contaminating the receiving groundwater. However, the risks may be mitigated by the nature of the aquifer, unusable ambient groundwater quality, or lack of reasonable access to or demand for the water. Groundwater contamination may occur in aquifers not specifically targeted for mining via leakage from injection or recovery wells or migration of solutions along structures. Solution mining may also pose risks to surface water because of the need to manage pregnant solutions and raffinates on the surface. Solutions unfit for reconditioning and reinjection must be disposed of, and land-application and infiltration galleries are often among the preferred options. Careful consideration must be given to the fate of all dissolved constituents present in solutions intended for land application. Depending on their mobility and attenuation, they may be taken up by plants, emerge as constituents in seep and spring waters, or become incorporated into shallow groundwaters.

1893

General Land Disturbance and Watershed Impacts Disturbance of the land surface is inherent in mining, and the associated water quality and water supply issues can be extremely important to stakeholders over large areas. Generally, the water quality issues arising from simple road building and other construction are limited to increased suspended sediments and turbidity. More serious impacts may be realized by diversion of runoff from one watershed to another or by interrupting the path of canals and ditches that supply water to downgradient areas. Intervention Strategies 0 Establish a complete inventory of existing water supplies, sinks, and conveyances in the area of a planned disturbance. Evaluate impacts to downstream users and apply appropriate industry-standard methods for sediment control. PREDICTION TECHNIQUES Recognition that water contamination issues associated with mining, especially ARD, required management and should be prevented if possible became commonplace about the middle of the 20” century. It was probably inevitable that laboratory methods would be sought to quantify the hture behavior of geologic materials in response to weathering and to predict the effects on natural waters. Industry and regulators have come to rely upon those methods that are collectively considered the tools of “waste characterization” studies. Unfortunately, problems abound, and there are distinct conceptual and operational biases toward one aspect of MIWs - prediction of the generation of low-pH waters resulting from the oxidation of pyrite. It is important to realize those biases and thus the limitations lnherent in “standard” waste characterization. This chapter only addresses the major categories of laboratory tests. The references provide a much broader treatment. White 111, et al. (1999) traces several methods in common use today to 1974 and the eastern U S . bituminous coal region. The fact that those methods were developed for coal and are now routinely used to evaluate all mining waste materials, often with little or no recognition of the underlying assumptions, is indicative of shortcomings in current industry-standard practice, This seems to be the case despite extensive and credible scientific work exposing those shortcomings, and wide distribution of the findings. Detailed descriptions of testing methods are outside the scope of this chapter, but references are provided. The material addressed here is sufficiently complex that the advice of an experienced practitioner is recommended before committing to a program of waste characterization or interpreting the results of a previous one. The following paragraph entitled “Using Economic Geology to Anticipate Water-related Problems” discusses a non-laboratory approach that may well provide the most reliable guidance but, unfortunately, would be considered unconventional and lacks the appearance of quantitative credibility offered by laboratory tests. Using Economic Geology to Anticipate Water-related Problems A concept possibly introduced by Kwong (1993), expanded upon by Plumlee, et al. (1994) and Plumlee, (1999), and applied to a wide variety of mineral deposits by du Bray (1995) is remarkable in its logic and usefulness. This method relies on the extensive economic geology and environmental knowledge of existing mineral deposits around the world and typical mining and beneficiation methods associated with those deposits. That knowledge is then used to predict the behavior of other analogous deposits and mines. Each class of mineral deposit has been associated with a “geoenvironmental model.” The basic models invite customization that considers sitespecific climates, geologic variations, and specifics of the beneficiation process. The data and observations that form the basis for these geoenvironmental models have clear advantages over laboratory data because they inherently include all variables, are full-scale, and accurately reflect the influence of time. Geoenvironmental models of mineral deposits may be applied at all stages of development, ranging from regional conceptual exploration to expansion or closure planning for an active mine.

1894

However, early application would seen to offer the greatest benefit if it applies to decisionmaking schemes as outlined by Lord, et al. (2001). Application of geoenvironmental models of ore deposits is essentially free and should prevent most environmental surprises. Static Laboratory Methods for Predicting Acid Production and Neutralization Considered the most fundamental of the waste characterization laboratory tests is a group of procedures known as the “static tests” or “acid-base accounting” (ABA) tests. In general, the procedures can be divided into two types - those that estimate acid generation or production potential (AGP or APP) and those that estimate acid neutralization potential (ANP). Results are typically expressed in terms of tonnes of CaC03 per 1,000 tonnes of rock and often assessed by calculating the difference or net acid neutralizing potential (ANP - APP = NANP or NNP) or the ratio (ANPIAPP). Paste pH Although simple and very inexpensive paste pH can provide much useful, albeit qualitative, information, COASTECH Research Inc. (1991) provides details for the simple procedure. The resultant pH in the high solids/liquid reflects the influence of only the most rapidly reacting, typically most soluble, solids. Generally, hydrolysis associated with selected sulfate, carbonate, hydroxide, chloride, and nitrate salts will dominate such reactions. Redox reactions, especially heterogeneous ones, are expected to have no effect because they are slow relative to the duration of the test. Accordingly, fresh sulfide minerals have no direct effect, but reaction products from prior oxidation of sulfides may be responsible for low paste pHs and generally indicative of ARD potentials. Minerals in this group include secondary sulfate minerals as discussed by Nordstrom (1982). Where evaporative conditions are influential and result in abundant secondary sulfate minerals after sulfide mineral oxidation, the paste pH will predict the runoff pH from the next rainfall event. Paste pH values near and above neutrality clearly reflect the absence of acid sulfate minerals but may not accurately reflect the ultimate equilibrium pH because of relatively slower reaction rates for hydrolysis of carbonate minerals. Estimating APP. Estimating APP appears simply but is really quite complex. It relies on measures of total sulfur in a material and the sulfur contained in one or more chemical extractions designed to liberate specific fractions of the total sulfur. These are referred to as sulfur forms or species, the most relevant of which is sulfide-sulfur. The objective is to quantify the sulfur contained in those sulfide minerals that will liberate protons when oxidized. According to standard procedures, sulfide-sulfur concentration is assumed to be entirely contained in pyrite and is then mathematically converted to APP according to the following simple formula:

APP (T CaCOJl 000 T rock) = wt % SsU&de X 3 1.25 The multiplicative factor 31.25 is referred to below as “the factor.” The following points pertain to determining and interpreting APP values: Measures of sulfur species are only estimates defined by the extraction procedure. Independent confirmation should be sought. Multiple and potentially significantly different “standard” methods for estimating sulfur species exist, including Sobek, et al. (1978) and ASTM (1988). Unfortunately, results are not assured to be comparable, and many laboratories have modified the “standard“ methods making inter-laboratory comparisons questionable; not all laboratories are candid about such modifications. Both standard methods are specifically designed to discriminate between sulfur found in the common sulfide minerals in coal (pyrite or marcasite), organically-bound sulfur (probably only significant in coal), and those readily-soluble sulfate minerals found in coal. No provision is made for non-pyrite/marcasite sulfide minerals, and it is left to the consumer to request tests that might elucidate this issue. Not all sulfide minerals are subject to dissolution by the standard methods, SO not all true sulfide-sulfur may report to sulfide-sulfur as defined by the methods.

1895

A wide variety of oxidation stoichiometries apply to sulfide minerals found in base and precious metal deposits, but the 3 1.25 factor described above is strictly based on: 7 FeS,+-O,+H,O+ 2

CaCO, -+ H’

Fe2’+2SOi-+2Hi

Ca”

+ HCO,

In fact, it can be shown that the factor ranges from negative values in the case of sulfide minerals that actually consume acid upon oxidation (e.g., chalcocite and pentlandite), zero for chalcopyrite, to nearly 100 for realgar. The prevailing stoichiometry may also vary if oxidants other that oxygen are introduced into the system. If ferric iron is introduced in quantities greater than what might be achieved by the oxidation of locally available pyrite, the applicable factor could exceed several hundred. The basis of the factor in stoichiometry implies the assumption that the reactions go tQ completion, thus the calculated APP represents a maximum capacity that may not be realized or may be realized only very slowly.

Estimating ANP. Considerably more attention has been given to ANP determinations relative to APP determinations and, in fact, ANP is even more subject to the details of the testing procedure. Unfortunately, numerous methods exist with subtle operational differences that can produce notso-subtle differences in the results. Three fundamental approaches to ANP estimations exist: The most commonly used methods involve mixing a known mass of rock with a known quantity of acid, allowing them to react with or without boiling, followed by a backtitration with a standardized base to determine the amount of acid consumed. Results from this seemingly simple method are subject to variations sourced in the grain size of the specimen, the molarity of the acid, temperature and duration of the reaction period, redox conditions of the experiment, and the endpoint pH used for the back-titration. An especially sensitive variable involves the use of “the fizz test,” a highly subjective method of choosing the appropriate molarity and volume of acid. White 111, et al. (1999) describes five methods that follow this approach. COASTECH Research Inc. (1991) provides detailed methods for three. Lawrence and Wang (1 997) discuss the sometimes dramatic influence of method details on the outcome of ANP tests. Lawrence and Wang (1996) explore in great detail issues related to the various ANP techniques. Methods in this group are variously known as Standard Sobek, Modified Sobek, the Lawrence Method, B.C. Research Method, Standard and Modified NP (pH 6 ) . It is incumbent upon the consumer to obtain a complete description of the method as practiced by an individual laboratory and to not rely on the advertised name of the method. No one method is universal, and no method currently available is without merit in some application. In general, it is not possible to choose the best ANP method a priori. Rather, it is strongly recommended that any significant waste characterization program initially include a comparison of ANP methods and reconciliation of the results with the deposit mineralogy (see below). A second method tends to reinforce a popular oversimplification of ANP that presumes that simple carbonate minerals are the sole, or only important, basis for ANP. It employs measurements of total carbon and organic carbon to estimate totalinorganic carbon by difference. Calculations are then made based on the assumption that all inorganic carbon is present as simple calcium and magnesium carbonate minerals that are available for acid neutralization. Although, this approach may be applicable and highly efficient in specific cases where simple carbonates are predominant, it is not recommended for general use. This method has been submitted for inclusion by ASTM. See White 111, et al. (1999) for further discussion and references.

1896

A third approach involves direct, quantitative determination of mineral abundances using microscopy, X-ray diffraction in conjunction with X-ray fluorescence, mineral separations, or other applicable techniques. A theoretical acid consumption could then be assigned to each mineral. This method is probably only rarely used alone but should be performed on representatives of any unfamiliar group of materials for validation of assumptions about the identity of minerals responsible for acid neutralization. “Kinetic” Laboratory Methods for Anticipating Weathering Behavior The so-called “kinetic” procedures are intended to extend the information accumulated from the static tests. While static tests purport to measure the ultimate potential of a rock for acid generation, kinetic tests attempt to provide information about whether that potential will be realized. The format of most kinetic tests is to induce accelerated weathering by enhancing the conditions thought to increase weathering rates, specifically pyrite oxidation rates. No consideration is given to acid-neutralizing rates. Testing typically consists of alternately exposing crushed rock to flows of air with higher and lower relative humidity during a one-week period and flushing the material at the end of the week to remove soluble reaction products. The one-week cycle is then reinitiated. With few exceptions, the tests do not record andor control air flow rates, relative humidity, or temperature. In addition, grain size and chamber configurations are variable between laboratories. Test methods following the procedures outlined above are considered “conventional” and are conducted in various configurations of chambers but were originally conducted in shoebox-sized boxes (humidity cells). ASTM (1996) describes an analogous apparatus and method using a column and a far more rigorous procedure for controlling testing conditions. Other, less commonly used procedures have been devised but are in many ways more abstract relative to field conditions. These include a single-pass column extraction, the B.C Research Confirmation Test that forces bacterial (Thiobacillusferrooxiduns) sulfide mineral oxidation, Shake-Flask where all reactions take place in solution under agitation with or without bacteria, and Soxhlet Extractions that recirculates extraction fluids through a small rock sample at elevated temperature. COASTECH Research Inc. (1991) provides detailed procedures for humidity cells and the latter four types of kinetic tests. A conventional kinetic test can provide indications of the relative rates of acid-producing and acid-neutralizing reactions. However, it is a common misconception that kinetic tests provide information on absolute rates of reaction under field conditions. In general, they cannot. This can only be accomplished when extensive empirical correlations exist between true field-based measurements of rates and laboratory tests; in that case, why perform kinetic tests? Kinetic testing, as currently practiced, is far from being sufficiently sophisticated to predict absolute rates. One benefit of kinetic testing is that liberation of constituents (other than protons, iron, and sulfur) resulting from sulfide mineral destruction and acidic attack of other minerals can be evaluated. All the procedures described above generate solutions that may be analyzed for their dissolved constituents and essentially represent a leaching experiment under conditions affected by sulfide mineral oxidation. Kinetic tests, as originally proposed by Sobek, et al. (1978) took the form of “humidity cells.” Their use was recommended as a follow-up to acid-base accounting when the results of those tests were “inconclusive,” i.e., could not be interpreted as clearly predicting net acid-producing or acidneutralizing behavior. This recommendation for using kinetic testing as a “tie-breaker” seems to have given rise to another misapplication of the procedures by implying that the tests would demonstrate whether or not the material would “go acid.” Kinetic testing used in that mode will always be either indefensible or nearly worthless. Results are indefensible if testing reveals no net acid production because it can always be argued that the testing time was insufficient. Confirmation of this problem is offered by the fact that the recommended duration of the original tests (for coal-related materials) was ten weeks. The accepted time has steadily increased through 20 and 40 weeks, and examples of rocks finally “going-acid” after several years of continuous testing have been published. Results are nearly worthless if net acid production is finally achieved because static testing would have likely already predicted that the total acid-neutralizing capacity was less than the total acid-producing capacity. Further, as discussed above, the time required to achieve net acid 1897

production is of extremely limited use for predicting field behavior. Even comparing the apparent relative rates of acid production and acid neutralization can be misleading because the experiment was designed to increase sulfide mineral oxidation only. Routine use of any of the above “kinetic” tests in a waste characterization program should be carefully considered because of their relatively high costs, potentially long times required, and the ambiguous results. Only a very focused program with narrowly defined objectives and supporting field data is likely to make significant and definitive use of kinetic testing. Laboratory Methods for Predicting Metal Leachability In addition to an evaluation of a material’s potential to produce ARD or to liberate metals and other constituents in conjunction with ARD production, there is concern for constituent leachability under non-ARD conditions. Laboratory tests addressing this situation subject crushed rock to agitation with mildly acidic solution designed to simulate natural precipitation or solutions containing dilute organic acids from degradation of organic matter. Tests simulating leaching by natural precipitation include the Synthetic Precipitation Leaching Procedure (EPA, 1994a) and the very similar, old version of the Nevada Meteoric Water Mobility Procedure. Both use a bottle-roll apparatus to mix a relatively small mass of crushed rock and artificial precipitation for 18 to 24 hours. Leachates are decanted, filtered, and analyzed. A revised Nevada Meteoric Water Mobility Procedure (Standardized Column Percolation Test Procedure) re-circulates a mass of artificial precipitation through an equal and relatively large mass of coarsely crushed rock (NDEP, 1996). Another procedure commonly but potentially inappropriately used is Toxicity Characteristic Leaching Procedure, or TCLP (EPA, 1994b). Although leachability tests are straightforward in their execution and the results usehl for comparison of materials, a quantitative interpretation of the data for field applications is very difficult. The concentration of a given metal in the leachate from any one of the above procedures is a function of the concentration of the metal in the solid, its extractability, and the volume of the water relative to the mass of the solid. What is an appropriate solidliquid ratio for simulating field conditions? Furthermore, it is unlikely that a human receptor would attempt to drink the equivalent of a direct leachate from a mining waste material. How much dilution should be incorporated into the calculation of concentrations at the point where a receptor would drink the water? The only standards for evaluating a leachate directly are assigned to just eight “RCRA” metals and are associated with the TCLP test only. The standards in that case are inexplicably based on 100 X drinking water standards and may be irrelevant to a given mine situation. To be realistic and meaningful, leachability studies should address actual field conditions. Data should be obtained from on-site exposure experiments that determine actual ratios of rainfall to runoff for a given mass of rock at the actual size anticipated for a dump or other situation. Runoff concentrations should then be evaluated in the context of a site-wide water balance model that includes meteorological conditions, location, flow and mass of receiving water bodies, and locations of receptors. Site-specific Field Methods In addition to laboratory tests, field weathering or kinetic tests may be conducted at the mine site in advance of mining and continued throughout operations. Among the obvious advantages of onsite experiments over laboratory tests are that environmental conditions are far more representative of expected post-mining conditions, time is not compressed in an unquantifiable way, and physical scales (especially particle sizes and rock masses) can be more appropriate to actual conditions. Further advantages are the low cost and ability to use existing personnel to monitor the experiments. Disadvantages include the time required for acquisition of useful data and the requirement for greater amounts of sample material. Non-ideality remains and quantification and upscaling can sill be challenging, but laboratory data must be reconciled to such empirical data. Morin and Hutt (1997) discuss two basic types of on-site weathering experiments that can be operated under ambient conditions to provide non-laboratory information on ARD production and characteristics. These include piles, cribs, or bins of rock exposed to ambient weather and equipped with a drainage collection system to intercept all runoff. A second type is “minewall stations” that collect runoff from known areas of rock faces exposed in mine settings or natural

1898

outcrops. Depending on the location and setting of these rock faces, the runoff may be derived from atmospheric precipitation or groundwater sources from either the saturated or vadose zones. Extrapolation of information gained from either type of experiment can be problematic. Rock piles may differ from actual conditions of concern in terms of exposed surface areas and thus the efficiency of wetting by a given precipitation event. Not only is runoff volume a non-linear function of precipitation volumes, solute concentrations in runoff are related to runoff volumes, and concentrations in a given runoff event are influenced by the nature of the previous event. In the case of rock face stations, similar issues exist, and additional efforts must be made to separate and quantify solution contributions from various sources. Despite the shortcomings and interpretive complexities inherent in any experiment with a large number of variables, there can be no excuse or rationale for not establishing a robust on-site experimental program early in the operation and maintaining it throughout the mine life and possibly beyond. The program should include an annual review of the data with consideration of what impacts the results might have on future mining practices at the site. In addition, careful consideration should be given to the applicability of the experiment design - will it provide the required information for closure design, runoff prediction, long-term tailings behavior, or pit lake chemistry prediction? Considerable financial advantages may be derived from a well designed and operated program.

REFERENCES Alpers, C.N and D.K. Nordstrom. 1999. “Geochemical modeling of water-rock interactions in mining environments.” The Environmental Geochemistry of Mineral Deposits Part A: Processes, Techniques and Health Issues, Reviews in Econ. Geology, Vol. 6B, edited by G.S. Plumlee and M.J. Logsdon, Littleton, CO: SOC.Econ. Geologists, 289-323. ASTM. 1988. “Method 4239-85.” American Society for Testing and Materials Annual Book of ASTMStandards, Vol. Sec. 5 . , Vol. 05.05: Standard test methods for sulfur in the analysis of coal and coke using high temperature tubefurnace combustion method, 385-90. ASTM. 1996. Standard Test Method for Accelerated Weathering of Solid Materials Using a Modified Humidity Cell. Tech. Report No. D-5744-96. Am. SOC.Testing Materials. COASTECH Research Inc. 1991. Acid rock drainage Prediction Manual - A manual of chemical evaluation procedures for the determination of acid generation from mine wastes. Prepared for CANMET - MSL Div. and Dept. of Energy, Mines and Resources, Canada NO. MEND Project 1.16.lb. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage (MEND). Davis, G.B. and A.I.M. Ritchie. 1986a and 1986b. “A model of oxidation in pyritic mine wastes: Part 1: Equations and approximate solutions.” Appl. Math. Modeling, 10 (October): 314-22. du Bray, E.A. (ed.). 1995. “Preliminary compilation of descriptive geoenvironmental mineral deposit models.” U S . Geol. Survey. Elberling, B., R.V. Nicholson, E.J. Reardon, and P. Tibble. 1994. “Evaluation of sulfide oxidation rates: A study comparing oxygen fluxes and rates of oxidation product release.” Can. Geotech. J.,31: 375-83. EPA. 1994a. ‘“Toxicity Characteristic Leaching Procedure.” Test Methods for Evaluating Solid Wastes, Physical/Chemical Methods Laboratory Manual, SW-846. U.S. Environmental Protection Agency. EPA. 1994b. “Synthetic Precipitation Leaching Procedure.” Test Methods for Evaluating Solid Wastes, Physical/Chemical Methods Laboratory Manual, S W-846. U.S. Environmental Protection Agency. Evans, B. 1996. An environmental perspective on the national and global mining industry. In Maintaining Compatibility of Mining and the Environment, edited by G.H. Brimhall and L.B. Gustsfson, 17-2 1. Littleton,CO: SOC.Econ. Geologists. Fennemore, G.G., W.C. Neller, and A. Davis. 1998. “Modeling pyrite oxidation in arid environments.”Environ. Sci. Technol, 32, No. 18: 2680-87. Filipek, L.H., and G.S. Plumlee, Editors. 1998. “The Environmental Geochemistry of Mineral Deposits.” Reviews in Economic Geology, Vol. 6B. Littleton, CO: SOC.Econ. Geologists. Kwong, Y.T.J. 1993. Prediction and prevention of acid rock drainage from a geological and mineralogical perspective. Tech. Report No. 1.32.1. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage (MEND). 1899

Lawrence, R.W. and Y . Wang. 1996. Determination of neutralization potential for acid rock drainage prediction. Tech. Report No. 1.16.3. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage (MEND). Lawrence, R.W. and Y. Wang. 1997. “Determination of neutralization potential in the prediction of acid rock drainage.” Prediction. Fourth International Conference on Acid Rock Drainage, Proceedings. Vancouver, B.C., Canada: 449-464. Lord, D., M. Ethridge, M. Wilson, G. Hall, and P. Uttley. 2001. “Measuring exploration success: An alternative to the discovery-cost-per-ounce method of quantifying exploration effectiveness.” SEG Newsletter, 45 (April). MEND. 1995a. MINEWALL 2.0: User’s Manual. Tech. Rept. No. MEND Project 1.15.2a. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage. 55 pp. plus diskette. MEND. 1995b. MINEWALL 2.0: Literature review and conceptual models. Tech. Rept. N o . MEND Project 1.15.2b. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage. 97 PPMEND. 1995c. Application of MINEWALL 2.0 to three minesites. Tech. Rept. No. MEND Project 1.15.2~.Ottawa, Ontario, Canada: Mine Environment Neutral Drainage. 193 pp. Molson, J.W., D.W. Blowes, E.O. Frind, J.G. Bain, and M.D. Wunderly. 1997. Metal transport and immobilization at mine tailings impoundments, Waterloo Center for Groundwater Research, Univ. Waterloo no. MEND Associate Proj. PA-2. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage. Morin, K.A., and N.M. Hutt. 1997. Environmental Geochemistry of Minesite Drainage: Practical Theory and Case Studies. Vancouver: Minesite Drainage Assessment Group (MDAG Publishing). Morin, K.A., and N.M. Hutt. 1999. “Prediction of drainage chemistry in post-mining landscapes using operational monitoring data.” Paper presented at the Ecology of Post Mining Landscapes, Cottbus, Germany. NDEP. 1996. Meteoric Water Mobility Procedure (MWMP), Standardized Column Percolation Rest Procedure. Nevada Division of Env. Protection. Nicholson, A. D., M.J. Rinker, , P. Tibble, G Williams, and M. Wiseman. 2000. “An integrated approach to assess acid generation and metal release from sulfide tailings using oxygen consumption measurements, porewater chemistry, and geochemical modeling.” Paper presented at the Fifth International Conf. on Acid Rock Drainage (ICARD 2000). Denver, Colorado: SOC.Mining, Metal and Explor. (SME). Nordstrom, D.K. 1982. “Aqueous pyrite oxidation and the consequent formation of secondary iron minerals.” Acid Surfate Weathering, SSSA Spec. Pub. No. 10, edited by D.M. Kral, 37-56. Madison, WI: Soil Science SOC.Am. Nordstrom, D.K., and C.N. Alpers. 1999. “Geochemistry of acid mine water.” The Environmental Geochemistry of Mineral Deposits Part A: Processes, Techniques and Health Issues, Reviews in Econ. Geology, Vol. 6B, edited by G.S. Plumlee and M.J. Logsdon, 133-60. Littleton, CO: SOC.Econ. Geologists. Perkins, E.H., H.W. Nesbitt, W.D. Gunter, L.C. St-Arnaud, and J.R. Mycroft. 1995. Critical review of geochemical processes and geochemical models adaptable for prediction of acidic drainage from waste rock, Noranda Technology Center, MEND Report 1.42.1. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage (MEND). Plumlee, G.S. 1999. “The environmental geology of mineral deposits.” The Environmental Geochemistry of Mineral Deposits Part A: Processes, Techniques and Health Issues, Reviews in Econ. Geology, Vol. 6B, edited by G.S. Plumlee and M.J. Logsdon, 71-1 16. Littleton, CO: SOC.Econ. Geologists. Plumlee, G.S., and M.J. Logsdon, Editors. 1998. The Environmental Geochemistry of Mineral Deposits. Reviews in Economic Geology, Vol. 6A. Littleton, Colorado: SOC.Econ. Geologists. Plumlee, G.S., K.S. Smith, and W.H. Ficklin. 1994. “Geoenvironmental models of mineral deposits, and geology-based mineral-environmental assessments of public lands.” U S . Geol. Survey Open File Report, 94-203.

1900

Robertson, W.D. 1994. “The physical hydrology of mill-tailings impoundments.” The Environmental Geochemistry of Sulfide Mine-wastesShort Course Handbook, Vol. 22, Waterloo, Ontario, May, 1994, edited by J.L. Jambor and D.W. Blowes, 1-17. Nepean, Ontario: Mineralogical Assoc. Canada. Scharer, J.M., W.K. Annable, and R.V. Nicholson. 1993. “WATAIL - A tailings basin model to evaluate transient water quality of acid mine drainage.” Report to Falconbridge, Ltd. and MEND-Ontario. Waterloo, Ontario: Institute for Groundwater Research, University of Waterloo. Scharer, J.M., R.V. Nicholson, B. Halbert, and W.J. Snodgrass. 1994. “A computer program to assess acid generation in pyritic tailings.” Environmental Geochemistry of Sulfide Oxidation, ACS Symp. Ser. 550, edited by C.N. Alpers and D.W. Blowes, 132-52. Washington, D.C.: Am. Chemical SOC. Schmiermund, R.L., and M.A. Drozd. 1997. “Acid mine drainage and other mining-influenced waters (MIW).” Mining Environmental Handbook, edited by J.J. Marcus, 599-617. London: Imperial College Press. Smith, L., D. Lopez, R. Beckie, K.A. Morin, R. Dawson, and W. Price. 1995. Hydrogeology of waste rock dumps. Tech. Rept. No. MEND Associate Proj. PA-1. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage. Sobek, A.A., W.A. Schuller, J.R. Freeman, and R.M Smith. 1978. Field and laboratoly methods applicable to overburden and minesoils. U.S.E.P.A. Tremblay, G.A., and C.M. Hogan, Editors. 2001. MEND Manual. Ottawa, Ontario, Canada: Mine Environment Neutral Drainage (MEND). Whlte, 111, W.W., LapakkoKA., and R.L. Cox. 1999. “Static-test methods most commonly used to predict acid-mine drainage: Practical guidelines for use and interpretation.” The Environmental Geochemistry of Mineral Deposits, Part A : Processes, Techniques and Health Issues, Reviews in Econ. Geology, Vol. 6B, edited by G.S. Plumlee and M.J. Logsdon, 352-38. Littleton, CO: SOC.Econ. Geologists. Wunderly, M.D., D.W. Blowes, E.O. Frind, and C.J. Ptacek. 1996. “Sulfide mineral oxidation and subsequent reactive transport of oxidation products in mine tailings impoundments: A numerical model.” Water Resources Res., 32, No. 10: 3 173-78.

1901

Environmental and Social Considerations in Facility Siting Barbara A. Filas, Robert W. Reisinger, and Cynthia C. Parnow’

ABSTRACT The environmental and social setting of a project plays an increasingly significant role in the siting of contemporary mining and processing facilities. While the mine location and development options are relatively fixed, there is far more flexibility in the siting of the plant site, waste repositories, and ancillary facilities. This paper explores the various environmental and social concerns to be taken into account in the siting of various project components. It discusses the analytical and evaluation processes by which impacts are quantified and priorities are set. These elements are integrated to arrive at a final project layout that is not only technically and economically feasible but also sensitive to important environmental and social variables. INTRODUCTION Selecting the best sites for the major support facilities at a mine is critical to the long-term success of the project. Moving material from one location to another for processing, shipping, and waste disposal is expensive, and the further the distance that material will need to be moved, the more costly the operating expenses will typically be. However, proximity of support facilities to the mine itself is not the only siting consideration that will affect the economics of a project. Environmental and social considerations can also have a profound effect on’the overall costs and profitability of a project if their effects are not balanced with other siting considerations. While the nature and extent of the mineral resource dictates where the mine will be located, there are usually many options available for siting support facilities. Support facilities may include primary crushing facilities, concentrating and refining facilities, coarse and fine waste disposal sites, fresh water supply reservoirs, and a variety of support buildings including offices, truck shops, and warehouses. None of these support facilities is fixed by the location of the mineralization, and, as such, site selection must consider the environmental effects, social ramifications, and project economics - commonly referred to as the “triple bottom line” - as an integrated and balanced decision. Every project is unique. There is no single set of rules to define the perfect layout to accommodate every design instance. It is important to prioritize the elements that contribute to a siting decision, fully understand the cost ramifications of each element, and then balance those factors into the final siting decisions. It is not enough to simply look at the engineering aspects any more as the environmental and social costs of a project are as much a part of the contemporary mining project feasibility as the act of mining itself. The remainder of this paper discusses how strategically locating mining and processing facilities can enhance the environmental and social aspects of a project. Environmental factors are presented first, followed by social factors. ENVIRONMENTAL FACTORS The location of major project components can have a significant effect on air, water, and land resources. Because every site and every project is unique, each of the environmental disciplines Knight PiCsold and Co., Denver, Colorado

1902

must be considered on their own merit and relative to the regulatory purview under which the project will operate. For example, the most important consideration for a cement-producing operation may be the ability to meet air quality requirements at major point sources such as the cement kiln. For a gold mine, it may be the control of cyanide solutions from a heap leach pad or tailings storage facility. For a mountaintop removal coal mine, the ability to reclaim the site to an acceptable landform may be the most important consideration. In each of these examples, other environmental factors may contribute to the decision balance, but the critical issues will most likely drive the decisions in terms of public and/or regulatory acceptance of a particular project design. Most regulatory jurisdictions in the world now require some sort of environmental impact evaluation for mining projects. Even the major international lending institutions have promulgated policies and guidelines for conducting environmental impact assessments of the projects they finance. In the past, it was common for the engineering planning and facility siting to occur exclusive of environmental considerations and the final layout then supplied to environmental analysts for impact evaluation. Recently, as various project alternatives have been evaluated in the impact assessment process and the public has become more involved in the regulatory or lending institution decisions, changes to site selection and design decisions due to environmental factors are becoming increasingly common. As a result, factoring environmental issues into the upfront planning and design process has become more the norm than the exception. The environmental impact assessment process is relatively straightforward and well understood. It is a rigorous process of establishing a detailed profile of the existing site conditions in the project area prior to project development. Baselines are typically established that profile physical resources such as climate, air quality, landform, geology, soils, surface water, and groundwater. Ecological disciplines typically profile the terrestrial and aquatic flora and fauna in the project area and how the habitats factor into local and regional ecosystems as a whole. These conditions establish the background against which project development plans and designs are overlain and evaluated. The evaluation considers, resource by resource, the impacts that project development will have on each receiving media. As significant impacts are identified, mitigation measures are considered as needed to reduce the level of impact to each receiving media to acceptable levels. Environmental scoping is a process by which the key environmental considerations are identified in advance of the detailed baseline studies and environmental impact assessment. While scoping does not circumvent the need for a detailed environmental impact analysis for a project, it is often used to determine the level of significance that the project impacts may have on the receiving media in the early planning stages. It can also be an important tool in determining whether any environmental, permitting, or regulatory fatal flaws exist before significant project expenditures are made. Scoping typically consists of a site reconnaissance by a qualified environmental professional and a preliminary assessment of the level of significance that project development will have on the various environmental receiving media. This preliminary assessment of impacts and their significance can then be factored into the planning and design process to limit the number of surprises that may be identified in the detailed environmental assessment process. Elements of the scoping process are typically evaluated by resource much as they are for the impact assessment. In performing the environmental reconnaissance of a project area, the environmental professional must ascertain relevant site information about each resource, the regulatory purview, and the project plans, for they will all fit together. The following presents ideas that may be significant to the environmental scoping for siting project facilities. It is not intended to be a comprehensive listing but to give the reader a sense of the types of information that should be drawn out in the scoping process in order to meaningfdly participate in the facility siting and design effort.

1903

Climate/Meteorology Significant climatic events such as hurricanes or extreme cold can affect the types of facilities that can be safely operated at a particular site. These events should be considered not only in terms of how they will affect the engineering aspects of a project but also in terms of how access could be restricted, thereby increasing the probability of risk situations during periods of inclement weather. Major projects may disturb significant areas of land and/or bum significant volumes of carbon fuels. For these projects, a more global perspective to climate impacts may need to be considered in light of greenhouse gas effects. Air Quality The presence of other industrial emission sources proximate to the project area can affect the airborne pollutant load. Some regulatory jurisdictions, such as the United States, allow air pollutant loads up to a particular level, after which no additional pollutant sources are allowed. Understanding whether the project is located in a designated or specially regulated air quality attainment area is important to determining the volume of pollutant that will be allowed in any point source discharge. Air quality control system performance is generally measured at the property boundary or at the closest point of public access. Wind patterns and dispersion characteristics may influence a project’s ability to meet air quality expectations. Siting facilities with due consideration of the compliance points relative to wind patterns and dispersion characteristics can save on control equipment and compliance issues. Landform Landforms that constitute a mining project area may include topographic features such as mountains, valleys, lakes, creeks, rivers, and wetlands. The nature of the landform can affect where facilities can safety be sited during mine operations and closure. For example, an undulating landform can be used to advantageously shield facilities from major view sheds by locating them in lower areas. In locating facilities in such depressions, however, it is important to consider the potential for temperature inversions that may entrap air pollutants. Geology and Seismology Geology and seismology can play significant roles in facility siting. Unstable geologic formations may lead to the shifting, cracking, or collapse of structures. For example, karst terrain may lead to increased sinkhole formation that in turn may lead to structural failures. If possible, facilities should not be located in areas of unstable geologic features and seismic activity. For facilities that must be located in these areas, design and construction practices that are more stringent must be taken. The geochemical characteristics of the ore, waste rock, and mine exposures (for example, pit walls and underground workings) also can impact facility siting. The basic question to be addressed concerning geochemistry is whether the geochemical properties of a project-generated material will potentially result in acid generation and leaching of metals. If the material is potentially acid generating, emphasis should be placed on isolating the facility to the extent possible to avoid or reduce the potential for environmental impacts. This is generally done by situating the facility away from surface water drainages and by best practices that may include, for example, coveringkapping the facility. Soils The availability of soils needed for operations and closure is often limited. As such, the conservation and efficient use of site soils is required. The siting of facilities should consider the need for and the availability of soils to serve as plant growth media, low-permeability materials, and other soil materials that may be needed throughout the life of the project. A soil material balance will provide a plan for meeting the soil requirements for the project.

1904

Locating facilities in strategic areas can help optimize the project soil material balance. For example, an area that includes non-beneficial soils may offer a preferred location for a facility. As another example, it may be feasible in some instances to backfill mine wastes into pits or underground workings, thereby reducing surface and soil disturbance as well as the amount of soil needed for reclamation covers. Certain soil profiles offer increased protection of groundwater resources and should be considered in siting mine waste facilities. Low-permeability soils, soils with a high attenuating capacity for metals, and an increased depth to the water table all are advantageous to the protection of groundwater resources. Hydrology The potential disturbance of surface water and groundwater resources is a significant environmental issue associated with mining projects. The site hydrologic characterization serves as a tool to locate facilities in areas that would avoid or minimize potential environmental impacts associated with water resources. The primary means for protecting a water resource is to prevent the resource from coming into contact with potential contaminants. Only after performing a comprehensive evaluation of the potential options for isolating the water resource should treatment options be evaluated as they tend to be more costly than prevention. Logically, facilities should be located in areas that are not susceptible to flooding. The site hydrologic characterization should include locating the extent of the design flood event as it may apply to the site. In addition, mine waste facilities should not be located in areas where surface water drainages exist in order to limit the potential for geochemical reactions. In cases where development of a project in mountainous topography may require the placement of waste materials in valleys containing intermittent or perennial streams, run-on diversions and underdrains should be considered to prevent water from contacting the facilities. Depth to groundwater is an important environmental consideration for locating waste facilities when a variation in the depth to the water table in the project area exists. Locating waste facilities in an area where the water table is deeper is preferable to an area where the water table is shallower in that the overlying soil column may provide for the attenuation of potentially adverse contaminants, thereby reducing the environmental risk of the project. An adequate source of water for the various aspects of the mine project may be developed from surface water and groundwater resources. The location of facilities that require substantial water inputs - for example, a concentrating facility - should consider the location of water supplies. Ecosystems Facilities should be located to minimize their impacts on sensitive ecosystems as identified during the site reconnaissance. For example, mine access roads may be located to minimize the impact associated with wildlife migration routes. Where impacts to ecosystems are unavoidable, the creation of conservation areas may provide a viable solution to mitigating degradation of habitat and the potential loss of biodiversity.

SOCIAL FACTORS Public acceptance of a mining project is fundamental to its success. A project that is technically sound may fail without public acceptance. The best engineering and environmental evaluations cannot overpower public opinion, which is usually driven by public concerns and how they are factored into project siting and design decisions. Social factors affecting project component siting decisions can take many forms. Social impact assessments have long been included as part of the environmental impact evaluation process; however, the concept of sustainable development has only in recent years become a major focal point of many mining initiatives. Where development banks like the World Bank are involved, social issues and sustainable development are crucial as they represent the institution’s fundamental reason for project investment.

1905

The social impact assessment has evolved from being a component of the environmental assessment into a stand-alone evaluation. Baselines typically consist of profiling the local population and demographics, social and political structures, land use and natural resources management, livelihood systems and employment opportunities, social services and infrastructure, public health care, vulnerable groups and indigenous peoples, and cultural and historic resources. Like the environmental baseline for the project, the social baseline establishes the background against which project development plans and designs are evaluated. The social impact assessment considers the impacts that project development will have on individuals, communities, customs, and historical features of the area. For any identified significant social impact, mitigation measures are considered to reduce the level of impact to acceptable levels. Public consultation and disclosure is a fundamental part of effective project planning. Informing the local community and interested stakeholders of the tentative project plans early and often in the planning and design process is an important step in establishing effective communications between the project proponent and stakeholders. The initial project scoping will typically identify key social issues that might affect project component siting. Preliminary siting and design concepts for major project components should be presented to the public through information meetings in the earliest phases of the project in order to solicit public input regarding performance concerns or site preferences. This public input must be factored into siting and design decisions along with the economic and environmental considerationsto assure project success. The following presents several social considerations as they relate to siting project facilities. As with the preceding discussion on environmental considerations, the list is not intended to be exhaustive, but rather should to give the reader a sense of the types of information that should be collected during the project scoping phase. Land Use The baseline evaluation should identify traditional land uses in the project area. Many new mining projects occur in third world countries in areas that may have historically been used for agricultural or pasture lands. For this reason, it is necessary to site facilities taking into consideration what may not appear to be an important agricultural area but may indeed have a significant impact on local income'sources. Resettlement Among the more important social impact issues that a mining project may need to undertake is the relocation of households or entire communities to accommodate project development plans. Resettlement issues are extremely sensitive and can cause significant project opposition if not handled fairly, expeditiously, efficiently, and transparently. While resettlement should be avoided to the extent possible, when it is necessary, its goal should be to provide each affected household with an equivalent or better living situation. In the event of resettlement, a community consultation program should be implemented in the early stages of the project, possibly during the exploration phase, so that the needs and concerns of affected populations in the project area can be properly assessed. Early input from the community often results in a better public opinion of the project and eases project acceptance along ensuing phases. In-Migration Another major social impact that may result from the mining project is in-migration. In-migration occurs for a variety of reasons, not the least of which is an increase in employment opportunities. These employment opportunities are the result of direct project employment as well as the increased need for goods and services to accommodate project employees and their families. Local communities may not have the capacity to accommodate the personnel requirements of a new project. In this case, there may be an influx of job seekers as the project gets underway. The social assessment must include an evaluation of the ability of local communities to provide the required work force and to accommodate newcomers. Communities that are targeted to receive an influx of newcomers must be identified and a set of actions developed to ensure that the

1906

newcomers are preferentially directed to locations where they can be adequately accommodated. This is typically accomplished by a combination of “push” and “pull” factors. “Push” factors will discourage settlement in certain areas and may include security provisions. “Pull” factors will encourage job seekers to settle in targeted locations by providing attractive features such as lowcost housing opportunities or improved goods and services. Public Health and Safety Mining projects often supply infrastructure in the form of clinics and hospitals that are made accessible to the general population. Many mining projects will provide these services if they are unavailable in near proximity to the project site. Providing this infrastructure in locations that provide easy public access can result in favorable public opinion of the project. Sustainability Often mining projects are viewed as a short-term use of non-renewable natural resources. This has resulted in a poor public opinion of mining in areas that have historically been non-industrial and whose inhabitants are often in lower income categories. Promoting a long-term, responsible use of natural resources will help ensure that there are resources available for sustained industrial growth far into the future. Mining projects can provide opportunities for long-term sustainable development in areas where there traditionally has been no infrastructure or employment related to industry. Training programs included in the community development program as well as on-thejob employee training provide skills in areas where training might otherwise not be available. An important aspect of the initial project scoping process is to assess the needs of the community and the sustainable development programs that would be most beneficial. For instance, offering agricultural classes that instruct communities on modem agricultural practices as part of the community development program can lead to higher regional productivity long after the mining operation has ceased. Cultural and Historic Resources Prior to preliminary siting and design concepts for major project components, a baseline evaluation of the cultural and historic resources in the project area must be performed. Identifying cultural and historic resources as part of the baseline evaluation will reduce the possibility of accidentally disturbing these resources during project activities, which could result in adverse public reaction. As many archeological or cultural resources cannot be quantitatively evaluated for significance due to either lack of supporting evidence or abundance of sites, alternative methods of protection of these resources are becoming an important mitigation method. If the baseline evaluation results in cultural or historical discoveries, in situ protection is generally preferred; however, data recovery or advancement of cultural interest programs can often serve as adequate mitigation.

CONCLUSIONS The siting of facilities with due consideration for the environmental and social factors can greatly enhance the overall environmental and social performance of the project. While facility siting offers an opportunity to mitigate environmental and social impacts, it is recognized that best practices are integral to the overall mitigation of environmental and social impacts. Some examples of best practices include mitigating surface water impacts through the strategic placement of run-on diversion structures, mitigating air impacts by watering haul roads, and reducing surface disturbance by salvaging soils where facilities will be located. A properly designed geographic information system (GIs) offers an excellent tool for visually integrating important environmental and social factors into the siting of project facilities. Ultimately, an evaluation of the environmental and social risks associated with each facility siting alternative should be performed to optimize the siting of mining and processing facilities.

1907

There is rarely a downside to factoring the environmental and social issues into facility siting decisions. Environmental aqd social consideration will usually support site selection in close proximity to the mine. This is because the more spread out the project components are, the greater the potential for environmental impacts and risk. By integrating project economics, environmental considerations, and social issues in a balanced siting decision, the layout options that emerge will likely prove to be the most economic in the long term.

1908

Selection of Metallic Materials for the MininghMetallurgicalIndustry Gary Coated

ABSTRACT The paper is broad-based and describes properties, applications and limitations of various metals (irons and steels, alloy steels, stainless steels, nickel, aluminum, and copper alloys, etc.) used in the mining and metallurgical industries. Mechanical properties, corrosion resistance, wear resistance, and high temperature properties will be discussed. Areas of applications will be discussed. Some guidelines for materials selection and application, and resources are included. INTRODUCTION Metallic materials are used in the mining/metallurgical industry often with very little thought. We are generally familiar with them, knowing their advantages and disadvantages. We may use them for their strength, corrosion or wear resistance, for their ability to withstand high temperatures, as a barrier for solids, liquids or gases, etc. What is normally used are alloys of metals. The use of high purity iron, nickel, aluminum etc. is very limited. Alloys, mixtures of at least one metal and one other element, are the commonly available metals. Steels and irons are iron-based, but usually with intentional additions of carbon, manganese, silicon, etc. “Stainless” steels contain a minimum of about 11% chromium, usually with other elements such as nickel, molybdenum, etc. to give them special properties. Some guidelines will be given to aid in materials selection. Selecting the right material also involves issues such as price and availability considerations. The higher alloyed materials generally require more sophisticated fabrication and especially welding techniques. This means that some attention should be paid to the selection of the fabricator, the inspection of the equipment and the supporting documentation. Some examples will be covered. In this paper, most of the alloys will be classified by their UNS (Unified Numbering System) number, an attempt to classify the chemical composition of all metals and alloys by a unique identifier composed one letter and 5 numbers. These numbers are used today by ASTM (American Society of Testing and Materials) for identifying alloys in their specifications. However in many cases the older designations are better known, so that for example, AISI (American Iron and Steel Institute) numbers will be also used. It is not possible in a paper such as this to cover all metals and types of applications. More exotic metals such as zirconium, tantalum, platinum, etc. which do find occasional use are not covered. Coatings (metallic, plastics, paint, rubber) will not be covered in any great detail each is a major topic in itself. As metallurgical processes often involve aggressive and perhaps toxic or explosive chemicals, materials selection needs to concern itself also with health, safety and environmental issues. A pipe leaking toxic chemicals can have severe legal consequences on the operation of a plant, and for the management of the plant too. The paper is divided into 3 parts. The first section called Alloys deals with a broad description of the classes of alloys with general properties and applications. The second part gives a few key principles that are important when selecting and working with alloys. In the final section, the use of available resources to aid in selecting materials is discussed.

Nickel Development Institute, Toronto, Canada

1911

METALS AND ALLOYS STEELS AND IRONS A number of the common names given to steels and irons are more traditional than accurately reflecting metallurgical classifications. In the UNS classification, carbon and alloys steels begin with either G, H or K, carbon or alloy steel castings with J, and cast irons begin with F. Steels are available in a wide range of strength levels, both inherent and from heat treatment. In many applications, this group of materials are the first that should be considered, because of their high availability, low price, ease of fabrication, familiarity, etc. Despite their comparatively low corrosion resistance, it is sometimes possible to build a piece of equipment using thick material to compensate for a high corrosion rate. However, other factors have to be taken into account, especially when safety and environmental issues are at stake. Difficulty in inspecting a critical component to know when it is nearing its end of life is a significant factor that can lead to a decision to use a more corrosion resistant but more expensive material. The corrosion resistance of these materials can be improved through various methods. Various coatings can be applied, both galvanic types such as zinc, or barrier types such as paint, plastic, rubbers, etc. (In many cases, it is more appropriate to view the coatings, especially rubber and plastics, as the prime material supported by a substrate of steel.) Steels underground or in water can be cathodically protected by application of a electrical current either directly from a power source or via special anodes of magnesium, aluminum, etc. These subjects are all outside the scope of this paper. Carbon steels. These are the workhorse alloys of constructional materials, piping, plate and sheet. Carbon steels are iron with small amounts of carbon, manganese, and a few other steelmaking elements such as silicon. A high carbon steel such as AISI 1045 (UNS G10450) may be easily hardenable with heat treatment. If the carbon content is relatively low, the steel will not easily harden, so for example, welding can be done without any special precautions. These steels do not become brittle at cold temperatures, so that a killed steel might be suitable down to -30°C (-20°F). Carbon steels do offer a certain degree of high temperature and fire resistance. However the weathering steels (below) are more commonly used at temperatures above about 300°C (570°F) because of their improved oxidation resistance and higher strength. The main alloying elements of a few common carbon steels are included in Table 1. With iron-based alloys, the balance of the composition not stated is iron. Table 1 Nominal composition (weight %) of some common carbon and alloy steels C Mn Cr Ni NAME Type UNS number AISI 1010 Low carbon steel GlOlOO 0.10 0.45 AISI 1045 High carbon steel G10450 0.45 0.75 ASTM A 36 Carbon structural steel KO2598 0.20 1.O 2% Cr I %Mo Alloy steel for high T K21930 0.10 0.45 2.2 3%% nickel (LF3) Alloy steel for low T K32025 0.15 0.7 3.5 AISI 4340 Alloy steel hardenable G43400 0.40 0.7 0.8 1.8 B-l cast Austenitic Manganese J91119 1.O 13

Mo

1.o 0.2

Alloy steels. This class of steel falls into different categories. There is alloying to improve the strength of the material, both at ambient temperature and at high temperature, and especially the heat treated strength and hardness. The chromium-molybdenum grades, e.g. 2% Cr I Y M o (UNS K21930) have greatly improved high temperature strength. The addition of nickel to a steel improves low temperature ductility, so that an alloy such as K32025 with 3%% nickel is good for many cryogenic services. Abrasion resistant steels, many of which are based on the AISI 4340 (UNS G43400) composition, are quenched and tempered to give a very high hardness. They tend to also be quite brittle, and can fracture under impact loading. Such steels are often used as abrasion resistant liner plates. Where water or other corrosives are involved, it should be

1912

remembered that these materials are not particularly corrosion resistant, with the net result that high wear rates can occur. In water, the steel forms a thin oxide layer, which is easily removed by abrasive particles, and the process is repeated again and again resulting in high rates of metal loss. In that sense, corrosion and abrasive wear are synergistic. An improvement in the corrosion resistance of the metal may improve the total wear resistance considerably. Weathering steels. These are low-alloy high strength structural steels that are often separated out from the alloy steel category because they have markedly increased corrosion resistance. They are alloyed with copper and usually a few other elements, primarily chromium, nickel, and vanadium, usually less than 1.5% total, and have good corrosion resistance in normal atmospheric exposure conditions. They are not hardenable by heat treatment. They form a semiprotective layer of oxide or “rust”. This layer is not protective under abrasive conditions, nor totally protective in marine or heavily polluted industrial atmospheres. A typical weathering steel (4 grades are covered in A 588) will have a 345 MPa (50 ksi) minimum yield strength and a 485 MPa (70 ksi) minimum tensile strength. They have been used for building exteriors, structural elements on bridges, and other structural applications where a rusty appearance is tolerated. Mining/ metallurgical applications include structural applications, and also certain equipment like fans, blowers, and ductwork, especially where sulfurous gases are present. Weathering steels are suitable for hot gas ducting with appropriate temperature limitations and other precautions (Andrew 1997). Austenitic manganese steels. These are a series of cast or wrought steels with 10% or more manganese. They are soft, unlike most abrasion-resistant alloy steels, but cold work rapidly to resist abrasion. Thereby the surface is wear-resistant, while the interior remains ductile. Cast irons. This is a family of materials broken out into unalloyed cast irons and alloyed cast irons. The carbon content varies from as low as 0.7% to as high as 4.5%. They are mostly used as valves, pumps, pipes, fittings, and occasionally as bar. Table 2 lists some representative cast irons.

Table 2 Nominal composition (weight %) of some representative cast irons Type UNS C Mn Si Cr Ni Mo number Gray cast iron F10004 3.5 0.7 2.5 White cast iron F45001 1.0 0.6 2.5 4.0 0.5 2.7 Malleable cast iron F20000 2.5 0.8 1.5 Silicon cast iron F47003 0.9 1.2 14.5 0.3 0.3 Nickel Cast Iron D2B F43001 1.0 2.3 3.5 20 2.6

Cu

0.3

Unalloyed cast irons include gray cast iron, ductile cast iron, malleable iron, and white cast iron. Gray cast iron is inexpensive, easily cast, and machinable. Ductile cast iron, made with a small nickel-magnesium addition, is often preferred for improved resistance to shock. They can also be modified to have excellent low temperature ductility. White cast iron is a controlled chemical composition gray iron that is rapidly cooled to make it very hard and very brittle. It is virtually unmachinable, and extremely difficult to weld. It has excellent abrasion resistance. Chilled iron is so produced to have hard white cast iron on the surface, and a more ductile gray cast iron interior. An extended heat treatment of white cast iron results in a type called malleable iron. Malleable iron has improved ductility and strength to gray cast iron, and is commonly used in tools, machinery, automotive parts, etc. There are also a number of proprietary chemical addition treatments that are made to improve the properties of the unalloyed cast irons. Alloyed cast irons have significant alloying additions of silicon, nickel, molybdenum, chromium or copper. A silicon cast iron which contains about 14.5% silicon performs well in hot concentrated sulfuric acid. Variations of this alloy with chromium and molybdenum give high corrosion resistance in chloride-containing media. A number of proprietary versions exist. Nickel can also be added in amounts of about 0.5% to 6% to improve common versions of gray cast iron.

1913

At about 4.5%, a martensitic alloy is produced with outstanding resistance to wear. Austenitic gray cast irons contain 14% to 38% nickel. Some of these alloys, such as F43001 have very good corrosion resistance and are suitable for moderately high temperatures.

STAINLESS STEELS Stainless steels are a family of steels which all show passive behavior, resulting from a minimum chromium content of about 11%. Other elements such as molybdenum, nickel, copper and nitrogen are added to increase corrosion resistance to different environments. The passive oxide layer on the exposed surface is chromium rich, but also contains the other alloying elements. The stainless steel family is composed of several different sub-families based on the crystal structure of the grains. Each of the families is discussed separately - austenitic, duplex, , ferritic, martensitic and precipitation hardenable. Stainless alloys span the range of corrosion resistance from low to very high, and also strength levels from relatively low to very high. Austenitic stainless steel. The majority of the worldwide production of stainless steels is of the austenitic type, also known as the “300 series” stainless steels. These grades contain nickel and other alloying elements that promote an austenitic structure. These alloys offer many different advantages, such as ease of fabrication and weldability, excellent high temperature strength, and excellent low temperature ductility. Those alloys with a fully austenitic structure are essentially non-magnetic, an important property in some applications. This family of alloys offers the widest range of corrosion resistance, from the base grade 304 up to high performance alloys such as S32654 and N08031, which are bordering on the nickel-base family. Table 3 lists a number of more common austenitic stainless steels, but there are numerous more alloys produced, some for very restricted applications. Table 3 Nominal composition (weight %) of some representative austenitic stainless steels Family / Alloy C, H, or UNS C,, Cr Ni Mo N Other both* number AUSTENITIC 304 304L 304H 316L 317L 3 17LMN 321 347 309s 310s Alloy 253 330

C&H C H

C C C H H H H H H

19 19 19

9 9.5 9

-

0.06 0.06 0.05

S31603 S3 1703 S31726 S32100 s34700 S30908 S3 1008 S30815

0.08 0.030 0.04 0.10 0.030 0.030 0.030 0.08 0.08 0.08 0.08 0.05 -

17 19 18.5 18 18 23 25 21

12 13 15 10 10.5 13 20 11

2.2 3.3 4.3

-

0.06 0.06 0.15 0.04 0.04 0.04 0.03 0.17

NO8330

0.10 0.08

18.5

36

-

0.03

S30400 S30403 S30409

-

Ti Nb

Ce, Si Si

High Performance

Austenitic Alloy 20 Alloy 904L

C C C

NO8020 0.07 20 35 2.5 0.04 NO8904 0.02 20 25 4.3 0.04 Alloy 825 NO8825 21 40 0.05 3.0 0.04 6% Mo (18%Ni) C S31254 0.020 20 18 6.2 0.20 6% Mo (25% Ni) C NO8367 24 6.3 0.030 21 0.22 Alloy 654 C S32654 0.020 25 22 7.5 0.5 ,410; 31 C NO8031 0.015 27 31 6.5 0.2 * Primary end use: C = corrosion resistant, H = high temperature resistant, C&H = both

1914

Nb+Ta,Cu

cu

Cu, Ti, Al

cu cu

Cu, Mn

Cu

Nomenclature. The UNS system uses S for stainless steels and N for nickel alloys. That is based on a definition that if an alloy has 50% or more iron, and meets the minimum chromium content requirement, then it is a stainless alloy, and is given an S number. If the alloy has less than 50% iron, and there is more nickel than any other element other than iron, it will then be given an N number. If less than 50% iron and more cobalt than any other element, it would be classified with the cobalt alloys in the R series. Another definition, used today by ASTM and in this paper, is based on the highest element in the alloy. If the minimum chromium content restriction is met, and there is more iron that any other element, then it is classified as a stainless steel. If there is more nickel than iron, than it is a nickel alloy. In the old AISI classification system, the suffix “L” (as in 304L) stands for low carbon, most often 0.030% carbon maximum. Low carbon alloys should always be used whenever welding is involved for corrosive service. The UNS number for these old standard grades should end in “03”, indicating the maximum carbon level. This system is not followed for the newer or proprietary alloys, for which almost all exist only with low carbon. An alternative to using low carbon is to stabilize with elements such as titanium, niobium, or tantalum. AISI 321 (S32100) or AISI 321 is essentially 304 with a titanium addition to compensate for the higher carbon content to achieve proper corrosion resistance after welding. These stabilizing elements have other drawbacks both in steelmaking and in welding, so the low carbon stainless steels have predominated for many years, at least in North America. Not so long ago, the stabilized versions predominated in many countries of Europe, but more recently they have also adopted the low carbon stainless steels for corrosive services. However, when on an old drawing a stabilized stainless grade is being specified for a corrosive service, rather than replace it with the same, it is best to ask whether a low carbon stainless could be used. Alloy 20(N08020) is an example though of an older corrosion-resisting alloy produced only in a stabilized version. It is very important to use the low carbon or stabilized stainless steels in corrosive oxidizing conditions, such as weak sulfuric acid solutions with oxidizing ions such as cupric or ferric. The “H’ (as in 304H) specifies both a minimum and maximum carbon content, normally 0.04 to 0.10%. The last 2 digits of the UNS number are normally “09”. H grades are meant to be used at elevated temperatures, where carbon gives the alloy increased high temperature strength. The ASME Boiler and Pressure Vessel Code gives stainless steels with 0.04% minimum carbon significantly higher allowable stresses at virtually all elevated temperatures. At one time, it was very expensive to lower the carbon content in stainless steels to the low carbon levels, and consequently, if one did not specify a low carbon stainless steel, one was virtually assured of getting one with a higher carbon level. Today it is easily and inexpensively achieved and even advantageous for most modern steelmakers, so one is most likely to receive a low carbon stainless if one does not otherwise specify. The prudent way however is to always specify the exact grade that one needs, depending on the application. “Dual-grade” stainless steels are ones that meet the specifications of two alloys. It is possible to produce dual grade 304/304L (UNS S30400/S30403). It is certifiable to both grades, and acceptable to the pressure vessel authorities. That allows one to design a pressure vessel using the allowable stresses of 304, but ensuring low carbon for corrosion resistance after welding. In the older AISI designations, the suffix “S” was a way of designating a low carbon version of some of the high carbon heat resistant alloys. For example, 310s (UNS S31008) contains 0.08% maximum carbon, a limit set to have improved weldability. Since these alloys are only intended for high temperature usage, steelmakers do not usually produce them with low carbon. However, it is possible to specify 310H (S31010) with 0.04% to 0.10% carbon. Corrosion resistance. Unlike many other materials, it is not common for stainless steels to fail by general or uniform corrosion. It is more likely that they fail by localized corrosion, e.g. pitting, crevice corrosion, stress corrosion cracking, or intergranular corrosion. As mentioned earlier, the corrosion resistance of austenitic stainless steels spans a wide range. This includes strong acids to strong bases and most chemicals in-between. It should not be assumed that the higher the alloying content, the more corrosion resistant it is. In nitric acid and in 98% sulfuric, 304L is superior to 316L and most high molybdenum stainless steels. That is

1915

because the chromium content is the most important factor in these highly oxidizing conditions and molybdenum can be detrimental. It is important to check a particular alloy’s corrosion resistance to the specific environment. In sulfuric acid at lower concentrations which tends to be reducing in nature, molybdenum alloys are preferred. 3 16L is normally considered the base grade, i.e. one would never use 304L in dilute sulfuric acid. Even higher alloyed materials are often required. However, there is no guarantee that a 6% molybdenum stainless is better than a 4% molybdenum stainless steel. Alloys like 904L and Alloy 20 are often used, and in high chloride containing acid solutions, the 6% Mo grades give good performance. Much good corrosion data exists for stainless steels; references will be discussed in a later section. A small copper alloying addition has been found especially useful to improve corrosion resistance in intermediate concentration ranges of sulfuric acid. Austenitic stainless steels are also useful in chloride media, with molybdenum and nitrogen being two powerful alloying elements. Many users are aware that austenitic stainless steels are susceptible to chloride stress corrosion cracking (SCC). This is especially true of the 304L and 316L, whereas the high performance grades with significant nickel and molybdenum additions are highly resistant, although not immune (Amvig et al 1998). Immunity normally is assured only when the nickel content exceeds 40%. The necessary preconditions for SCC are chlorides, tensile stresses, and a certain minimum threshold temperature. In most cases the minimum temperature is around 60°C ( 140°F), although under special conditions, SCC is possible even at room temperature. It is fairly easy to roughly compare the pitting resistance of various austenitic stainless steels. The formula is based on the effectiveness of the important alloying elements on resisting the initiation to pitting, and gives a number called the pitting resistance equivalent, or PRE.

PRE = %Cr + 3.3x%Mo + 16x%N There are a couple of variations in this formula, so it is important when comparing numbers to know which formula was used. The formula doesn’t take into account many factors such as surface finish, impurities in the steel, welds, and other critical factors, and it does not allow one to determine whether pitting will or will not occur in a particular alloy. Nonetheless it is useful for comparing materials. It is clear from the formula that the high performance austenitic stainless steels with high molybdenum and nitrogen contents have far greater pitting resistance. For example, the 6% Mo grades are far better than Alloy 904L, which in turn is better than Alloy 20. This ranking holds up in real industrial environments where pitting is a factor. Do not try to use it for other environments. There are some environments where one generally avoids using stainless steels. Chemicals such as wet chlorine, hydrochloric acid, sodium hypochlorite, and hydrofluoric acid are ones where stainless steels have only a very limited usability. The thin passive oxide layer on stainless steels has an amazing tolerance to velocity. In waters and other less corrosive chemicals, velocities up to 30 metres per second (100 feet per minute) present no problem. In chloride-containing waters, stainless steels are most likely to be attacked in stagnant conditions. A high flow velocity prevents pitting from occurring, therefore a typical minimum velocity is specified in heat exchanger tubes using brackish or sea water of 1.5 metres per second (5 feet per second). One problem often encountered is a reduced corrosion resistance of the machinable grades of stainless steel. For example, 303 (S30300) is a version of 304 with high sulfur (0.15% minimum), and machines very easily. It may not perform as well as 304 in some environments. Mechanical properties. Austenitic stainless steels are for most applications supplied in the annealed condition, and as such have reasonable ambient temperature strength levels and excellent ductility, as shown in Table 4.

1916

Table 4 Minimum mechanical properties for some austenitic stainless steels according to ASTM A 240 Alloy UNS Yield Strength Tensile Strength Min. Elongation % Number MPa (hi) MPa (hi) 304L S30403 170 (25) 485 (70) 40 205 (30) 304 40 S30400 515 (75) 205 (30) 304H S30409 515 (75) 40 316L S31603 485 (70) 170 (25) 40 205 (30) 3 17L 40 S31703 515 (75) 215 (31) NO8904 Alloy 904L 490 (7 1) 35 241 (35) NO8020 30 Alloy 20 551 (80) 6% Mo (18% Ni) S31254 310 (45) 35 655 (95) S32654 40 Allov 654 430 (62) 750 (109) Low carbon stainless steels have slightly lower strengths than their higher carbon counterparts, although today this can be improved by a intentional, small (less than 0.10%) nitrogen addition. That is what in effect has happen to obtain dual grade 3041304L. Higher nitrogen levels in some of the high performance alloys improves their strength levels, although the nitrogen is primarily present for other reasons, primarily corrosion resistance and structural stability. Although austenitic stainless steels cannot be heat treated to improve their strength, they can be easily cold worked to a very high strength and hardness. For example, 301 stainless in fully hard condition has a minimum yield strength of 965 MPa (140 ksi or 140,000 psi) and tensile strength of 1270 MPa (185 ksi). In such a condition, the material is a powerful spring. There are some special applications for strip steel in this condition, for example a thin sole plate in safety shoes. Austenitic stainless steels maintain their good ductility down to very low temperatures. The ASME code permits the use of some grades down to as low as -253°C (-423°F) without even requiring impact tests to check their ductility, others to only -196°C (-320°F). All these grades also get slightly stronger at cryogenic temperatures. Welds and cast material may not have quite the same ductility, and it is necessary for example to impact test welds for service below -101°C (150°F). High temperature properties. The austenitic stainless steels in general offer excellent oxidation resistance and high temperature strength. Other grades such as 330 (N08330) are excellent in carburization environments. In the standard high temperature austenitic stainless steels, 304 has useful oxidation resistance up to about 900°C (1650"F), whereas 309s and 310s are suitable up to about 1030°C (1900'F) and 1100°C (20OOOF) respectively. Special grades such as alloy 253 (S30815) are formulated to not only have excellent oxidation resistance up to 1100°C (2000"F), but combine that with improved high temperature strength. One disadvantage of the austenitic grades is a relatively high thermal expansion coefficient. In designing equipment for high temperature use, allowance must be made for movement of the material. At temperatures over about 550°C (1000"F), 304, 321, or 347 type alloys are often preferred over other ferritic materials because they are substantially stronger in creep strength. The standard yield strength measurements do not describe the slow yielding that goes on above this temperature. Fabrication. Most shops are familiar with fabricating the standard austenitic stainless steels. Welding at first is a bit more complicated than carbon steel, machining calls for different tools and feeds and speeds, and there is more springback in forming operations, but the learning curve is not very difficult. There are several good handbooks available with good information on this subject. When it comes to the high performance stainless steels, there can be significant differences related to welding and fabrication. These alloys are relatively expensive and are being used for very corrosive environments, so it is imperative that the shop be familiar with any necessary special practices. Most steel mills and some suppliers have booklets on fabricating their special alloys. It would be wise to choose a fabricator that has had experience with that particular

1917

or a similar alloy. If not, then it would be prudent to make sure the fabricator has the information and that the information has been disseminated to the shop floor. It is also important that during installation of the equipment, the installers should be aware of what they can and cannot do. And of course, a mill’s own maintenance department should be aware of how to properly weld and repair such alloys. Each alloy basically has its own special filler metal, although in some cases, a more universal filler metal may be possible. When a piece of equipment in austenitic stainless steel is delivered to site, or sits out in the open for awhile, there often appears rust marks on the surface. These marks were usually not visible when it left the fabricators shop. The cause is almost always surface contamination with iron, from tools, lifting chains, clamps, etc., sometimes during fabrication, sometimes during transportation and off-loading. In the cases where the free iron on the surface has originated in fabrication, it does not become visible until it comes in contact with water. That can be rain, night time condensation, etc. Although often only a cosmetic issue, in certain corrosion applications, this rusting can initiate other types of corrosion in service, especially pitting. It is always best to try to remove the contamination. Brushing and grinding are not particularly effective ways, as the free iron tends to smear on the stainless surface. It is better to chemically remove the contamination, using either phosphoric acid, or a special pickling mixture of nitrichydrofluoric acid. The latter is available as a paste, and while quite nasty, is very effective. Heat tint from welding is also best removed this way.

Duplex stainless steels. A duplex stainless steel is one that has a structure made up of two different phases, in this case austenite and ferrite. This is usually accomplished by reducing the content of nickel, a strong austenite former, to intermediate levels. Although the first duplex stainless steel, AISI 329, was developed in the 1930’s, duplex stainless steels fell out of favor in the 1950’s because of some shortcomings. These have been mostly overcome today, and there has been a resurgence in use as industry has found them to be very useful engineering materials, economically solving problems with some of the austenitic stainless steels. There are many different duplex stainless steels grades available today, all containing nitrogen which solves many of the earlier shortcomings. They range from low alloyed to high alloyed, with many of them being proprietary. This can present a problem with availability. The most common duplex stainless steel today is called 2205. It originally had a UNS number of S31803, but it has been upgraded slightly, and only the new version, UNS S32205, should be used. Table 4 gives the chemical composition of 4 representative duplex grades. Table 5 Nominal composition (weight %) of some representative duplex stainless steels Mo N Other Alloy C,H,o UNS C,, Cr Ni both* number 2304 (low alloy) C S32304 0.030 23 4 0.3 0.12 0.17 2205 (intermediate) C S32205 0.030 22.5 5.5 3.2 C S32550 0.040 25 5.5 3.3 0.18 Cu Alloy 255 (superduplex) S32760 0.030 25 7 3.5 0.25 Cu, W Alloy ZlOO (superduplex) C * Primary end use: C = corrosion resistant, H = high temperature resistant, C&H = both All of these grades are meant for corrosion-resisting applications. In fact, welded constructions in duplex stainless steels should only be used up to about 240°C (460°F) and unwelded up to 270°C (520°F). Above those respective temperatures, welds and parent material can suffer from impaired ductility with time. The guidelines given about fabrication of high performance austenitic stainless steels are equally if not more relevant with duplex stainless steels. They are generally cost effective materials, but their corrosion resistance and ductility can be easily impaired. Usually it is only after being in-service that these “mistakes” are discovered. ASTM A 923 is a specification for ensuring wrought mill products in a couple of the duplex grades are delivered with the high level of quality that is expected. It has been also applied

1918

to cast products and fabrications, although since the standard wasn't written with those products in mind, care must be taken to properly interpret the tests and results. Duplex stainless steels have a lower thermal expansion coefficient than the austenitic grades, closer to carbon steel. This property can be used, for example, if replacing carbon heat exchanger tubes. Duplex tubes have been often used this way, welding them to the carbon steel tubesheet. Corrosion resistance. One of the major uses of duplex stainless steels has been to replace 304L, 316L, or 317L that has suffered chloride stress corrosion cracking (SCC). All of the duplex alloys have better chloride SCC resistance than the above three 300 series alloys. Many of the high performance austenitic stainless steels have similar stress corrosion cracking resistance to the duplex - the differences are very complicated. In terms of pitting and crevice corrosion resistance, the 2304 grade is often compared to 316L, the 2205 is close to 904L, and the superduplex stainless steels e.g. 255, ZlOO and others, compare to the 6% Mo stainless steels. The PRE formula discussed for austenitic stainless steels is also valid for duplex stainless steels. There are certain acids where duplex stainless steels outperform their austenitic counterparts, e.g. organic acids such as acetic and formic. In dilute sulphuric acid solutions mixed with oxidizing metal ions such as cupric and ferric, there have been some mixed reports. In some cases the duplex alloys perform very well, but in other cases, the ferrite phase is selectively attacked. It is strongly advised that in-plant corrosion testing be performed prior to any large scale installation. The superduplex alloys do seem to work reasonably well in certain autoclave applications, although the actual corrosion rates vary from very low to very high. Again, care is advised and in-plant testing is virtually mandatory. It has been said that when the duplex alloys are good, they are very good, and when they are bad, they are very bad. Mechanical properties. The strength levels at ambient temperature of the duplex grades are generally higher than the austenitic stainless steels, but the ductility suffers slightly as shown in Table 6. These high strengths are quite important when designing pressure vessels, pressure piping systems, etc. as much of the wall thickness is only there for pressure-retaining purposes, and not for the corrosion resistance.

Table 6 Minimum mechanical properties for some duplex stainless steels according to ASTM A 240 Alloy UNS Yield Strength Tensile Strength Min. Elongation MPa ( h i ) % Number MPa (hi) 2304 S32304 400 (58) 600 (87) 25 2205 25 S32205 450 (65) 620 (90) Alloy 255 S32550 550 (80) 760 (110) 15 Alloy ZlOO S32760 550 (80) 750 (108) 25 Duplex stainless steels will become brittle at low temperature. With proper precautions, duplex stainless steels can be used down to -40°C (-40'F) or slightly lower, however only with impact testing. Fabrication. Welding has often been the critical issue, however special filler metals combined with proper quality wrought products have solved most of the potential problems. It is still possible through ignorance for a fabricator to cause serious impairment to duplex stainless steel.

Ferritic stainless steels. These alloys form part of the AISI 400 series stainless steels, to which martensitic stainless steels also belong. The ferrite phase is achieved to removing all or most of the nickel content. In theory, ferritic stainless steels are not hardenable by heat treatment, or at least not intended to be hardened. Table 7 lists some of the ferritic stainless steels. Like the other families, there is a wide range of ferritic stainless steels with corrosion resistance varying from

1919

very low to very high. The 409 grade, used in very large quantities in automobile exhaust systems, is barely a stainless steel with only 11.2% chromium. The 3CR12 alloy, originally developed in South Africa, has just slightly higher corrosion resistance. There exist different variations of this grade from different producers. 430 is just slightly less corrosion resistant than 304, and is commonly used in less expensive housewares. The 444 alloy compares roughly corrosion-wise with 316L. The 29-4-2 alloy compares roughly with the 6% Mo grades. The ferritic stainless steels all are virtually immune to chloride stress corrosion cracking.

Table 7 Nominal composition (weight 9%) of some representative ferritic stainless steels Alloy C, H, or UNS C,, Cr Ni Mo N Other both* number 409 C&H S40900 0.08 11.2 0.1 Ti 0.030 11.5 0.5 Ti 3CR12 C S41003 430 C S43000 0.12 17 444 C S44400 0.025 18.5 0.5 2.1 Nb+Ti Alloy 29-4-2 C S44800 0.010 29 2.2 3.8 446 H S44600 0.20 25 * Primary end use: C = corrosion resistant, H = high temperature resistant, C&H = both One of the difficulties with ferritic stainless steels relates to grain size and poor ductility. As the chromium content increases, the ductility for the same thickness decreases. In addition, significant weldability issues arise where welds can have very poor ductility in the heat affected zone. In practice the higher the alloy, the less the maximum thickness available. Some of the highest alloy ferritic alloys are available in a maximum thickness of only 0.6 mm (0.024 inch). This still makes these alloys suitable for applications such as heat exchanger tubes and sheets. Ferritic stainless steels have better thermal conductivity than austenitic stainless steels. The lower chromium ferritic stainless steels find application where an improvement is required over carbon steel, but perhaps cannot justify the cost of 304, or where 304 is not suitable, e.g. for reasons of chloride stress corrosion cracking. Ferritic stainless steels can be embrittled by hydrogen however. They have been used for structural applications as well as some wear applications, e.g. for ore chutes, buckets, rail cars, conveyor decking plates, structural applications in and around water, and in fume and dust extraction equipment. 409 is also used in hot exhaust ducts - the very low chromium content means that it does not embrittle in the 370-510°C (700900°F) range like the other ferritic stainless steels. 446 is a high chromium ferritic stainless steel sometimes used in smelters for high temperature high sulfur-containing gases, or whether there is some splashing contact with liquid copper. Not only does 446 embrittle in the 370-510°C (700-900'F) temperature range, it also embrittles due to sigma phase formation in the 540-870°C ( 1000-1600'F) temperature range, where the material can become as brittle as glass fairly quickly. The mechanical properties of the ferritic stainless steels are quite reasonable compared to regular carbon steels, although ductility is not as good as for the austenitic stainless steels, as shown in Table 8.

Table 8 Minimum mechanical -properties for some ferritic stainless steels according to ASTM A 240 or A 176 Alloy UNS Yield Strength Tensile Strength Min. Elongation Number MPa ( h i ) MPa ( h i ) % 409 S40900 205 (30) 380 (55) 22 430 S43000 205 (30) 450 (65) 22 444 S44400 275 (40) 415 (60) 20 29-4-2 S44800 415 (60) 550 (80) 20 Y

1920

Martensitic stainless steels. The main characteristic of this family of alloys is that they are hardenable by heat treatment, typically a quench and temper type. Some grades obtain a very high hardness, with 440C achieving typically 56-60 on the Rockwell C scale. Of course the ductility of the material suffers at such high hardness. Properties can be modified by a change i n tempering temperature, a compromise between hardness and ductility. The best of the martensitic stainless steels has a corrosion resistance far inferior to 304 stainless. These alloys are stocked in the annealed condition for easy machining. Not only are these alloys soft in this condition, but will “rust” with exposure to water and air. In the hardened condition, they have slightly better corrosion resistance and will not rust. This author has visited several end users who had specified a martensitic stainless steel for a particular application, but forgot to specify that they wanted it in the hardened condition! Table 9 lists two martensitic stainless steels as representing the range of alloys. There are numerous others, many of which are proprietary. Table 9 Nominal composition (weight %) of two representative martensitic stainless steels Alloy C, H, or UNS C,, Cr Ni Mo N Other both* number 410 C S41000 0.15 12.5 0.2 17 440C C S44004 1.1 * Primary end use: C = corrosion resistant, H = high temperature resistant, CbH = both These alloys are primarily used for their hardness, or especially 410S, for their strength. They offer some abrasion resistance, but like the quenched and tempered alloy steels, they are not so successful in corrosive environments. They are produced mainly as round bar for shafting, but some plate and sheet are also available. Table 10 lists some representative heat treatments and resulting properties.

Table 10 Nominal mechanical properties after hardening and tempering for two martensitic stainless steels, 25 mm (1 inch) diameter round bar Alloy UNS Heat treatment Yield Tensile Elong. HardStrength Strength % ness Number T = tempered MPa (hi) MPa (hi) Rc 410 S41000 As Hardened 43 T 200°C (400OF) 1000 (145) 1310 (190) 15 41 965 (140) 1240 (180) 15 39 T 315°C (600°F) T 650°C (1200°F) 590 (85) 760 (110) 23 B97 440C S44004 As Hardened C60 1970 (285) 1900 (275) 2 c57 T 315°C (600’F) Precipitation hardenable stainless steels. The outstanding characteristic of this group of alloys is that they have the best corrosion resistance of any hardenable stainless steel. In this group, there are actually different families (martensitic, austenitic, semi-austenitic) as well as grades. The hardening mechanism is different from the martensitic stainless steels - these alloys form precipitates. Instead of a quench and temper heat treatment, these alloys are heated to a suitable temperature for a certain period of time. The hardenability is not dependent on cooling or quenching rate, but on time at temperature. Since there is no quenching, there is little risk for distortion. Parts do change size slightly (shrink) at a calculable rate. Table 11 gives the composition of two more common grades. Alloy 17-4 is by far the most commonly available and most used within the mining/metallurgical industry. Alloy 450 is a proprietary grade. In another oddity, these grades are hard in the annealed (as-delivered) condition, and in the hardening heat treatment, they can be made either harder or softer, according to the desire. These grades can be used in the annealed condition, although they will have better ductility in the precipitation

1921

hardened condition. Mechanical properties for both grades in some heat treated conditions are included in Table 12. Some other precipitation hardenable grades are soft in the as-delivered condition. Table 11 Nominal composition (weight %) of two precipitation hardenable stainless steels Alloy C, H, or UNS Cmx Cr Ni Mo N Other both* number Alloy 17-4 C S 17400 0.07 16 4 cu, Nb Alloy 450 C S45000 0.05 15 6 Cu, Nb *Primary end use: C = corrosion resistant, H = high temperature resistant, C&H = both Table 12 Minimum mechanical properties after different heat treatments for two precipitation hardenable stainless steels, according to ASTM A 564 Alloy UNS Condition Yield Tensile Elong. Number Strength Strength % MPa ( h i ) MPa ( hi) L/T 10 17-4 S17400 H900 1170 (170) 1310 (190) H1025 1000 (145) 1070 (155) 12 930 (135) 16 H1150 720 (105) Annealed 660 (95) 900 (130) 10 450 S45000 H900 1170(170) 1240 (180) 6/10 H1150 520 (75) 860 (125) 12/18

Hardness Rc 40 38 28

39 26

Applications include shafting, bolts, pins, and other articles made from round bar. Alloy 17-4 is also available as plate and sheet, and is weldable. Although a grade like 17-4 might cost more initially than a martensitic stainless steel, the cost of heat treatment is less, and likelihood of distortion is minimal. The improved corrosion resistance over the martensitic alloys is often a major advantage. Cast Stainless Steels. The above discussion was about wrought stainless steels, but for the most part, comparable cast alloys are produced. They are given different designations, by both the American Casting Institute (ACI) and by UNS. The latter uses different numbers because the chemical composition does differ. See the section “Cast versus Wrought” under Selection of Materials - A Few Key Principles. In Table 13, the names and family groupings of some more common cast stainless steels with similar wrought alloys are given. Some cast alloys have no ACI number, but have a UNS number, and vice versa. In both the duplex stainless steels and the high temperature austenitic stainless steels, there are far more cast alloys that don’t have a wrought equivalent. It should also be noticed that only one ferritic stainless steel is represented in the table below. Due to their low ductility in heavy sections, totally ferritic castings are not common.

1922

Table 13 Identification of selected similar wrought and cast stainless steels Wrought Alloy Wrought UNS Similar Cast Cast UNS Number Alloy Number Austenitic 304 S30400 CF8 J92600 304L S30403 592500 CF3 304H S30409 592590 (CF8) 316 CF8M S3 1600 J92900 316L CF3M S31603 J92800 317L S31703 CG3M J29999 --32 1 S32100 J92630 347 S34700 CF8C 592660 309 593402 S30900 CH20 310 J94202 S31000 CK20 High Performance Austenitic Alloy 20 NO8007 NO8020 CN7M 6% Mo (18%Ni) S3 1254 593254 CK3MCuN 6% Mo (25% Ni) NO8367 J9465 1 CN3MN Duplex 2205 CD3MN S32205 J92205 Alloy ZlOO S32760 J93380 CD3MWCuN Ferritic 446 S44600 J926 15 CC50 Martensitic 410 CA15 S4 1000 J91150 420 S42000 CA40 591153 Prec. Hardenable 17-4 S 17400 CB7Cul J92180 NICKEL ALLOYS. Nickel alloys for the mining/metallurgical industry can be broken down into 4 main categories - commercially pure nickel, chromium-containingalloys (which will be broken further into 2 subgroups), nickel-copper alloys, and nickel-molybdenum alloys. The chemical compositions of some of the more significant alloys are included in Table 14. Commercially pure nickeI. Alloy 200 and its variations are used commercially used for high temperature caustic, but find little use in the mining industry. Nickel plating, whether by the electroless or electrochemical methods, is outside of the scope of this paper, but does find use in this industry. Chromium containing alloys. This group has the largest number of alloys and the broadest usage in the nickel alloy family. It can be subdivided into 2 groups, the nickel-chromium alloys and the nickel-chromium-molybdenum alloys. Nickel-chromium alloys. The main alloy in this group is Alloy 600, with Alloy 601 being an important variation strictly for high temperature applications. Alloy 600 is used for corrosive services, mostly in caustic, but also for high temperature halogen service. With 76% nickel, Alloy 600 is immune to chloride stress corrosion cracking, but will pit. Nickel-chromium-molybdenum alloys. This group of alloys is the most useful to the mining/metallurgical industry because of their high resistance to acids, both oxidizing and reducing. They perform well in highly acidic chloride containing environments, and are used extensively in wet flue gas scrubbers. Alloy 625 is useful in a broad range of applications, but is being replaced in many applications by the "C" family of alloys. C-276 is still the most common, but various new improvements have been commercialized with C22, Alloy 59, and now C2000. There is some data to suggest that these alloys suffer from transpassive corrosion in some

1923

oxidizing autoclave environments, and more work needs to be done with some of the higher chromium alloys such as G-30, which may perform better.

Table 14 Nominal composition (weight %) of some selected nickel alloys Alloy C,H,or UNS Ni Cr Mo Fe Cu Other both* number Alloy 200 C NO2200 99.0 mi n C&H NO6600 Alloy 600 76 15.5 7 H NO660 1 Alloy 601 61 23 13 A1 NO6625 C&H Alloy 625 61 21 9 4 Nb, Ti, A1 N10276 C Alloy C-276 57 16 16 5 W, V C NO6022 Alloy C-22 56 22 14 4 w , v, C Alloy G-30 NO6030 43 30 5 15 1.5 N b , W C Alloy 400 NO4400 66 31 C Alloy K-500 A1,Ti 66 29 NO5500 Alloy B-2 C N10665 69 28 * Primary end use: C = corrosion resistant, H = high temperature resistant, C&H = both Nickel-copper alloys. The two main alloys in this group are Alloy 400 and a precipitation hardenable variation K-500. The latter is useful for shafting, and for parts requiring increased hardness or resistance to galling. These alloys are used extensively in brine systems, and certain applications in fresh water systems. Like most copper-containing alloys, they are subject to corrosion by ammonia. They have poor resistance in oxidizing media. Nickel-molybdenum alloys. This group consists of Alloy B-2 and other variations. These alloys show excellent performance in pure hydrochloric acid up to the boiling point, but very small quantities of metallic oxidizing ions, even a few parts per million, or even oxygenated acid, can result in very high corrosion rates. This author has heard of a few occasions where these alloys have been used where they shouldn’t have, and where this most expensive alloy had lasted only weeks. Where it is good, it is usually very good. In-plant testing is strongly advised before specifying such alloys. Cast nickel alloys. There exist roughly similar cast versions of the wrought alloys, but often there are sufficient differences. Cast nickel alloys are found in ASTM A 494. ALUMINUM Aluminum falls into the “passive” group of materials, meaning that they depend on a passive oxide layer for their corrosion resistance. They can exhibit good corrosion resistance in the middle pH ranges, but not in strong acids or bases. This makes them a useful material in many air and water environments. Although pure aluminum is soft, most commercial alloys are either strain hardened (indicated with an H suffix) or heat treatable (indicated with a T suffix). This gives them a high strength to mass ratio, an advantage used in many structural applications. Aluminum alloys have good low temperature ductility, making them useful for certain cryogenic applications. Aluminum alloys are fairly easily fabricated. Table 15 lists the UNS classification of the alloy types.

1924

Table 15 Wrought aluminum allov UNS number structure UNS Number Major Alloying Element A9 1 xxx None, 99.0% A1 minimum A92xxx Copper A93xxx Manganese A94xxx Silicon A95xxx Magnesium A96xxx Magnesium and silicon A97xxx Zinc A98xxx Miscellaneous Note: Cast alloys are numbered similarly, but with the first number as 0, e.g. A02000 Corrosion resistance. Most aluminum alloys have similar corrosion resistance, although the copper containing alloys are the least corrosion resistant. A rough pH guide for use is 4.5 to 9.5. This qualifies for most natural waters and external atmospheric conditions. Mine waters will often fall outside this range, and the presence of metallic ions such as iron, copper, lead, and mercury will cause accelerated corrosion. Aluminum is used for specific chemicals, e.g. hydrogen peroxide, highly concentrated nitric acid (>go%), and many organic chemicals. However, certain chloro-organic compounds such as ethylene dichloride or certain alcohols can rapidly corrode aluminum generating hydrogen, creating a fire hazard. Applications. Aluminum is used for many structural applications, such as roofing, building panels, hand railings, ladders, fencing, and shaft man cages. Sometimes pre-painted aluminum is used for buildings not only to improve lifetime, but to lower reflectivity. Other applications include underground ventilation ducting, heat exchangers for mine refrigeration plants, ore hoppers, and air and conduit piping (Andrews 1997). Use of aluminum as well as magnesium and titanium underground in certain mines, especially coal mines, may be forbidden or restricted, due to the possibility of fires or explosions related to a thermite reaction (Forrester and Bonnell, 2001). When these light metals come in frictional contact with iron oxide e.g. rust on steel, a violent reaction occurs that can ignite a flammable atmosphere. COPPER AND COPPER ALLOYS Some use of copper and copper alloys is made in the mining/metallurgical industry. The copper alloys can be divided roughly into categories of brasses (copper-zinc alloys), copper-nickel alloys, and bronzes (copper primarily alloyed with any elements other than the previous two, e.g., tin, silicon, aluminum). Table 16 lists a number of these grades. Nickel-copper alloys were covered under nickel alloys.

1925

Table 16 Nominal composition (weight %) of some common copper and copper alloys Cu Zn Sn A1 Ni Other Name UNS number Wrought Alloys c11000 99.9 Copper (ETP) C23000 85 15 Red Brass C27000 69 31 Yellow Brass C44300 Admiralty Brass 72 27 1 As 0.1 C52400 Phosphor Bronze 90 10 P 0.3 C61400 Aluminum Bronze D 90 0.2 7 Fe 3 C65500 High Silicon Bronze 95 1.5 0.6 Si 3 C70600 86 90-10 Cu-Ni 1 10 Fe 1.5 C7 1500 70-30 Cu-Ni 68 1 30 Fe 1 Castings Leaded Red Brass Manganese Bronze Tin or G Bronze

C83600 C86500 C90500

85 57 87

5 40 2

5 1 10

1 1

Pb 5 Mnl

Corrosion resistance. Copper is resistant to most natural waters, except for soft acidic waters, and within certain velocity limitations. It is especially useful in stagnant potable waters. The brasses which contain tin and usually another element such as arsenic (admiralty brasses) are also used in more aggressive waters. The copper nickels are useful also in more aggressive waters, and can tolerate a higher flow rate than other copper alloys. Some of the bronzes can give adequate corrosion resistance in weak sulphuric acid solutions. A good source of information about copper and its alloys is the Copper Development Association (www.copper.org ) Applications. In addition to use in various types of water for piping, heat exchangers, etc., there are some copper alloys used in special applications where certain properties are required. Beryllium bronzes are hard and used for their non-sparking properties in tools. Other copper alloys have excellent anti-galling properties. TITANIUM Titanium and its alloys were rarely used in the miningkydrometallurgical industry previously, except where very aggressive chemicals such as hydrochloric acid were used. It was also used to some extent for seawater applications. Titanium belongs to the reactive class of metals, to which zirconium and tantalum also belong. It has become very useful in the newer hydrometallurgical processes involving high temperature and high pressure. Titanium is often alloyed with small amounts of other elements (primarily palladium and ruthenium) to improve its corrosion resistance, especially in reducing acids. Table 17 lists some of the more corrosion resistant titanium alloys. The very expensive Titanium Grade 7 with 0.15% Pd used to be the only step up from Titanium Grade 2 (unalloyed titanium), but now a large number of alloys exist which are lower in cost than Grade 7. Table 17 Specific alloying additions (weight %) of some titanium alloys UNS Pd Ru Mo Ni Name number Grade 2 (unalloyed) R50400 Grade 7 R52400 0.15 R53400 0.3 0.7 Grade 12 Grade 16 R.52402 0.06 Grade 17 R52252 0.06 Grade 26 R.52404 0.11 Grade 21 R52254 0.11

1926

Titanium alloys can offer excellent corrosion resistance in a wide variety of oxidizing and reducing environments, which cannot be covered in this paper. There are many excellent sources of information, such as the Titanium Development Institute (www.titanium.org). A few words of caution about titanium. Titanium can burn in strongly oxidizing conditions. Titanium is successfully used in wet chlorine, but can burn in dry chlorine. It can burn in highly oxygenated conditions, which has occurred in some autoclave circuits. Titanium should also not be used in liquid oxygen nor red fuming nitric acid. Titanium becomes brittle when exposed to hydrogen ions, which can even originate from corrosion, causing hydriding. Other seeming innocuous chemicals such as methanol (which may be used as a cleaning solution) can result in degradation of the metal. Even small traces of fluorides may cause hydriding and accelerated corrosion rates.

OTHER METALS AND ALLOYS Other metals such as lead, cobalt, zinc, tin, magnesium, tantalum, zirconium, silver, gold, and platinum may be used for specific purposes, but are not covered in this paper. SELECTION OF ALLOYS -A FEW KEY PRINCIPLES Operating Conditions In order to select the proper material, it is necessary to know the exact conditions. In the author’s experience, most of the time there is not sufficient data given by engineering staff or plant staff to make anything other than general statements. Following is a list of things to consider in order to help to determine what the metal will actual experience and will hopefully be able to handle. 1. Impurities. Plant and process personnel often know what chemicals are in the process that are important to the process, but are not always aware of the total makeup of the solutions. Knowing levels of impurities such as chlorides and fluorides are critical. Even parts per million of certain oxidizing ions can either destroy some alloys or help keep passive other ones. A full analysis is a very good starting point. It is important to know if a solution is aerated or de-aerated. 2.

Heat flow. Often the temperature in a tank will be set at one level, but there are areas that may have a much higher temperature.

3.

Upset conditions. Although it is not possible to consider every possibility, nor to guard against them, it is a good exercise to look at the various scenarios. For example, if a valve is jammed open or shut, or if a tank is only partially filled, or if one part of the process has to be bypassed, what is the affect on the temperature, concentrations, etc. This is especially important if hazardous chemicals are involved.

4.

Process changes. Again, one cannot look at every possibility, but often process engineers know that they are likely to modify the process, often to produce more product. Will higher temperatures or higher velocities be used. What about different ores - will they result in a change in contaminants and will that have a positive or negative affect on corrosion rates.

Design Each material has its strengths and weaknesses. Often one is tempted to replace materials using exactly the same sizes. Although sometimes this works, sometimes it doesn’t and often money is spent needlessly. Carbon steel components are often made thick to allow for a measurable corrosion rate. If replacing with stainless steel, it may be that there will be no measurable corrosion and the material thickness can be reduced significantly. Carbon steel and copper often have velocity limitations, so heat exchangers and equipment are designed for low

1927

velocity flow. Stainless steels often suffer localized attack in stagnant conditions, so a high velocity is preferred to keep the surface clean and avoid pitting. Detailed discussion of design considerations for stainless steels is given in NiDI Publication No. 9014. Cast versus Wrought product form There are distinct differences between nominally the same alloy grade in cast and wrought forms, although sometimes the difference doesn’t mean very much, and one is not necessarily better than another. However it is important to be aware that there are differences, and when they may become a factor. Differences based on production methods. Typically wrought stainless steel alloys are melted and refined in modern equipment in batches of 20 to 150 tonnes. The refining is well controlled, impurity levels generally very low, and the difference between aim and actual chemical composition is very small. One should remember that this aim composition is the aim of the mill, not necessarily what is best for e.g. the corrosion resistance of the steel. As an example, the molybdenum content of 316L according to UNS S31603 is 2.0 - 3.0%. Steel makers today can control the molybdenum content quite tightly to between 1.95 and 2.05% (and remember that 1.95% meets the requirement of 2.0% minimum!) Wrought products are produced in series on highly automated lines, meaning less chance for human error and a high degree of consistency. In contrast, stainless steel foundries produce castings virtually individually by the piece. The furnace size can vary from tens of kilos to 50 tonnes. There is often no refining stage for the steel at a foundry, the impurity levels are determined by the choice of raw materials, and the quality can vary considerably from foundry to foundry, and to some extent, from piece to piece. With castings, there can be considerable weld repair at the manufacturing stage to correct problems from the pouring stage. Some castings specifications call for heat treatment after the weld repairs, e.g. ASTM A 744, others e.g. ASTM A 743 do not, although often there is an option for that as a supplementary requirement. In critical applications, heat treatment of the casting after all welding may be necessary. In castings, some degree of porosity is normal, whereas in wrought products, there should be no porosity. Various non-destructive examination methods are available for castings, but it can be difficult to know which ones to specify. A new ASTM specification, ASTM A 990, “Specially Controlled for Pressure Retaining Parts for Corrosive Service” is intended for the most severe services. Differences in alloys. Because castings are often melted in smaller quantities, there is the potential to customize the chemical composition for a particular application. Where it might be virtually impossible to interest a wrought mill in producing 10 tonnes of a special chemistry sheet, there would be many foundries eager and willing to quote on the same quantity of castings. There are many unique casting alloys that are virtually impossible to make via the wrought route. This includes alloys that have poor hot or cold ductility, and alloys that cannot be easily welded into components, but can be cast to their final shape. With wrought alloys, with the exception of a number of specialty alloys, the names of the grades are standardized and easily identifiable. With cast alloys, there are a host of trade names that can make identifying alloys and finding substitutes a chore. There can be some significant differences in performance between nominally identical wrought and cast alloys, for example between wrought and cast “Alloy 20”. The cast version contains less nickel and is niobium-free, which means that it should not be welded. Other differences relate to metallographic structure in the austenitic stainless alloys. Wrought 304 stainless steel is normally close to 100% austenitic in structure, whereas the cast version, CF8, usually contains between 10 and 20% ferrite. These differences in structure are intentional, controlled via both chemistry and heat treatment. The ferrite component of most austenitic castings is intended to reduce defects during casting, especially hot tears. A magnet is strongly attracted to the ferrite in these alloys, whereas a magnet has only faint or no attraction to wrought austenitic alloys. However, for the same reasons as castings, welds have some ferrite and a magnet will be fairly strongly attracted. However, certain chemicals can preferentially corrode this deltaferrite, and especially in the higher molybdenum alloys such as CG3M, a 317L (3.0-4.0% Mo)

1928

type alloy. Some of the even higher molybdenum cast stainless alloys are fully austenitic, e.g. CK3CuMN and CN3MN, both with over 6% Mo. The fully austenitic structure creates some casting challenges, and as a word of warning, discussions should be had with the selected foundry regarding heat treatment procedures for these alloys. The ASTM specifications for these grades specify too low a minimum heat treatment temperature. The larger grain size of castings can be an advantage in high temperature applications with respect to creep strength. Castings grades typically have a higher carbon content, also an advantage for high temperature strength, although a disadvantage in corrosive applications. For example, 310s wrought product has a carbon maximum of 0.08%, whereas the rough cast equivalent, CK20, has 0.20% carbon maximum. There is also a higher carbon cast version, HK40, which is more common, and that has 0.45% carbon max. The higher carbon does decrease room temperature ductility, but castings are usually cast to shape, so this is not usually important. Cost and Availability One engineer the author knows is constantly reminding people that availability of a material is a material property just like strength or modulus of elasticity. Certainly price and availability do go hand in hand. 304L and 316L are the most commonly available stainless steels, produced by dozens of producers around the world. There is a wide variety of product forms and much competition which keeps the price down. Conversely, proprietary grades available from only one supplier can not only be difficult to obtain, they can be very expensive. Try to specify alternative alloys where possible. For example, there are 4 or 5 different variations of the 6% Mo stainless steels, all with roughly (but not exactly) the same corrosion resistance and mechanical properties. One supplier might be good on plate, another on sheet, and another on bar. By giving options, it is possible for fabricators to put together a reasonably priced package by having the suppliers compete against each other. On the other hand, if one supplier is actively working with you in supplying information and test coupons etc., it is only fair to try to work with them to resolve the commercial issues that may arise. Put a value on the technical support that they give you.

RESOURCES Prior experience If the process is not new, but exists at another plant, much can be learned there about materials selection. One should be aware that there can be significant differences based on compositional differences in the ore body, differences in climate, water quality, availability of skilled labor, and even differences in the operating philosophy of the company. Therefore it is impossible to say that if Alloy Q worked at Plant A, then Alloy Q will work at Plant B. It is not sufficient to just look at the current and past history of materials’ performance; it is important to thoroughly evaluate the solution chemistries, abrasive qualities, and other factors such as changes to the process, process upsets that may have occurred, and maintenance practices. Materials selection data for existing plants needs to be used with caution. It is valuable to know that a pipe has lasted a certain number of years, but it is far more valuable to know if it has its original wall thickness left, or if it is ready to be replaced. Much valuable information can be gathered from test coupons exposed to the actual solutions. It takes time and effort to put together a test program and then collect and evaluate the data afterwards. It is almost becoming a lost art. Test coupons should be exposed for a minimum of 3 months, and preferably a full year. When a corrosion failure occurs, most engineers wish they had installed test coupons a year earlier. The comparison process requires dedicated teamwork from process specialists, design specialists, maintenance specialists, materials specialists, fabrication specialists, etc. Selection of materials for a new plant using a new process is an extremely difficult task. So many factors have to be taken into account, with little experience to rely upon. However, important clues can be gained from similar if not identical plants if they exist, pilot plants, and perhaps even chemical plants that might have similarities.

1929

Corrosion Handbooks There are many good Corrosion Tables and Handbooks available. In the hands of an experienced materials specialist, they provide clues to what materials might work. In the hands of the inexperienced, they have often led to very bad materials selection errors. It works both ways sometimes an expensive material is chosen when a more cost-effective one would have worked as well, whereas most of the time, a material is chosen that fails quickly, and sometimes catastrophically. Corrosion tables have many limitations. They are most often based on single, reagent grade chemicals, not what one has in a metallurgical plant. Most tables report only a general corrosion rate, valid if uniform corrosion is the corrosion mechanism. With stainless steels and nickel alloys, localized corrosion is more likely to occur and lead to failure. Corrosion tables rarely take into account velocity effects, abrasive effects, effect of oxygen concentration in the corrosive, vapor phase affects, sensitivity to contaminants, etc. However, they are valuable and can give us clues to which type of alloys to consider. Internet Much useful information is available directly from the Internet from reliable sources. Reliable sources are sometimes the same organizations and companies that you would go to anyhow for information, but there are new only net-based suppliers. A good search engine such as Google is useful in identifying unknown grades and trade names.

SUMMARY This paper attempts to discuss in general terms the properties of various metals used in the mining/metallurgical industry, with both advantages and disadvantages of the various types. A few tips are given about selection of materials, as well as some resources for further information. REFERENCES R.H.C. Andrew. 1997. Practical Guidelines for Corrosion Protection in the Mining and Metallurgical Industry. NACE International, Houston TX. P.E. Arnvig, H. Andersen, et al. 1998. SCC of Stainless steel Under Evaporative Conditions. Corrosion 98, Paper 25 1. NACE International, Houston, TX.

D.J. Forrester, G.W. Bonnell. 2001. The use of light metals and their alloys in underground coal mines. CIM Bulletin, Vol. 94, No. 1054, p76ff STANDARDS AND SPECIFICATIONS ASTM Standards. Books of standards are published annually by ASTM International, West Conshohocken, PA. Individual standards are updated on a regular basis, some as many as several times a year. Individual standards are available on a download bases from the ASTM website, www.astm.org UNS. Metals and Alloys in the Unified Numbering System. 9* Edition. 2001. Society of Automotive Engineers, Warrendale, PA SELECT RECOMMEND HANDBOOKS Printed paper editions are given, but note that many of these are available in electronic version, either on CD or downloadable from web sites. CASTI Handbook of Stainless Steels & Nickel Alloys. S. Lamb, editor. CASTI Publishing, Edmonton, AB 1999 Steel Products Manual: Stainless Steels. H. Cobb, editor. The Iron and Steel Society, Warrendale, PA 1999

1930

Metals Handbook, Desk Edition. 2"d Edition. J.R. Davis, editor. ASM International, Metals Park, OH. 1998 The Metals Black Book, Ferrous Metals. 4* Edition. J.E. Bringas & M.L. Wayman, editors. CASTI Publishing, Edmonton, AB. 2000 The Metals Red Book, Volume 2, Nonferrous Metals. 3rdEdition. J.E. Bringas & M.L. Wayman, editors. CASTI Publishing, Edmonton, AB. 2000 Avesta Sheffield Corrosion Handbook for Stainless Steels. 8" Edition. Then Avesta Sheffield, now AvestaPolarit, Avesta Sweden 1999 Corrosion Data Survey, Metals Section. 6* Edition. D.L. Graver, editor. NACE International, Houston, TX. 1986 Woldman's Engineering Alloys. 9~ Edition. J.P. Frick, editor. ASM International, Metals Park, OH. 2001 Stahlschlussel (Key to Steel). 2001 edition. Verlag Stahlschlussel Wegst GmbH, Marbach, Germany. 2001

SOME USEFUL BROCHURES AVAILABLE FROM THE NICKEL DEVELOPMENT INSTITUTE Available either as pdf downloads from www.nidi.org or in print. Many others available. No. 12 014 Copper-Nickel Fabrication No. 12 007 Copper-Nickel Alloys - Properties and Applications No. 11 022 Castings - Stainless Steel and Nickel-Base No. 11 021 High-Performance Stainless Steels No. 11 01 8 Properties and Applications of Ni-Resist and Ductile Ni-Resist Alloys No. 11 017 Ni-Hard Material Data and Applications No. 11 012 Guidelines for the Welded Fabrication of Nickel Alloys for Corrosion-Resistant Service No. 11 01 1 Machining Nickel Alloys No. 11 007 Guidelines for the Welded Fabrication of Nickel-Containing Stainless Steels for Corrosion-Resistant Service No. 11 006 Nickel in Powder Metallurgy Steels No. 10 088 Nickel Plating and Electroforming - Essential Industries for today and the Future No. 10 086 Corrosion and Heat-Resistant Nickel Alloys , Guidelines for Selection and Application No. 10 081 Properties and Applications of Electroless Nickel No. 10 075 Selection and Use of Stainless Steels and Nickel Bearing Alloys in Nitric Acid No. 10 074 The Corrosion Resistance of Nickel-containing Alloys in Hydrofluoric Acid, Hydrogen Fluoride, and Fluorine No. 10 068 Specifying Stainless Steel Surface Treatments No. 10 064 Clad Engineering No. 10 063 Selection and Use of Stainless Steels and Nickel Bearing Alloys in Organic Acids No. 10 057 Selection and Performance of Stainless Steels and Other Nickel Bearing Alloys in Sulphuric Acid No. 10 021 Procurement of Quality Stainless Steel Castings No. 9014 Design Guidelines for the Selection and Use of Stainless Steels

1931

Elastomers in the Mineral Processing Industry Paul Schnarr’, Leon E.Schaeffe?,and Hans J. Weinand

ABSTRACT Elastomers have been used extensively in the mineral processing industry since the 19” century. They are used, cost effectively, mostly because of their resistance to wear, impact, flexing, and corrosion, but also for their ability to control vibration and noise. Elastomers are used in process operations from mining, comminution and separation through to final product handling. The most appropriate elastomer for a given application is selected for its particular properties.

INTRODUCTION The definition of an elastomer is “a polymeric material, such as a synthetic rubber or plastic, which at room temperature can be stretched under low stress to at least twice its original length and, upon immediate release of the stress, will return with force to its approximate original length.”’ This property allows elastomers to be used for many products which are subject to deformation or compression and must not be destroyed by such forces. Abrasion resistance is often the main feature in choosing an elastomer based product over alternate products. In the mineral processing industry, abrasion often results from a slurry’s suspended particles coming in contact with the elastomer in a combination of impinging and sliding action. The erosive wear of a metal alloy depends on the alloy’s microstructure and hardness; an elastomer’s resistance to abrasion, especially of the impinging type, depends on the resilience of the elastomer. When a particle contacts the wear surface, an elastomer deflects, then returns to its original position, with little or no wear. This unique property of elastomers generally gives them a substantial increase in abrasion resistance over metals, provided the particle size and speed are not too great.’ Another main reason for using an elastomer product is the chemical resistance of the elastomer. In general, the rule “like dissolves like” applies. This means that nonpolar elastomers such as natural rubber will handle most chemicals used in the mineral processing industry, which are typically water based polar solutions. When nonpolar solvents or oils are encountered, elastomers such as nitrile-butadiene rubber, must be used. Field experience, collected over many years, provides the best basis for successful material selection. Chemical resistance charts are also useful. In difficult cases, especially those involving mixed and/or unknown chemicals at varying temperatures, immersion testing in the customer’s actual plant is the best basis for a choice of materials. Most elastomers used in mineral processing are actually used in composites. They can be bonded to textile fibers such as nylon, polyester, polyaramid and others to increase strength and stiffness, with a loss in elongation. They can be bonded to metals to combine the strength and rigidity of the metal with the elasticity of the elastomer. They can be bonded to various plastics, often to get a surface with a very low coefficient of friction. An elastomer compound development is usually a compromise, which can be shown as an equilateral triangle.

’ Technical Director-RubberEngineering, Salt Lake City, UT Rubber Engineering, Salt Lake City, UT Rubber Engineering, Germany

1932

Performance

3 P’s The rubber compounder tries to achieve the best compromise by choosing the elastomer, or blends of elastomers, and then adding various fillers and chemicals, of which there are an infinite number of combinations. The end customer sees only the Performance and Price part of the triangle but the compounder has to deal with processing through various mixing, forming and vulcanization steps. HISTORY Ancient mesoamerican people were processing rubber by 1600 B.C. The latex was combined with a species of morning glory vine to give products with enhanced elastic behavior, such as rubber bands to haft stone ax heads to wooden handles! It is easy to speculate that such an ax could have been used in a mining operation. Christopher Columbus was probably the first European to see natural rubber. The mention of rubber trees is to be found in the eighth (decade) of De Orbe Novo by Pietro Martire d Angliera, published in Latin in 1516. Very little use was made of rubber until 1823 when Charles MacIntosh patented the use of coal tar na tha to dissolve natural rubber and use it to produce waterproof garments or “macintoshes.’” These useful garments were undoubtedly used in wet mining conditions. About 1839 Charles Goodyear discovered that rubber chains could be linked together, i.e. vulcanized, by heating with sulfur and white lead. In this process, sulfur linkages form bridges between the rubber chains.6 Thus the first true elastomer was produced. Goodyear made little commercial progress. In 1843 Thomas Hancock of London took out a patent similar to Goodyear’s and was more commercially successful. One of his first products was thin sheets of rubber and these were probably used in some form of mineral processing. By 1850 a whole range of rubber articles was available5 and though it is doubtful many of these were specifically designed for mineral processing, it is likely many did find use. Advances in elastomer technology have been essentially continuous since the early discoveries. Most were made with the tire industry in mind, as it is by far the largest user of elastomers. Since many of these advances involve basic properties such as abrasion, tear, and flex resistance they can be readily adapted by compounders, who are designing elastomer products for the mineral processing industry.

POLYMER TYPES-ASTM D2000 ASTM D2000 is the standard classification system for rubber products in automotive applications. ASTM D2000 3.1: Purpose: states The purpose of this classification is to provide guidance to the engineer in the selection of practical, commercially available rubber materials, and further to provide a method for specifying these materials by the use of a simple “line call-out’’ designation.” It also serves the needs of other industries in arranging rubber products into characteristic material designations. These designations are determined by types, based on resistance to heat aging, and classes, based on resistance to swelling in oil. Table I categorizes some of the more common types used in mineral processing. The abbreviations are per ASTM D1418-01. “

1933

Heat and Oil Resistances of Elastomers ASTM D-2000 Designations

Table I H

250

G

225

F

200

E

175

D

150

C

125

IIR

’O01

SBW BR

FKM VMQ EPDM

70

A % Swell

Class

CM

I

HNBR CSM

XNBWNBR

CR

NR NIA

170

120

100

80

60

40

30

EU,AU 20

10

A

B

C

D

E

F

G

H

J

K

There are also grade numbers, suffix letters, and numbers to describe necessary qualities beyond the basic requirements. There are also standards groups in various other countries.

COMMONLY USED ELASTOMERS Natural Rubber (NR) This is probably the biggest single polymer used in mineral processing. The latex from the trees of Heveas basiliensis is the only important commercial source. The hydrocarbon portion of NR is cis1,4 polyisoprene, which comprises more than 10% of a mixture with naturally occurring resins, proteins, sugars, etc.5 As to the Price part of the 3 P’s, in early 2002, (all future price comparison in this paper will be based on early 2002 prices) NR is the lowest priced polymer. It is even lower priced than carbon black, the most important filler, on a volume basis, so that high polymer content “pure gum” compounds are relatively inexpensive. Because of the supply and demand, prices can fluctuate quite widely year by year. Several years ago the prices were approximately three times current prices. It was also more expensive than most common synthetic “tire grade” polymers. As to the Processing part of the 3 P’s, NR is very good. It mixes, calendars and extrudes well. It has good green (uncured) strength and very good building “tack,” which means it adheres to itself and other rubber compounds very well and tends not to stick to the metal surfaces of the processing equipment. This building tack and green strength makes it very suitable for composites such as tires, hose, belts, etc where different components have to be adhered to each other before vulcanization. It also makes for good seam quality in tank lining. NR is often used as a tie gum in halobutyl lining for that purpose. It also makes it relatively easy to line the ID of pipes, especially small diameter ones, where a tube is formed and expanded and must adhere to the pipe before and after curing. As to the Performance part of the 3 P’s NR is often the preferred polymer. “Pure gum” compounds (often defined as those with a specific gravity of less than 1.0) with a hardness of approx. 40 Shore A, are often used in sheeting, pump liners, pipe lining, etc. because of their very high resilience, combined with good strength. Strength can also be characterized as tear resistance

1934

or as cut growth resistance. This high strength of soft natural rubber is certainly due to its ability to undergo strain-induced crystallization. The physical properties of NR compounds (and synthetic rubber compounds) are affected by the type and amount of fillers used. Carbon black is the most commonly used filler. Increasing amounts of carbon black increases the hardness and modulus of the vulcanizates. Resilience and resistance to impinging type abrasion decreases along with elongation. Tensile and tear strength and resistance to sliding type abrasion increase, with increased loading, up to certain point, and then decrease. The black loading is adjusted to maximize the desired properties. Fine particle sized silica can also be used with much the same response as carbon black. The relatively new highly dispersible forms, in combination with organo-silane coupling agents are the basis of the “green tire” tread, which combines good abrasion resistance along with low rolling resistance.8 These advantages also translate to products used in mineral processing. In NR compounds, this technology can be used to produce compounds with the abrasion resistance of a “pure gum” compound but with the cut and tear resistance of a harder black loaded compound. In applications, such as mill circuit discharge pumps handling slurries with up to 12mm diameter particles, the service life is increased considerably for the pump liners.’ The favorable elastic properties of NR manifest themselves in very low damping (low hysteresis) and low heat build up in dynamic deformation^.^ This makes NR especially useful in dynamic applications such as vibration elements used in centrifuges and most importantly in very large off-the-road tires, which also use mostly NR because of its low heat build up, dynamic fatigue resistance and excellent resistance to cutting and chipping. NR vulcanizates are not as heat resistant as most synthetics. 70 C is normally considered the maximum service temperature in dry applications. Special compounding techniques using very low sulfur or no sulfur cure systems, along with protective antioxidants can extend this limit to over 100°C, especially in wet service. Non-protected NR vulcanizates are very prone to ozone cracking, especially in non-black compounds. If they are stored improperly, under stress, they may even crack before being put into service. Black compounds which contain proper amounts of wax and modern straining antiozonants are far less prone to cracking. Since NR is nonpolar, its vulcanizates can swell to several times their original volume in nonpolar solvents such as mineral oil, toluene, gasoline, diesel, etc. NR has good resistance to polar fluids such as many mild acids and bases typically found in the mineral processing industry. The compression set resistance of NR can be very good, with proper compounding, at relatively low temperatures (c70°C), so it is widely used in load bearing applications. The low temperature properties of NR vulcanizate are excellent without any special compounding and exceeded only by polybutadiene or silicone. NR can be blended with various synthetic rubbers with a diene component. Synthetics can be added to NR to improve heat resistance, sliding abrasion resistance, and give moderate oil resistance. Natural rubber can be added to synthetics to improve building tack and green strength and to lower cost.

Polyisoprene (IR) Initial attempts to make synthetic rubber date back to the mid 1800’s. Two samples labeled “Isoprene” and “Artificial Rubber” were recently discovered at the University in England, where Sir William Telden conducted his experiments in 1892.’’ Commercialization of synthetic polyisoprene was started in 1960. The performance of synthetic polyisoprene (IR) is somewhat similar to its natural counterpart. Processing is better in some ways in that it does not require mastication and generally mixes, calendars and extrudes faster and with more consistency. Processing is poorer in that it lacks the green strength and tack of NR. The main reason it is now not used more is that it is currently priced much higher than NR. A number of plants processing it have been shut down and a lot of material is coming out of Russia, which had a large capacity for strategic reasons.”

1935

Styrene Butadiene Rubbers (SBR) Large-scale production of SBR started in WWII because of the shortage of NR. It was known in the U.S., as GR-S (Government Rubber-Styrene). It is a major polymer in passenger and light truck tires but is little used in large off-road tires because it’s compounds have higher heat build up than NR and less resistance to cutting and tearing. It does have good resistance to sliding abrasion and impact so it can be used in products such as belts, sheet rubber, truck box liners, and mill liners. The heat resistance is somewhat better than NR and special compounding can take the maximum temperature rating up to approximately 110°C. Processing is generally good, but SBR lacks the tack and green strength of NR. It does have lower nerve and flows better than NR and can be blended with NR to improve flow in large transfer molded products. The price of SBR is currently a little higher than NR, but historically has averaged a little less, and does not fluctuate as widely. Butadiene Rubber (BR) Most of the butadiene rubber used is solution polymerized with a very high cis-l,4 configuration. BR is seldom used as the sole polymer in any mineral processing application as the strength is very low and processing very difficult. BR is most commonly blended with NR or SBR, with most of the technology coming from the tire industry. When blended with NR, or SBR in compounds containing filler to give hardness, usually in the 50-80 Shore A range, BR improves abrasion resistance, heat aging, resilience, reversion resistance, fatigue resistance and low temperature flexability.12 Sidewalls of passenger tires are commonly based on blends of natural rubber and Cispolybutadiene. They have outstanding resistance to catastrophic crack growth under severe conditions, such as when a tire runs into a curb or hits a deep chuckhole; but also a sidewall must resist crack growth over long times at smaller strains experienced during “normal” rolling.l 3 This resistance to different types of crack growth is very important for many products used in mineral processing other than tires, such as belts, especially those with flanges, hose and mill liners, which often use blends of NR and BR. The processing of BR blends is generally good, NR/BR blends have good tack and green strength so they can be used for composites. The price of BR is currently lower than that of NR and the same as SBR, but historically it has commanded a premium over both. Butyl Rubber (IIR) IIR is a copolymer of mainly isobutylene and a small portion of isoprene. This largely saturated chain determines its main properties; good resistance to oxidative and ozone degradation, to chemicals, and a low gas permeability3 It is very nonpolar and therefore resistant to polar chemicals, especially acids and bases at higher temperature and concentration than NR compounds. Mechanical properties are fair. Resilience at room temperature is very low and this leads to low resistance to impinging abrasion. Its main use in mineral processing is the lining of tanks, pipes, and pumps. Processing is fair in general but it is very incompatible with diene rubbers such as NR, SBR, and BR so small amounts of an IIR compound can contaminate other compounds or be contaminated by other compounds, so the manufacturing facility must be careful to avoid such contamination. IIR polymers and compounds are considerably more expensive than NR, BR or SBR. Chlorobutyl (CIIR) and Bromobutyl (BIIR) CIIR and BIIR are prepared by the halogenation of IIR. Either halogen gives increased cure reactivity. As a result improvements occur in vulcanization rates, the state of cure, and reversion resistance and covulcanization with other diene rubbers is also possible.

1936

CIIR and even more so BIIR vulcanizates have lower gas permeability, better weather and ozone resistance, higher hysteresis, better resistance to chemicals, better heat resistance, better adhesion to other rubbers than those of IIR.3 CIIR and BIIR vulcanizates are used in much the same places as IIR and have replaced IIR in many applications such as belts, hose, and tank lining. There is very little difference in processing BIIR and CIIR or IIR except BIIR or CIIR do not contaminate or are contaminated, by other diene rubbers. The price of BIIR and CIIR is only slightly more than IIR. Ethylene - Propylene Rubber ( EPM and EPDM ) The first polymers commercialized were merely the copolymers of ethylene and propylene (EPM). Because the polymers are totally saturated, the compounds could only be cured with peroxides. The limitations of peroxide cures did not satisfy the needs of the rubber industry. If a third monomer, a diene, is added during polymerization the resulting rubber (EPDM) can be vulcanized with sulfur, giving more flexibility in curing, without a significant loss in the stability of the original copolymers! The rest of this paper will deal with only EPDM, as it is much more commonly used in mineral processing. As EPDM’s have fully saturated backbones, (the unsaturation of the diene is in a side chain) the resistance to ozone and oxygen is excellent. As nonpolar hydrocarbon elastomers, with an amorphous nature, EPDM polymers have good electrical properties. The nonpolar nature gives resistance to polar materials such as phosphate esters, many ketones and alcohols; many acids and bases, water and steam. Resistance to chlorinated solvents is fair. Resistance to nonpolar oils, gasoline etc is very poor, although compounds with high loading of black and oil have lower volume swell compared to other hydrocarbon elastomers. Resilience is only fair so resistance to impinging abrasion is much less than natural rubber vulcanizates but much better than butyl vulcanizates and therefore EPDM has replaced butyl in many applications requiring such abrasion resistance. Resistance to sliding abrasion is good, as well as tear resistance. Compression set resistance, particularly at high temperatures is good, if properly compounded. Heat resistance of up to 200°C can be achieved for special applications. Processing is generally good except for a lack of building tack. This severely limits EPDM’s use in hand lay applications such as tank lining. The price of EPDM compounds varies widely. High filler and oil loading are more widely used than in most polymers to achieve relatively low cost compounds, but these compounds tend to have poor strength and abrasion resistance. High quality stocks especially those with peroxide cures, are considerably more expensive than NR, BR, or SBR stocks. EPDM is used in hot material belts, and hose because of its excellent heat resistance. Excellent electrical properties often make it the preferred polymer for cable insulation and electrical connectors. Its chemical resistance often makes it the preferred polymer for specialty belts, hose and pump liners. Polychloroprene (CR) Neoprene is the generic name for polychloroprene. It has been produced commercially since 1931 and had rapid acceptance because it is much superior to natural rubber for heat resistance and oil resistance. Heat resistance is better than NR, BR, or SBR but cannot approach that of EPDM. When heated in the absence of air neoprene withstands degradation better than many elastomers normally considered more heat resistant and retains its properties 15 times longer than the presence of air. Compression set at higher temperatures is better than natural rubber and 100°C is typically the test temperature rather than 70°C.

1937

Abrasion resistance is not as good as natural rubber but generally better than most heat and oil resistance polymers. This is also true for tear strength and flex resistance. The resilience of gum neoprene vulcanizates is a little lower than natural rubber but it decreases less with increased filler loading,I4 so the resilience of most practical neoprene compounds is higher than that of natural rubber with comparable volume loading. Because of the chlorine in the polymer, products made from neoprene resist combustion to a greater degree than products made from non-halogen bearing p01ymers.I~This means neoprene can be compounded to meet the flammability requirements of the Mine Safety and Health Administration (MSHA), Conveyor Belt Flammability Program, and various other flammability tests, without massive loadings of soft mineral fillers, and special flame-resistant plasticizers. Such loading in non-halogen bearing polymers leads to much poorer properties, such as abrasion resistance and tear strength. While the oil resistance is not as good as some highly polar elastomers such as nitrile butadiene, neoprene compounds are generally considered moderately oil resistant. Compounded with red lead, neoprene compounds are very water resistant, as well as moderately oil resistant. This leads to use in belts, hose, pump liners, and other products which are handling oily water, especially at temperatures approaching boiling, and also require good abrasion resistance. These same types of compounds are also resistant to many acids at higher temperature than natural rubber can handle. Neoprene shouldn't be used in parts, which are bonded to metal for hydrochloric acid service because acid migration can cause bond failures. Processing is generally fairly good but compounds can be relatively fast curing at lower temperature, (scorchy) so care must be made in mixing, calendering and extrusion to prevent premature vulcanization. Tack is generally quite good so building composites is very practical in belts, hose and tank lining. Good tack and green strength also make neoprene very common in contact cements and in 2-part room temperature vulcanizing cements, which can be used to bond various substances such as rubbedrubber or rubbedmetal. The cost of neoprene is quite high compared to NR, BR, or SBR, and even other heat and ozone resistant polymers such as EPDM, so where EPDM can be used, it is usually preferred because of price considerations. Nitrile Elastomers Simple nitrile elastomers are copolymers of acrylonitrile (ACN) and butadiene in monomer ratios ranging from 18/82 to 50/50. (NBR) The basis for selection of an elastomer having a particular monomer ratio is usually the oiVsolvent resistance, as well as the low temperature performance required in the final vulcanized article. Higher ACN gives better oil resistance but poorer low temperature properties. Using the same base formula a 45% ACN polymer will give a compound with 0% swell in ASTM #3 after immersion for 70hrs at 149"C, while a 18% ACN polymer leads to 30% swell. The 45% CAN compound has a brittle point of -8°C while the 18% ACN compound is -58°C. As the ACN content increases, tensile strength and hardness increase while the resilience and compression set resistance decrease! Oil resistance is usually the reason for use of nitrile compounds in belts, hose (especially tubes), O-rings, seals, gaskets and various other products. Heat resistance is generally good and can be enhanced with special compounding techniques, and by using polymers with built in antioxidants. For ultimate heat resistance a hydrogenated nitrile (HNBR) is used. Compounds made from HNBR have been used successfully in flapper seals in a smelter operation at 200"C, handling very abrasive and corrosive fly ash.15 Resistant to impinging abrasion is only fair, as resilience is generally low with normal ACN levels. Resistance to sliding abrasion is also fair but can be dramatically improved by using a carboxylated nitrile (XNBR); HNBR is also excellent. Flex resistance for NBR products is fair but can be improved with special antioxidants. HNBR elastomers have outstanding flex resistance.

1938

Resistance to polar fluids decreases with increasing ACN levels and there are usually better choices for resistance to acid, bases and water than NBR. HNBR has much better resistance to oxidizing fluids and can be used in hot acids and bases. HNBR has been used in water service, under high pressure, at up to 175°C. Ozone resistance, except for HNBR, is generally poor but can be dramatically improved by blending with polyvinl chloride (PVC). This leads to use in such products as hose covers, cable covers and conveyer belts where oil resistance and fuel resistances are required. NBRRVC can also be blended with NR or SBR to give moderate oil resistance and abrasion resistance. The tack of the NR portion allows such blends to be used to line tanks, pipes and other hand lay products. Processing characteristics are good for all the nitriles, except for low tack. This limits their use in land lay applications. The cost of NBR is between NR and neoprene. XNBR is a little more. HNBR compounds are roughly 10 times the cost of NBR, which limits their use to products which absolutely require their unique properties.

Chlorosufonated Polyethylene (CSM) HYPALONO chlorosulfonated polyethylene was first introduced by Du Pont in 1952.6 CSM compounds have good heat, oxygen and ozone resistance, moderate oil resistance and excellent electrical properties but their main feature for use in mineral processing is their resistance to strong oxidizing acids. Tanks, pipes, and pumps can be lined with CSM and hose tubes can handle corrosive chemicals including 66“ Baumd sulfuric acid.6 Good colorability, along with the other good properties, make CSM compounds a good choice for colored cable jackets. CSM compounds can be used for geomembranes for lining reservoirs. They are installed in an uncured state for simple seaming and repairs, if necessary, and then slowly cure during subsequent aging, giving an increase in toughness and durability.6 Processing is a little different than most polymers, in that CSM is more thermoplastic and softens more with heat. Compounds tend to be “scorchy.” Tack is not very good but hand lay items can be made with care. When making items such as hose, neoprene is often used as a cord coat, because its better tack makes building easier, and adhesion to the cord and the CSM tube and/or cover is good. The cost of CSM compounds is generally a little more than neoprene compounds. Silicone Elastomers (Q) Silicone rubber has both excellent low (-65°C) temperature properties and can stand exposure to 315°C. Poor properties such as tear strength and abrasion resistance limit silicone rubber in most mineral processing application. Liquid RTV (room temperature vulcanizing) compounds are useful for small repairs and sealing applications and have been used for poured-in-place gaskets. Polyurethanes (AU or EU) Most polyurethanes are different from other elastomers in that they are cast. Two components are mixed together. One of the components is a prepolymer, which consists of two major chemical structures. One is an isocyanate, usually MDI (methylenebisdiphenyl diisocyanate) or TDI (tolulene diisocyanate). The other is a polyol, either a polyether (EU) or a polyester (AU). The other compound is a curative, which contains hydroxyl or amine groups. Urethanes are used in general because they have excellent physical properties including high tensile and tear strength, high resilience, abrasion resistance, and excellent resistant to nonpolar oils and fuel and ozone. Compared to other elastomers polyurethanes tend to have their optimum properties at much higher hardness and modulus. This leads to higher load bearing capabilities and higher tip speeds in pump impellers, than softer elastomers. As a general guideline, esters are better for tensile strength, tear strength, sliding abrasion oil resistance and heat resistance. MDI-ethers are better for rebound, low temperature properties, impingement abrasion and hydrolysis resistance.I6

1939

The limitation of polyurethanes are chiefly three.lb Owing to a certain thermoplasticity their upper temperature limit is around 110°C. Polyurethanes are subject to hydrolysis in the presence of moisture and high temperature. At low temperatures most polyurethanes can withstand continuous contact with water for years. None can withstand steam for prolonged periods. In between these extremes polyurethanes may or may not be suitable for use. The MDI-ethers are much preferred for hydrolysis resistance. Sometimes a polyurethane part used in the wrong conditions will appear to be performing better than the same part made from a more waterresistant elastomer, such as natural rubber, and then it will fail rapidly. Certain chemical environments (strong acids and bases and polar solvents such as ketones or esters) are also unsuitable for polyurethanes because of their polar nature. Processing is different than other elastomers in that most polyurethanes are cast into a mold, as previously discussed. The mixing of the prepolymer and curative can be done with specially designed machines in large shops, or by hand in small shops, or for very small runs. The polyurethane is often bonded to metal in the casting operation. Fabrication errors may induce failure even in properly designed parts. Errors include incorrect proportions of prepolymers and curative, overheating the prepolymers during storage and improper application of adhe~ive.'~ The cost of polyurethanes varies with the type but is generally considerably more than general purpose elastomers such as natural rubber. This can be made up for, especially for small runs, by the rather low cost of the tooling and equipment used. Polyurethane is also available in millable gum form, which can be handled on conventional rubber processing equipment and cured in similar molds. It comes in various grades much like the castable types and offers similar properties at a little higher price. Lack of knowledge, poor technical advice, and consequently the wrong choice of grade or inadequate processing are the main causes of comparatively low use of millable gum polyurethane in the world wide rubber industry." Fluoroelastomers Practical fluoroelastomers, introduced in the 1950's, provide extraordinary levels of resistance to chemicals, oils and heat. VITONB),made by DuPont, is probably the name most familiar to the mineral processing industry, although other companies in the U.S.A., Europe and Asia produce a wide variety of different fluoroestomers. Few fluoroestomers product are specifically designed for mineral processing but some are used, in hose and pump liners, to handle aromatic solvents such as toluene and very strong acids, and O-rings, gaskets and seals to handle aggressive chemicals and solvents, and extreme heat. Processing tends to be difficult. Tack is very poor, contamination can be a problem and high temperature post cures are often required. The cost of even the lowest price fluoroestomers polymer is roughly sixty times that of natural rubber. The most expensive parts are costed by the gram. Other Elastomers There are many other elastomers available but products from them are seldom designed for the mineral processing industry, even though some are used. A sampling of them are polyacrylates, (ACM, ANM, AEM,) chloro-polyethylene, (CM) ethylenehinyl acetate, (EVM) Epichlorohydrin (CO or ECO), epoxidized natural rubber (ENR), acrylonitrile- isoprene (NIR) and stryeneisoprene rubber (SIR). There is also a whole class of thermoplastic elastomers (TPE). These behave more or less like other elastomers at room temperature but like plastics above their melt temperatures. They are processed on equipment typically used in the plastic industry. Many or even most rubber processors don't tend to think of TPE's as true elastomers, but more like flexible plastics, although they do have some use in mineral processing, in items such as hose and cables.

1940

USES As previously stated elastomer products are used extensively in mineral processing, mostly because of resistance to wear, flexing and corrosion. Some of the specific uses are: Truck Box Liners These are often the first elastomer product to see ore in an open pit mining operation. They are molded in large, thick sections and fastened by T-bolts or welded studs. They often outwear metal liners con~iderably.'~Their elasticity during loading reduces chassis shock. This can lead to reduced wear of the tires and lower maintenance of drive train components of the of the truck. Driver acceptance is also good because of this reduced shock. Disadvantages, compared to metal liners, include reduced volume, but elastomers have the advantage if the total weight being carried is the limiting factor. In cold climates where the engine exhaust is run through the tank bed the heat is too great for elastomer liners. The most common elastomers used are SBR, NR, and BR compounded for maximum resistance to impact, cutting and abrasion. Tires

Most of the tires specifically designed for the mineral processing industry are off-the-road (OTR) types. These tires are exposed to very rough, sharp surfaces with the result being chipping/ chunking of treads.20In haulage equipment, where large loads are carried at relatively high speeds, heat generation is a major factor in designing the compounds used in the tires. The needs of the OTR tires are usually satisfied by natural rubber compounds, or blends of N M R , NWSBR or NR/BR.

Belts The needs of elastomer compounds used in standard flat belting are similar to many other applications. Cover compounds must be abrasion resistant, resistant to cutting and chipping, and resistant to cracking caused by repeated flexing and exposure to ozone and low temperatures. The textile or steel reinforcement must be coated with a compound, which must bond to the textile or steel and also to the cover compound and not delaminate after continuous flexing, much like a tire. For use in normal conditions these needs are met by NR, SBR, or blends of NR, SBR, and BR. There are also many special situations where resistance to oils or solvents of various kinds may be required and appropriate compounds must be chosen. This can be complicated in extremely cold climates, as found in the oilsand mining industry, where special low ACN NBR compounds are often chosen as giving the best compromise of low temperature flexibility and oil resistance. Flame resistance is often required by various regulators, such as MHSA and special compounds must be used to meet their requirement. Oil resistance and flame resistance are often needed, along with normal belt requirements. Neoprene or NBR / PVC special compounds are often used. High temperatures are sometimes encountered and specially compounded EPDM's are often used. Some belts are used to convey material up steep grades and these belts will often have molded in cleats or nubs to provide the lift, and molded in flanges to contain the material. These flanges are subject to more strain, than the main cover of the belt, and must be designed to prevent flex cracking. Special requirements of other conveyor belts are also encountered and must be compounded for. EPDM has been used successfully at 150°C.2' These and other belts may be required to be electrically conductive to prevent static build up and sparking. This is usually accomplished by using special carbon blacks. Belts are also used for filtering applications. Horizontal filter belts are endless belts with grooves leading to holes in the center of the belt. The belt supports a filter cloth and a vacuum is used to draw fluid through the filter cloth and the belt. The fluids are often fairly strong acids or bases, at elevated temperatures, so chemical resistance is often the main criteria for choice of an elastomer. Belts are usually designed for a particular application and immersion testing of various elastomers is often performed at the users site before an elastomer compound is chosen.23

1941

Cables Low voltage mining cables generally contain two elastomeric components, namely, an insulation and a jacket. The insulation must be specially compounded for electrical properties, usually using mineral fillers, such as calcined clay treated with silane coupling agents. The jacket must protect the rest of the cable components. Compounds tend to be similar to oil and flame resistant belt compounds, except that they also must have surface resistivity (non-conducting). High voltage mining cables also utilize two electrically conducting elastomeric compounds that are extruded over the conductor and insulation and are part of the electrical shielding system.” Pumps There are many uses for elastomer lined pumps in mineral processing. Vertical sump pumps use hand lay compounds extensively on the submerged portion of the pumps, such as the pedestal and the piping. Chemical resistance is usually the main basis for the selection of elastomers. Inside the pumps, the liners and impeller are often molded elastomers, chosen for abrasion resistance as well as chemical resistance. The mining industry’s continual request for larger processing equipment has placed new design demands on suppliers of heavy-duty slurry pumps. Such pumps (mostly horizontal) are often selected for some as the toughest jobs in a hard rock concentrator, such as mill discharge, cyclone feed or tailing transport.” Soft natural rubber is usually selected in purely abrasive application. Larger particles are handled even better with compounds which combine the tear strength of a harder compound with the resilience of a pure gum.’ Slurry pumps usually have thicker liners than other elastomer lined pumps as experience has shown that when lining thickness is increased by a factor, wear life is increased by several times that factor. Elastomer compounds can be bonded to materials such as fiber glass reinforced plastic (FRP) or thermosetting phenolic / nylon cloth to stiffen liners to prevent collapsing during disruptions such as cavitation or surges. They can be bonded to ceramics to take advantage of the best of both materials. Materials other than natural rubber may be used, usually for fluid resistance, much the same as previously described for other applications. Positive displacement pumps are used in feeding autoclaves in the pressure oxidation of gold and other ores. The valves are a critical item in these pumps and subject to abrasive and corrosive slurries, and flexing at very high temperatures (up to 210°C). Work is being done (by this author and others) to develop proprietary compounds for this extremely demanding application. Wear Liners Various items that can loosely be called wear liners are extensively used. These include liners for chutes, mine cars, skips, crushers, feeders, launders, and others. Abrasion resistance is usually the most important consideration. This is affected by factors previously discussed and also by the impact angle. Most abrasion damage is dome by the shearing force whose vector is parallel to the rubbers impact surface - particularly at impact angles under 30”.” Screens Rubber screens are used for scalping and dewatering. Scalping screens are usually molded with steel reinforcement. Other screens may have a textile reinforcement. They can be molded in large sheets and the holes can be cut using a high pressure water jet to prevent “hourglass” shaped holes, which can occur from die cutting. Rubber’s major advantage over steel screens is greater life span, while other advantages include reduced “binding” or clogging, a noise level reduction of up to 15 dB and increased life of the screening unit.’’

1942

Hose Much hose is specifically designed for mineral processing, especially those for handling slurries. The tubes tend to be similar to pump liners and other abrasion resistant products. Covers tend to be very similar to those used as cable covers. Froth Flotation Elastomer compounds are used extensively in froth flotation units which allow mining of low grade and complex ore bodies, which would have otherwise been regarded as uneconomic. In earlier practice the tailings of gravity plants were of a higher grade than ore treated in modern flotation Launders and tanks are rubber lined, as well as star rotors, dispersers and hoods. Natural rubber is usually used and the life of the parts is usually measured in years because abrasion in the relativity slow moving rotors is not too severe. Other elastomers are chosen, usually for oil resistance. These include neoprene, NBR, and urethane although urethane has had some problems with hydolysis in long term service. The units tend to get bigger for greater efficiency. Star rotors are now being transfer molded over 1.lm using over 200kg of rubber compound.

Handlay Linings Important applications where rubber is hand laid, in uncured sheets, and bonded to metal during vulcanization, include pipe, fittings, filter drums, and tanks. Numerous other products can also be hand laid. For pipes and fittings, pure gum natural rubber is usually the choice, especially for tailings, where it typically gives service for many years. It is very easy to install. Thickness can vary and up to 25 mm is common. Double layers or more can be applied as almost every handlay job is a custom design and is very easy to change, because no expensive molds are involved. Moderate oil resistance can be achieved in natural rubber compounds by blending with oil resistant polymers. Other polymers such as neoprene, CSM or CIIR are used for chemical resistance. Tanks and filter drums are often lined with chemical resistance in mind. Soft natural rubber is still preferred if conditions warrant, as it is the most economical. Hard rubbers (ebonite) or combinations of soft and hard rubbers are sometimes used for heat resistance and chemical resistance, sometimes with graphite fillers, specifically for chlorine resistance. Since ebonite is hard, and not a true elastomer anymore, it can crack due to impact or drastic temperature changes. Various other elastomers are used for chemical (often acids) resistance, with the most common being neoprene, CSM and CIIR. CIIR is often used with natural rubber tiegum to make the lay up easier and the bond better. Modem adhesive systems have been specially designed for hand lining.26They usually consist of a metal primer, an intermediate coat, and a tacky coat to hold things together before and during cure. Rubber tearing bonds are usually obtainable. If possible, an autoclave cure using steam pressure to give temperatures of approximately 140-150"C, is preferred for maximum adhesion, both rubber to metal and rubber to rubber. Demand for pipe over 18m long has led to the construction of very long autoclaves. If the parts are too large for an autoclave an exhaust steam cure is usually the next choice. The cure time is long and there is more tendency to get blisters, and therefore to have minor repairs to the lining. For on site lining where steam is not available, a chemical cure (which involves toxic and flammable chemicals) or a very fast curing compound, which cures at ambient temperatures, can be used. Those systems tend to be much more difficult and therefore expensive to apply. Grinding Circuit Applications The following sections, based on a recent paper by Leon E. Schaeffer and Hans Weinand?' will discuss the detailed applications for rubber in a typical grinding mill circuit with emphasis on the use of rubber in the mill itself.

1943

In the 50’s and 60’s rubber was installed in grinding mills on a trial and error method. This has changed in the last three decades of the 1900’s with use of computers. In the future rubber will replace more traditional metal liners. There are industries other than mining which use grinding mill circuits, such as the power industry. Computers will assist in assuring that rubber is successful in the expanded use of rubber in these circuits in the year 2000 and beyond. Grinding Mill Circuit (closed circuit) A grinding mill circuit normally starts with a coarse crushed feed material and finishes with a predetermined particle size, which is referred to as the liberation size in a mineral processing plant. The complete circuit uses rubber as an acceptable lining material. The coarse material that is referred to as ‘feed’ is directed to the grinding mill via a rubber lined feed pipe or feed chute. Water is added at this point or prior to this point to make a slurry. The feed pipe enters the center of the feed end of the grinding mill and requires a sealing arrangement to prevent slurry from leaking from the mill. There are several designs available utilizing rubber. From this point the feed material goes through a rubber-lined trunnion. There are also several designs to consider here. Some are better for rubber. After this the feed material enters the rubber lined grinding chamber of the mill. After some retention time the ground feed material, which is now called ‘product’, discharges through a rubber lined discharge trunnion liner. From here the product passes through a discharge trommel screen with nominal lOmm slots. The product passes through the opening in the trommel screen and only a small amount of tramp or oversized material is discharge from the circuit. The minus lOmm product then passes by gravity into a rubber-lined sump and is pumped to a rubber-lined cyclone via a rubber-lined pump. The fine material or product passed on for use or further processing while the coarse fractions pass back to the feed material of the circuit for further reduction. The connecting piping is also rubber lined. In this typical circuit, as outlined above and shown in figure #1, all components are rubber lined but other lining materials could be used. Feed chute or pipe Feed material is fed to the grinding mill via a feed pipe or chute. In a fine or medium grinding application a feed pipe is used and is lined with 40 to 50 durometer natural gum rubber. The rubber lining has a nominal thickness of 12mm with a double thickness on the outer radius of the pipe to take impact and sliding abrasion. In some cases a molded urethane liner is used especially if it is a secondary or tertiary grinding circuit. In primary grinding circuits, such as autogenous or semi-autogenous grinding, a heavyduty feed chute is used. Here molded rubber bars in the 60 to 70 durometer range are used. Partially worn mill lining lifter bars can be used for this application. A rubber hose can also be used for this application. Mill Feed pipe or chute seal During the operation of a grinding mill, there is some slurry, which splashes at the feed entry area. Some type of sealing device is required here to prevent the slurry from being discharge from this feed area of the mill. There are several designs that have been used. One design uses a small bucket wheel to lift the slurry up and dump it into a top opening of the feed pipe where it is fed back into the mill. A drip ring is attached to the pipe so slurry in this area drops into the bucket wheel area. A 6mm flat rubber gasket is attached to the outside area preventing any splashing from getting out of the mill. (See Fig. #2) Another method is to attach a rubber or urethane molded wedge shape to the feed pipe. This would match with another wedge surface and create a seal. (See Fig. #3) Still another method is to attach a flat 12mm thick gum rubber gasket with a metal backup ring to the outside of the mill. The inside diameter of flat rubber would be 12 mm smaller than

1944

the outside diameter of the feed pipe. When the feed pipe is pushed through the smaller opening, a seal is created. Since the mill is running at a relatively slow speed, the wear on the outside of the pipe is minimal. However, some mill maintenance people have elected to put rubber on the outside of the pipe as well for extended life. (See Fig. #4)

,

Mill feed trunnion area The feed trunnion area is used to transport the slurry from the feed pipe or chute to the inside of the grinding mill chamber. For years cast metal liners in a horizontal or conical shape have been used with advancing spirals in this area. (See Fig. #5 & #6) A rubber lined trunnion liner of the same design in some cases has lasted 4 to 5 times longer than metal. Rubber in the 40 to 50 durometer range would be used in light and medium duty mills. However, in the heavy-duty autogenous and semi-autogenous mill a 60 to 70 durometer rubber would be used. In some cases because of the mill and piping configurations, a horizontal trunnion liner with advancing spiral has to be used. The spirals create turbulence, which increase the wear. If the arrangement permits, a smooth conical trunnion would be preferred. A rubber lined smooth conical feed trunnion liner would have reduced wear and provide a longer life. (See Fig. #7) Grinding mill liner Rubber is an excellent lining material for grinding mills. It can be used in most grinding mills. The heads of the grinding mills do not contribute to the actual grinding, except if the heads have a large angle. So the rubber head linings are used to protect the actual mill head from wear. Various designs are being used to get the longest possible life. A bar and plate design is common where the bar gets most of the wear. But the bar does create turbulence and it is therefore desirable to have smooth linings around the trunnion areas. The shell or cylinder lining design is very important since it transmits the power into the charge. It is partially responsible for the type of grinding action in the mill. Here again a bar and plate design is commonly used where the shape of each can be modified to get the optimum performance of the grinding mill. (See Fig. #S) The trial and error method has been replaced with computer generated simulations to show the effects of a liner design. Towards the end of the 1900’s research groups and universities got more involved with theoretical and graphical explanations of what actually happens inside a grinding mill. The use of DEM, Discrete Element Method, has helped to determine the forces and trajectories of individual components of the mill charge. This results into a fairly accurate simulation of the actions inside a grinding mill. By using the various programs available, a mill lining can be designed with some assurance of a successful operation, thereby eliminating the previous trial and error method. Currently these programs work in one plane of the mill or 2D, two-dimensional. However, 3D simulations are rapidly being developed and will be used in the years ahead. (See Fig. # 9) The rubber used for these lining is between 55 and 70 durometer. The lower durometer rubber is used for lighter duties while the higher durometer for heavy-duty applications. Discharge trunnion liner The discharge trunnion liner is normally horizontal with a reverse spiral to return grinding media to the inside of the mill. A rubber covered trunnion liner can last 3 to 4 times longer than metal. (See Fig. #lo) Discharge trommel screen assembly The discharge trommel screen assembly consists of a steel frame, some screening material and an advancing retention spiral. In order to protect the trommel frame for corrosion and erosion, it is rubber covered with a 40-durometer rubber in a nominal 6mm thickness or less. The discharge trommel screen is used to protect the mill discharge pump from tramp or oversized material. In some cases it is used to separate the oversized material which is returned for

1945

further reduction. The screen material can be rubber, however urethane has also been used successfully in many applications. The advancing spiral can be a rubber covered steel or it can be made from urethane in segmented components. (See Fig. #11)

Product discharge sump The sump receives all of the product material that is discharged through the trommel screen. The sump is sized to have a certain amount of retention time or surge capacity depending on the system design. A sump will typically have a 12 mm thick 40-durometer rubber lining. A trommel screen normally discharges most of the material at the beginning of the screening area. It is desirable to double the rubber thickness in the sump area where this large volume of material hits the sump surface. (See Fig. #12) Mill discharge pump The mill discharge pump pumps the mill discharge product to hydro-cyclones. The pump can have many different lining materials. The pump casing can be solid metal or have metal, rubber or urethane linings. The pump impeller can be metal, rubber or urethane. There may be a combination of material used inside the pump. For instance, the pump could have a rubber lining in the casing and a rubber, metal or urethane impeller. Each application has to be analyzed for the appropriate materials to get the best economy. (See Fig. #13) Hydrocyclone The cyclone is used to continuously separate the product according to the desired cut in the particle size distribution. The overflow or fine material goes to flotation or processing outside of the grinding circuit. The underflow or coarse particles are recirculated back though the mill with new feed material. The lining materials used in cyclones are rubber, urethane and ceramic. These materials may also be used in combinations. The inlet head could use rubber, the body and cone could use urethane, and the apex could be made from ceramics. The choice of materials is based upon the best wear life and economy. (See Fig. #14) Piping The piping system connects the pumps, cyclones, and return together. Because the slurry is abrasive, the pipe should be rubber lined. This is also an excellent application for rubber slurry hose because of its flexibility. The rubber lining material for both the pipe and hose is normally 40 to 45 durometer natural gum rubber. Some circuits may have chemicals or oils which would require a different type of rubber. Normal Applications for grinding mill circuits The normal application for a grinding mill circuit is the mineral or ore processing plants, which handle iron, lead, zinc gold, copper, silver, phosphate etc. There are many processing plants throughout the world. Metal is still the dominant mill lining material used throughout the world. A lot of these are good applications for rubber. Even thought the growth of the grinding mill circuits is minimal because of the low metal prices, there are many existing metal lined mills, which could use rubber. The big grinding mills which go up to 40 foot diameter are using metal but there are still some portions of these which can use rubber. Applications other than mining Normally, when someone mentions a grinding mill circuit it is thought to be in mineral processing or mining operation. However, with the current and future trends to clean up the environment, power plants are using a grinding mill circuit to reduce the particle size of limestone and mix it with water to produce slurry. The slurry is pumped to a scrubber, which reduce the sulfuric acid in

1946

the flue gas emissions in coal burning plants. Even though the limestone grinding circuit is not in a mining operation, it is a processing operation for reducing limestone and is an excellent application for rubber. These systems are referred to as FGD (Flue Gas Desulfurization) systems. The United States has installed many of these scrubbing systems at coal burning plants over the last three decades. Europe and the rest of the world have installed some systems over the last decade. However, as local air pollution laws become more stringent, more of these systems will be installed in the future and there will be more applications for rubber.

Conclusion With the proper design and rubber selection, rubber can be economical used in many grinding mill circuits. With the use of computers with such programs, as mill simulations, the designs will be economical from the beginning. In cases where the mill grinding circuit is existing, it is advisable to examine the areas where the major wear takes place and redesign the area to reduce wear. This redesigning and the proper use of rubber can make an installation very successful. Mining is the major use of the grinding mill circuit but there are other industries that use this to reduce particle size of materials. The power plant with the FGD system is one of these. These applications will provide some future growth for rubber in the years to come.

ACKNOWLEDGMENTS The author would like to thank the various people in the references for their helpful discussions; Joseph Wagner for help in obtaining literature, Leon Schaeffer and Hans Weinand for permission to use their paper and Tony J. Valdez for putting this all together. REFERENCES 1. McGraw Hill Dictionary of Scientific and Technical Terms, Forth Edition, 1989. 2.

B. Betts and P. Schnarr, 1997, Metals and Elastomers for Pump Application, AICHE, Lakeland FL.

3.

Werner Hofmann, 1980, Rubber Technology Handbook, Oxford University Press.

4.

Science 06/19/99 Vol. 284 Issue 5422, ~ 1 9 8 8 Prehistoric : Polymers: Rubber Processing in Ancient Mesoamerica.

5.

C.M. Blow, 1971, Rubber Technology and Manufacture, Butterworth and Co. (Publishers) Ltd.

6. Robert F Ohm, 1990, The Vanderbilt Rubber Handbook, RT Vanderbilt Co. Inc. 7.

Maurice Morton, 1987 Rubber Technology, Van Nostran Reinhold Company Inc.

8. Product Information, UltrasiICB 7000GR, 3/1999, Degussa-Huls Corporation. 9.

Conversations with Ron Bourgeois and Alex Roudnev, Weir Slurry Group.

10. Joan C. Long, July - August 2001, Rubber Chemistry and Technology, Rubber Division, American Chemical Society Inc. The History of Rubber. A Survey of Sources about the History of Rubber. 11. Conversation with Douglas N.Hartley, Interex World Resources, Ltd.

1947

12. H. Fries, B. Stollfub, Oct 18-21, Rubber Division, American Chemical Society, “Structure and Properties of Butadiene Rubber.” 13. G.R. Hamed, HS Kim, September- October 2000, Rubber Chemistry and Technology, Rubber Division, American Chemical Society Inc. On the Reason That Passenger Tire Sidewalls are based on Blends of Natural Rubber and Cis-Polybutadiene. 14. R.M. Murray, D.C. Thomson, 1963, The Neoprenes, E. I. Dupont de Nemours and Company 15. Conversations with Keith J. Hart, Rubber Engineering. 16. Dr. Ronald R. Fuest, What Polyurethane? Where? Selecting the Right Polyurethane for

Various Applications, Uniroyal Chemical Co. 17. Kenneth R. Oster, March 27, 1995, Rubber and Plastic Nerves, Polyurethanes: Achieving Top Performance, Air Products and Chemicals Inc. 18. Jim Ahnemiller, Nov 1999, Rubber World, An Introduction to the Chemistry of Polyurethane Rubbers, TSE Industries. 19. Richard A. Thomas, Diverse Use in an Expanding Market for Rubber in Mining and Processing, reprinted from Engineering and Mining Journal. 20. John B. Habec, Floyd A. Walker, Oct 23-26, 1984, Rubber Division, American Chemical Society, Improved Durability in OTR Mining Tires. 21. Conversations with Mark Schmidt, Cambelt 22. Daniel H. Jessop Jr., Steve Bunish, Robert M. Wade, Nov. 1985, Rubber World, Elastomers for Power Cables in Mining. 23. Conversations with John Alexander, Rubber Engineering; Jerry Hunt, Eimco Process Equipment. 24. Russel A. Cartes, March 1999, Engineering and Mining Journal, Vol. 200 25. B.A. Wills, 1992, Mineral Processing Technology, Pergam Press. 26. Product information, ChemlokO, Adhesive Guide for the Rubber Lining Industry, Chemical Products Group I Lord Corporation 27. Leon E.Schaeffer and hans J. Weinand, Rubber Applications in the Mining Industry, IRC 2000 Rubber Conference, Helsinki, Finalnd, June 2000

1948

Grinding mill circuit Product

- Recirculating

L

Feed

%

6. Trommel Screen 2. Feed Pipe Seal 7. Sump 3. Feed Trunnion 8. Slurry Pump 4. Mill Linings 9. Piping 5 . Discharge Trunnion I0.Cyclone Figure 1 Feed Pipe Seal

Bucket Wheel Method Figure 2

Vedge Seal Method

Undelsr :edGasket Seal Method Figure 4

Figure 3

1949

Feed Trunnion Liners

Straight Spiral Trunnion Liner Figure 5

onical Spiral Trunnion Liner Figure 6

Typical Mill Liner Section

Figure 8

1950

Smooth Cone Trunnion Liner Figure-7

MILL SIMULATION

Typical Discharge

Typical Discharge Trommel Screen

Trunnion Liner Rubber-covered Frame

Urethane or Rubber Screen Panels

Figure 10

Figure 11

1951

Rubber Lined Sump

Figure 12 Slurry Pump

Figure 13 Slurry cyclone

Rubber Lined, Also Urethane, Metal or Ceramic

FIGURE 14

1952

Plastics For Process Plants & Equipment Guyle W.McCuaig'

ABSTRACT Thermoplastic Thermoset and Dual Laminate piping, ducting, and vessels have been used worldwide, primarily in chemical and electrochemical processes for mineral processing, for the past 40 years. Use of thermoplastics and dual laminates as lightweight and corrosion resistant alternatives to metals particularly in electrochemical and by-product gas cleaning operations, is now standard practice for many liquid, slurry, and gas-liquid mass transfer operations in mineral processing. INTRODUCTION The design and fabrication of Thermoplastic, Thermoset, and Dual Laminate equipment in hydrometallurgy is a major growth industry worldwide (Bertelmann 1992). New Thermoplastic welding standards, such as the North America AWS.G.l.10 and the new Dual Laminate Vessel Standard, American Society of Mechanical Engineering (ASME)-Reinforced Thermoset Plastics (RTP- l), Dual Laminate Appendix M-14, combined with existing European standards British Standards Organization (BS) 4994 and Deutsche International Norm (DIN) 16965 Part Two, allows today's designers and engineers to be confident with their choice of plastic materials for hydrometallurgy processes (See Table 1). Table 1 Existing International Standards for Dual Laminate equipment 1 DIN 16965 Part 2: Type B pipes dimensions 2 DIN 16964: General~qualityrequirements and testing 3 DIN 53769 Part 1: Determination of adhesive shear strength of Type B pipelihe components 4 DIN 16966 Part 2, DIN 16966 Part 4, DIM 16966 Part 5 5 DIN 16966 Part 8: Laminated joints dimensions for Type B piping 6 BS 6464: 1984 - British Standard specification for reinforced plastics pipes, fittings and joints for process plants 7 PRN 88: Swedish pressure piping code for plastics 8 DIN 16962 Part 4-12: Polypropylene pipe and fittings, dimensions, fusion and general quality requirements 9 DIN 16963 Part 4-10: Polyethylene pipe and fittings, dimensions, and fusion 10 DVS 2207 Part 15: Heating element butt welding of PVDF and ECTFE pipe and fittings 11 CGSB 4 1 -GP-22: Canadian Standard for: process equipment; reinforced polyester; chemical resistant, custom contact moulded pipe and fittings (NBSPS-15-69 American equivalent) 12 ASME B31.3-1996: Process piping, ASME code for pressure piping B31.3 13 ASTM C1147-95: Standard practice for determining the short term tensile weld strength of chemical resistant thermoplastics 14 ANSVAWS GI. 10: 2000 -Evaluation of hot gas and heated tool thermoplastics welds 15 CEN (Europe): In process standard for approval testing plastics welding personnel in Europe

1 Prolite Plastics Ltd., Vancouver, Canada

1953

GLOSSARY Thermoplastic Liner

Thermoset

-

-

Dual Laminate

Thermoplastic sheet, pipe or rod used as corrosion resistant liner in Dual Laminate constructions. Injection moulded, extruded or presslaminated from Thermoplastic resins. Thermosetting resin of polyester or epoxy vinylester used to manufacture tanks or piping; mainly used as the structural shell.

-

Appliance manufactured with a Thermoplastic liner fully bonded to Thermoset structural layer.

ASME: American Society of Mechanical Engineers AWS: American Welding Society for Plastics G.1.A Committee. BS:

British Standards Organization

DIN:

Deutsche International Norm

ECTFE: Ethylene Chlorotrifluoroethylene FRP:

Fiberglass Reinforced Plastic

Furan: Thermosetting resin of furfuryl alcohol HDPE: High Density Polyethylene MFA: Methylfluoroalkoxy Fluoropolymer PE:

Polyethylene

PFA:

Perfluoralkoxy Fluoropolymer

PP:

Polypropylene

PP-H: Polypropylene Homopolymer PTFE: Polytetrafluorethylene(teflon) PVDF: Polyvinylidene Fluoride RTP:

Reinforced Thermoset Plastic

PLASTICS FOR PROCESS PLANTS AND EQUIPMENT Thermoplastics, thermosets, and dual laminates (Wegener 1991) are widely used in 3 major areas of solution metallurgy, including a) Hydrometallurgical acid leaching, b) Electrochemical metallurgy, and c) By-product gas cleaning (see Table 2).

1954

Table 2 Process Plant Applications of Plastic Materials Metal Process Conditions Calcine Leach 90" C, H2SO4 Zinc: Electrowinning 70°C, H2SO4

Materials PP-WFRP PP/FRP

70"C, H2SO4 ambient, dilute H2SO4

PVC/FRP HDPE

Leaching

11O"C, HCl

PVDF/FRP

Electrolysis Selenium reduction

80" C HN02 SO2/H2SO4

PVDF/FRP ECTFE/FRP

HCVBenzene

Furan; MFAlFRP

H2SOdJH3P04

FRP;P P m

SOz/H2S04

PP/FRP; P V C m

HC1O.d H7SOA

PVC-FRP

Copper:

Electrowinning Heap Leaching

Magnesium: Silver

Columbium & Zirconium: Solvent Extraction Fertilizer

Phosphoric Acid

Gold Ore Roasting

Flue-Gas Scrubbing

Fume HoodsDucts

Electrochemical processes lend themselves to dual laminate construction due to the ability of the thermoplastic liner to resist attack by the high concentrations of acids used, the ability to resist the permeation of acids (Van Amerongen 1964), and the flexibility to resist elevated temperatures up to 95" C in acid solutions. Dual laminate piping (see Figure 1) can use the strength of the Fiberglass Reinforced Plastic (FRP), which is at the same order of magnitude of some metals to withstand long pipe span distances without extra support, large surge pressures (based on the 1O:l safety factor of the FRP casing), and a minimum expansiodcontraction in a restricted environment such as is found in anchored pipe racks (see Figure 2).

Thermoplastic Liner

Bonding Layer

Fiberglass Structural Layer

Figure 1- Cut-out of dual laminate pipe section to include thermoplastic liner, bonding mechanism (chemical bond or fabric embedded mechanical bond) and FRP structural

1955

MIR FILL LAYERS

THERMOPLASTIC

JOINTLINE

NOTE% 1. Material to be PVC/CPVC, PP/PE. PVDF, ECTFE, MFA 1.1. Liner to be Thermoplastic. 1.2. Structural laminates: 1.2.1. Resin: Premium grade vinylester resin 1.2.2. Cure system: MEKP-CONAP-DMA 1.2.3. Reinforcement: Type E-Glass ECR-Glass 1-112 oz/sq.ft. chopped strand mat (M) Type E-Glass24 ozlsq.yd. woven roving (R) Type E-Glasscontinuous roving for filament winding 1.2.4. Surface coat resin: d w CVeil and UV resistance agent 2. Structural FRP Lamination: 2.1 Filament wound pipe: FW. CV Filament winding continuous roving wound at +/- 55’ to 2.1 .l. thicknesses required 2.2 Hand Laid-up Pipe, Fit!ings and Joints: 2.2.1. Structural hand laid-up layer to be composed of layers of mat and roving to approximatethicknessesrequired

Figure 2: Dual Laminate Pipe Joining Methods Dual laminate vessels (see Figure 3 ) are lightweight when compared to brick or ceramic linings, and can be built in one piece to withstand the corrosion of hydrometallurgical processes at roughly half the capital cost. Recent advances in world standards have brought credibility and safety to many higher temperature and load bearing applications that were once the domain of metals, alloys, and acid brick lined or ceramic components..

1956

Critical services for corrosive media cover an extreme range of chemical compositions in inorganics.

Figure 3: PVC/FRP Dual Laminate Tower One common use of dual laminates in hydrometallurgy is electrolytic zinc manufacture using PolypropyleneFRP (see Fig. 4), which gives excellent corrosion resistance and mechanical properties up to 95" C and is resistant to chlorides (often a problem with stainless steel). PolypropyleneFiberglass Reinforced Plastic (PPFRP), is unaffected by any electrochemical currents in the cells and will handle the abrasiodcorrosion problems of slurries and solids due to the nature of polypropylene liners (which are similar to polyethylene - Conde 2000). The liner thickness can be varied to provide appropriate allowance for erosionlabrasion.

Figure 4: PP/FRP Zinc Cell Feed Tank

1957

Silver electrolysis (and gold parting) in nitric acid again lends itself to Polyvinylidene Fluoridefiiberglass Reinforced Plastic (PVDFFRP) construction with the corrosion of the concentrated nitric acid handled by the PVDF liner and the structural FRP able to withstand the elevated temperature. Strong acid chloride solutions at temperatures up to 110°C are handled by various fluoropolymerFRP combinations as noted in Table 3. At lower temperatures, HDPE and PP are satisfactory. Acid leaching and solvent extraction technologies use a variety of inorganic acids and aggressive organic solvents which are best handled by thermoplastics at lower temperatures or dual laminates at elevated temperatures (see Table 3).

Table 3 Suggested Thermoplastic & Dual Laminate pipe temperature and service ranges Max. Temp Min. Temp. DIN 16964 Bond Max. Temp Min. Material Spec. Service Dual Service Dual Strength N / M M ~ Liner Plastic only Temp. Lam. Lam. PVC 60°C -15°C ASTM 1784 80°C -15°C 7 CPVC 90°C -15°C ASTM 1784 100°C - 15°C 7 PP-H 90°C -15°C ASTM D4101 100°C -30°C 3.5 PVDF 140°C -40°C ASTMD3222 121°C -40°C 5 ECTFE 180°C -76°C ASTM D3275 121°C 5 -50°C MFA 260°C -190°C ASTM D6314 121°C 5 -50°C HDPE 80°C -50°C D1248 85°C 3.5 -50°C PFA 260°C -190°C D3307 121°C -50°C 5 The most widely practiced application of low concentration sulphuric acid distribution leaching is for copper leaching pads, where sulphuric acid solution is distributed over a wide surface area using high density polypropylene pipe and collection of the liquid with a flexible liner of PVC or PE underneath the mass of copper ore. Plastic piping is cost effective in this service (see Figure 5). Plastics find application under reducing acid conditions, such as in SO2 reduction of selenium in copper and silver refining (Table 2) where alternatives (ceramics, exotic alloys) are significantly more expensive. Another major application for PPfiRP vessels is leaching of uranium ore that is generally extracted in sulphuric acid. Historically, wood stave tanks were used in this service. Old plants have been retrofitted with HDPE or PP liners, bolted to the walls. One-piece vessels in PPIFRP construction can now handle this process with a liner thickness designed for the abrasion requirements of the vessel.

Figure 5: Process Piping Cost Comparison for Thermoplastic, FRP, Dual Laminate, Steel, Lined Steel, and Alloys (ls*Quarter 2002 in U.S. Dollars)

1958

In the processing of ores to produce zirconium and columbium a leaching and extraction process using mixed acids and organics such as benzene and toluene is used. This is a very difficult corrosion system for any thermoplastic or thermoset, but the most commonly used material in this case is furan, which is the only thermoset with suitable resistance to aromatic organic solvents as well as inorganic acids. Recent trials have shown fluoropolymers such as MFA, teflon or PFA in dual laminate construction to effectively resist permeation in many lining techniques (see Figure 6).

Figure 6: Categories of Lining Systems (Hall, 1994) Off-gas handling of roasting processes comprises the third major use of plastics in process plants. The most common applications would be zinc or gold processes where SO2 is collected and converted to H2S04 in acid plants. Recently constructed acid plants designed in Europe have used polypropyleneFRP gas ducting and scrubbing towers. Some North American technologies use thermosets of straight FRP construction, with limited lifespan due mainly to SO2 attack. These materials have replaced acid brick and ceramic lined steel towers in most cases. In Noranda's new magnesium production process, magnesium is leached from asbestos waste using azeotropic HCI near 110°C; PVDFFRP ducting is a good material selection with care to be taken to avoid excess permeation. In design of high temperature acid or 2 phase (e.g. H2S04+ SO2) systems care must be taken to review not only the fluoropolymer which would be most corrosion resistant but also the permeation effects on the bonding system. If hydrochloric acid or a mixture of severe gases permeates through the fluoropolymer sheet, care must be taken to choose the correct backing or mechanical bonding cloth. For high temperatures the best choice may be a glass backing rather than a synthetic backing such as polyacrylic nitrile or polyester. Recent research has found this to be an important issue in the long-term service of fluoropolymers in laminate construction as in lining of steel reactor vessels for pressure or vacuum service (see Figure 7).

1959

P

Chemical Resistance Figure 7: Thermoplastic Products- Service Temperature vs. Chemical Resistance CONCLUSIONS Thermoplastic, thermoset, and dual laminate use for equipment and piping has been growing in the world-wide mineral processing industry. Successful applications range from liquid sluny lines to gas transfer lines to high pressure piping and vessels. Plastic materials have non-metallichon-electromagneticproperties, which is important in relation to electrochemical corrosion, grounding, and safety, as well as for eliminating the need for costly instrumentation. In leaching processes, plastics have lightweight properties so that piping can be moved easily in outdoor operations, particularly in difficult terrain. Plastics can resist temperatures as low as -5OoC,and will withstand corrosion and ultra-violet attack for long periods of time. In gas cleaning operations, high temperature fluoropolymers are becoming competitive with lead lined, acid brick lined, or alloy materials. New applications will develop and in response to both developments in material technology and wider recognition of engineering properties of plastic based materials. Selection of plastics for mineral processing applications can now be based on a combination of international standards and industrial practice. REFERENCES Bertelmann, L. 1992: Pipework in Plant Construction Kunstoffe 82 No.6, pp.5 10-514. Dow Chemical 1994: 7" University of Witwatersrand Composites Conference, Johannesburg, Rep. of South Africa. Conde, Maria 2000: Swedish Corrosion Institute - Hydrochloric Acid and Water Permeability in Fluoropolymer Tubes, NACE Conference. Glein, Gary 1996: Metals, Fiberglass, and Thermoplastic Tanks and piping, a review of costs and other decision factors, NACE Conference. Hall, Nelson L.:Dupont Engineering, Using Fluoropolymers to Resist Permeation of Corrosives. Lueghamer, Albert 1977: Polypropylene Comparison of Types PP-H,--PP-B, PP-R - Private Communication, AGRU. McCuaig, G. 2000: Prolite 2000 Prolite Plastics Ltd., Prolam Piping Handbook.

1960

Pankow, Virginia R. 1986: Dredging applications of high density Polyethylene Pipe - Hydraulics Laboratory, U.S. Army Engineer Waterways Experiment Station, Vicksburg, MS 39 180063 1. Schommer, R. 1987: Thermoplastic Liner in Tank Construction - Troplast AG, Troisdorf, Germany. Van Amerongen, G.J. 1964: Diffusion in elastomers, Rubber Chemistry and Technology 37 pp. 1064-1152. Wegener, M. 1991: Strategies for the Correct Design or Components in Fiber Composite Materials, Thesis, Aschen. Weib, Dr. Johann 1999: PVC-C Materials Purity and Stress Relieving - Private Communication, Troplast.

1961

Commercial Acceptance and Applications of Masonry and Membrane Systems for the Process Industries Robert E. Aliusso, Jr.’ Thomas E. CrundulL2David M.Mulone3 and Robert J. Storms4

ABSTRACT The use of ceramic and membrane materials is a well-established engineering approach to corrosion control, with over 100 years of commercial history. The concepts of a composite, corrosion proof, ceramic lining system are reviewed to allow the mechanics of a proper design to be understood. The primary function of each of the composite parts is reviewed in relation to properties of the composite vessel or lining system. A review of design factors to be considered when choosing applications for ceramic and membrane lined process vessels is presented. Past and recent commercial applications of composite ceramic lining systems are presented to illustrate life cycle reliability. The economics of ceramic and membrane lining systems are briefly discussed. INTRODUCTION The use of masonry and membrane lining systems for process vessels is a well-established engineering approach to corrosion control. This approach was first applied in the late 1800s to the digesters of pulp manufacturers in the developing paper industry. The initial and current applications are based on the excellent engineering properties of masonry and membrane materials, including: Chemical resistance Thermal shock resistance Abrasion resistance Good thermal and electrical insulating properties Strength in compression Extreme mechanical durability Low life cycle costs Versatility in physical shape and application. In many cases, a masonry and membrane lining system is used when no other materials of construction can be found to resist the corrosive/erosive environment. The properties of masonry and membrane lining systems have contributed to the commercial viability and growth of many new applications throughout various industries.

COMPOSITE CERAMIC LINING SYSTEMS Before technical and economic discussions can begin, a composite vessel or ceramic lining system or structure must first be defined. The corrosion or abrasion resistant vessel or ceramic lining system consists of three main components: Structural body - the containment vessel; usually metal (always for elevated pressure applications) or concrete for atmospheric operation

The Stebbins Engineering and Manufacturing Company and Stebbins Africa (Pty) Ltd The Stebbins Engineering and Manufacturing Company The Stebbins Engineering and Manufacturing Company and Stebbins Australia Pty Ltd The Stebbins Engineering and Manufacturing Company

1962

Primary corrosion barrier - an organic (e.g. FRPFRF or rubber) or inorganic (e.g. lead or alloy) membrane; and Masonry barrier - the masonry lining (e.g. carbon, caustic resistant, or acid brick). Each layer in a composite vessel ceramic lining system has a primary function that is essential to the success of the corrosion protection. To more fully understand the concept of the composite system, we need to understand the functional purpose of each part. A properly designed ceramic lining system and structure accounts for each component to be designed into a composite system. A complex analysis is completed to confirm that each component of the system operates safely within its mechanical properties. The containment vessel is designed to withstand the hydraulic and physical loads imposed by the process conditions while the ceramic lining system is designed to withstand the chemical, thermal and abrasion conditions of the process. It is important to design the ceramic lining system to account for the interactive forces between the containment vessel and the ceramic lining. Component # 1 - Structural Body The structural body is designed to carry the physical load of the process and of any ancillary equipment. Most commercially practiced design codes for the structural body ignore the structural requirements of a masonry and membrane lining system. Since the ceramic lining system is interactive, it is essential to involve the masonry and membrane designer in the design of the structural body. Details of design and fabrication for the structural body must be altered to accommodate the requirements of the masonry and membrane lining system. These include:

Elimination of flat surfaces and ledges; Restrictions in yieldstrain to comply with masonry and membrane mechanical properties; and Elimination of surface defects and geometry that is not compatible with developing intimate and total contact of the membrane and masonry lining with the substrate.

-

Component #2 Primary Corrosion Barrier Corrosion resistance is predicated on the theory of a tight composite lining system. This theory is dependent on the membrane being in intimate contact with the structural body and the masonry barrier being in intimate contact with the membrane. Any voids between the membrane and structural body that occur provide cells that will result in a pool of corrosive liquid against the structural body, due to osmotic pressure. The primary corrosion barrier is selected to prevent corrosion of the structural body. The membrane can prevent corrosive reactions from occurring by:

Acting as an impermeable layer that keeps the corrosive chemicals from the substrate; Acting as an electrical insulator to prevent the classical Galvanic Cell from forming; Acting as a sacrificial barrier which neutralizes any corrosive chemicals before it reaches the substrate; and Acting as a catalyst that sets up a passivating atmosphere adjacent to the substrate. The membrane choice must be optimized for thickness, cost and durability. In addition, the membrane must be physically tough enough for the masonry lining to be installed without worry of damage. Occasionally, membranes are chosen that require such carefully controlled installation conditions for proper application that the choice becomes impractical. In addition, it can disrupt the work of others nearby due to safety or environmental reasons. It is typical that the nonmetallic membranes are spark tested before ceramic lining application. In addition to known metallic membrane materials such as lead, many nonmetallic materials can function as primary corrosion barriers. Materials such as fibre reinforced plastic (FRP), fibre reinforced furan (FRF), rubber and even Portland cement can have outstanding corrosion resistance when used behind a masonry barrier.

1963

Component #3 - Masonry Barriers The masonry barrier can consist of many materials ranging from very expensive silicon carbide, high alumina, or carbon to cost effective fireclay or red shale. The selection of the masonry material and the thickness depends primarily upon the thermal, chemical and abrasion conditions that exist within the vessel. The masonry barrier is used to limivalter the exposure of the primary corrosion barrier (membrane) from temperature, process chemistry, and abrasion. The masonry barrier performs many functions to help the primary corrosion barrier perform its designed function” 0

0

0

Masonry barriers should be designed to provide thermal insulation. This allows the primary corrosion barrier to operate at temperature levels required to achieve optimum corrosion resistance. Most masonry barriers have low porosity and can be designed to decrease the corrosive chemical exposure of the primary corrosion barrier. The masonry barrier prevents direct and free exchange of the corrosive chemicals with the primary barrier. Most masonry barriers develop controlled forces that push against the primary corrosion barrier (membrane), holding it in intimate contact with the substrate.

Properties of most masonry materials exhibit non-reversible physical growth due to chemical swell. This chemical growth is completely independent of thermal expansion. The chemical growth, in a properly designed composite ceramic lining system, aids in lining stability, when coupled with proper configuration. Chemical growth also eliminates bond dependency of the lining to the substrate, since it provides a radial thrust moving the lining into more intimate contact with the structural body.

HISTORY/APPLICATIONS First used in North America in the 1880s, masonry and membrane lining systems have a long and successful history. Pulp manufacture for papermaking utilized sulfite digesters which operated at high temperature, elevated pressure and contained very corrosive chemicals. In the 1880s, limited corrosion control choices existed to protect mild steel process vessels such as pulp digesters. Lead sheets were the typical choice to protect process vessels from corrosion. However, fatigue often caused failure of the lead. Also, the exposed lead sheets commonly sagged, causing failure. Henry Stebbins, a pioneer in the field of masonry and membrane systems for corrosion control first applied masonry over the lead for support. It was later discovered that this composite lining system offered greater reliability than exposed lead. From these first applications, masonry and membrane lining systems played a role in the growing process industries. In the 1930s, a monolithic chemical resistant masonry and concrete design for construction of process tanks was first applied, by Henry Stebbins’ company. The first monolithic reinforced concrete and masonry tanks were rectangular and had concrete covers. The walls and covers of these tanks became part of the plants’ walls and operating floors. This construction technique is unique in that the concrete shell and the corrosion-resistant lining are built simultaneously without forms. This unique formless construction is suitable for a wide range of atmospheric chemical processes. During the first half of the 1900s process chemistry increasingly became more aggressive as more corrosion control methods became available. Masonry and membrane linings already had more than 50 years of success at this time. Masonry and membrane lining systems allowed many processes to become commercially viable during this period by offering a reliable means of corrosion protection of critical process equipment where no other materials would offer proven reliable protection. The (non-aluminum) extractive metallurgy industry first applied pressure leach autoclaves in 1947 for tin reactors at Wah Chang in Texas City, TX. Again, Henry Stebbins’ company was called upon to perform a design for these “new-age” reactors. Innovative material selection and design techniques were employed based on the 60 years of pressure vessel experience from their ongoing involvement with batch pulp digesters. 1964

A ten-year lapse preceded the next large step in metal refining pressure leach application. In the 1950s, a large complex for nickel laterite processing was planned in Moa Bay, Cuba and Braithwaite, LA by Freeport Nickel. These plants brought new challenges for masonry membrane lining design. All critical process equipment and nearly all process vessels were equipped with masonry and membrane lining systems for corrosion control. Autoclaves operating at temperatures in excess of 400" F and pressures over 500 psi were designed. The original masonry linings were installed in 1959. Many of the vessels have their original base course linings and operate today! Needless to say, the design and material selection was a great success. Over the last 20 years, many pressure leach processes have been employed for gold, nickel, copper, cobalt, and other metals. In many of these applications, masonry and membrane lining systems have been chosen, and have shown very high reliability when properly designed. During the same period, atmospheric metal leaching has also grown. Masonry and membrane linings or monolithic construction continue to play an increasing role in corrosion control techniques for these atmospheric applications. There are existing applications where 20+ years of maintenance free service have been demonstrated.

PRESSURE VESSEL COMPOSITE DESIGN Tight vs. loose design concepts Now that we have established all of the components of the composite, it is important to note the different design concepts and the reliability comparisons for different applications. For brick lined vessels in general, two design methods exist - tight and loose linings. The tight lining design is the most reliable corrosion control design for hydrometallurgical systems. A loose lining is common to dry, high temperature refractory, where voids are made part of the lining design to allow for thermal expansion. These voids are detrimental to long term lining success for pressure vessels containing corrosive liquids. Unlike high temperature refractory materials, corrosion resistant masonry exhibits chemical swelling, in addition to reversible thermal expansion. Chemical swelling is irreversible and is a function of the raw material and manufacturing process of the finished masonry product. The most common masonry used for corrosion protection in hydrometallurgical process vessels is a fireclay acid brick, although graphite, carbon and specialty masonry are also specified and used. Fireclay acid brick can be classified as standard duty acid (SDA) brick or pressure vessel grade acid (PVGA) brick. SDA bricks are usually tight-bodied material that exhibit very good acid resistance and high chemical swell rates. These materials are very resistant to wear but commonly spall when used in pressure vessels due to the high chemical swelling rates. PVGA bricks offer very good acid resistance but have low chemical swelling rates. When used in pressure vessels, PVGA brick materials have shown excellent spall resistance when compared to SDA brick. PVGA bricks are used only in a tight-bonded lining system in which controlled swelling allows the entire masonry to be supported under low stress and remain stable. Membranes are typically noncompressible and do not flow under pressure. Brick spalling is typically insignificant. Membranes, since they are better protected, offer longer service. Process inputs to pressure vessel design Successful applications of pressure hydrometallurgy require effective collaboration between the process design team and the vesselflining design team. Process inputs to design of a pressure leach vessel, often in the general form of a performance specification, will typically include: Slurry flowrate and retention time - to determine process dimensions of vessel. Operating pressure and temperature range - to determine mechanical design of vessel. Chemical conditions - to determine lining components and nozzles design. Number and relative size of compartments General and specific gas-liquid mass transfer requirements (e.g. tonnesh of oxygen, agitator power, dimensions, weight) Temperature control strategy (e.g. flashhecycle vs. dilution water cooling for exothermic feeds). 1965

Once established by the process design team, this data provides the fundamental design criteria that the vesseVlining design team must integrate into the design of the vessel, lining, partitions, and nozzles.

Pressure vessel design requirements When lining a mild steel pressure vessel with masonry and membrane systems, owners and consulting engineers must pay close attention to coordinating the design requirements of the masonry and membrane lining with those of the steel vessel. Depending on the lining contractor’s material selection and process conditions, various special requirements may exist. It is very important that the steel is not purchased ahead of lining contractor’s completion of the masonry, membrane and nozzle designs. The designs should be completed by a lining contractor that specializes in corrosion consulting which is focused on the design, production and installation of masonry and membrane systems. This approach has a proven high success rate in overall lining reliability, performance and economic optimization. Until the lining thickness can be determined and the nozzle designs are completed by the lining designer, the steel dimensions will not be known or optimized. DEVELOPMENTS Continuing areas of general development in pressure hydrometallurgy are identified and briefly discussed as follows: Masonry The most significant general developments have been in the production of consistent quality PVGA. This material was initially developed over 70 years ago for batch pressurized processes in pulp manufacturing for papermaking. There is an extremely high success rate when utilizing PVGA for pressure vessel applications in mineral processing. Nozzle Designs A variety of changes have taken place in this critical area. New seal arrangements have been developed for high pressure sealing. New materials have been tested and used to allow better corrosion resistance and thermal protection of the membrane under newer, high process temperatures. Organic Membranes Lead has been a common choice as the membrane in pressure vessels, with some alloy used in nozzle areas. However, new high temperature, high chloride or other high severity processes may make lead an inappropriate selection. New masonry and membrane lining designs which utilize organic membranes with excellent resistance to high temperature and chlorides are being utilized in pressure vessels. The organic membranes should be carefully selected to conform to a tight lining design so as to remain hard, well bonded and not flow under hot conditions to form voids. Vapor Zone Mortars In pressure oxidation autoclaves, silicate mortars have a short life in the vapor zone before repointing is required. The use of inorganic lead based mortars has solved the repointing problem, but poses new challenges for safe handling during installation and especially removal when dust is unavoidably generated. Owners and consulting engineers must look carefully at a lining contractor’s safety record and performance history when handling lead. If handled properly and installed correctly, low maintenance can be expected through the life of lead based mortar applications With newer higher temperature processes, especially for nickel laterite production, more temperature resistance is required for mortars. Commercial applications exist where new vapor zone materials are used in acidic conditions in excess of 550” F.

1966

COMMERCIAL APPLICATIONS When an owner or consulting engineer performs a feasibility study, various potential applications exist for masonry or membrane systems. Applications requiring consideration of masonry options include: Pressure Leach Applications High (>230”C) pressure acid leach (i.e. laterites) Sulfide pressure oxidation (high, intermediate or low severity) Steam recovery (“flash and splash”) and final (flash) pulp discharge Precipitation reactors Atmospheric Operations Mixed acid or otherwise highly aggressive chemical conditions Insulation for endothermic reactions Example - Pressure Oxidation of Refractory Gold OresKoncentrates Pressure oxidation of refractory gold ores and concentrates has been a major source of new gold production since commissioning of Homestake’s McLaughlin mine autoclave circuit in 1988; over 20 autoclaves have been installed for similar service worldwide. All of these vessels (and many of their auxiliary flash and heat recovery units) employ masonry/membrane/steelshell construction For competitive reasons, the authors are not prepared to present current design information. Readers requiring further descriptive information are referred to the open literature (Thomas, 1994 and cited references) with the caution that all areas of masonry/membrane technology continue to evolve rapidly.

-

Example TiOz Ore Autoclave In the beneficiation of ore for one manufacturer’s chloride process for the production of pigment grade TiOz, the ore is upgraded from 60 to 95% TiOz by leaching with hot azeotropic hydrochloric acid. The process unit for this reaction is a rotating, nonmetallic membrane and ceramic lined mild steel autoclave. The reduced ore is fed to the rotating autoclave followed by a charge of hot HCl (18 %). The contents are steamed up to 148.9”C (300°F) at 206.9-275.9 kPag (30-40 psig). The ceramic lining has lifters (brick ledges) that cascade the ore as the autoclave rotates. This action causes a slow mixing to promote the dissolution of soluble metallic compounds that are removed in solution to upgrade the ore. The movement of the ore creates a highly abrasive and corrosive condition. Few metals can survive in this process. This process has been successfully contained in a spherical ball autoclave, lined with an acid resistant nonmetallic membrane and a two course brick lining. In this application we have all of the basic components: The carbon steel spherical pressure vessel (structural body); The membrane (primary corrosion barrier); and The ceramic acting as a thermal, chemical, and abrasive shield for the membrane. None of these materials could limit corrosion by itself. In combination, they function as an economical solution to contain a highly corrosive and erosive pulp, which allows this process to be commercially viable.

General Examples Over the last thirty years new processes involving that composite masonry and membrane lining systems have shown great success. Notable processes include flue gas desulphurization (FGD) systems for fossil fired power plants, atmospheric leaching reactors for metal refining and pressure leach autoclave circuits for metal extraction. The majority of pressure leach applications throughout the world have used masonry and membrane lining systems with excellent success. Due to this success, many of the same companies are now utilizing masonry and membrane lining systems in their atmospheric leach vessels.

1967

COMMERCIAL BENEFITS Several significant factors make ceramic lining systems commercially beneficial to various process industries. Process Flexibility During the pre-feasibility and feasibility stage of a project, process characteristics and equipment costs are initially reviewed. It is typically at this stage of the project where various types of corrosion resistant liners for the critical vessels are considered. Of the various lining options available, composite ceramic lining systems offer a high degree of process flexibility. This is mainly due to the ceramics ability to be tolerant of significant changes in process conditions. Fundamentally, ceramics have excellent engineering properties that allow them to perform in many different process applications. Ceramics are: abrasion resistant, chemically resistant, thermal shock resistant, very strong in compression, extremely durable, economical compared to other materials and provide good thermal and electrical insulation. The unique physical characteristics of many types of ceramics such as acid brick, caustic resistant brick, carbon brick, silica brick, silicon carbide brick, zirconia brick, borosilicate brick, high alumina brick, porcelain brick, and alumina silicate brick make ceramic linings extremely versatile. Due to the wide range of masonry and membrane material choices, there are very few situations for which a composite masonry and membrane combination cannot be engineered to allow the operator a high degree of process flexibility. Also due to the engineering properties, a masonry and membrane combination can be reliably applied to contain very severe corrosive/erosive conditions. Vessel Reliability Ceramics tolerance of variable process conditions is a significant factor in providing reliability. Most ceramics, even if they are not completely resistant to a specific set of chemical conditions, function as a sacrificial lining with a useful life while preventing catastrophic damage to the substrate. This is due to the large mass of chemical resistant material present and the multicomponent nature of the lining.

Economic Impact Similar to process requirements, which are established during the pre-feasibility and feasibility stage of a project, the economics are also investigated and defined. During this stage the cost of the major equipment and vessels is estimated. The feasibility of a project depends greatly on favorable economics for capital costs, and expected operating and maintenance costs. In reviewing cost information, ceramic lining systems can range from under US$20/ft2 to over US$250/ft2depending upon the vessel and its operating conditions. Overall economics have been strongly in favor of ceramic lining systems for many applications. CURRENT OPPORTUNITIES Many opportunities exist for plant designers and the operators to save capital and maintenance costs by using composite ceramic linings systems more extensively in their process vessels. A comprehensive technical and economic review of alloy, ceramic, and other corrosion resistant technologies available for each project is a sound business approach. Considerable savings in life cycle costs of capital equipment may be identified during the early stages of a project by scrutiny of the materials of construction. CONCLUSION It should be considered basic philosophy that all materials including ceramics, alloys, rubber and coatings are designed and selected based on their ability to resist the chemical and physical conditions of a particular process. We hope we have provided basic information to understand the engineering design and material selection parameters for composite ceramic linings and structures.

1968

The selection, and structural analysis of composite ceramic lining systems for atmospheric and pressure vessels is a well-established and proven science. Design criteria have been proven by the test of time and the actual in-service performance has been demonstrated. Because of many years of development, there is now in service a wide range of composite masonry and membrane lining system applications. However, today’s engineering graduates have not been exposed to the above types of materials and construction, due to the lack of standard textbooks that deal with these subjects. We must continually strive to increase our knowledge of all the available choices to develop the most economical solutions to our corrosion problems. The use of composite ceramic lining systems provides the engineer with a reliable method for providing an economical means to contain the most aggressive chemical processes. Ceramic linings and structures are very versatile, and usually competitive with alloys, rubber linings and high grade coatings, when evaluated on a life cycle basis.

REFERENCES E.F.Tucker. January 1950. Digester Linings for Soluble-Base Sulphite Pulping. TAPPI Vol. 33 No. 1 Beaumont Thomas. March 1950. Non-Metallic Lining Materials for Process Vessels in the Pulp Paper Industry. Corrosion R. Hancock, D. Malone, and G. Charlebois. November 1990. Design and Quality Control of Composite Ceramic Lining System for Corrosion Control. Proceedings NACE, Canadian Region, Eastern Conference Gary W. Charlebois. January 1991. Chemical Resistant Ceramics for the Process Industries. Materials Performance Vol. 30, No. 1, p. 71-75. David J. Malone, Robert J. Storms, and Thomas E. Crandall. June 1995. Commercial benefits of ceramic lining systems used in atmospheric and pressure leach vessels. Proceedings ALTA Conference. Hydrometallurgy 39, p. 163-167. The Stebbins Engineering and Manufacturing Company, et al, unpublished internal research and memorandum. The Stebbins Engineering and Manufacturing Company, Watertown, NY USA R. Flynn. 1981. Construction, Inspection and Maintenance of Tile Tanks and Linings. Proceedings TAPPI Engineering Conference. Book I1 Gary W. Charlebois. January 1991. Chemical Resistant Ceramics for the Process Industries. Materials Performance Vol. 30, No. 1, p. 71-75. Robert E. Aliasso. September 1996. The Use of Ceramics as a Highly Reliable Means of Protecting FGD Equipment. Proceedings NACE Northeast Region, 341hAnnual Corrosion Conference K. G. Thomas, Research, Engineering Design and Operation of a Pressure Hydrometallurgy Facility for Gold Extraction; CIP Gegevens Koninkluke Bibliotheck, Den Haag; ISBN 0-96980670-1, 1994.

1969

The Development of an Electric Power Distribution System Malcolm N. Brodie. P. Eng.'

ABSTRACT The planning of an industrial electric power distribution system involves several stages of study. The first stage is establishing the peak power demand of the projected facility. The second stage is establishing an acceptable source for the power. The third stage is the development of the economic trade-offs for first costs and losses, first costs and the value of any loss of production following a first electrical failure, and first cost and the value of any loss of production that may be attributable to unsatisfactory power quality. The final stage is the development of a composite that incorporates all of the requirements and choices that have been established in the initial stages.

INTRODUCTION The successful exploitation of an ore body depends on a number of factors including the development of a suitable electrical source and distribution system. This in turn requires knowledge and consideration of the size of the intended operation, the planned operating life, the cost of power, the cost of down time, and the availability and cost of capital. These different items should be evaluated first in isolation then in concert.

DEVELOPMENT OF THE ELECTRIC POWER DISTRIBUTION SYSTEM The electric power distribution system must satisfy a number of criteria. Those to be examined include: the capacity to allow all process equipment to operate in concert as well as independently conformance with the requirements of the inspection authority having jurisdiction (ANSI, CSA, etc.) minimization of the energy losses ease with which the system can be maintained minimization of the down time following a first failure The concept of a single source and a multitude of diverse loads leads to the development of a radial distribution system. This is the normal system starting point and it is modified to suit individual loads, the physical location of the loads, and to enhance the security of selected parts of the system. The balancing of the incremental cost of increased reliability, and the probability of an outage with the attendant loss of production, is commonly done using statistical data and weighting factors based on experience and judgement. These can be assigned for each system component to achieve a numerical comparison. As part of such a study, the agreement on values to be used for short interruptions, scheduled interruptions, and deferred production is important because production lost during or because of

' Sr. Staff Consultant, Electrical, Fluor Daniel Wright Ltd.

1973

any interruption is effectively deferred until the end of mine life. In addition, where mine life is expected to be short in terms of the normal life of electrical equipment, the requirement for preventative maintenance is very limited. As an example, a high voltage power circuit breaker has a mechanical operation expectation that could easily exceed the expected life of the mine. The same circuit breaker has a fault operating limit that restricts it to relatively few operations at rated fault current. In this application, a circuit breaker might not see a rated fault current operation during the life of the mine. The incorporation of stand-by generators for critical process areas and Uninterruptible Power Supply (UP S) equipment for critical control, protection, and communication functions facilitates re-starting following an interruption. The next stage of development of the system must look at the effect of a first fault anywhere in the system, and the acceptance of the consequences or the selection of suitable measures for mitigation. As an example, a transformer failure could, of itself, cause an outage of a couple of months in the area it supplies but the availability of a spare would reduce this to a day. Where a day is considered unacceptable, the next option is a parallel unit that can carry the combined load at its ONAF 65OC rating. This would bring an outage down to a few hours with a small increase in losses. If this is still considered to be unacceptable, the next stage of development requires automatic switching which should reduce the interruption to momentary. The assignment of the costs of each type and duration of outage must recognize both the loss of income and the avoidable and unavoidable costs. This information is usually provided by the mine in question. The statistical availability of the various items of electrical equipment in the envisaged configuration can be taken from historical data and adjusted to suit the particular application. Where guidance from the mine is limited, the selection of a system configuration usually starts with a simple radial system then, as required by judgement, sufficient extra equipment is added to limit any loss from a first fault to an amount considered appropriate for the service. To ensure that the basis and the expected result are understood and agreed, it is most desirable to have the pertinent factors tabled. There are two system frequencies in common use in different parts of the world. The north American practice of using 60 Hertz (Hz) results in motors and transformers that are smaller, lighter, and less costly than the corresponding 50 Hz equipment that is used in Europe. Other parts of the world have variously followed one practice or the other. In many cases, consideration should be given to the use of 60 Hz equipment even in an area where 50 Hz is present, except when political influence mandates otherwise. There are standardized voltage levels in both systems that are related to both the size of an individual load and to the aggregate load in a small area. In general, both systems describe V<1000 volts as low, 100015 kV as high. In both systems loads will be served at the lowest voltage that does not result in more than 500 A. Similarly, when a load would exceed 1000 A it would normally be transferred to the next higher voltage. In between it could be either depending on the type and size of the adjacent loads. The most common 50 Hz low voltage is 400 V, although some installations are now using 690 V. This does not affect either motors or control but allows a significant saving in conductors. The most common 60 Hz low voltages are 480 V in the United States and 600 V in Canada. Again the 600 V level results in a significant saving in conductors. For very small loads that can be served by a single phase supply, the common 50 Hz voltage is 220 V and the 60 Hz voltage is 1201240 V. The most common 50 Hz intermediate voltages are 3.3 kV, 6.6 kV, and 11 kV. The corresponding 60 Hz voltages are 4.16Y12.4 kV, 7.2Y14.2 kV, and 13.8Y18 kV. In both systems, there are existing systems that operate at voltages different from these, but the contemporary approach for minimizing current is to use a voltage near the upper limit for each class of equipment. The sub-transmission and transmission voltages range through 25 kV, 34 kV, 46 kV, 69 kV, 138 kV and 230 kV. The selection of voltages in these ranges is sensitive to the maximum load

1974

and the distance over which it must be carried. As an example, it would be possible to carry 100 MVA for I mile at 69 kV, but to carry it for 100 miles consideration would have to be given to 230 kV.

SIZE (CAPACITY OF THE ELECTRICAL SYSTEM) The size is directly related to the desired rates for mining and mineral processing. It is affected by the process selected, the mining plan; e.g.,-open pit or underground, the availability of water, the selected location for storage of tailings, and the need for residential accommodation. A common starting point is an allowance of I kVA per ton processed per day for a conventional crushing, grinding and flotation plant. Other plant types; e.g., crushing, leaching, SXEW, can have significantly different values. The process designer will refine the plant requirements to recognize harder or softer ore, simple or complex processes, single or multiple separations and the associated materials handling; e.g., feed, concentrate, and tailings. The requirements for power to provide water, tailings disposal, and residential accommodation will all depend on the site. The full range from least required to most required is of the order of -lo%, to +15%. The need to handle material over a significant difference in elevation can make a further difference. The utilization of variable frequency drives and/or large rectifiers has caused serious power quality concerns because of the generation of harmonics. The effects can be significantly reduced by phase shifting to simulate a high pulse number. If the basic unit is a full wave three phase switching assembly, the current will have the classic 6 pulse form. Selected phase shifting causes the associated harmonics to add vectorially instead of arithmetically. The effect of this shifting with similar loads is the capping of the sum near the value of a single unit for any combination from one unit to all units. The options are:

Table 1 Options for phase shifting No. of units Pulses 1 6 2 12 3 18 4 24 5 30 6 36 7 42 8 48 9 54 10 60

Shift, O 0 30 20 15 12 10 8.6 7.5 6.7 6

As an example, a system based on 48 pulse ultimate development could start with two units at 0 f 7.5" and have 0" as a common spare. It could add 30" and 30 7.5" and finish with 0 & 15" which would give 7 units with a common 0" spare. If 7 units or less are installed, the spare can replace any transformer with little effect on the system. If 8 units are installed, the spare will match one other unit and, unless that is the unit being replaced, will cause an increase in the harmonic level. This will necessitate replacement by the original unit to bring the harmonic level back to the original level.

OPERATING LIFE This is the basic design life in years. It can range from as little as five years to more than fifty years. It is normally selected by the Owner to optimize his return, having in mind his knowledge of the ore body and his assessment of the potential for future political interference with his operation.

1975

COST OF POWER The cost considered here is the cost of power available at the intended plant. It may be either purchased from an existing source or produced on site. If purchased, the cost will be the purchased cost plus the cost to amortize the cost of connection, plus the cost of the losses that occur in the connection. If produced on site, it will include the cost of fuel at the site plus the cost to amortize the generating equipment cost and the cost to operate it. In either case, consideration must be given to the effect of inflation over the life of the plant. COST OF DOWN TIME This cost is the cost that results from a first failure. It is not the cost of routine scheduled down time that is a factor in plant annual capacity. The cost to be used is usually taken as a combination of foregone income, plus the unavoidable continuing costs, plus the costs associated with the restoration of the damaged component. COST AND AVAILABILITY OF CAPITAL The availability of capital is normally the Owner's concern. The cost of capital is normally defined as the rate of return desired by the Owner. CONSIDERATION OF FIRST FAILURE Electrical equipment when properly selected, installed, and maintained, routinely exhibits acceptably long life. In spite of this, failures do occur and the consequences must be evaluated so that appropriate insurance spares can be kept on hand. The times to restore service that follow are offered only as a guide: the particular circumstances of each installation, including site conditions, site labour skills, and the presence of spares must be part of the evaluation if the financial consequence of any first failure is to be minimized. The worst failure is the loss of the source. Where power is generated on site, the source should include sufficient engines to carry the load plus two additional units. This allows for a unit as spare while a unit is being overhauled. Where power is purchased from a remote source, the receiving transformation should be able to carry the full plant with one unit out of service. As a minimum, with two transformers each should be able to carry the full load at 65OC rise with one stage of forced cooling. The working transformer in Figure 1A will be fully loaded and working at rated temperature. A spare must be rated for the full capacity so the combined capacity is 2.0 p.u. Failure of the unit in service will cause an outage of a few days. The transformers in Figure 1B will each be loaded to 0.75 p.u. of its self-cooled rating and will be working at about 0.6 p.u. of its rated temperature rise. This will statistically increase the life of each by four times. No additional spare is required so the combined capacity is 1.5 p.u. Each unit would be rated at 1.0/1.5 p.u. kVA, ONAN/ONAF, 55/65"C rise. Failure of either unit will cause an outage of about one day. The transformers in Figure 1C will be operating as in Figure 1B but the addition of the secondary switchgear would reduce an outage to about 4 hours. These approaches do not repair the fault but limit the downtime to that required for switching. Repair of a unique transformer could take 6 or more months depending on the damage and access to plant, materials and labour.

1976

-

Figure 1 Transformer high voltage Overhead lines can be spliced in hours, a pole can be replaced in a day, or a transmission tower replaced in weeks. In most cases a single tower can be replaced temporarily by a wooden structure in days. A failure of a power cable can usually be located and spliced in a couple of days depending upon whether it is direct buried or in ducts. Provision for interconnection can frequently be used to advantage. Secondary substations may be double-ended, switchable between two feeders, or interconnected by secondary cables, depending upon the desired level of protection. The configuration in Figure 2A is a typical radial arrangement. The availability of a suitable spare unit would limit an outage to a couple of days. The configuration in Figure 2B is a typical double-ended arrangement. A transformer failure should be isolated in a few hours. The configuration in Figure 2C is similar to that in Figure 2B but is dependent on the interconnecting cable. A transformer failure should be isolated in a few hours. Both Figures 2B and 2C use extra transformer capacity to reduce the duration of an outage that would result from the failure of a transformer. The arrangement in Figure 2C might allow beneficial location of the two units if the loads form two separate groups.

1977

i1

1.0 p.u.

1.0 p.u.

1.0 p.u. (A)

&.o

p.u.

L

1.0 p.u.

8 5

2.0 p.u.

u J.

1.0 p.u.

(B)

(C)

-

Figure 2 Transformer medium voltage Analysis has shown that cables are the most failure-prone component of a system but also the most easily repaired. Transformers are not readily repaired so they must be backed up by the capacity of another, the availability of a spare, or by interconnection. The consequences of each potential failure must be weighed, the appropriate response selected, and the necessary insurance spares be made available. The most likely failure in a motor is a bearing. If an impending failure is recognized before the bearing collapses, the time to replace the bearing should be a matter of hours. If a bearing is left until it has collapsed, there is a real risk that rotor and stator will meet and the laminations may be damaged beyond repair. The next likely failure is electrical breakdown of the winding insulation. In some cases, a knowledgeable technician can isolate the failure by cutting out the damaged coil and returning the motor to service. This does not repair the damage but allows the motor to continue in service while arrangements are made to either rewind or replace it. It has to be recognized that the modern motor with vacuum impregnated windings is hard to repair with a partial rewind: the coils are so firmly secured in the slots that they are almost impossible to remove without damage. Most control is an assembly of small components. Except in case of a power component failure that causes extensive arc damage, the smaller control components can be replaced in hours. It must be noted that, if there are no appropriate insurance spares on hand at the time of failure, the down time will increase to that required to effect a repair or to arrange for a replacement. For transformers and motors this is considered to be far too long to even be considered. Any first failure should be covered by either excess capacity in another unit or backed up by a spare of equal or greater capacity. The more unique the item is, the more unlikely it is that a reasonable replacement will be available on short notice.

1978

EVALUATION OF LOSSES All energized electrical equipment suffers from losses. These losses cannot be eliminated but can be adjusted for least owning cost for any load factor and period of time. The most direct approach to evaluating the losses is the establishment of the present value of a 1 kW loss under agreed conditions. The factors are: the incremental cost of electricity (e) ($/kWh) the anticipated inflation in the cost of electricity (f) (%/annum) the selected interest rate(i) (%/annum) the period under consideration (ie., project life) (N)(years) the anticipated annual load factor (ELF) (P.u. rated) the anticipated annual load factor increment (g) (P.u. rated) any associated load related losses (e.g., forced cooling) (kW).

The formulae used to give effect to these factors are shown in Appendix 1. An example of the result is shown in Table No. 2 which gives typical default values for the factors and the corresponding present value of a kW for three different power costs.

Table 2 - Present value of losses under selected conditions Condition Default Value Inflation rate 5%/annum (0.05 P.u.) Interest rate lO%/annum (0.10 P.u.) Project life 12 years 0.8 Annual load factor Annual load increment 2%/annum (0.02 P.u.) Cost of Electricity $ /kWh $0.05 $0.07 $0.09

Present Value of 1 kW $3,926 $5,496 $7,067

SELECTION OF CONDUCTOR SIZE There are a number of criteria to consider when selecting conductors including the requirements for insulation and handling. Insulation is rated by the maximum temperature to which it can be continuously exposed, the environment in which it can be used (dry only or wet and dry), and its ability to withstand the activities associated with installation and subsequent use. Minimum first cost tends to favour minimal conductor size which, in turn, favours insulation with a high temperature rating. The insulations most commonly used at this time are cross-linked polyethylene (X-link) and ethylenepropylene- rubber (EPR), both of which may be used in wet or dry locations at temperatures up to 9OoC. The prime requirement of the normal installation codes (ANSI NFPA 70, CSA C22.1) is that the continuous service temperature of a conductor must not exceed that for which the insulation is rated. This gives the minimum conductor size that can be considered. There are three other factors that can have a significant effect on the selection of the size of any particular conductor. The first factor is the length of the run. Voltage drop is proportional to the length of the run and is usually limited to 3% of the nominal voltage for the circuit. As the length increases it becomes necessary to use a larger conductor. The second factor is the value of the losses that will occur in the cable. This is proportional to the length of the run also, but is evaluated separately from the voltage drop because it is affected

1979

by the cost of power as well as by the length. It is commonly found that this alone will justify at least one conductor size larger than would be required to satisfy the temperature rating. The third factor is related to an assessment of the probability of having to increase motor size on any particular drive. An owner may require each motor starter to be equipped with load conductors that correspond to the maximum rating of the starter rather than the initial motor where it is smaller so that each starter can be used with any motor up to its capacity without other change except overload setting. Installations made using armoured cables; e.g., Teck, usually make these comparisons by considering only the incremental cost of the cables because the associated incremental costs for supports and labour are insignificant. Installations made using conduit and wire must consider the incremental cost of the conduit each time a size change is necessary.

GROUNDING AND BONDING Grounding refers to a permanent continuous conductive path to the earth. It must be achieved in such a manner that the path can carry any current that can be imposed on it while limiting the consequent voltage rise. Where the current is intentionally limited, the permitted current must be able to actuate the associated protective devices. It provides a reference for all system voltages when established at the source. Bonding refers to the extension of the ground connection to all non-current-carrying metal parts under the same general requirements. Grounding and bonding conductors are both to carry current only in case of a failure. This is quite separate from a neutral, which normally carries residual (unbalance) current, although the neutral will be connected to ground at its source. The other condition to be recognized is the potential requirement for isolated grounding of some electronic equipment. This can apply to Programmable Logic Controllers (PLCs) and to associated computer installations. It can be effected by using isolating UPS equipment and/or marked outlets, and carrying an insulated grounding conductor from the outlets all the way to the system ground electrode without any other interconnection with the power system grounding or bonding conductors. In effect, this is similar to the required separation of lightning arrester grounding conductors and power system grounding conductors except at the ground electrode. EMERGENCY POWER Emergency power may be required to be available in any case of loss of the normal supply to provide lighting for safety, to prevent equipment damage, or to facilitate restarting of the process. UPS equipment may be used to provide for the continuation of control functions but not the operation of process equipment (see Figure 3(A)). The supply for process equipment is normally provided by engine driven generators that are started automatically on sensing loss of the normal supply. The engines can be expected to pick up load within 10 seconds of start initiation. To do this dependably, each engine must be exercised regularly; e.g., for one hour once a week, with a load of not less than 1/3 of its rating. For small engines the exercise load is usually independent of the normal power system (see Figure 3(B)). Large engines can parallel with the normal system so that the fuel is not wasted. These engines can be started by loss of the normal source, by a peak shaving controller, or by an exercise timer. In the first case, load must be shed to suit the engine(s). In the other cases, it will operate in parallel with the normal source at a preset load level (see Figure 3(C)).

1980

Figure 3 Emergency power INSTALLATION The various codes covering electrical installations and the interpretations by the authorities having jurisdiction allow some optional approaches to the installation of equipment. Motor enclosures range from open to totally enclosed. It is common practice to have all squirrel cage induction motors totally enclosed without regard for the environment. The other motors (wound rotor induction and synchronous) tend to be the larger units and will generally have site specific enclosures. Motor control, both LV and MV, is most commonly arranged into assemblies of NEMA 1 enclosures and placed in electrical rooms that are kept clean and cool. It has to be expected that maintenance will be compromised when control units are mounted outside of an electrical room even when the enclosures are changed to NEMA 12, NEMA 3R, or NEMA 4 as may be appropriate. Switchgear is affected in a similar manner. It has to be expected that maintenance will be best in an electrical room, next best when assembled with a walk-in enclosure, and most difficult with a weatherproof enclosure. Transformers that are large will usually be oil-filled and installed outdoors. These should have valved, removeable radiators and be braced for vacuum filling. They should have a control compartment with a thermostatically controlled heater, a light, and a convenience outlet. They should be installed with provision to contain and cool all of the insulating oil in case of a rupture. Transformers that are small may be oil-filled or dry types. This choice commonly depends upon whether they can be mounted outdoors, the site elevation, and the environment. High current interconnections as between a transformer secondary and the associated switchgear may use insulated bus, cable duct, or cables. All other connections may use armoured cables supported by tray or individual conductors in conduit. The armoured cable known as Teck was pioneered by a mine but has spread to all the other industries including pulp and paper, and petro-chemical, and to commercial work. There are still installations being made with wire and conduit but they are generally limited to either additions or some special circumstances.

1981

Appendix 1 Evaluation of Losses

The incremental cost of electricity is a combination of standby costs and energy costs whether the power is purchased or generated on site. If the power is purchased, the incremental cost is a combination of the applicable rate schedule demand charge, the applicable rate schedule energy charge, and the associated load factor. The relationship is: Incremental cost (e) = (Demand charge) ((Load Factor) (720))-’+ energy charge $/kWh where Load Factor = hours in service per month 720 An example might be:

e = 4.IOl$/kW 0.9 (720)

+ 0.0437$/kWh = $0.0500/kWh

The inflation rate (9 is assumed to be constant and should be rounded to an agreed figure (e.g., 5 %/annum). The selected interest rate (i) may be either the current cost of money or a designated rate of return (e.g., lO%/annum). The project life (N) will be the period selected for the study (e.g., 12 years). The anticipated annual load factor (ELF) affects the combination of no-load and load losses. It is also affected by the design margin for growth. A typical value for an industrial plant might be 0.8. The anticipated annual load increment (g) is assumed to be small but a constant positive value. It might be 2 %. If the device being evaluated is continuously energized but intermittently loaded it becomes necessary to allow for the no-load losses and for the incremental load losses. This is typical of a transformer where the No-Load losses and the incremental Load Losses are identified. To simplify the calculations it is advantageous to combine the interest rate and the inflation factor into the net discount rate (r) where: r = (l+i) ( l + ~ - l1 -

where i and fare per unit

The value of No-Load losses for a continuously energized unit then becomes: 8760 (e) (r)-l( 1 - ( I +r)-N)(NL) The value of Load related Losses becomes: 8760 (e) (ELF)’(( I+&”

( I+r)-N -I) (1 - (l+r) (l+g)-”I (LL)

As an example, the present value of a I kW loss in a power cable using the typical values suggested above would be calculated as follows: r = (1 + 0.10) (1 + 0.05)” - 1 = 0.048

PV of 1 kW = 8760 (0.05)(0.048)-’(1- (I+ 0.048)-’*)(1) = $3926

1982

Selection of Motors and Drive Systems for Comminution Circuits Peter F. Thomas', P. Eng.

ABSTRACT The selection of a mill drive has been complicated by three distinct trends. All mills (Autogenous Grinding (AG), Semi-Autogenous Grinding (SAG), Ball, and Pebble) are getting larger. All process control is becoming more complex. The desired electrical system capacity is frequently not available. AG and SAG mills up to 40 ft (12.2 m) in diameter are in service and designs are available up to 44 ft in diameter (13.4 m). Process requirements usually require adjustable speed drives for these mills. Ball mills up to 26 ft (7.9 m) in diameter are in service. The majority of the smaller sizes operate at a fixed speed but adjustable speed is becoming more common for the larger mills. There are many options available for both fixed and variable speed drives. These options are defined, the features explained, the limitations discussed, and the costs compared to assist with the selection for any particular case. Initial cost, cost of installation, cost of commissioning, and cost of operation and maintenance have to be taken into account for each drive option considered. The impact of each drive system on the design, cost and operation of the overall plant electrical system is discussed. Also, consideration must be given to mill drive availability, and the potential cost resulting from delays in plant production. INTRODUCTION Over the past 30 years a number of papers have appeared discussing mill drives (see references). The primary emphasis has been on AG and SAG mills, particularly the variable speed options for larger mills. Recently, an interest in variable speed for large diameter ball mills has developed. The intent of this review is to present the latest technology considered for AG and SAG mills, and ball mill drives, plus provide some guidelines on when to select fixed and variable speed drives. For the variable speed options, gear and gearless drives are included. Single-pinion fixed-speed gear drives offer the simplest design and the lowest capital and operating costs. In the past these drives were restricted to smaller mills, however recent technological changes in gear manufacturing have extended the range of mill diameters. Similarly, dual-pinion fixed-speed drives offer reliable and simple designs with proven load sharing capabilities. These fixed-speed drive systems can be used on the largest ball mills considered to date. With variable speed, particularly for AG and SAG mills, the drive system became more complex and expensive. In the not too distant past the variable-speed choices were few (namely dc motors), but improvements in solid state technology and controls has resulted in a number of higher-efficiency variable-frequency alternatives that have pushed dc motors out of the picture. Likewise, modern slip-energy recovery systems for wound rotor motors have overcome highenergy losses generally associated with liquid rheostats. 1

Marketing Manager. Mineral Processing, GE Industrial Systems, Peterborough, ON, Canada

1983

A guide to contemporary application practice based on the common North American diametedlength ratios of 2:l for SAG mills and 2:3 for ball mills is shown in Figure 1. MILL POWER VS. DIAMETER

30

-

25

-

I

1 Gearless Range I

3

z

20-

3

15

a-

I

10 5-

o! 10

15

20

25

30

35

40

45

50

DIAMETER, F T

Figure 1 Mill power vs mill diameter

Taken as a whole, the engineer is faced with many choices for a drive system. The selection process involves quantitative (economic) and qualitative (subjective) factors. Choosing the correct factors is difficult and, for most projects, represents one of the most important choices in selecting the proper system. Fortunately, the engineer has a number of analytical tools to resolve the economic issues. Given the correct (and sometimes incorrect) assumptions, some subjective factors can be transformed into economic factors. However, in the end it may be the unresolved subjective factors that will determine the selection, even after exhaustive analysis of the capital and operating costs.

PROCESS CONSIDERATIONS

Why Variable Speed? The operator (or control system) can rapidly react to changes in ore characteristics, be it ore hardness or feed size distribution. Soft andlor fine ore can result in a low total charge volume leading to liner damage and accelerated ball and liner wear. This condition can be corrected by increasing the circuit feed rate only when downstream conditions permit such changes. Otherwise, it must be corrected by reducing the speed of the mill to force the balls to impinge on the charge and not the liners. When grinding out a SAG mill, variable speed is valuable for the same reasons. Variable speed drives also provide the advantages of slow starts (stops) of the mill and, for some systems, inching of the mill and protection against damage by movement of a cemented mill charge. However, for most overflow ball mill circuits, variable speed is only valuable when circuit feed rate control is required downstream. Without downstream constraints, ball mills are typically operated with a maximum ball charge and a fixed speed, thus negating any need for variable

1984

speed. SAG and ball mill circuits are designed such that the circuit capacity is ball mill limited over most of the range of ore hardness expected from the ore body. Most variations are compensated for by the variable speed drive on the SAG mill. In rare cases, the ore may be SO variable that the speed (or power) range for SAG mill cannot compensate for an extremely hard component, resulting in a need to reduce the power (or speed) of the ball mill to avoid over grinding and affecting downstream processes. In a few circuits, the operator can balance the SAG mill and ball mill by directing a portion or all of the crushed SAG-mill pebbles to the ball mill. Other plants allow partial cyclone underflow recycle to the SAG mill to further improve the balance of power. It should be noted that mine planning, ore blending and blasting practices play an important role in the design of milling circuits. Modern plant designs recognize that feed characteristic control is not perfect and that variable-speed SAG mills are necessary for any successful operation. In nearly all cases, regardless of the SAG mill diameter, variable speed is not an issue for evaluation and is accepted as the standard design. On the other hand, variable-speed ball mills require careful evaluation and in most cases cannot be justified based on process considerations. It is only the cases where gearless drives are considered that trade-off evaluations are required between variable-speed and fixed-speed drives for ball mills and between gearless and gear-driven variable-speed drives for SAG mills.

ELECTRICAL CONSIDERATIONS Table 1 gives 16 different configurations for the range of mills covered in this paper. Table 1 Basic arrangement of mill power transmissions Drive Driver No. of Motor Type Primary Pinions Speed 1 Fixed Ring Gear Single Reducer 900-1200 2 Reducer 900-1200 3 NIA 180-200 4 Dual Reducer 900-1200 5. NIA 180-200 6 Vari. Ring Gear Single Reducer 900-1200 7 Reducer 900-1200 8 NIA 180-200 9 NIA 180-200 10 NIA 180-200 11 Dual Reducer 900-1200 12 Reducer 900-1200 13 NIA 180-200 14 NIA 180-200 15 NIA 180-200 16 Gearless NIA NIA 9-15

Motor Type W.R. Ind . Syn. W.R. Syn. W.R. Ind. Syn. Syn. Syn. W.R. Ind. Syn. Syn. Syn. Syn.

Power Supply

Quadra. SER PWM LCI PWM

ccv SER PWM LCI PWM

ccv ccv

The induction motors, both squirrel cage and wound rotor type, will be high speed design with 6 or 8 poles, depending on the choice of gears. The synchronous motors will be low speed with 30 to 40 poles depending on frequency and gear ratio. Synchronous motors used with cycloconverters on geared systems will have 8 - 10 poles. In selecting the type of drive to be used, several factors have to be considered before corning to a final decision, and these are discussed below.

1985

Enclosure and Ventilation The shape of a motor is subject to a number of constraints. The normal starting point for a design of an induction motor makes the axial length of the active material equal to the diameter of the rotor. This will also apply to high speed synchronous motors.(i.e., motors with 4 - 14 poles), Typical length to diameter ratios range between 0.6: 1 to 1.5:1. With induction motors the rotor diameter is'constrained by the centrifugal force that can be allowed without exerting undue stress on the cage or wound rotor winding, and/or without lowering the interference fit between the shaft or spider, and the laminations to an unstable level. With high speed Synchronous motors, the rotor diameter is again limited by the centrifugal force acting on the poles, the field winding and the fit between the shaft and the spider. Another constraint on rotor diameter may occur when WK2 must be limited to achieve fast response to speed regulation. Low speed synchronous motors have to provide sufficient space to accommodate the multiplicity of poles so the physical dimensions generally end up with the rotor diameter being much greater than the axial length of active material. Design criteria for Synchronous motors is based on the ratio of rotor axial length of active material to the distance between the centres of ajoining pole tips (pole pitch). With low speed machines the optimum value is 1.6 to 1, and can be as high as 4 to 1. High speed machines will be between 1 to 1 and 2 to 1. The low speed synchronous motor will have a length to diameter ratio of 0.15:l to 0.35:l. Manufacturing practice will use a series of diameters, typically increasing in 10% intervals, with a range of axial lengths for each diameter. These proportions gives rise to two distinctly different requirements for enclosure and ventilation. The high speed machine will have relatively high pressure drop through its ventilation passages such that dirt particles entering the machine will tend to precipitate within the machine, thereby restricting the airflow, and causing increase in temperature in the machine. This condition is further aggravated by the relatively small area available for air inlet, thereby creating a high air velocity and drawing heavier dirt particles into the airstream. By contrast the low speed motor with 30 - 40 poles, enjoys a low pressure drop through its ventilation passages, and the large area available for air inlet, gives rise to a low inlet air velocity. This results in only the finer dirt particles being drawn into the motor, and because of the low pressure drop, most of these are blown through the motor with very little left within the machine. Fixed Speed Operation For the low speed motor it is common practice to enclose the air inlets down to the motor base, and allow the ventilating air to be drawn from the pit below the motor through openings in the pit wall, further reducing the air velocity entering the motor enclosure, and allowing it to discharge into the mill bay at the top of the motor frame. (See Fig. 2)

1986

AIRCUT

*qp* I

MOTOR ENCLOSURE

I

AIROUT

Figure 2 Updraft enclosure and ventilation system However the high speed motor must have added protection from the entry of dirt, and the choice will lie between an open motor with filtered air inlet, a weather protected type I1 enclosure, a totally enclosed motor with an air to air heat exchanger, or a totally enclosed motor with an air to water heat exchanger. The WP II enclosure is usually specified to also have inlet air filters. Whichever of these options is selected, it will increase the initial cost and add increased maintenance expense to the installation.

Adjustable Speed Operation Grinding mills are constant torque devices, however rotor driven fans on electric motors follow a cube law with the result that at reduced speed the motor is unable to ventilate satisfactorily if the speed reduction is greater than 10%-15%, and/or if it is required to use the motor for inching the mill. Because of this, it becomes necessary to add forced ventilation to the motors. If the motors are of open construction for ventilation from the surrounding air, then a filtered air supply should be ducted to the motor air inlets, with sufficient static pressure to overcome the system resistance of the ductwork, and the motor pressure drop requirement. It should be noted that forced ventilation cannot be applied to a WPII enclosure. Where machine geometry permits, the ventilation system can be equipped with gravity dampers, to allow self-ventilation when the motor is running at or close to full speed. If the motors are totally enclosed with some form of heat exchanger, then separately motor-driven internal air circulating fans should be employed. For motors with air-to-air heat exchangers, the external fans should be separately driven. Gearless

1987

motors are usually totally enclosed and equipped with a number of heat exchangers and internal air circulating fans. When calculating losses (to determine overall efficiency of a drive system) the power to drive the fans should be added to the system losses. The gearless motor will also be equipped with a small filtered air pressurizing system to maintain positive pressure in the motor enclosure to assist in keeping dust, and liquid contaminants from the motor enclosure, along with carefully designed running seals.

Motor Construction High speed induction and synchronous motors are usually supplied with end bracket mounted bearings. Low speed synchronous motors, however, are usually supplied with pedestal-supported bearings and a fabricated steel base having sufficient axial length to permit axial shifting of the stator for cleaning and maintenance purposes. In order to provide adjustment of the motor for alignment purposes, it has become practical to install soleplates between the motor base and the foundation, thereby allowing adjustment of the motor position without disturbing its own alignment to the base. With wound rotor induction motors, the sliprings should be located outside of the motor enclosure to prevent brush dust from entering the motor windings, alternatively a separate internal ventilated enclosure for the sliprings may be supplied. The gearless motor stator will be split into three, four or six pieces depending upon shipping clearances, and lift capacity. Heat exchangers are either distributed around the stator frame, or are located in banks below the stator, each accompanied by an internal air circulating fan. The rotor of the gearless motor is comprised of field poles which are attached to the mill, usually at the shell/head interface onto a flanged extension of the mill head, this being the stiffest part of the mill structure. In some instances, with shell-type bearing construction, the rotor components may be mounted on a torque tube extension of the mill shell, or onto a flange ring on the shell located close to the shell bearing. Bearings Both high and low speed motors, induction and synchronous, will have sleeve type bearings. For low speed motors the bearings will usually be self-oil lubricated. High speed bearings will be flood lubricated, and flood lube units with heat exchangers will be required to cool and circulate the oil. The power consumption of the flood lube unit should be added to the motor losses. The gearless motor is, of course, supplied without bearings, and relies on the mill bearings for support of the rotor. To assist in installation and commissioning, high-pressure lift pumps are supplied with selflubricated bearings. These are also required for inching if the motor is used for this purpose. Soleplates For ease of installation and alignment soleplates are supplied with both bracket bearing and pedestal bearing design motors. These are particularly helpful when installing twin pinion drives. Clutches The air clutch is used extensively with single and twin pinion low speed drives when using synchronous motors. This device permits the motor to be started uncoupled, thereby enabling the use of low inrush current design motors, and also avoiding any torque amplifications during the acceleration of the motor. With weak power systems, reduced voltage starting of the motors is possible, and on twin pinion drives each motor can be started separately, thereby reducing the impact on the power system. A major benefit of the air clutch is that it will act as a shearpin in the event of a power fault at the motor. A sudden short circuit on the terminals' of any motor can produce very high instantaneous torques, and depending on the type and design of the motor this torque can be up to

1988

10-12 times rated torque. This can have a devastating effect on the coupling, and/or gearing connected to the motor shaft, but once the air clutch has accelerated the load, and locked up, its air pressure can be slightly reduced to limit its torque capability to a value less than 200%. If a fault should occur the clutch will slip, thereby protecting the connected equipment from damaging torques. With adjustable speed drives using low speed motors, the air clutch again comes into use a) as a shear pin against excessive short circuit torques, b) in the case of a twin LCI or PWM drive, both single pinion or twin pinion, it is required to provide bypass operation of the mill at fixed speed, by starting the single pinion motor, or one of the twin pinion motors across the line, and c) in the case of LCI, in order to avoid torque amplifications in the zero to 10% speed range, where the inverter bridge is forced commutated, the clutch is left open and only closed when the inverter is load commutated and torque pulsations have reduced to non threatening values. It should be noted that the air clutch is not used to accelerate the load with high speed motors due to the excessive wear which takes place during acceleration. However the air clutch can be installed as a shear pin on the high speed motor, or between the low speed output shaft of a gearbox and the pinion driving the ring gear to allow an unloaded motor start.

Installation and Alignment Correct alignment of drive train components is critical to good performance of an installation. Also it is important to check alignment regularly since foundations can settle differently, and an initially sound motor to pinion, and pinion to gear alignment can move significantly during the first few months of operation. Misalignment between motor and pinion is a major contributor to wear in a clutch. Once a drive train has been aligned statically, it must be checked in the dynamic state of driving a loaded mill. This has often proved difficult to achieve because of pressure to keep up production once the mill has been put into service, with resultant excessive clutch and gear wear. Continuous infra red monitoring of pinion tooth mesh is recommended since it would warn against development of a meshing error. Motor to pinion alignment can also be checked with the mill running and records kept which will show any change over time. One major cause for concern in alignment occurs with a mill designed for operation in both directions of rotation. With a single pinion drive, the location of the pinion is such that the pinion tooth is lifting the girth gear and the pinion therefore presses down onto its besrings. If the rotation is reversed, the pinion tooth is now pushing down on the girth gear and the pinion is pushing upwards on its bearings. As a result, the pinion will lift upwards by the amount of clearance designed into the pinion bearings. The minimum clearance is typically 0.005” (0.127 mm), but can be as large as 0.012” (0.304 mm), which is enough to increase wear in the clutch. Also by reversing rotation, the axial centre line of the mill will shift as the centre of gravity of the charge moves from one side of centre to the other. This problem also occurs with twin pinion drives where one pinion moves up whichever rotation is selected. The solution is to set the alignment at the mid point of the bearing clearance, which reduces the error by half and brings the alignment error to a more acceptable level. With gearless mills, which are invariably reversible, the axial centre line of the mill will move towards the centre of gravity of the charge, thereby creating an error in the airgap between the stator and the rotor poles. The worst-case condition of this occurs when the mill is starting to turn, and the charge is being lifted prior to cascading. As the airgap changes, the imbalance in the gap is reflected in the magnitude of the magnetic pull between the stator and the rotor. Consequently, the stiffness of the stator frame, mill bearings and the foundation must be designed to achieve stable operation with this magnetic pull acting on the components. The motor designer can reduce the maximum pull and the required structural strength and also reduce the total weight of the machine if he can use multiple paths or circuits in the design of the stator winding. (e.g., with a four circuit winding the pull will be reduced to about 40% of that exerted in a machine with a single circuit).

1989

Drive Configurations The use of squirrel cage induction motors is limited to small single pinion drives. This is due primarily to the high inrush current of the motor. The squirrel cage motor has not been considered for the twin drive option primarily due to the high inrush current problem, and the lack of adequate controls to maintain reasonable levels of load sharing between the drives. Fixed Speed Single Pinion Single pinion drives utilizing one pinion are currently available up to 10.0 MW. A high speed motor will drive the pinion through a main gearbox, and will be either a 6 or 8 pole machine. A low speed, leading power factor synchronous motor, driving the pinion directly through an air clutch is the preferred drive configuration since it offers high efficiency, reactive power capability, overload capacity, rugged construction, low inrush current, minimum maintenance, and competetive evaluated cost. The single input to a power-splitting pinion stand drive using a high speed wound rotor motor can be used up to approximately 13.5 MW. The motor will utilize a liquid rheostat for starting purposes, and will drive through the gear reducer to two pinions driving a single ring gear. A synchronous motor can be applied to this drive configuration using a hydroviscous clutch to allow an unloaded start for the motor, thereby avoiding torque oscillations during acceleration, and enabling the motor inrush current to be controlled within the constraints of the power system. Because of the small installed base for this drive system, performance data regarding availability and maintainability is not available. Fixed Speed Twin Pinion Twin high speed wound rotor induction motors can be used to drive through gearboxes to twin pinions. The motors will share load on average within about 5% and can share a common starting resistor. Gear runout can give rise to load swings between the two motors during a revolution, which can lead to accelerated gear wear. Also motor characteristics may not match perfectly, which can cause an offset in load share between the two motors. To overcome these conditions, a permanent slip resistor may be installed between the two rotors. This will improve the load share capability, but it will be at the expense of drive efficiency, which will be reduced by 1.2%-1.5%. A twin low speed synchronous motor drive has been successfully developed utilizing air clutches to allow the motors to start individually. Additionally, this system has a special winding built into the quadrature axis of the motor rotors (see Figure 3) to allow precise load sharing at all times; i.e., on average, and during each revolution of the mill, compensating for gear runout. Air clutches are installed between each motot and its associated pinion, which allows the motors to be started one at a time thereby lowering the impact on the power system. Once the motors are synchronized, the mill is accelerated using the air clutches. After the clutches lock up, the load share between the two motors is monitored, and any error is adjusted by applying current to the quadrature axis winding in the motor rotors. If this error exceeds a pre determined amount, an automatic clutch pulsing system comes into play and brings the two rotors closer to exact load sharing. The regulator controlling the quadrature axis current then brings the two motors into exact load sharing, not only on average, but during each revolution of the mill compensating for run out of the gears. This system is known as Quadramatic'"'. A typical layout of the motor pit is shown in Figure 4 where all cable connections to the motor are made in the motor pit, leaving clear access at the mill floor operating level. It should be noted that this drive imposes less stress on the pinions and ring gear than any other geared drive, regardless whether they are single or twin pinion. This twin pinion approach enjoys all the benefits and features of the low speed single pinion synchronous motor described above.

1990

DmcT AXIS

OlREff AXIS QUAmATURE AXIS RESULTANT A W L € W C H FLUX IS s)(IfTEO W H RESPECT TO FIELD STRUCTURE

1991

ILL M N U - I A W S MOTOR

REMOTE CUMACT

CUBICLE

Figure 4 Typical detail of mill motor pit

VARIABLE SPEED DRIVE OPTIONS From Table I, it will be seen that eleven adjustable speed options have been identified. These cover single pinion geared, twin pinion geared, and gearless.

Wound Rotor Motor(s) with Slip Energy Recovery (SER) This drive system can be applied to both single and twin pinion geared drives. Since wound rotor induction motors are not cost competitive at low speed, the scheme generally employs high speed motors driving through reduction gears to the pinion(s). The drive is started with the use of liquid rheostats providing adequate starting and accelerating torque to bring the mill up to speed, with a low inrush current of approximately 200% to 250%. When the motor reaches full speed it will be driving the mill at its maximum speed, typically 80% of critical. In order to run at base speed (76% critical) and further to 60% critical, the motor slip must be increased accordingly. This can be achieved by inserting resistance in the rotor circuit, and dissipating this energy into the liquid rheostat. This is very inefficient, so rather than using the rheostat, the slip energy is converted to direct current, inverted to the frequency of the

1992

power system feeding the motor, and then fed back into the power system through a step up transformer. Most mill configurations will have natural frequencies in the region of 2 Hz-4 Hz, 8 Hz-10 Hz, 20 Hz, and 40 Hz. In reducing the speed of the wound rotor motor, the slip energy recovery equipment will generate forcing frequencies at multiples of 6 times the slip frequency, depending on the number of pulses built into the equipment. A higher number of pulses equates to increase in cost. For example in reducing the motor speed by 5%, the SER will excite all natural frequencies between zero and 18 Hz (6 pulse) or 36 Hz (12 pulse), etc. Because of this there is a high probability that certain speeds in the operating range of the mill will need to be deadbanded (unable to operate within a particular speed range). Wound rotor motors have low electromagnetic damping, and will tend to accelerate gear wear. Advantages of SER drive: 0

Low first cost (unless 18 or 24 pulse SER is used) Motors will operate (at 80% C.S.) if SER is out of service Low speed inching available with liquid rheostat Smaller filter sized for slip energy only Can load share twin drive.

Disadvantages of SER drive: Deadbanding required to avoid resonance Accelerated gear wear Poor overall efficiency Need for special attention to motor enclosure and ventilation Need to maintain precise alignment of motor/gear/pinion. No shear protection against transient torques caused by short circuit faults. High maintenance of sliprings, brushes and liquid rheostats No automatic protection against a cemented charge.

A variation of this drive involves connecting the rotor windings to a low frequency source. This variation allows the motor speed to be reduced or increased. If PWM power is used for this purpose, then smooth speed variation is possible. This system is limited to relatively small speed variations. Pulse Width Modulated (PWM) Drive This drive system is a relatively new system which can be used for variable speed mill drives. It can be applied to both Induction and Synchronous motors. The PWM was originally introduced for small low voltage drives, but has now been developed in various configurations for medium voltage applications. There are now several configurations and combinations of semiconductors available to produce the power required for mill drives up to at least 20 MW. These include: IGBT (Integrated Gate Bipolar Transistor) IGCT (Integrated Gate Cornmutated Thyristor) IEGT (Injection Enhanced Gate Transistor). The device draws its power from a dc bus and inverts it to alternating current by chopping blocks of dc at high frequency.

1993

The IGBT is the original transistor power supply and multiple modules are connected in series to increase voltage, and in parallel to increase current. Switching frequency is typically 1600 Hz The IGCT is a higher powered thyristor which can handle much higher current than the IGBT, thereby reducing the number of devices required for a given power output. Switching frequency is typically 500 Hz. The IEGT is a higher powered transistor than the IGBT, with a very low gating voltage and with a high rate of change in gating voltage (dV/dt) requiring minimal snubber circuitry. It is switched at 500 Hz. This device enjoys the power level of the IGCT with far lower complexity and power requirements in the gating circuit. Because of the high dc content in the PWM output, a step up or down transformer should not be placed between the PWM and the motor. Unless the Utility voltage matches the PWM output voltage, the motor cannot be transferred to the bus without a transformer that matches the Utility voltage to that of the motor. The PWM can be configured for two or four quadrant operation. Since grinding mills do not regenerate, two quadrant operation will suffice. This will allow the converter side to be a straight diode bridge. The input transformer will be designed for a minimum of 12 pulse, or alternatively 18 or 24 pulse, thereby reducing the harmonic content to the Utility to below that required by IEEE 519. This configuration will operate at a power factor of 0.95-0.96 lag. If a PWM converter bridge is substituted for the diode bridge, then the power factor can be improved further to unity or leading. The PWM requires an input transformer with a large number of windings to feed the relatively large number of circuits. The PWM is a voltage source inverter and can be used on induction motors, as well as on synchronous motors, so where initial cost is of prime importance, then the combination of PWM, squirrel cage induction motor, and gearbox can provide the lowest initial cost. To protect gears from the high transient torques that can occur in the unlikely event of a cable or terminal fault at the motor, a disengaging device should be installed between the motor and the gearbox. The PWM power supply is usually rated at its maximum capability, so in order to provide sufficient torque to accelerate the mill through its cascade point, the PWM, as with all electronic power supplies, must be capable of producing sufficient power to accelerate the mill through the cascade point. If the mill is accelerated slowly, then the maximum amount of torque required should not exceed 130%.Similarly, this or greater torque can be required when inching the mill to check for a cemented charge. For inching, the maximum torque could be required for approximately 30 seconds. This capability should be confirmed by the supplier for a range between 150% and 160% of the required steady state mill load for a minimum of 30 seconds. This capability should be cmfirmed by the supplier. With its high inrush current, the squirrel cage induction motor may have difficulty starting across the line should the PWM be out of service. Larger drives, in conjunction with low speed synchronous motors for both single and twin pinion, have been considered, and like all other drive types there are benefits and disadvantages. They compare quite closely to the LCI drive and, where the power requirement matches the output capability of the drive, they are competitive with the LCI. At the time of writing, there is insufficient installed base to comment on reliability and availability, but they do offer another choice. The PWM can also be considered for the power supply for gearless drives where it has the distinct advantage of having a high front-end power factor, harmonic content to meet IEEE 5 19, and very low risk of creating damaging short circuits. The PWM can also be used on a twin pinion drive. Advantages of the PWM system: 0

0

Can work with either induction or synchronous motor Synchronous motor operates at 1 .O power factor

1994

0

High power factor to the Utility Eliminates need for a harmonic filter Does not generate significant torque pulsations Can inch at very low speed Cemented charge protection available Reversible Line side power interruptions do not cause faults on load side.

Disadvantages of the PWM system: Relatively complex electronics Cannot transform output voltage Motor may need transformer in order to run on the Utility.

Synchronous Motor(s) with Load Commutated Inverter (LCI) Drive The LCI can be configured in,any one of three ways: 12/6pulse 12/12 pulse with two winding motor 12/12 pulse with summing transformer and single winding motor. Most LCI systems will utilize a three winding transformer to connect to the Utility, with a 30" phase shift (Wye - Delta) between the secondary windings and with each secondary feeding into a 6 pulse converter bridge. This configuration will cancel out most of the 5' and 7'h harmonics, thereby presenting a 12 pulse load to the Utility, and reducing the size of the required filter to meet IEEE 519. The LCI is a current source inverter, and it uses the back EMF generated within a synchronous motor to commutate the inverter bridge so that it will invert at the frequency at which the motor is running. Between zero and 10% of rated speed, there is insufficient back EMF to commutate the inverter, so the bridge is force-commutated over this range. The curves in Figure 5 show the ripple torque effect for both 6 pulse and 12 pulse inverters, during forced commutation and load commutation.

1995

Figure 5 Torque pulsation curve

1996

The 12/6 Pulse LCI. This configuration (see Figure 6) will cancel most of the 5'h and 71h harmonics fed into the system as stated previously. However, the inverter will feed 5Ih and 7Ih harmonics (as well as higher order 1lthand 131h;etc.) into the motor requiring it to be made larger to accommodate the heating effect of these harmonics; and for satisfactory commutation, the motor subtransient reactance must be kept below 18%-20%. Other than with relatively small mills, below 2200 kW, the 6 pulse system is not recommended.

-I

i EXCITATION TRANSFORMER

EXCITER FIELD SUPPLY

T

L

LINESIDE TRANSFORMER

V P

RECTIFIER

-1

INVERTER

REACTOR

SUMMINGTRANSFORMER

T )

BYPASS BREAKER

I

SYNCHRONOUS MOTOR

Figure 6 1216 pulse LCI with bypass capability

The 12/12 Pulse LCI. As well as cancelling the 51h & 7'h harmonics on the line side, this configuration (see Figure 7) will also cancel out most of the 5'h and 7th harmonics on the inverter side. With high speed motors, it may be convenient to feed the two 6 pulse outputs from the inverter into two phase shifted windings in the motor stator. If this is done, it would not be possible to operate the motor on the Utility at fixed speed in the event of an outage of the LCI. With low speed motors, a two winding configuration would not be practical due to the large number of stator slots that would be required to accommodate the winding. To overcome this, and in the case of the high speed motor, the outputs of the two 6 pulse inverter bridges may be summed together through a three winding transformer to provide 3 phase 12 pulse power to the motor. The output voltage of this transformer can be selected to match the distribution voltage SO that the motor could bypass the LCI and run at fixed speed if required. By cancelling out the 5* and 71h harmonics, the motor will run much cooler and the ripple torque produced by the motor will reduce to a much lower value than that produced with a 6 pulse supply. The motor can be designed with much higher reactances, which results in a low inrush current if the motor is required to start across the line.

1997

LINE-SIDE TRANSFORMER

T )

SUMMING TRANSFORMER

BYPASS BREAKER

I

synchronous motor

Figure 7 12/12 pulse LCI with bypass capability This configuration would allow one design of motor to drive both fixed and variable speed mills of like power and base speed, thereby saving on spares and cost of foundation design. A disconnect switch between the output of the summing transformer and the motor, supplemented by a bypass motor starter, would enable the motor to run at either adjustable speed or fixed speed. In order to protect the gears from the damaging effect of transient torques which can be generated in electric motors under cable or terminal fault conditions, air clutches are provided between the motor(s) and the pinion(s). The air clutches are disengaged while the motors accelerate through the forced commutation speed range (0%-lo%), thereby avoiding any risk of torque amplification at natural frequencies. Further, if the drive is to operate in bypass (see Figure S), the air clutch allows both motors to be brought up to synchronous speed uncoupled.

1998

i

-

& ................................... o - v r r c ”

.................................

1

EXCITER FIELD SUPPLY ...................................

SUUM

L

i SET UP SWITCH

...................................

L

I

d

/

i-

SYNCHRONOUS SYNCHRONOUS MOTOR

MOTOR

Figure 8 Twin 12/12 pulse LCI with bypass capability The LCI operates at a lagging power factor of 0.85 at full load, but this will be improved to approximately 0.95 when a filter is added. The filter will increase the system losses by approx. 0.5%. Advantagesof the LCI system: Current source inverter. Inverter fault currents are controlled by the dc reactor in the LCI With summing transformer, motor can match Utility voltage, allowing interchangeability with fixed speed motors on Ball Mills Motors can run at fixed speed on Utility in the event of drive outage Low cost, low loss filter can reduce harmonics Inching available at 10%of rated speed Inching by position available Cemented charge protection available Initial cost competitive with other systems above 3000 hp Proven system Drive is reversible and regenerative Very high availability.

1999

Disadvantages of the LCI system: 0

0

Increase in footprint to accommodate summing transformers, and bypass set-up switches Unsuitable for continuous operation below 10%speed 12/12 with summing transformer premium priced on smaller mills.

Cycloconverter (CCV) Fed Systems The CCV drive is a low frequency drive supplying power at frequencies up to 30 % of line frequency. It has been the drive of choice for gearless drives where the motors operate at frequencies between zero and 7 Hz. However, with the introduction of the PWM power supply, the two drives will compete with each other for the engineers favour. The CCV can also be applied to geared drives both single and twin pinion but, these configurations, since the output frequency is less than 30% of line frequency, the motor will have a fewer number of poles in order to run at pinion speed. The CCV is a voltage source inverter, and since the frequency conversion occurs without going through a conversion to dc and inversion to ac, a fault occurring on the line side can result in a sudden short circuit on the motor side. Special techniques can be provided to ensure an orderly shutdown of the CCV bridge prior to the drive breaker opening, however opening an upstream breaker can cause such a fault (see Figure 9). To cater to this condition, it is necessary to provide extra bracing to the motor windings, andor add high speed interrupters between the CCV and the motor. I

i

BREAKER

) TRANSFORMERS

MOTORS

Figure 9 Twin CCV geared drive

2000

To protect gears from the high transient torques that can occur with the CCV, disengaging devices such as shear pins or air clutches should be inserted between the motors and the pinions. Because the motor(s) must run at low frequency, it is not possible to operate a CCV drive with bypass capability. The CCV normally requires an encoder for operation at very low speed, with vector control above about 5% of rated speed. The CCV is fed from three, three winding transformers, with their secondaries connected wye and delta to provide some harmonic cancellation. This works adequately when the output of the CCV is fed into a single motor (as in the case of the gearless drive). However in the case of a twin pinion drive, the cancellation will only be partial because the motor rotors will depart from true alignment to each other during a revolution of the mill due to gear run out. Since the harmonics generated by the CCV are multiples of the operating output frequency of the drive, it follows that the total amount of harmonics generated will be greater than with a LCI, but these harmonics are of smaller magnitude and are spread over a larger range of frequencies. The need for a filter, and the size of the filter are both dependent upon the short circuit capacity of the power system relative to the drive capacity. Generally speaking, if the system short circuit MVA is less than 20 times the converter MVA, then it is probable that a broad band filter will be required to control harmonics. The design of the filter is more costly than the tuned filter that is used with a LCI, and its losses are about three times those of a tuned filter for the same size drive. The CCV provides a clean quasi sinusoidal current waveform to the motor(s), and can be operated at frequencies down to zero without torque pulsations. The output voltage of the CCV for a twin drive is in the order of 1500 volts. Advantages of the CCV drive:

0

0 0

0

Inching available down to zero rpm Cemented charge protection available Inching by position available No torque pulsations Marginally better motor efficiency than other drives Reversible and regenerative Charge let down capability.

Disadvantages of the CCV drive: Motor(s) dedicated to the CCV, no bypass capability Susceptible to power supply transients Broad band filter required, with higher cost and higher losses Higher first cost Low voltage output increases cable costs Need “shear protection “.

CCV and PWM Driven Gearless Drives With the gearless drive, the rotor of a synchronous motor is mounted directly on to the mill, usually on a flanged extension of the mill head. The stator is made in segments, generally three for smaller diameter mills and four for larger diameter mills. Where lifting capability and /or shipping clearances demand, six segments may be considered. The stator winding may comprise one or two windings, using single- or multi-turn coils that are deployed in one or more circuits, depending o n the design philosophy of the manufacturer (see Figure 10). Special attention should be given to this design since it will have significant impact on the reliability, availability, and repairability of the machine.

2001

PRIMARY BREAKER

TRANSFORMERS

T

SYNCHRONOUS MOTOR

Figure 10 Gearless drive with single 3 phase CCV supply Motor voltage can vary from less than 1500 volts to close to 5000 volts, depending on the design of the stator winding. As with the twin CCV drive, the gearless motor is subject to sudden short circuits arising from unscheduled interruption in the power supply due to upstream breaker trips. These occurrences have caused significant damage to installed equipment and its foundations. These forces must be taken into consideration in the design of the motor and its foundation. This kind of event will not occur with a PWM fed mill (see Figure 11).

2002

A

.............................

A

.......... .......... .........

Figure 11 Gearless drive with single 3 phase PWM IEGT supply Because of the low frequency at which these motors operate, typically 0 Hz-7 Hz, the core losses only comprise approximately 10% of the total motor losses, leaving the stator and rotor copper losses as the prime source of machine losses. By adding copper to the machine, the winding resistance can be lowered, thereby achieving remarkably high motor efficiency. However it must be understood that nothing is free, and the cost increment in achieving an efficiency improvement of 1% may well exceed the loss evaluation in dollars for the kW saved. Once the motor is properly aligned the airgap is continuously monitored by sensors, which will alarm and shutdown the machine if the airgap error exceeds the allowed tolerance. Special running seals are provided between stator and rotor to protect against the ingress of water, mud, and dust. Advantages of the gearless drive: Elimination of gears reduces quantity of components Gear losses are eliminated Can achieve high efficiency Inching available down to zero rpm Cemented charge protection available Inching by position available No torque pulsations Reversible and regenerative Charge let down capability No filter required with PWM.

2003

Disadvantages of the gearless drive: Dedicated to CCV or PWM power supply Susceptible to power supply disturbances CCV only Broad band filter required with CCV, with higher cost and higher losses Higher first cost on smaller units High installation cost Long outage if motor damaged. OTHER CONSIDERATIONS When a mill comes to rest there is a tendency for the charge in the mill to settle and cement into a solid or semi-solid mass. If an attempt is made to start the mill in this condition, there is a great risk of carrying the charge through close to 18O9 at which point the charge will break away from the shell and crash across the mill with potentially destructive results. To avoid this kind of occurrence, it is good practice to inch the mill to ensure that the charge is loose and able to cascade. Mechanical inching drives are normally portable so that one machine can serve several mills. A disadvantage of the mechanical incher is the time taken to install and remove it before running the mill. Electric inching has been available for some time. Wound rotor motors can be turned slowly using their liquid rheostats. This will enable the mill to be positioned for maintenance purposes as well as check for a cemented charge prior to starting the mill. To inch synchronous motors with sliprings, the field is excited, and a dc source is applied through a set of commutating contactors to the stator winding. These contactors synthesize a low frequency ac voltage. A newer version of this system replaces the contactors with static switches to achieve the same result. If this system is to be considered with synchronous motors having brushless exciters, then the motors must be specified to have an ac fed exciter to allow full excitation at zero speed. This system also allows mill positioning for maintenance as well as checking for a cemented charge prior to starting the mill. If the motor has a conventional dc fed exciter, then this kind of inching cannot be used. Methods have been devised to check for a cemented charge by:

Reducing air pressure into the clutch to a level just above that required to cascade the mill, and closing the clutch. Either the mill will accelerate, showing that the charge was not cemented, or the mill will stall with clutch slippage showing that the charge was in fact cemented. This method has two flaws: a) there is a lot of clutch wear, and b) if the mill is only partially loaded, then there may be sufficient torque available from the clutch to carry the cemented charge up to the top of the mill and allow it to fall under gravity. Noting the position of a liner bolt, applying the clutch for 2-3 seconds, then tripping the clutch. If, when the mill rolls back to the rest position, the liner bolt has moved forward from its original position, then the charge is unlikely to be cemented. Conversely, if the position of the liner bolt is unchanged, then the charge is likely to be cemented. This method also adds to clutch wear. Using a combination of degrees of rotation of the mill, and acoustics and/or vibration, it is possible to determine whether the charge has cascaded within an allowed number of degrees of rotation. This method is independent of the amount of fill in the mill, and allows the mill to continue to accelerate if cascade is detected, thereby minimizing wear in the clutch. None of these methods should be used for positioning the mill.

2004

If the mill has a variable speed drive on it, then the drive can be used for both inching and detection of a cemented charge. It should be noted that if the motor is connected to the drive train during inching, then high pressure lift pumps must be installed at the motor bearings because the inching speed will be too low for the bearings to maintain an oil film. Also if the motor is used for inching electrically, then care must be taken to ensure that the time taken for inching is within the thermal capacity of the motor, unless forced air ventilation is used.

Power Factor Large electrical power users purchase their electrical energy according to a tariff negotiated with the Utility. This usually takes the form of $,’X’ per kVA (kilo volt-ampere) of maximum demand, plus $“Y” per kWh of energy consumed. In order to keep the demand charge as low as possible, it is desirable to keep the power factor of the site as close to unity as possible. Traditionally mine sites have relied on their relatively large installed base of synchronous motors to supply the required power factor correction. With the introduction of adjustable speed drives, some of the inherent reactive power availability has been removed, and it has become necessary to rely on the power factor improvement obtained from the capacitors in the drive filter, supported by any synchronous motors used in the installation. In some cases, banks of capacitors have to be added to limit any power factor penalties. Example: I SAG Mill 15 MW with uncorrected power factor of 0.84 lag plus two ball mills each 7.5 MW fixed speed 0.8 pf lead single pinion geared synchronous motors. SAG mill corrected to 0.95 lag using filter capacitors. Total MVAr’s (Mega volt amperes reactive) contributed to the system = 6.32 MVAr. lead. If such an installation is repeated with adjustable speed drives on the ball rnill(s) then the installation will draw a total of 9.86 MVAr’s from the system. Therefore, to supply the same amount of power factor correction for the plant 16 leading MVAr’s are required. The cost of this must be added to the total cost of the second case. In cases like this, it may be better to consider a 15 MVAr synchronous condenser to supplement the MVAr’s obtained from the required filters. Overload Capability Unless it is specified, and included in the price of the equipment, no overload capability is available from any adjustable speed drive system, or from a fixed speed induction motor drive. Fixed speed synchronous motors, however, do have inherent overload capacity since it is possible to exchange their reactive power capability for active power, so a 0.8 leading power factor synchronous motor can supply a 25% increase in its kW output without exceeding its temperature guarantees. Power Distribution Considerations For large installations, it is sometimes desirable to use a primary distribution voltage higher than 13.8 kV and 24 kV is commonly selected. This voltage can be fed directly into the primary transformers of adjustable speed drives using LCI, PWM or CCV power. As previously shown, twin LCI drives can obtain bypass capability by installing a 24 kV/motor voltage transformer. This transformer can also be used as a spare for the LCI drive. Wound rotor fixed and adjustable speed drives along with all fixed speed drives will require a lower voltage and a separate medium voltage distribution system. This can be at 13.8, 6.6, or 4.16 kV and will supply all the medium voltage loads at the site. This approach to the power distribution does not cause additional transformer loss penalties for medium voltage equipment below 24 kV.

2005

Altitude Considerations As altitude increases the air density decreases. The heat generated by electrical equipment becomes more and more difficult to dissipate with altitude increases. Equipment that depends directly on cooling air to remove its losses has to be overdesigned so that the cooling air can remove the heat while not exceeding the equipment’s thermal capacity. Where air-to-water heat exchangers are employed for equipment cooling, the equipment will require less derating if the heat exchangers are made larger to increase the heat collection surface. Also voltage, BIL and partial discharge margins are reduced. Maintenance Considerations Electrical equipment today is designed and built to be as free from maintenance as possible. Electronic solid state equipment has virtually no wearing parts, and is not expected to fail or wear out in normal operation. However it is customary to carry an inventory of components for an adjustable speed power supply. For motors of a geared drive, spare bearings, brushes, brushholders, brushless exciter components, and often a spare wound stator and rotor are mandated. Where more than one motor of the same rating is installed, one set of spares will be shared between them. For a gearless drive, the cost of a spare stator would be prohibitive so this is not called for. Spare rotor poles can be carried, and with some designs stator winding components can also be carried. Because of design differences, the costs of these capital spares have not been shown. Generally speaking, routine maintenance on geared drives can be done during normal planned mill outages. Unplanned outages due to component failure can usually be handled in a relatively short time. Back up systems can provide for continued operation until a planned outage can be arranged. With gearless motors, a stator winding fault will cause an extended outage which can range from 2-3 days to several weeks, depending on the type of construction used by the motor manufacturer. Equipment Cost In compiling costs, the following criteria has been used: An allowance for the losses is added to the first cost in an attemp to demonstrate the effect of efficiency for each drive option Fixed speed single and twin pinion drives do not include an electrical house since their control equipment is normally located in an electrical room supplied for this and other equipment. All adjustable speed drives include an electrical house, with the usual complement of equipment such as drive controls, motor control center for mill motor and mill auxiliaries, graphics package, etc. Fixed speed drives are fed from a medium voltage distribution system Adjustable speed drives are fed at the design distribution voltage, except wound rotor drives can only accept power at 15 kV or less. No transformer penalty (cost or loss evaluation) has been applied to the wound rotor drive since a distribution voltage of 15 kV or less will normally be available as well as higher voltage which can feed directly into CCV, PWM, or LCI drives Primary circuit breakers for each drive are included The equipment costs are reflections of the efficiencies used to compile the loss evaluations. In general, the efficiency values used are at the level where the cost of improving the losses matches the loss evaluation number of $3,000 per kW Motor speeds of 200 rpm for geared low speed synchronous motors and 900 rpm for wound rotor induction motors are used The bypass equipment (switchgear, transformer, and set ups) as shown earlier in Figure 8, increases the first cost of the electrical equipment by approximately 5%.

2006

Figures 12, 13, 14, and 15 show the cost relationship between the various drive systems. Note: These costs are shown on a “per unit” base; i.e., they relate to the cost of one drive relative to another without using actual costs frozen in time. Figure 12 covers adjustable speed drives for SAG mills using first cost, and this is followed by Figure 13, which takes into account the drive efficiency. Since the geared LCI PWM and CCV cost and performance are very close to each, other, a single curve for low speed synchronous (LSS) is used.

Variable Speed S A G

M ill D r i v e “ F i r s t C o s t “ C o m p a r i s o n

9

7

6

3

2

Figure 12 Variable speed SAG mill drive “first cost” comparison

2007

E v a l u a t e d

V a r i a b l e S p e e d (Losses Evaluated

6

2 5

5

2 5

b

2 5

S A G

M ill D r i v e C o s t

a t $ 3 0 0 0 1 k W

)

3 2 5

2

2 5

1

2 5

4

5

6

7

8

9

10

1 2

1 4

D r i v e P o w e r (M W

1 6

I 6

2 0

2 2

2 4

2 6

)

Figure 13 Evaluated variable speed SAG mill drive cost (losses evaluated at $3000/kW)

2008

Figures 14 and 15 show drive first and evaluated costs for fixed speed ball mills.

Fixed Speed Ball Mill "First Cost" Drive Pricing 7 ,

3

4

5

6

1 +LSS

7

8

10

9

12

14

16

Drive Power (MW)

+WR

+Twin

LSS +Twin

Figure 14 Fixed speed ball mill "first cost" drive pricing

2009

WR +Gearless

18

1

20

22

Evaluated Ball Mill Drive Cost Comparision (Loss evaluated at $3000 per kW Loss) 8

7

6

*

5

u8

s4

3 h

&

3

2

1

0

3

4

5

6

7

8

9

12

10

14

16

18

20

Drive Power (MW) +LSS

+WR

+Twin

LSS

+Twin

WR

+Gearless

Figure 15 Evaluated ball mill drive cost comparison (loss evaluated at $3000 per k W loss) In all cases the loss evaluation has a significant impact on the total cost. A more detailed analysis of costs can be found in "Selection and Evaluation of Grinding Mill Drives " C.D. Danecki, G.A. Grandy and P.F. Thomas Paper DB-14 CIMM-SME Symposium Vancouver Oct 2002.

ACKNOWLEDGEMENTS The author wishes to thank Stuart Walters for his help in formulating this document, Mac Brodie for his helpful and encouraging critique as the document developed, George Grandy, and Craig Danecki for providing process and mechanical supporting information. In conclusion, I would like to thank GE Canada for their support and assistance during the preparation of this paper. REFERENCES Barratt D.J., Brodie M.N., and Pfeifer M. 1996. SAG Milling Design Trends, Comparative Economics, Mill Sizes and Drives. Proceedings International Autogenous and Semi-Augotenoug Grinding Technology 1996, eds. A.L. Mular, D.J. Barratt, and D.N. Knight, I11 : 1228. Bassarear, J.H., and Thomas, P.F. 1985. Variable Speed Drives for Semiautogenous Mills. SME Meeting. Danecki C.D., Grandy G.A., and Thomas P.F. 2001. Mill Drives in the Third Millenium. SME Annual Meeting.

2010

22

Selection of Metallurgical Laboratory and Assay Equipment; Laboratory Designs and Layouts Peter F. Wells,’

ABSTRACT

The design of laboratories for metallurgical and chemical studies in mineral processing operations has to consider 1) the needs for safety and health of the staff operating the laboratory,

2) the requirement to reduce the samples to a mass and particle size that is suitable for metallurgical test work or chemical analysis,

3) the need to ensure that samples do not become oxidized or contaminated during processing, 4) the requirements to be able to efficiently carry out the test work or analysis and transmit the results to the required parties. The necessary equipment, the layout of laboratories and the data acquisition and transmittal systems are considered based on the above requirements.

INTRODUCTION Design of metallurgical control and development laboratories has received very little attention in the metallurgical literature with only the description of the Western Australian Mineral Research Centre (Bagshaw 1996) being published in the last ten years. The Laboratory Design Handbook (Crawley Cooper 1994) provides a wealth of information on research laboratory design but, while essential reading to provide general background, has no specific information for the designer of a metallurgical laboratory. Prior to that, the paper in the previous edition of this manual - “Design and Installation of Concentration and Dewatering circuits” (McKenzie and Haig 1986) provided a comprehensive survey of the state of the art at that time. Other works, which should be consulted, are “Design of a Mine-site Laboratory Facility”(She1ton 1981), and “A Guide to Laboratory Design”(Everett and Hughes 1981). The U.S. National Research Council book “Laboratory Design” (Coleman 1951) is very comprehensivebut is now mainly of historical interest. While the essentials of work flow and efficiency of operation have been well covered in the literature, the important change that has now become apparent is the need to emphasize the safety and health of the metallurgist and his or her staff and the great need to make the metallurgical laboratory a pleasant place to work. It is increasingly difficult to attract and retain highly competent people in a field that is, in certain quarters, viewed as a sunset industry and often involves a fly idfly out work schedule. Since concentrators are intrinsically noisy and dusty and metallurgical personnel must, of necessity, spend a substantial part of their working day in that environment, it is essential that the metallurgical laboratory, in contrast, is a pleasant place to work in to maximize the operation’s ability to keep professional staff.

’ Section Head, Mineral Processing ,INCO Technical Services Limited, Mississauga, Ont. L5K 1Z9, email: [email protected]

201 1

Most of the discussion in this paper is directed towards the design of metallurgical laboratories associated with operating plants. However many of the criteria are also applicable to research and development facilities and the requirements for the latter will be covered in a separate section. LABORATORY PLANNING

Involvement of the metallurgist, who will be responsible for the operation of the metallurgical laboratory, in the design process will be the most critical parameter in ensuring that a satisfactory design is achieved provided that the individual has the breadth of experience in various operations to know what does and doesn’t work. The metallurgist’s first task will be to determine the expectations of management with respect to the functions and goals of the metallurgical department and then state them explicitly in the form of a mission statement. This will then lead to agreement on the number of staff and the equipment and, only when these parameters have been established, should the architect and engineering company become involved. In most metallurgical plants, the location of the metallurgical laboratory will be dictated by the following considerations: l)noise, dust and vibration levels in the immediate area 2)the retrieval and processing of the daily (shift) metallurgical accounting samples 3)the retrieval and processing of metallurgical development samples 4)the need for the metallurgical personnel to work closely with operations management and personnel 5)the need to monitor, calibrate and correct the operation of an on-stream analyser In base metal plants, the on stream analyzer and gang sampler should be centrally located and having the metallurgical laboratory adjacent to this equipment helps ensure that the analyzer receives timely attention. The metallurgical laboratory should also be close to the control room thus allowing the easy interaction of the metallurgist and metallurgical technicians with operations staff. A central location may make providing outside windows to the laboratory a challenge but the latter is an important consideration when trying to maximize the aesthetic appeal of the working environment. In addition to choosing a location not immediately adjacent to sources of dust and noise the laboratory environment can be improved by the use of adequate sound insulation and double doors to create an airlock with boot cleaning facilities by the outer door. SAFETY EQUIPMENT

The safety station will include an emergency shower and eyewash, fire blanket, “kill” switch, portable fire extinguishers and protective glasses, monogoggles, face shields and gloves. Gas testing equipment, cyanide and other antidotes, resuscitation equipment and a spill kit should be located in the same place. Vented cupboards (usually incorporated into dust hoods) are required for acid and organic liquid storage. All electrical equipment that could come into contact with these vapours must be rated explosion-proof. A separate cupboard for personal protective equipment adjacent to the main exit should be provided. A second exit fkom the laboratory is required for reason of safety as well as convenience. Sprinklers for fire protection should be added although Fire Code requirements will likely make them mandatory in any case. LABORATORY SERVICES

Ideally inter-floor space of at least 1.2 meters will be allowed for services. Air make-up is a large item as the laboratory must be maintained at a positive pressure with respect to the concentrator to

2012

prevent infiltration of dust. A typical 1.4 meter long dust hood will withdraw 0.57 m3/sec (1200 cfm) of air resulting a significant heating or air conditioning load and to prevent noise in the feed duct system, oversize air feed systems and inlet ducts will be needed. Ceilings at least 3.1 meters (loft) high allow for use of flotation columns in the laboratory as well as giving an uncluttered feel to the workspace. Natural lighting fiom a full-length window on one side of the laboratory with a well designed artificial lighting plan will further enhance the working environment. Water requirements are for hot and cold domestic water, demineralized water and addition of a carefully marked faucet to deliver process water will assist the metallurgist when carrying out flotation testing. A tempered water supply is required for the shower and eyewash. Dried compressed air should be supplied for flotation air, for hoses in dust hoods and for pressure filters if fitted. The vacuum pump and trap for the filters must be located where it can be easily serviced but far enough fiom the laboratory that the noise from the pump will not be heard. A water sealed pump is normally specified. Drains can report to the final tailings sump provided that there is a separate drain for acidic residues that reports to a holding tank for neutralization before discharge. Power available in the laboratory should include some multiphase lines for powering crushers, mills and magnetic separators . REQUIRED EQUIPMENT Sample Preparation

The equipment for a metallurgical laboratory will be somewhat dependent on the particular industry but in almost all cases the requirement to efficiently process accounting and development samples will dictate the laboratory layout. A rotary sample splitter with vacuum filters and sample bucket cleaning facilities will be the first station. Samples will require drying in an oven, which must have very close temperature regulation for sulfide samples to prevent oxidation. The ability to control the temperature between 40 and 100 degrees Celsius will cover the range required for most dry operations. Hot plates may be used to dry samples provided that the technician agitates the cakes to prevent local overheating. The dried filter cakes must be weighed and then screened with the oversize reduced to -75 microns in a roller mill or similar in preparation for most assay procedures. The roller mill or pulveriser should be well insulated to reduce noise and the soundproofing system should be convenient for the technician to install and remove. Modem roller mills are extremely heavy and a pneumatically operated cantilevered arm should be used to move the mill from the grinding cabinet to the dust hood. Rolling, riffling to further reduce the samples size, bagging and addition of an identifying bar code strip on the sample bag complete the process. Hot plate drying, if used, and final sample preparation must be carried out in dust hoods. Ideally, there will be separate hoods and pulverizing equipment for feed, concentrates and tailings to minimize contamination. Face velocities in dust hoods should exceed 0.5dsec (100 fVmin) and the design must cause the dust-laden air to descend to a trap. Care must also be taken to ensure that the noise level created by the air movement and fan does not exceed 75 DBA to provide a satisfactory working environment for the metallurgical technicians. Flooring should be sloped slightly to a drain to allow for wash down and should be “cushion floor” tiled. If ceramic tiles or concrete floors are specified, rubber cushion mats must be provided. A storage area for “reject” samples should be included in the immediate area. Not keeping additional samples fiom both production and development work has caused much uncertainty when there is a failure to balance and a well organized reject sample storage area allows the metallurgist to retrieve samples for mineralogical study or re-assay with minimal expense. A similar storage room should be included for storage of equipment and supplies.

2013

Test Samples Whatever the process used to concentrate the ore in the plant, the metallurgist will require a laboratory equivalent to carry out plant performance monitoring and development studies. An area for air drying samples of ore and equipment for crushing and screening the ore to the size acceptable for a laboratory rod mill (normally between 3.35 -1.7 mm (6 and 10 mesh)) is essential. This should be in a separate room in order to minimize contamination and localize the noise. The room should have a roll up door or at least high double doors to the plant area to allow samples to be brought in on a fork lift truck. Riffles or preferably a rotary splitter will be needed to split the ore into aliquots for grinding and a sample for head analysis. Bar coding of all samples assists in maintaining an easily retrievable inventory. Scales for weighing large samples and the aliquots and a freezer to store samples complete the equipment requirements for the preparation room. Comminution Grinding of ores for flotation is usually carried out with a 2 kilogram charge in a 200mm (8 inch) diameter rod mill with provision in the mill for addition of air or oxygen in order to keep the Redox potential of the ore similar to that in the full-scale mill. It may be necessary to use a mixture of stainless and mild steel rods to adjust the potential to that observed in the plant cyclone overflow. If the plant operation includes a SAG or AG mill, use of ceramic balls as media may be necessary to obtain the correct Redox potential. Mills should be in a counter balanced framework so that no lifting is involved when charging or emptying the mill and complete soundproofing around the mill with a convenient means of removal is a must. Ultrafine grinding such as often required in plants where leaching is the main separation technique are best carried out with an attrition grinders such as those supplied by Metso Minerals (Davey 2002). Care that the Redox potential of the ground product is similar to that obtained in the plant is even more important with such grinding devices. Attritors (sometimes called detritors) should also be used for simulating fine regrind operations.

Flotation/Leaching/GravitySeparation It is preferable to carry out test flotation under a dust/fume hood because of the fine aerosol containing solids produced by a flotation cell. The dust hood must be of adequate height so that the metallurgist will not be leaning forward when using the flotation machine. A vacuum filter station should be located on a bench next to or opposite the dust hood to allow for rapid dewatering of products. Continuous monitoring of physical and chemical parameters such as temperature, pH, Redox and conductivity during flotation testing has become an accepted part of the procedure and facilities to hook the instruments to a datalogger or to the plant network should be provided. There are now several automated laboratory flotation cells available commercially and such an investment will pay back quickly fiom the improved reproducibility obtained. For leach testing, a miniplant line is preferable to the traditional enclosed container on rolls as it is possible to measure and control the leaching environment much more closely. The miniplant system should be hooded and have large enough feed and product tanks to allow the system to duplicate the residence time of the operation. Where plants employ gravity separation, setting up a laboratory scale unit may be feasible depending on the particle size of the feed. In case where spirals are used, a full-scale spiral, agitated stock tank and feed pump can be set up if sufficient ceiling height is allowed. Additional Equipment Sizing equipment comprising wet and dry screening facilities and a cyclosizer will be required. Sieve shakers should be well sound-proofed and be located next to a dust hood for sample weighing. An ultrasonic screen cleaner should be located close to the shaker.

2014

Often missing from modem laboratories is a polished-section making station and optical microscope. While the use of such instruments as the QEM Scan and the Mineral Liberation Analyser (MLA) has injected a much needed quantitative element into mineralogy, the importance of the plant metallurgist being able to identi@ the average state of liberation of the ore and then being able to identify when changes in mineralogy cause a variation in metallurgical response cannot be overemphasized. In remote locations the polished sections can be prepared, photographed and the photographs transmitted electronically to a consulting mineralogist if more specialized assistance is required. ANALYTICAL LABORATORY

Assay laboratory requirements vary widely depending on the industry. Operations with sulfide ores will likely rely on Ion Carbon Plasma (ICP) or Atomic Adsorption (AA) spectroscopy which means that the dissolution of samples in perchloric acid will be required. Hoods where perchloric acid is used must be completely washed down periodically and the wash down solution and other acidic waste from the laboratory must be disposed of in such a way that it does not come into contact with sulfide because of the danger of hydrogen sulfide or sulfur dioxide formation. The same safety equipment is required as for the metallurgical laboratory. Vibration is of major concern for analytical balances and although anti-vibration platforms can minimize the problem, it is still advisable to locate balances close to a major support column and preferably close to an exterior wall. If high grade concentrates, 50% Zinc or 70% lead for instance, are to be analyzed, very great care has to be exercised because of the imprecision introduced by a large number of dilutions or very small sample size. In some instances, especially where an inexperienced staff is required to carry out these determinations, titrimetric or gravimetric procedures should be used. LABORATORY INFORMATION MANAGEMENT SYSTEM

The most significant way in which metallurgical laboratories have changed since 1987 when the previous design manual was published, is in the way the data generated in the plant and laboratory is processed and transmitted to management. In most concentrators the operating data are now captured and stored on a server using software such as the PI system (OSIsoft of San Leandro, California, USA) (Bascur and Kennedy 2002) which allows downloading of any data into spreadsheets or presentation of the data on a personal computer in real time. This software has been used to process laboratory derived data such as assays which are input directly fiom the instrument to a ‘Tag” on the server. The tonnages, also obtained directly fiom scales, and the assays are verified by the metallurgist and transferred into a new set of tags with the daily or shift balance being calculated from the verified data. The balance can be transmitted to management electronically as a spreadsheet so that the process requires no clerical assistance; entry and transcription errors are eliminated but there is a substantial investment in computer software. LABORATORY LAYOUT AND LOCATION

The suggested layout for a metallurgical laboratory shown in the Figure 1 would be expected to service up to a 10,000 tonne per day operation employing flotation as the primary separation process with an uncomplicated flowsheet. The expected complement of a chief metallurgist, metallurgist and two metallurgical technicians would have sufficient space for all routine metallurgical procedures. Office space for the metallurgist and metallurgical technicians should be close to the laboratory; either between the metallurgical laboratory and the analytical laboratory or on the floor above next to the control room. Figure 2 shows a suggested laboratory location below the control room and metallurgist’s offices and next to the analytical laboratory. The on-stream analyzer would be located in the center of the plant on the flotation floor adjacent to the metallurgical laboratory.

2015

METALLURGICAL RESEARCH AND DEVELOPMENT LABORATORY

The major difference between a laboratory for research and development as compared to that in a production environment is the provision of facilities for all of the major separation techniques and also sufficient space and services to operate a continuous miniplant. The requirement to provide estimates of circulating loads in new or revised circuits, which has up until recently been attempted, with limited success, by use of locked cycle tests, is now achieved using a miniplant with feed rates in the 10 to 20 k g h range. Such plants have equipment which is similar in size to the typical laboratory batch rod mill or flotation cell. At the low feed rates, it is possible to obtain satisfactory design data for feasibility costing of a process with perhaps 2 to 3 tonnes of ore, a quantity that can be obtained from drill core. Environmental requirements dictate that discharges fi-om the laboratory must be retained in a tank or lagoon. Drainage reports to a holdings tank were it is treated and assayed before release or disposal. Solids are removed by allowing them to settle in drums with the latter transported to a suitable disposal site -normally a company tailings area. A requirement of any metallurgical laboratory is a fi-eezer and research laboratories should have sufficient storage to keep several tonnes of samples. This necessitates the installation of a walk-in freezer which should have the capability to maintain -3OOC. A design for a mineral processing laboratory that has been found to work well is shown in Figure 3. The layout for this laboratory resulted fi-om the input of all of the mineral processing staff who had dealt with the limitations of their former laboratory in some cases for close to 30 years. The company where this laboratory was installed gave permission for publication of the design but asked that their name not be revealed. DESIGN AND CONSTRUCTION

During the design phase, the metallurgist as the customer, should be expected to monitor the engineering closely to ensure that hisher specifications are adhered to. Similarly, there should be fi-equentvisits to site to ensure that changes initiated during construction do not impair the overall operation of the laboratory. Commissioning of the laboratory must include checking of hoods to ensure that specifications are met, faucets have the correct water, that water is not contaminated and that drains report to the correct sump or sewer. ACKNOWLEDGEMENTS

I would like to thank Andrew Kerr of Inco, Clarabelle Mill for reviewing the document and making many helpful suggestions. I wish to thank Inco Limited for permission to publish this paper. REFERENCES A.N. Bagshaw 1996. The Western Australian Mineral Research Centre: A Resource for Metallurgical Testwork and R&D, Proceedings of the AwZMM, Annual Conference 24-28 March 1996, Perth, Western Australia, Australia 253-257

E Crawley Cooper 1994. The Laboratory Design Handbook, CRC Press, Inc, Boca Raton, Florida, USA C.L. McKenzie and R.A. Haig 1986. Metallurgical Laboratory Design and Operation, Design and Installation of Concentrationand Dewatering Circuits, Eds A.L. Mular and M.A. Anderson, SOC of Mining Engineers Inc, Littleton, Colorado, USA, 655-666 D.C. Shelton 1981. Design of A Mine-Site Laboratory Facility, Proceedings of Asian Mining '81, Singapore, 23-25 November, Institution of Mining and Metallurgy, London, England, 275-284 K Everett and D. Hughes 1981 A Guide to Laboratoty Design ,Butterworths, London, England 2016

6) H.S. Coleman 1951. Laboratory Design, Ed. Reinhold Publishing Corporation, New York,USA 7) Graham Davey 2002, Ultrafine and fine grinding using the METSO stirred Media Detritor (SMD), Proceedings of 34IhAnnual Meeting of the Canadian Mineral Processors, January 2002 , Editor Jan Nesset, Canadian Mineral Processors Division of Canadian Institute of Mining Metallurgy and Petroleum, Montreal, Canada

Osvaldo A. Bascur and J. Patrick Kennedy 2002. Web Enabled Industrial Desktop To Increase Overall Process Effectiveness In Metallurgical Plants, SME Annual Meeting, Feb. 25-27,. PrePrint 02-134, SOCof Mining Engineers Inc, Littleton, Colorado, USA Key for Figure 1

B

Roll up door to ore lay down area Jaw crusher

C

Gyratory crusher

D

Screen

E

Roll Crusher

F

Rotary Splitter for dry coarse solids

G

Scale

H

Freezer

J

Rotary splitter for slurry samples

K

Filter stations

L

Hooded hot plates

M

Drying Oven

N

Sample Preparation hoods

0 P

Pulverizer

Q

Dual Screen Shaker (sound-proofed)

R

Cyclosizer

S

Laboratory rod mill (sound-proofed)

T

Flotation Hoods

U

Attritor

v

Store and Reject Sample Storage

W

Instrument bench

X

Computer work station

Y

Polished section polishing and light microscope

Z

Exterior window

1

Floor drains

2 3

Sinks Door to Analytical Laboratory or Metallurgists’ offices

4

Safety supply cabinet

A

Sample weighing and Screen Cleaning

2017

Figure 1, Metallurgical Laboratory layout

A 4 =I

I------0 (3 El =' 0 0

1; X

0

V

W

0 0

Figure 2, Suggested Location for Metallurgical Facilities in Concentrator

Plan

Metallurgists' and Process Control Offices

Control Room

I

I

I

Elevation

1

I

I

Control Room

I

ORices Y r

Flotation Floor

2018

F i g m 3, Design of a M i n d processing Research and Development Leboratory

2019

On-Line Composition Analysis of Mineral Slurries T.F. Braden, Outokumpu Technology USA Inc. M. Kongas, Outokumpu Mintec Oyj K . Saloheimo, Outokumpu Mintec Oyj

ABSTRACT

In the past 25 years, on-line composition analysis (OCA) of mineral slurries has become a vital part of not only design but also the operation and control of base metal mineral processing plants. Operating plants need OCA information in order to optimize their profitability under varying ore feed and metal market conditions. A description and brief status of OCA methods such as prompt gamma neutron activation (PGNAA), nuclear magnetic resonance (NMR), laser induced breakdown spectroscopy (LIBS), froth image analysis and x-ray diffraction will be presented with a more detailed discussion developed for x-ray fluorescence (XRF). Return on investment for implementation of OCA utilizing XRF is presented. INTRODUCTION

The prime objective when managing a mineral beneficiation process is to ensure that all sections of the available ore body are processed in a manner that maximizes plant profitability. On-line composition analysis (OCA) plays a crucial role in helping to acheve maximum profitability. It must be noted, up-front, that OCA by itself cannot accomplish the goal of maximum profitability, but requires a very close partnership with a process controYmanagement system and other instrumentation. However, irrespective of whether a plant is manually or computer controlled the availability of adequate real-time mformation on the dynamic behavior of a process is of vital importance to its operating efficiency. When control decisions are based on only a limited amount of process information, it can seriously impair the ability of the production personnel to formulate and implement sound control strategies. Ldcewise, when reliance has to be placed on outdated assay data that fails to accurately characterize the short-term trends that exist within a process, it is safe to predict that money is being lost (Barker et al. 1987). BACKGROUND

In years past, it was common practice as a plant level operator or foreman to have a vanning plaque and microscope handy to help illuminate changes in ore type, losses into the tails, or bad actors in the final concentrate. The use of a vanning plaque and microscope required a knowledgeable head, a steady (shaky for the purpose of vanning) hand and a calibrated eyeball. Operating decisions based on the use of a vanning plaque and microscope required an obvious, not subtle, change in some aspect of a process stream. If the eyeball and head were properly calibrated then a change in the process could be recognized and plant operation modified to account for the change. Unfortunately, ore bodies have become lower and lower grade making many changes almost imperceptible to the calibrated eyeball even though the knowledgeable head can almost make-out the culprit. Individual opinions on the plant behavior, which were not based on facts, made consistent operation impossible.

2020

Likewise, it was common in many plants to train operating personnel to accomplish “slop assays” by cutting grab samples from the process, filtering and drying the sample and then performing a simplified mimic of the wet chemical method employed in the analytical laboratory to determine assay. At best this method gave a limited trend of the actual assay and at worst was wrong and misled the operator about a change in the process. Since the day that mining was first begun, it was apparent to all operating and management personnel that there was a need for fast and accurate analysis of the metal composition of a process stream. If such an instrument could be found then not only would plants be able to improve their process control but also the economics of processing lower and lower ore grades would improve. Suffice it to say, as Hales and Marchant (Hales and Marchant 1980) so aptly stated, “the single most important on-line instrument is one that will measure metal concentrations.” The need to measure has always existed but it was not until the last 25 years that we were able to find an instrument that could successfully measure metal concentrations.

DISCUSSION OF ON-LINE COMPOSITION ANALYSIS METHODS OCA instrumentation come in very many types and are based on many different measurement methods. Since the last review of OCA methods, (Cooper 1984), most of these instruments have not changed nor have their methods. Few have found real application in mineral plants around the world. However, it is worthwhile mentioning some of the old and some of the new because of their potential bright future with further development and refinement.

PGNAA Prompt gamma neutron activation analysis (PGNAA) is based on a nuclear reaction occurring between neutrons and atomic nuclei in the sample. Absorption of neutrons by the nuclei releases one or several gamma-ray photons. The gamma radiation is measured using an energy dispersive detector. The radiation intensity at the element specific characteristic energies is dependent on the concentration of that element in the sample. (Tran & Evans, 2001) In principle PGNAA is suitable for analyzing the whole periodic table of elements, with a few exceptions. There is, however, quite large variation in the sensitivity of different elements to the activation process. The neutrons travel long distances through matter without absorption and thus the rate of the prompt gamma reactions is low. The deep penetration of the neutrons and measured gamma rays allows analysis of bulk material on a conveyor belt or in a chute. On the other hand, the sample volume to be measured must be relatively large as compared to other methods, in order to get sufficient sensitivity. With mineral slurries, maintaining a homogeneous suspension in a large volume presents a technical challenge. At low concentrations PGNAA requires relatively long measurement times to achieve sufficient accuracy. Minimum detection limits can reach down to 0.01% for the most detectable elements at a 10 min counting time with 0.1% being a typical value. The most common neutron source used in commercial systems is a Californium-252 isotope source. The 2.6 years’ half-life creates a need for periodic replacement of the sources to maintain the analyhcal performance. In recent years there has been quite intensive development on neutron tubes as an alternative source (Lebrun et al. 1998). However, the lifetime of neutron tubes is still fairly short for continuous on-line operation. The radiation shields required by operational safety make the systems quite large in size and weight.

2021

PGNAA applications in OCA of concentrator plants is limited to cases where XRF does not provide all required analysis information. This is usually the situation when assays of light elements (silicon (Si), aluminum (Al), potassium (K), sodium (Na), magnesium (Mg), phosphorus (P) and sulfur (S)) are of crucial importance to the flotation control. Typical PGNAA systems cost upwards of $500,000 not including the cost of installation.

NMR Nuclear magnetic resonance (Nh4R) is based on absorption of radio frequency electromagnetic waves by atoms that are polarized in a strong electromagnetic field. The method is sensitive to any element isotopes having an odd number of protons and neutrons in the nucleus. The absorption of the resonant fi-equency,or emission after a short RF pulse, in a sample chamber is dependent on the concentration of the absorbing element. The structure of the absorption is sensitive to the molecular form of the sample material. N M R spectroscopy is widely used for studying molecular structures in the laboratory. The only working application for N M R analysis in concentrator processes is phosphorous assaying in phosphate plants (Shoniker et al., 1998). Fluorine analysis has been tested.

LIBS Laser Induced Breakdown Spectroscopy (LIBS) uses a focused Laser light beam pulse directed to the sample surface, causing a small short-lived plasma spark. The plasma a t o m emit characteristic photons in the visible light and near-W region (200-800 nm). The emitted light is collected and transmitted via a fiber optic cable to a spectrophotometer and detected using a CCD element. The laser pulses can be repeated at typically 10 Hz rate, and several spectra are usually averaged to reduce the random variation in the measurement results. Like all optical emission methods, LIBS can in principle detect any elements in the periodic table. (Laughlin et al, 1999; Rosenwasser et al, 2000) The reported minimum detection limits for different elements in liquid or solid samples range from 0.1 to 1000 ppm depending on the application, sample matrix and laser source/spectrometer setup. One of the basic characteristics of the technique is its limitation to a very small sample volume at a time. In a practical setup, the atomized sample in a single spark is of the order of micrograms. The small sample volume can be overcome by scanning the sample surface with a large number of laser sparks. When applied directly to slurry, LIBS has practical representivity problem since the spark occurs only at a very shallow depth. Aqueous systems are difficult for the technique due to strong absorption of the laser energy by water. LIBS has evolved in environmental applications to measure low heavy metal concentrations in soil and metallic particle pollutants in air. So far, the OCA applications have been field tests for iron ore concentrate (Barrette et al., 1999) and copper concentrate.

Froth Image Analysis The visible color, or reflectance of light from the froth surface has been applied to OCA of concentrates in connection with some flotation froth image analysis research. The method is indirect and in practice requires an elemental on-line analysis for calibration. The applications include the prediction of molybdenum (Mo) concentration in a copper ( a ) - M o concentrate as well as zinc (Zn) content in a zinc rougher concentrate (Ylinen et al, 2000).

2022

X-ray Diffraction X-ray diffraction (XRD) measures the concentrations of minerals not elements. Minerals are made up of atoms and molecules arranged in an orderly three-dimensional array - a crystalline lattice. When a x-ray beam is directed onto the mineral surface, radiation is diffracted at known angles characteristic of the crystal structure. Each mineral has its own unique crystal structure and therefore the angle of the diffracted radiation tells what mineral is present. The intensity of the diffracted radiation is directly proportional to the concentration of the mineral. Detection limits for XRD analyzers installed online are typically 250 ppm. XRD accuracy is extremely sensitive to particle size variation since XRD is primarily a surface phenomenon. X-ray diffraction (XRD) has been used to measure mineral or indirectly calculate elemental concentrations in phosphate and potash processes (Saarhelo et al, 1990). X-ray Fluorescence The OCA methods previously in this section have not found general acceptance in mineral processing plants. Whether it is due to their high capital investment or their poor performance, none of these methods have been accepted as being able to successfully measure metal concentrations on a on-line analysis basis. Not mentioned, so far, is the OCA method of X-ray fluorescence (XRF). X-ray fluorescence has found general acceptance in base metal mineral processing, especially in plants utilizing flotation for ore concentration. It has become the dominant and decidedly more successful OCA method. OCA-XRF deserves a more detailed explanation and description in order to do it justice.

W€IY OCA-XRF ANALYSIS? When reliance is placed upon traditional sampling and laboratory analysis techmques, delays of 2-24 hours are common between sample collection and the availability of an assay. The value of the laboratory assay information is further diluted when it is based upon a composite shift sample; as composite samples completely mask the effect of short-term process changes. In contrast, when an on-line x-ray fluorescence (XRF) analyzer is used for continuous realtime, multi-stream surveillance, adverse process trends are detected in minutes rather than hours. This permits the timely implementation of corrective control measures, based upon assays that accurately reflect the prevailing process conditions. Figure 1 compares the performance of the Laboratory versus an On-line XRF Analyzer.

Figure 1.0 Comparison of Laboratory versus On-line XRF Analyzer

2023

On-line XRF analysis of slurry samples is based on some of the the same measurement principles as the assaying of manually prepared samples in the laboratory. The major differences are: - On-line samples are taken in the same way day and night. Laboratory samples are

inconsistent because of manual sampling and preparation. - With an on-line XRF analyzer the sample contains the full range of original particles.

Thus, particle size affects the accuracy of coarse slurry measurements when compared to a homogenized sample in the 1aboratory.On-line XRF analyzer assays reflect the present status of the process. Laboratory assays are out of date and thus may have large errors, if used for process control.

-

When considering all the effects, on-line XRF analysis is more appropriate for process control purposes than laboratory assays. The vast majority of OCA installations in the last 25 years have been XRF analyzers. Based on personal observation, it can be said that there remain very few base metal operating plants that do not have an OCA-XRF analyzer HISTORY OF OCA-XRF ANALYSIS

The history of the development of the OCA-XRF analyzer is well documented by (Cooper, H.R. 1976). In the 25 years since t h s last history was published, the installation of OCA in mineral plants (primarily base metal plants) has started to approach the saturation point, where almost every open and operating plant has an OCA-XRF analyzer installed. The past and recent history of OCA-XRF analysis can be summarized in the following Table 1.O

PERIOD Pioneering (1955 - 1959) Development (1960 - 1969)

Growth (1970 - 1985)

Maturity (1986 - 2002) Future (2003 - ????)

TECHNOLOGY Changes rapidly due to research Changes less rapidly, research and development lead to new discoveries Stable, fewer new innovations, focus on packages that add value for money Stable, no new innovations, quality becomes primary issue New breakthrough offers restart to Pioneering or Development Period

Speculation leads to few beta test sites Less speculative, no industry wide acceptance only plant by plant acceptance Confidence increases, industry wide acceptance, companies start to standardize Approaches saturation with obsolescence becoming issue Speculation leads to few beta test sites

XRF TECHNOLOGY DESCRIPTION X-ray fluorescence (XRF) technology involves the use of an excitation source (x-ray tube or radioactive isotope) to provide radiation used to excite a sample and cause the elements in the sample to fluoresce. Fluorescence (see Figure 2.0) is caused by the radiation from the excitation source, called X-ray quanta, which knock out inner shell electrons in the sample atoms ("excitation"). An electron fiom a higher shell fills the electron vacancy to keep the atom stable. The energy difference between shells is emitted as an X-ray fluorescence photon. The energy of the emitted radiation photon is characteristic to each individual element in the

2024

periodic table. An XRF analyzer contains a detector or series of detectors to measure and convert the emitted radiation photons into characteristic energy intensities corresponding to nickel (Ni) or zinc (Zn). This intensity information is elements, for example, copper (a), then converted to assays using a calibration model.

incident

: :Y % z:@ ? ::J $radiation

fluorescence

.--7

.

.M M

-XK-T&-

--.

Figure 2.0 X-ray Fluorescence Principle The x-ray fluorescence intensities can be measured using fixed crystal wavelength dispersive spectrometers (WDX) for each element and solids content. The alternative is to use energy dispersive (EDX) detectors, which measure the whole fluorescence spectrum. A comparison of WDX and EDX has previously been presented in, “On-stream X-ray Analysis” by Harrison Cooper (Cooper H.R. 1976). The assays from the analyzer indicate the percentage by weight of the concentration of an element of the total solids. To compensate for solids content variation in the sample, the measured element specific intensities are corrected by measuring the backscatter signal from water in the sample. Owing to the absorption of water and solids, the measured fluorescence signal comes only from a thin lmm layer just in front of the plastic window separating the continuously flowing sample from the x-ray source and detector (See Figure 3.0). This window must stay clean and the sample in the sensitive area must be representative and free from air. Sample flow against the window prevents scaling and fills the sensitive layer continuously with fresh representative sample.

Figure 3.0 Description of XRF Analyzer Flow Cell and Analyzer Probe

2025

Analysis of light elements in slurry using XRF (sulfur ( S ) for example) is not possible because the fluorescent radiation from the light element is absorbed by water, the analyzer windows and fmally the air path to the detector.

On-stream versus In-stream Systems On-stream and in-stream analysis refer to the method in which the process stream to be analyzed is presented to the x-ray system. The choice as to whether to install an on-stream or in-stream system requires special consideration. The viewpoint as to which type of system to pursue will largely depend upon what is demanded from the system and how the assay data is utilized. In a plant where the analyzer system is employed to automatically implement assaybased control, the operating staff will demand accurate assays and be committed to keeping the system properly calibrated. However, when a system will be used simply to furnish the operating staff with analytical trends, there is often a tendency to place less importance on accuracy and more emphasis on minimizing operating costs. Three important factors should be taken into consideration when deciding which system to install: (1) representativeness of the sample that is analyzed, (2) ease and quality of calibration, (3) analytical performance of the XRF technology.

Representativeness of Sample When the in-stream method is used, a separate analyzer probe is installed at each process stream; refer to Figure 4.0. The active area of the probe is immersed in the stream and analyzes the material flowing past it. In contrast the modem on-stream system shown in Figure 5.0 can employ one or more analyzer probes, each of which is capable of measuring 1 to 24 streams.

Ruggedly constructed devices manufactured out of steel, lined with rubbed or ceramic, are used to cut continuous samples of 100-300 liters per minute from process flows as large as 5,000 liters per second. These devices are know as primary samplers and are designed to extract a truly representative sample that is characteristic of the whole process flow. Each primary sampler is specifically designed to handle a particular process application and flow conditions. The sample flow is then transported via gravity or pumped to a device known as a multiplexer.

2026

The multiplexer reduces the primary flow to the analyzer’s measuring cell to approximately 20 liters per minute and controls which of the 6 to 18 streams is passing through the system for measurement. The multiplexer, piping and the measuring cell are automatically flushed with water between sample measurements. The multiplexer also ensures that the material being analyzed is free of air bubbles, oversized particles and flotsam such as plastic and wood chips; any of whch produce inaccuracy in the assay results. A level control device within the multiplexer ensures that the measuring cell receives a constant flow. The flow to the cell, and its design, are of fundamental importance to the overall accuracy of the system. The sample is continuously analyzed as it passes through the measuring cell. A thin polymer (mylar) film (window) separates the x-ray tube and detectors from the sample. XRF is a surface measuring techmque in that the analysis region is restricted to a volume of sample very close to the window. (More detailed discussion of this phenomenon will come later in this document.) The cell design must, therefore, ensure that all sizes of particle within the slurry pass through the narrow region at the surface of the window, otherwise, inaccuracy will result due to biased sampling. The cell geometry is designed to produce a high degree of sample mixing by creating turbulence within the sample flow. Scale build-up on the window can create major inaccuracy. One major source of scale can be the reagent chemicals used in the process. Good cell design helps to minimize this serious problem. The turbulence w i h the cell not only creates good sample mixing, it also results in the window being cleaned as a result of the particles abrading the surface. Window ruptures are sensed by the system and alarmed for operators to replace. The latest technology employs a moving window film like a camera that moves continuously between the measuring cell and the analyzer probe. The in-stream method as shown in figure 4.0 reduces the capital and operating costs associated with sampling, sample transport and pumping. However, care should be taken to minimize slurry particle segregation in the process stream. Some type of external mixing device should be introduced into the process to eliminate, or minimize, particle segregation. As in the case of the previously described on-stream method, an in-stream probe has similar sized window and only analyzes a thin layer of sample. The window separates the process stream from an inner protective window and the radioactive source and detector.

2027

In order to generate an accurate assay, it is essential that the material that passes the surface of the window is representative of the total process flow. In the case of a slurry, accuracy is affected by variations in particle size, sample composition, mineralization, the slurry density and variable flow conditions. When immersing a probe into a process stream, there is always a large degree of uncertainty as to the representativeness of the material passing through the thin region at the surface of the window. In the case of a wide launder, the various particle size fractions tend to separate into layers and the stream may frequently meander. Sample bias can also occur in a large vessel where segregation through settling, flotation and/or poor mixing results in non-representative material passing the probe window. Because variations in process flow change the conditions at the window and this impacts upon the analyzer calibration, it is often necessary to minimize these effects by passing the process stream, or part of it, through a steel construction known as an analysis zone, refer to Figure 6.0. When large flows are involved, a primary sampler may also be needed.

Whereas factors such as inadequate sample mixing and variations in process flow, entrained air, coarse particles, flotsam and window buildup can be easily controlled in an on-stream system; they present fundamental problems for an in-stream system requiring careful engineering to avoid or eliminate.

Ease and Quality of Calibration Calibration will be described in more detail later in this document. For the purposes of this discussion calibration is the process of calculating the parameters which the analyzer will use to calculate assays from the x-ray intensities measured. The initial calibration procedure requires a suite of samples to be collected from each point in the process that is assayed. The number of samples required will depend upon the number of elements being measured and the sample matrix. In the case of an on-stream analyzer, the process flow that is being sampled for calibration is the actual sample flow passing through the measurement cell. After leaving the measurement cell, a calibration sample device diverts the sample into a bucket. The calibration sample is then analyzed in the laboratory, where it is weighed, filtered and dried to zero moisture. After sample preparation it is then chemically analyzed. The x-ray intensities measured by the analyzer can then be compared with the laboratory results. The analyzer's on-line modeling and regression programs are used to establish the most suitable calculation model for each assay; i.e. conversion of x-ray intensities into metal concentrations.

2028

A fundamental difference exists in the procedure for collecting calibration samples for an instream system. Due to the longer measurement time of an EDX probe, it is necessary to measure a tailing stream for 5-10 minutes in order to obtain acceptable statistics. During this time period, a series of individual grab samples or a continuous sample must be cut from the process. If the composition or particle size distribution of the sample that is sent to the laboratory differs from the material that flowed within the thin region of the probe window, then the quality of the calibration will be substandard. If the probe is used with an analysis zone, figure 6.0, the calibration sample would be extracted at point “A”, while the analysis is measured at point “B”. Hence the need to insure proper mixing at the probe window and the taking of a complete cut across the overflow at point “A”.

Quality of Analysis EDX versus WDX Most in-stream probes utilize a radioisotope source for generation of x-rays and use the measurement technique known as Energy Dispersive X-ray Fluorescence (EDX). This analysis technique has historically used cryogenically cooled solid state detectors that require liquid nitrogen or room temperature Germanium detectors. However, a new Peltier cooled silicon PIN diode detector is available that does not require liquid nitrogen. The wavelength dispersive x-ray fluorescence (WDX) technique typically uses a x-ray tube to generate x-rays and uses a crystal spectrometer with room temperature proportional counter detector for each element to be measured. Most WDX systems require either air-cooling or water-cooling for the x-ray tube and a high voltage power supply. High resolution is very important when the x-ray lines of the measured elements are closely spaced, as in the case of cobalt (Co) and iron (Fe). Figure 7.0 compares the resolving power of an EDX solid state detector with that of a WDX system. It can be seen from the top spectrum that an EDX detector cannot separate the Co and Fe peaks; whereas, the WDX system resolves the x-ray peaks. At 5.9 KeV the resolution of a WDX detector is 30 eV; whereas, the liquid nitrogen cooled solid state EDX detector is 170 eV. The PIN type EDX detector’s resolution has been reported to be 280 eV.

2029

Consideration should be given to the relative accuracy and sensitivity differences between WDX and EDX systems. For example, Figure 8.0 shows the measured spectrum for a zinc (Zn) tails stream containing .12% Zn. From the figure you can see the Zn peak is barely visible and corresponds to 2 counts per second of Zn (less background). In comparison, Figure 9.0 shows a WDX measurement for the same material where the Zn peak corresponds to 52 counts per second. For the EDX system to achieve a 10% relative accuracy (based on counting statistics only) 1,000 counts must be accumulated. This will require 500 seconds or 8.3 minutes measurement time in order to accumulate enough counts to offer a 10% relative accuracy. For the WDX system to achieve a 10% relative accuracy would require 19.2 seconds measurement. As the concentrations to be measured increase, the difference in the ability of the EDX system to offer an accurate analysis in a reasonable time becomes more achievable.

ENERGY DISPERSIVE TECHNOLOGY (EDX), 0.12% Zn TAILS

20

I 0

0

30

LOV

-

The whole XRF spectrum is measured by energy dispersive detector EDX detector can only measure total of 20 kcounts per second signal, in this case less than 2 counts per second are due to Zn Sensitivity for assays below 0.1 %and small changes is lost because most of the measured pulses are useless

.

Figure 8.0 EDX Spectrum for Zinc tails WAVELENGTH DISPERSIVE TECHNOLOGY (WDX), 0.12% Zn TAILS 2°K-

P

I!

40

0

20

10

30

heV

nly the relevant element peak is measured by wavelength dispersive detectoi *Wavelength dispersive (WDX)detector can measure a total of 30 kcounts per second. In this case there are 52 counts per second directly related to the Zn content @Sensitivityis high for assays below 0.1% and small changes because all measured pulses are valuable

Figure 9.0 WDX Measurement for Zinc tails

2030

Another consideration when comparing EDX and WDX systems is the sensitivity of the system as measured by the minimum detection limit (MDL). MDL is defined as the smallest concentration the analyzer can correctly discriminate from zero concentration. WDX technology can be used to analyze concentrations typically down to 10 ppm in slurries. In contrast, the relative performance of an EDX probe is a magnitude lower, i.e. 100 ppm. In the case of the PIN detector based EDX analyzer the MDL may be 200 ppm or as high as 500 ppm. The above factors need to be considered if a system is to measure very low concentrations. This is especially important if the samples contain elements whose atomic numbers are closely spaced. For analyzing elements lower in atomic number than titanium (Ti), EDX is required. EDX devices can analyze slurries containing elements as light as chlorine (Cl) without special geometry and nitrogen purging. Both EDX and WDX probes can analyze elements up to uranium (U). XRF ANALYZER SYSTEM COMPONENTS

A typical on-stream XRF analyzer system (refer to Figure 10.0) consists of: Primary sampling system Secondary sampling system Analyzer probe Calibration sampler Probe control set Analyzer management station Connection to process control system (digital control system, DCS)

2031

Primary Sampling System The primary sampling system consists of standard proven samplers installed in the process flow, and pipes and pumps for transporting the primary sample flow to the multiplexers and back to the process. The multiplexers are part of the secondary sampling system. The primary sampling system will be discussed in more detail later in this document. Secondary Sampling System The secondary sampling system includes both multiplexers and demultiplexers. Each multiplexer has 1-6 inlets, and there can be up to four multiplexers connected to one analyzer probe. The multiplexers also include an automatic composite shift sampler, which cuts shift samples from the sample streams. The user can define the time interval for the composite sampling. A sample filtering unit can be included as an optional device. For the sample measurement the multiplexers cut a smaller secondary sample, remove air and trash from the sample, stabilize the sample flow, and direct it to the analyzer probe and into the measurement cell. The demultiplexers direct the sample flow back from the calibration sampler to the return pipes. Analyzer Probe In the analyzer probe, the element intensities are measured by using the X-ray fluorescence method. The analyzer probe contains the measurement cell, measurement channels, X-ray tube, high voltage supply, pulse processing electronics, safety interlocks, a cooling device, and an Automatic Window Changer as an optional feature. From the measurement cell, the sample flows to the calibration sampler. Calibration Sampler The calibration sampler is a cross-cutting sampler used for collecting representative samples for laboratory analysis during the calibration intensity measurements. From the calibration sampler, the demultiplexers direct the sample flow back into the return pipes. Probe Control Set The Probe Control Set (PCS) is a field user interface for the analyzer system, which includes a display, indicator lights, and control switches. It also includes processors for controlling the measurement sequence, and the sampling equipment (on demand control is accomplished here), and for transferring data to external systems. The PCS offers a user-friendly approach for monitoring and controlling the system functions. Analyzer Management Station The Analyzer Management Station is a PC, which includes tools for calibrating the analyzer and managing the parameters of the analyzer. The AM Station displays the assays, operational state of the analyzer, and the active alarms. The AM Station also includes service and maintenance tools that allow the support personnel to view the logged diagnostic data from the analyzer.

2032

Connection to Process Control System (Digital Control System, DCS) The DCS connection is a point-to-point serial communication line, which gives information on the statuses, alarms, and concentrations. The user can also make a measurement request in the DCS system; i.e., you can program the analyzer to measure a certain sample. Plant Ethernet network is used more and more to communicate assay information for monitoring and control.

SLURRY SAMPLING TECHNOLOGY A well-engineered sampling system can result in the reliable generation of a representative sample of the process that will operate without blockages at all process conditions. In addition, a properly designed sampling system can result in low operating costs with minimal maintenance requirements. The capital investment required for a reliable and accurate sampling system can be minimized by investing in knowledgeable and experienced sample engineeringprior to the installation and startup of the system. The location of the analyzer in the plant has a decisive effect on the investment and maintenance costs of the sample transport system. Therefore, selecting a place for the analyzer is the most important aspect in the design of the sampling system. Likewise, consideration should be given to not only sample flow to the analyzer but also the sample reject flow from the analyzer back to the process. The turbulence of the slurry flow in process pipes and launders mixes the slurry well in horizontal directions. This has made it possible to develop simple and small stationary (static) samplers, which are placed in the process where the slurry is well mixed. The errors in assays are insigmfkant compared to the errors caused by changes in particle size and mineralogy. In a well-engineered sampling system the use of static samplers can give good quality samples for runtime material balances. For really accurate material balance calculations shift composite samples collected by cross-stream cutter samplers give samples correctly mirroring variable process flows and solids contents. Static samplers are chosen with the following criteria in mind:

1. Provide the correct sample flow rate for the analyzer secondary sampling system 2.

Provide a representative sample of the process flow.

For good representivity, flow velocity into a stationary cutter or nozzle should be engineered to be about the same as the velocity of the bulk flow around the cutter or nozzle. This principle is called isokinetic sampling. If the velocity of slurry into the cutter is too high, more water is sucked from around the cutter. Only the finer, lighter particles are preferentially captured with that flow, causing a deficiency of coarse particles in the sample. If the velocity of slurry into the cutter is too slow, the fmer, lighter particles follow the flow around the cutter causing an abundance of coarser, heavier particles in the sample. Static samplers can be divided into two groups. The first group uses the process pressure at the sampling point which is typically generated via a pump. The second static sampler group, the gravity flow static samplers, have no process pressure at the sampling point. Figure 1 1.O shows an example of a typical pressure pipe sampler (PSA) from the first group which uses process pressure to generate a sample. The PSA sampler, when properly engineered, can use the process pressure to transport the sample to the analyzer.

2033

The second type of static sampler is called a launder box cutter sampler (LSA). Figure 12.0 explains the principles behind this type of sampler.

Sample flow from the static samplers can be continuous or controlled “on demand”. Controlled sampling means automatic stoppage of the sample flow and flushing of the cutter or nozzle and sample pipes, when the analyzer does not need the sample for measurement. All static samplers should be equipped with a shut off valve for the sample flow and flushing valves for upstream and downstream flushing of the cutter or nozzle and sample pipe. In the case of on demand control all valves should have pneumatic or electric actuators.

EXAMPLE OF AN XFW ANALYZER PROJECT Purchasing, installing, and starting up an XRF analyzer is referred to as an XRF Project. This section describes the entire project in a chronological order, from collecting the required data and the feasibility study to making the Customer Support Agreement. The time schedule for this project can be found in Figure 13.0

The project consists of several predefined phases: - Collecting the application data - Checking the feasibility - Defining the project scope and tender Making the business contract Planning

2034

-

-

-

-

Delivering the analyzer Installing Commissioning and startup Calibration Training Customer Support Agreement

A prerequisite for the successful implementation of an XRF analyzer project is an effective co-operation between the supplier and the future users of the analyzer. The project should be led by an experienced project manager from the supplier, an XRF analyzer expert, who ensures that reliable support is provided throughout the XRF analyzer project.

Collecting The Application Data Correct application data is needed for checking the feasibility and ensuring a smooth start of the project. The collected data contains information on the samples to be measured such as: element concentrations, slurry densities, particle size distributions, and environmental information such as electric voltages used on site, ambient temperatures and altitude above sea level.

Checking The Feasibility Before making the business contract, the supplier should study how the requirements of the application can be fulfilled. On the basis of this feasibility check, the supplier and the customer determine how the analyzer should be set up to best suit the particular purpose.

Defining The Project Scope and Tender The project scope and tender are defined on the basis of application data and the feasibility check.

Contract The business contract is signed when both parties have agreed on the modules and options required for constructing an analyzer system suited for the purpose.

2035

Planning Before installing the system, the local infrastructure of the system needs to be planned carefully, including power supply, pressurized air, water supply, and sampling system. There are two important phases in the XRF analyzer project; installation of the sampling and calibration. A well-planned installation is crucial for the proper operation of the sampling system, and a careful calibration ensures accurate and reliable measurement results. The sampling needs to be planned with special care. It is recommended that a supplier expert takes care of the basic engineering. The supplier expert considers carefully what kind of samplers to install, and how to place them in the plant, because this has an effect on the sample quality. The sample pipes and their installation are crucial for the reliable sample flows. It should also be considered whether to use pumps or have the sample flow by gravity or process pressure. The planning takes place on site, and includes both supplier and customer personnel. A proper primary sampling system guarantees a reliable and representative sample, and a suitable sample flow rate for your analyzer system. There are a wide range of different primary sampling components for different sampling situations and process flows, which make it easy to find the most suitable solution for the plant.

Delivery The delivery of the system takes place about ten (10) weeks after the signing of the business contract. During this time, the analyzer system is engineered, manufactured, tested, and shipped to the customer location. The delivery time varies according to the content and extent of the delivery. Installation There are two installation practices used with an XRF Analyzer: The system can be installed entirely by supplier (this includes installation of both the analyzer and the sampling system) The customer can install the system on its own and then supplier checks the installation. Most XRF analyzer systems have a modular structure, and therefore, can be installed in several different configurations. When installing the system, consider the location of the analyzer carefully: The analyzer should be placed in the middle of the selected sampling points to allow easy sample transportation without using costly pumping systems. Insure that the analyzer is easily accessible: there should be enough working space in the multiplexer area and on both sides of the analyzer probe to enable trouble-free operator access and maintenance. Avoid placing the analyzer near the mills due to the noise and vibration, which can cause disturbances in the detectors and also makes for an unpleasant working environment.

2036

A typical on-stream XRF analyzer can be installed with three different layouts:

Two Level Layout (Figure 14.0) In the basic, two level installation layout, the multiplexers are located on the second level, and the analyzer probe and the Probe Control Set (PCS) on the f i s t level.

Low-Head Layout (Figure 15.0) In a low-head installation, all components of the XRF analyzer are on the same level.

Shelter Installation (Figure 16.0) For the convenience of the operating and maintenance personnel it is often better to have the analyzer probe and the probe control set (electronics) located inside an analyzer shelter, and the multiplexers on the roof of the shelter. The Analyzer Management Station can be placed in the shelter or the control room. The shelter can be delivered with an air cooler, which stabilizes the ambient

2037

temperature in the analyzer probe area. Please note, modem OCA-XRF analyzer systems have been designed to meet NEMA 4 and NEMA 4X standards with corrosion resistant materials of construction thereby eliminating the need for a shelter except in the case of operator convenience.

Commissioning and Startup When the XRF analyzer system has been installed, the supplier’s commissioning engineer should check that all parts of the system are properly installed and configured, which is a prerequisite for correct operation and accurate measurements. The commissioning engineer should also assist with the configuration. When the analyzer installation has been approved, the Commissioning engineer will start the system, and check that all parts of the analyzer function properly.

Calibration Calibration contains two simultaneous processes; i.e., collecting the information on the intensities and assays and determining the mathematical formulas for calculating the concentrations from the intensities. Each analysis of each sample flow requires its own calibration. In the calibration process, the measured intensities are registered, and the measured samples are analyzed in the laboratory. This is repeated approximately twenty times for each process line. The results of both the laboratory and the analyzer are compared to fmd a mathematical correlation formula using a calibration tool llke the one shown in Figure 17.0. The samples should be taken from a varying range of circumstances, i.e., from low and high element concentrations. In the implementation project, calibration is divided into two different phases; a preliminary and a final calibration. The preliminary calibration includes fmding the first calibration model for the analyzer. The final calibration includes collecting observations from all process situations of a normal range, and calculating the equations for the calibration.

2038

Preliminary Calibration: During the training, a preliminary calibration takes place. 1020 calibration samples are taken and analyzed by the customer. The time period between the startup of the analyzer and preliminary calibration is approximately 1 to 2 weeks. The preliminary calibration includes finding the first calibration model for the analyzer. From the preliminary calibration onwards, the customer staff is fully capable of taking the control of the analyzer, and the analyzer can already be used normally in the concentrator, although the measurement accuracy has not reached its peak yet.

Final Calibration: During the final calibration, the customer continues taking calibration samples. The final calibration includes collecting observations from all normal process situations (approximately 40-70 observations from each stream). The samples should cover the variation found in the process. On the basis of these observations, the equations for the final calibration are calculated. There is remote assistance available for these calculations. Final calibration can take several months to be completed, depending on the plant and the stability of the process. After the final calibration, the analyzer reaches the best measurement accuracy. The peak performance is maintained by checkmg the calibration for process changes. Maintaining the Calibration: In maintaining the calibration, new observations ought to be collected approximately once a week in order to ensure accurate measurement results. In a year, adequate observation material is collected to recalculate totally new coefficients of the calibration equations. In case of a reformed process or a new ore deposit, the recalibration is performed as necessary. An important part of maintaining the calibration is the availability of a dedicated software tool for calibration of the analyzer. Figure 17.0 shows some typical outputs of such a tool. As can be seen in the figure, all the regression variables are visible and a relative error plot shows the comparison of the laboratory and the analyzer assays.

2039

Training The supplier should provide XRF analyzer training either on-site or on supplier’s premises. The training is very practical; the analyzer is operated normally already during the training period. The training includes collecting the first samples for the calibration. The training is based on continuous interaction between users and the trainer, and so, users are able to influence the contents of the training. Customer Support Agreement The XRF analyzer customer support agreement insures that the analyzer system will be checked twice a year to prevent degradation of the analyzer system performance. This can be considered a preventive service to insure the analyzer performs the same after two years or two days of operation. The agreement should also include a Remote Diagnostic System (see Figure 18.0) which is provided by using a modem connection for direct communication with the Analyzer Management Station. Regular training conducted onsite should also be included in the contract. Support of calibration after completion of final calibration should be included in the customer support agreement to insure that the best possible calibration is maintained throughout the life of the analyzer system.

OPERATION AND MAINTENANCE OF XRF ANALYZERS The XRF analyzer has been designed for trouble-free operation and easy maintenance. An XRF analyzer does not require a full-time operating individual for daily support, but only regular check-ups to ensure that the analyzer is functioning properly. Operation and maintenance tasks are generally shared between four different user types in the following way: Metallurgist The metallurgist maintains the calibration of the analyzer. Responsibilities include reporting on the shift values, and observing as well as maintaining the calibration.

2040

Process operator The process operator monitors assays and uses the measurement results for running the process. The measurement results are available in the Analyzer Management Station and the Process Control System. The alarms and status of the analyzer are also available for the Process operator in the Analyzer Management Station. Floor operator The floor operator is responsible for the maintenance of sampling equipment. The tasks include cleaning the trash screens daily, maintaining primary sample flows, changing the measuring cell window and unplugging the sample lines and samplers, if necessary. Instrumentation staff The instrumentation staff is responsible for the maintenance and troubleshooting of any operating problems with the analyzer. Specific training on component replacement is an important part of training received from the supplier. The responsibility of maintaining the analyzer is typically shared between the supplier and the customer instrumentation staff.

OPERATION SAFETY A typical X R F analyzer includes a reliable security system. It has a number of internal sensors that indicate if abnormal situations occur in the analyzer. In the case of an abnormality, the measurement cell flow is automatically bypassed, the measurement functions are disabled, and the X-ray radiation is shut down. The operator safety is guaranteed if, the regulations and safety instructions are followed at all times.

Safety and licensing requirements differ between the EDX and WDX systems. When a radioactive source is used rather than a x-ray tube, special licensing and periodic inspection are required in most countries. With a x-ray tube, there is no radiation when there is no power to the tube, so no special licensing or inspection is typically required. It is important to check with local officials to insure that the correct licensing and inspection is done for the type of system used. Likewise, handling, replacement and disposal of radioactive sources requires special consideration. Check with the supplier of the source to quantify the requirements for disposal.

WHAT TO EXPECT FROM MODERN OCA-TRACE The current state-of-the-art in OCA, as we have been discussing so far, is the on-stream XRF analyzer system. Using the characteristic of this system we can build a picture of what is now possible in OCA. The factors that can be used to build this picture are:

-

Timely assay information

-

Reliable system performance

-

Accurate measurement of metal concentrations

-

Cost Effective investment as measured by capital costs, operating and maintenance costs and-balanced by payback or retum on the costs applied

2041

Timely Assay Information: Measurement time of the XRF technology is a key component of timely generation of assay information. Typical performance runs from 1 minute to 5 minutes per stream analyzed for the simultaneous measurement of up to 6 to 10 elements. The absolute best achievable at this time is 15 seconds for high concentration streams like the final concentrate. The feed and the tails stream would require 30 seconds measurement time. Please note, that the physics of the XRF measurement and the type of equipment used tie measurement time and accuracy together. For the sake of this discussion the measurement times used are those that can achieve the instantaneous accuracies mentioned later in this document. Another important component of this timely generation of assays is the multiplexing of more

than one stream past the same analyzer. Multiplexing makes the overall system more cost effective because for a minimal increase in capital cost more potential payback can be added to the system through measurement of additional streams. Multiplexing means that after each stream is measured there must be a draining time to allow the majority of the previous stream to remove itself from the multiplexer. Then a water flush must be made to clean-out the rest of the previous material, followed by a filling time required to change over to the next stream and reach a stable environment to make the next measurement. Typically this drain, flush and fill cycle requires from 20 seconds for tails and feed streams to 40 to 60 seconds for frothy concentrate streams. A 18 stream multiplexed system is capable of measuring all 18 streams in under 15 minutes and a 12 stream system capable of 10 minutes measurement time. All of these system times (measurement, drain, etc.) must be accomplished within the feedback needs of the overall plant control system in order to provide timely information for control purposes. If the analyzer does not complete the measurement fast enough then the control system has to either wait for the information or develop its own calculated measurement. Reliable System Performance:

Reliable system performance is measured by the total availability of the XRF analyzer system components. Components such as primary sampling, secondary sampling and analyzer probe have their own individual availability that influences the total system availability. In practice, total system availabilities of 95% or better have been achieved, (Jensen D.L. 1999). In order to achieve a total system availability of 95% the component system availabilities must also be very high. Experience has shown that analyzer probe availability is 99% over a period of one year. Secondary sampling system availability is also 99% due to the simplicity of the design and abrasion resistant materials of construction. The primary sampling system availability is the most variable. This is most often due to poor sample system design, improper installation, and the typical wear and scaling of lines and failure of components such as pinch valves, solenoid valves and coils to drive the valves. Once design and installation problems have been eliminated then the sampling system availability can approach 95%. “On demand” control of sample line flushing offers to maximize the sample line life and built-in system alarms for flowrate changes allow for quick determination and repair of sampling system failures and can minimize downtime for replacement of a failed part. XRF analyzer system availability of >95% is possible and achievable when such automatic flushing controls and flow alarms are utilized.

2042

Accurate Measurement of Metal Concentrations: The accuracy of a measurement is a function of the sample parameters such as matrix composition, mineralization and particle size. For the purpose of using the measured results for process control the accuracy’s being discussed are instantaneous accuracy’s. That means that if you were to take a calibration sample at the exact time the measurement is being made by the analyzer, then the relative error between the calibration sample and the analyzer would define the instantaneous accuracy of the analyzer. Under normal operation conditions, the following relative standard deviations can be achieved for individual slurry sample measurements of concentration levels well above the relevant minimum detection limits of the analyzer: Minor concentrations 3-6% Major concentrations 1-4%

Cost Effective Investment: The typical capital investment* for an on-line XRF analyzer system on a per stream basis is: 24 stream system = $22,000 - $40,000 6 stream system = $50,000 - $68,000 1 stream system = $82,000 - $102,000 Consumables costs* for an on-line XRF analyzer system on a per stream basis are: 24 stream system = $1,500 - $2,500

6 stream system = $1,000 - $2,000 1 stream system = $500 - $750

Operating costs* for an on-line analyzer system on a per stream, per year basis are: 24 stream system = $250 - $780 6 stream system = $900 - $1,125 1 stream system = $1,100 - $1,600 *Depending on the system type, physical plant layout, pumping requirements and sampling requirements. Consumables does not include the cost for liquid nitrogen.

BENEFITS OF XRF The benefits of having an OCA-XRF analyzer in a plant have been documented in a few cases (Jones et al. 1991; Jensen 1999; Kongas et al. 2001; Lahteenmaki et al. 1999). Remember it’s not the OCA-XRF alone that gives the benefits but a control system to automatically act on the assay information must be considered. Looking at these documented results shows, at least, in one case (Jones et al. 1991), the payback for the investment was on the order of 4.8 to 5.3 months. Benefits such as improvements in plant metallurgy (grade and recovery) are typical (see Figure 19.0) and are necessary to insure a good payback for the investment. Other benefits that become obvious after implementation are:

-

-

better understanding of the process, i.e. coarser grind leads to lower recovery or high cleaner circuit circulating loads result in significant losses calculation of material balances leading to determining process bottlenecks ability to on-line test chemicals or process changes in a before and after scenario

2043

-

-

helpful training tool for new operating personnel lower reagent costs when combined with an assay based control strategy provide on-line metallurgical accounting

When speaking about the benefits of having an OCA-XRF analyzer it is amazing that only a few documented cases of actual calculated payback can be found. Either most of the systems installed did not achieve any payback or else no one has bothered to calculate the payback or publish it. There is a large positive benefit of doing a payback analysis. It serves as a confinnation of the economic justification of the purchase and serves as a future reference when the OCA-XRF system becomes obsolete. SYSTEM OBSOLESCENCE

Since the introduction of on-stream XRF analyzers over 35 years ago many operating plants have experienced the phenomenon known as analyzer obsolescence. This phenomenon relates to the fact that technology, hardware and software changes quickly and often components cannot be replaced because manufacturing has moved on to the latest technology or hardware component. This leaves a plant in the position of having increased maintenance costs and having to replace not just a component but to upgrade the whole analyzer. Currently the economical lifetime of an XRF analyzer is between ten and fifteen years. It is important to have a proactive plan to either upgrade the obsolete parts or replace the analyzer system before reaching the point of analyzer death. The benefit of being proactive is to maintain the original benefit of the analyzer system and not lose it to higher operating costs and lower system availability.

WHERE ARE WE? A PHILOSOPHICALPERSPECTIVE We are at a point where the distribution of plants that have OCA-XRF analyzers and are using them for plant control can be described by the distribution shown in Figure 19.0. On the lower end where a plant does not have an OCA-XRF there are only a handful of plants. On the upper end of where plants are utilizing their OCA-XRF in a plant-wide optimizing or economic based control philosophy there are only a handful.

Figure 19.0 Histogram of Maximized Profitability versus Degree of Process Control

2044

In between these extremes are 90% or more that are not utilizing their OCA-XRF to the fullest extent for one reason or another. Some of those reasons are: 1. Ore body is simple and therefore requires only simple process control, only need assay trends for emergency actions 2. Process control strategylsystem too complicated or requires major investment in time andor manpower to change 3. OCA-XRF system project failed to deliver a properly designed sampling and analyzer system, resulting in maintenance problems and low system availability 4. OCA-XRF system calibration not maintained properly

5. The overall operating philosophy of plant management does not support a high level of process control

The Future of OCA There is still room to improve the reliability of existing sampling systems by utilizing “on demand” flushing and control. Likewise, calibration and analyzer probe maintenance can be improved via remote diagnostic support relieving the plant from having to rely on a high level of local expertise. New hybrid systems that combine the benefits of EDX and WDX are available. These systems allow each installation to optimize the cost versus benefits associated with whether WDX or EDX is used. If we polish our crystal ball and look into the near future, it will likely show that OCA will, still, be based primarily on XRF technology. However, XRF technology will be complimented with particle sue analyzers, video imaging, and specialized statistic and control packages. The combination of XRF technology and video imaging has already been demonstrated (vanOlst et al. 2001). Likewise, on-line mineral balances and net smelter return calculations have been accomplished with the right combination of technologies (Holdsworth 2002). Improving the measurement of OCA via combining OCA-XRF assays with other faster process measurements will probably be the wave of the near future. In the slightly more distant future the direct on-line measurement of minerals and perhaps even the on-line measurement of the degree of liberation of minerals will be possible. Already some success off-line using a digital scanning electron microscope to characterize ore and mineral types has helped improve plant efficiency (Lotter et al. 2002). In another case, results from the digital scanning electron microscope have been used to calculate the phasespecific surface area (PSSA) and have correlated it to metallurgical results in laboratory scale tests (Winckers A.H. 2002). Now we’re talking liberation! Wouldn’t it be ironic if the next breakthrough in on-line composition analysis would be based on using the old microscope on the operating plant floor?

2045

REFERENCES Barker, D.R., H.J. Melama, T.F. Braden 1987. The use of on-stream analysis and computerbased process management for concentrator control applications. Paper presented at American Mining Congress convention 1987. Barrette L., S. Tunnel, J.A. Boivin, M. Sabsabi, T. Martinovic, G. Quellet, On-line iron ore slurry monitoring using laser induced plasma spectroscopy, in Control and Optimization in Minerals, Metals and Materials Processing, 38&Annual Conference of Metallurgists of CIM, Quebec City, Quebec, Canada 1999. Cooper, H.R. 1976. On-stream x-ray analysis, Published in Flotation A.M. Gaudin Memorial Volume, ed. M.C. Fuerstenau, Chapter 30, New YorkAIME. Cooper, H.R. 1984. Recent development in on-line composition analysis of process streams. Published in Control '84 Mineralhletallurgical Processing, ed. J.A. Herbst, Chapter 4, New York:SME. Hales, L.B., G.R. Marchant 1979. Instrumentation. Published in Computer Methods for the 80's in the Mineral Industry, ed. A. Weiss, Chapter 5.2, New York:SME. Holdsworth M., M. Sadler, R. Sawyer 2002. Optimizing concentrate production at the Greens Creek mine, SME preprint 02-063. Jensen, D.L., Flotation supervisory control at Cyprus Bagdad. Published in Advances in Flotation Technology, ed. B.K. Parekh and J.D. Miller, Section 6,433-440. Jones, J.A., R.D. Deister 11, C.W. Hill, D.R. Barker, P.B. Cnunmie 1991. Process control at the Doe Run Company. Proceedings Plant Operator's Symposium, SME Annual Meeting 1991. Kongas M., K. Saloheimo 2001. When is the XRF assay good enough for process control. SME preprint 01-189. Lahteenmaki S., J. Miettunen, K. Saloheimo 1999. 30 years of on-stream analysis at the Pyhasalmi mine. SME preprint 99-147. Laughlin A. W., C. R. Mansfield, D. A. Cremers, M. J. Ferris , Real-time, in situ analysis of exploration samples using LIBS , Preprint 99-105, SME Annual Meeting, Denver, Colorado, 1999. Lebrun P., P. Le Toumeur, B. Poumarbde, H. Moller, P. Bach, On-line analysis of bulk materials using pulsed neutron interrogation , 15th Int. Conf. on Applications of Accelerators in Research and Industry, Denton, Texas , 1998. Lotter N.O., P.J. Whittaker, L. Kormos, J.S. Stickling, G.J. Wilkie 2002. The development of process mineralogy at Falconbridge Limited, and application to the Raglan mill. Proceedings 34' Annual Meeting of the Canadian Mineral Processors. Session I. Paper 1. vanOlst M., N. Brown, P. Bourke, S. Ronkainen 2001. Improving flotation plant performance at Cadia by controlling and optimising the rate of froth recovery using Outokumpu FrothMaster. Proceedings 33rdAnnual Meeting of the Canadian Mineral Processors. Session I. Paper 3.

2046

Rosenwasser S., G. Asimellis, B. Bromley, R. Hazlett, J. Martin and A. Zigler, Development of a Method for Automated Quantitative Analysis of Ores Using LIBS, LIBS 2000 1st International Conference on Laser Induced Plasma Spectroscopy and Applications, Pisa, Italy, 2000. Saarhelo K., U. Paakkinen and P. Pennanen, On-line Analysis in Industrial Mineral Applications,'8 Industrial Minerals International Congress, 1990. Shoniker Joe and Ronald Vedova Roger L. Vaughn, PCS Phosphate White Springs automatic control and on-stream analysis innovations have pay-off in big gains, Engineering Foundation Conference on Phosphate Beneficiation, White Springs, Florida, 1998. Tran K. C. and M. Evans, Intelligent instruments for process control of raw materials in the basic industries, Preprint 01-34, SME Annual Meeting, Denver, Colorado, 2001. Winckers A.H. 2002. Metallurgical mapping of the San Nicolas deposit. Proceedings 34" Annual Meeting of the Canadian Mineral Processors. Session I. Paper 3. Ylinen R., J. Miettunen, M. Molander, E.-R. Siliamaa, Vision- and model based control of flotation, Future Trends in Automation in Mineral and Metal Processing, F A C Workshop, Finland, 2000.

2047

Introduction to Process Control B. Flintof

ABSTRACT Process control for the stabilization and optimization of mineral processing circuits first emerged in the late 1960’s, and over the past three decades has evolved into one of the most capital effective investments available to mill management. Consequently, the art and science of applying the technology has emerged as a core competence in the industry. This chapter provides an introductory overview of process control, including elements of both practical and theoretical interest. The intent is to provide a systems oriented fi-ameworkfor tlunking about process control by exploring the constituent elements of: process; measurementlmodulation; hardware; strategies, users; and, maintenanceldevelopment.

PREAMBLE Process control in mineral concentration has been around for many decades. Over time, the balance between manual versus automatic regulation and/or optimization of the constituent processes has changed very significantly. While it is a little difficult to point to an event, or even a period that signaled the move by operators to accept automatic process control, the author believes that the early ‘70’s marked the point of a profound change in thinlung. The catalyst was the minicomputer, whch combined I/O and HMI devices with high level programming languages (e.g. BASIC, FORTRAN), all in a relatively inexpensive package. By today’s standards these machines were extraordinarily primitive, but for the first time process engineers had a means to easily develop, test, and modify process control strategies. Moreover, operators also had a tool that marshaled and presented relevant process data and information in a form that would enable them to make better, faster, and more systems-oriented decisions. An interesting byproduct of the introduction of this new technology was that it provided the common ground for operators and engineers to discuss and implement ideas aimed at process improvement - a pre-cursor to what later became know in the “Quality culture” as self-directed teams. The returns associated with these process control investments were often very impressive (see Chapter 2, Flintoff and Mular, 1992, for example). The emergence of the PLC in the late 1970’s and the DCS in the early 1980’s as design standards for new plants was perhaps the concrete indication of a wide scale embrace of automation by the mineral processing community. In this period, university curricula were also modified to include courses on process control along with related technologies such as modeling and simulation. These were necessary to prepare graduates to fill an implementation void that the rapid advance of instrumentation and control hardware had created at the process interface. Paradoxically, the old succession model was inappropriate, as for the first time the young staff were required to mentor the older staff on the art and science of digital control. The 1990’s and the first few years of the new millennium have seen steady progress in process control development. From a technology perspective, the principal drivers have been: (a) the PC, which has changed the thinlung and expectations around control hardware architecture; (b) advanced control , whch has emerged in robust and effective implementations; (c) digital bus technology, which is changing the way we hnk of implementing process control; and, (d) the Internet, which while it has not yet had any direct impact on process control, has dramatically and very quickly raised the level of computer literacy and HMI expectations of operators and engineers alike. There is little doubt that management now recognizes process control as one of the most capital effective investments available in the pursuit of shareholder value. Moreover, as automation technologies evolve in mining, one now sees growing interest in the digital integration

’Metso Minerals - Minerals Processing Business Line, York, PA 2051

of the enterprise, enabling the application of business controls, the next level of “automation.” Process control is a key enabling technology in this new field and it promises to be an interesting time for those challenged to bridge the plant floor with the board room. The process control journey is destined to continue, as we explore improvements in existing systems and approaches, as well as new applications. That makes t h s volume an important contribution, as it benchmarks the current status of a technology that is ever in flux.

INTRODUCTION TO PROCESS CONTROL This introduction is intended to provide a top-down conceptual overview to one of the most important technologies process engineers have at their disposal today - process control2. More specifically, it aims to present a framework for the reader to think about this subject. To do this, the author has chosen to begin by examining process control in the broadest context, and then to drill down to the fundamental elements that will be discussed in the companion chapters.

Figure 1 A Categorization of process control Most of us workmg in the area are familiar with control categorizations (segmentation by level) such as illustrated in Figure 1. That is, we think of process control as the measurement and pre-processing of field data, which is then presented to operators andor algorithms, that then decide whether manuaUautomatic changes are required to final control elements, or other variables outside of the control system. In other words, process control is focused on the regulation and optimization of equipment or circuits in ”normal” operation, and actions are based on objectives set periodically by management. Figure 2 is based on an adapted form of the Computer Integrated Manufacturing (CIM) model by Bhatt, 1992. It illustrates process control in the context of a much more holistic information technology hierarchy. In thls figure the information derived from, and actions taken by the control system are now synchronized with asset management (e.g. maintenance planning and scheduling, procurement, etc. functions) and with business systems (e.g. mine planning/dispatch, concentrate marketing & sales, shpping logistics, etc.).

The reader more interested in process control theory is referred to the standard college text books on the subject, including Stephanopoulus, 1984 and Seborg et al., 1989.

2052

Figure 2 Connecting process control with business controls

Succinctly, Figure 1 is more concerned with Process Management, while Figure 2 is representative Production Management, and for the purposes of this chapter the working definitions for these two terms are as follows. Process Management: That set of activities focused on the management of operating equipment and/or circuits that comprise a unique process in the production value chain. These activities are necessitated by changes in the processing characteristics of the feedstock(s) and/or changes in production targets. This is achieved by the regulation of certain material inputs, process operating conditions, and product quality through the optimal deployment ofpeople and technology. (Process control is an integral part ofprocess management.) Production Management: That set of activities focused on the overall management of the constituent processes in the production value chain. These activities are necessitated by such factors as the maintenance requirements of the process(es), and by changing markets andlor corporate business goals. This is achieved by the development of enterprise wide operations and maintenance schedules as well the establishment of process specific production targets and the communication of this information to the process and asset management subsystems.

The focus of this chapter is “process control,” as generally defined by Figure 1. Figure 3 provides a more detailed look at the elements that comprise an effective industrial process control system. As the phrase implies, “process control” begins with process. Process can have as significant an impact on the performance of a control system as any of the other (control) elements. The instrumentation layer includes sensors to make measurements of process variables, and final control elements used to manipulate variables in the field. The control hardware layer comprises all of the Input/Output (I/O) subsystems, computer gear and operating software, as well as the peripheral devices for data presentation and storage. Control strategies consist of a blend of more traditional regulatory and advanced controls to develop a robust and effective package. This layer often includes some ad hoc functionality unique to the process, and intended to enhance the robustness of the application. The user group layer comprises: the instrumentation and electrical technicians who supporf the instrumentation, hardware and regulatory controls; the process and control engineers wh support and develop the strategies; and, the operators who manage the application and pr ide expert input as required. Last, but certainly by no means least, are the maintenance, training and system development programs that are essentially intended to sustain and improve the various elements of the system. This latter element is probably the Achilles’ heel of a control system, and while things seem to be much better these days, the 1970’s and 1980’s included numerous examples of well performing systems that soon fell into disuse because of the

J’

2053

lack site champions and formal programs of support. Finally, the structure of Figure 3 is strongly correlated with the chronological development of the discipline in mineral processing. This starts with on-line composition analysis (beginning in the 1960's), moves through the minicomputer era (beginning in the 1970's), to the control strategy development (beginning in the 1980's), and on to the user group (beginning in the 1990's).

Academics generally focus on the first three layers (instrumentation + hardware + strategies), which collectively have been termed the Control Triad, depicted in Figure 4. It is clear from the collection of papers in this session that these are very important points. From an operator's perspective, perhaps the most important of these is the apex dealing with strategies, as these are arguably the most unique, depending upon the process, the field instrumentation complement, the nature of plant disturbances, and the business goals of the corporation. Just to illustrate this point, even the advanced control of similar grinding circuits with similar objectives shows at least 20% customization in the application.

Figure 4 The control triad (after Herbst and Bascur, 1984) Figure 5 drills down into strategies and there are several interesting points to be made. The first is that strategies are hierarchcal in nature, e.g. an effective supervisory control strategy can only be built on a foundation of effective regulatory controls, and so on. What is different is that the level of infrastructure investment (in field instrumentation and computer hardware, etc.)

2054

decreases as one moves up the hierarchy, whereas the intellectual investment increases. Secondly, there are performance related benefits realized by each level of control strategy. In the past, one of the more difficult choices was to determine when one was approaching the point of diminishing returns on the development input/output curve. That is, when does one abandon the tools at one level, and move on to the next. De fact0 standards are emerging based on the industry's collective experience.

Finally, process control (e.g. as seen through research publications, undergraduate courses, etc.) can sometimes give one the impression that t h s is a highly mathematical, typically complex, often counter-intuitive science. All of that can be true, particularly as one moves up the hierarchy shown in Figure 5. However, we would all do well to remember that the general goals of process control are really quite straightforward, and as illustrated in Figure 6 , they are simply to: (a) squeeze the variance - i.e. demonstrate that the process can be controlled; and then (b) shift the target - i.e., exploit the benefits of control by maintaining optimum targets. 50' 40'

(1) Squeeze the Variance

Q

g30~

G20~ 10' 0

10

I

-lo

15

20 25 Controlled Variable

30

15 20 25 Controlled Variable

30

Figure 6 A graphical interpretation of the goals of process control

2055

The balance of the paper takes a closer look at some of the issues and trends in process control, using the elements of Figure 3 as the framework. The maintenance and developments aspects will be addressed within the other layers.

PROCESS Over the past couple of decades there has been a move to simplify the design of mineral processing flow sheets. Two indicators of t h s trend would be very large equipment (e.g. SAG mills of 12.2 m diameter) and a move to more open circuits in flotation. The first indicator dictates the need for very effective process control because the operation depends more and more on fewer parallel circuits, and in many cases just a single very large process line. The second indicator results from what has been termed in industry as the “Back to the Basics” movement (e.g. Stowe, 1992), and an example is presented in this section.

To introduce the problem, consider the hypothetical flotation circuit illustrated in Figure 7. In this instance we suppose that it is desirable to regulate the rougher tailing grade, perhaps by the manipulation of airflow. The existence of a recirculating stream from the cleaner tails can induce control problems. The algebra in the figure demonstrates the relationship between tailings grade, feed grade and the recoveries in the rougher and cleaner banks. If the recirculating load is very low (R2 = 1), the response of the circuit to, say, a decrease in air flow in the rougher bank shows a “first order” like increase to the new steady state, as illustrated in the graphic. If, on the other hand, the recirculating load is very large, then a decrease in air flow will initially cause much of this material to be rejected in the tailings stream, but with increased rougher residence time the rougher tailing will eventually stabilize at a level quite close to the value prior to the change in air flow, again as shown on the graphic. Intuitively, one would expect that process control would be easier to implement and more effective in the case of a very low recirculating load. This simple contrived example illustrates a couple of points. Firstly, it helps to explain why older design procedures tended to favor flotation circuits with a fairly high degree of recirculation. They were inherently self-regulating, except in the presence of a sustained disturbance, which would ultimately necessitate a compromise on grade and/or recovery. Secondly, it explains the recent move to more open circuits which often provide superior metallurgical performance, but which demand a greater degree of process control to maintain operational efficiency. Morari, 1983, has used the term resilience (see definitions below) to capture this notion of the interaction between process and control. “Resilience - Describes the ability of the plant to move fast and smoothly from one operating condition to another (including start-up and shut-down) and to deal effectively with disturbances.” “Dynamic Resilience - the quality of the regulatory and servo behavior which can be obtained for the plant by feedback .”

2056

Not surprisingly, the impact of the process on the quality of process control has been addressed by leading control theoreticians: e.g.- “However, it has long been recognized by both industry and academia that modlfications of the physical system itself can sometimes afect the resilience significantly more than changes in the controller.” (Morari, 1983) or “The relation between process control and design is also important. Control systems have traditionally been introduced into a given process to simplifi or improve their operation. It has, however, become clear that much can be gained by considering process control and design in one context. The availability of a control system always gives the designer an extra degree offreedom, which frequently can be used to improve performance or economy. Similarly, there are many situations where dificult control problems arise because of improper design.” (Wstrom and Wittenmark, 1984) To conclude this section, and in keeping with the flotation theme, Figure 8 illustrates the circuit re-engineering effort at a Canadian copper-zinc concentrator (Stowe, 1992). As suggested above, the much more open circuit on the right was easier to operate and control, leading to significant performance improvements. It is important to note that in this case there was no new process equipment, sensors or hardware, simply a redeployment of existing assets.

Figure 8 Process re-engineering to improve (dynamic) resilience INSTRUMENTATION Field instrumentation includes sensors (e.g. flow, temperature, density, composition, pressure, level, etc.) and final control elements (e.g. valve position, variable speed drives, etc.), as depicted in Figure 9. Given the preponderance of devices in the sensor category, it is not uncommon to hear sensors and instrumentation used synonymously. Accordingly, this section will focus principally on sensors.

2057

Figure 9 A flow control loop starting and ending with instrumentation Many of the sensors required to develop a solid foundation of regulatory controls exist, as will be illustrated below. Although the slurry systems frequently encountered in mineral processing have presented some design challenges, quite a large number of the sensors (e.g. flow meters, level monitors, pressure gauges, etc.) in use are common to other process industries. Despite this broad usage, sensors are not always trouble free. The fact that there are still arguments over the general merit of slurry flow measurement underscores the point, i.e. you can still find people who claim success and others who claim failure using magnetic or ultrasonic flow meters. About the only thing one can be sure of is that t h s is generally not the fault of the instrument. Pareto's rule applies, as -80% of the problems arise from improper application and/or installation. The point being that one must take care in selecting, installing, calibrating and maintaining field instrumentation. Table 1 is intended to illustrate the range of sensor technologies that are commonly applied in grinding circuit process control. Clearly, there are many choices of technologies, and w i t h each technology there is generally a good choice of suppliers. However, what one really needs from t h.s. rather extensive list is dictated by the circuit disturbances and the strategies employed to reject them. ~~

'able 1 Sensor technology choices for comminution systems (source Herbst et aL, 2002) Technology Employed Measurement Bin Level (solids) ultrasonic devices, laser devices, load cells, mechanical devices 0 TankLevel ultrasonic devices, capacitance probes, differential pressure (slurry/water) devices, conductivity probes, mechanical devices current transducer (+ conversion), power transducer, torque meter 0 MotorPower Solids flow 0 electronic belt scale, nuclear belt scale, impact meter Slurry Flow 0 magnetic units, ultrasonic units Water Flow 0 vortex shedding devices, turbine meters, differential pressure devices microwave units Moisture (dry solids) radiation gauges, U tubes, differential pressure devices 0 Moisture (slurries) 0 diaphragm devices 0 Pressure VibratiodSound accelerometers/microphones 0 thermocouples, resistance thermal devices, IR imaging Temperature Particle Size (dry image analysis techniques solids) ultrasonic devices, mechanical (caliper) devices, soft sensors Particle Size (slurries)

2058

.. ..

I

pH Tramp Metal MillLoad Speed

I

I I I

I

0

specialized electrodes, conductivity probes magnetic field devices power based devices, acoustics, load cells, strain gauges, soft sensors, conductivity tachometer

One of the exciting developments related to the more generic instrumentation is the emergence of Foundation Fieldbus (or Profibus). Broadly speaking, this is a standard digital communications protocol that has permitted supply side f m to imbed intelligence in their sensors and final control elements. For example, one can buy a flow sensor that not only measures flow (the only signal available on the old analog [e.g. 4 - 2Oma output] gear), but using embedded diagnostics can perform condition monitoring functions to indicate whether the estimated flow has been corrupted by mechanical or process problems. Moreover, this device can have control functionality (sampling and signal filtering, PID controllers, loop diagnostics, etc.) and can communicate with neighboring smart devices. Significant reductions in project costs arise from savings in detailed engineering (-20%)3, configuration (-30%), and installation and commissioning (-40%). Operational savings arising from predictive maintenance and remote support (-70%) are also very significant. Perhaps motivated by Galileo’s (1564-1642) challenge - “We have to measure everydung that can be measured, and make measurable everything that is not yet measurable.” - sensor development continues to be a focal point in control research. One can differentiate between technology that is finding different applications, and applications seemingly still looking for a technology. For example, in the case of the former there has been a reblrth of interest in acoustics for monitoring AG/SAG mill performance, and in particular to avoid ball on liner collisions from a cataracting charge in SAG mills (e.g. Valderrama et al., 200, Campbell et al., 2001, Pax, 2001). Vision systems, and especially those based on image analysis have also seen relatively fast acceptance in coarse particle sizing (e.g. WipFrag - Maerz, 2001; Split - Girdner et al., 2001; and T-Vis - Herbst and Blust, 2000), and froth monitoring (e.g. JKFrothcam - Kittel et al., 2001; Frothmaster - Brown et al., 2001; Visiofroth4). With respect to applications looking for a technology, mill load estimation is a good example. Thls is a critically important variable in throughout optimization for AG and SAG mills, hence the research activity over the past decade or so. As one can see from Table 1, there are power-based devices (e.g. Pontt et al.,1997; Koivistoninen and Miettunen, 1989), acoustic devices (e.g. Barrientos and Telias, 1997; Spencer et al., 2000), load cells units (e.g. Jones and Wright, 2001), strain gauge systems (e.g. Herbst et al., 1990; Dupont and Vien, 2001), conductivity units (e.g. Marklund and Oja, 1996; van Nierop and Moys, 1995) and soft sensors (e.g. Herbst et al., 1989), to choose from. All have strengths and weaknesses, and with the possible exception of soft sensors, none have a large enough installed base or long enough operating history to be established as the technology leader. In fact, as more is learned, it seems likely that some combination of technologies will emerge as the best means to measure mill load and other in-mill properties. Finally, there have been some interesting statistical developments over the past few years looking at new and superior empirical modeling tools with which to develop more robust empirical calibrations (e.g. Clustering - Ginsberg and Whiten, 1992) or process models (e.g. Partial Least Squares - Hodouin et al., 1993 and Bartolacci and Boujila, 2000).

The cost savings have been distilled from numerous vendor presentations on this subject, and should probably be considered as best case estimates. Personal communication with Thierry Monredon, Cisa, Orleans, France

2059

I

HARDWARE Control hardware is a very important part of a process control system, and often accounts for 20% - 25% of the total capital cost in a relatively well-instrumented mineral processing plant. It is also an area that is quite well serviced by the major vendors, which generally helps to make selection a rather low risk step. Despite the long heralded convergence of functionality between the two main platforms - the Programmable Logic Controller (PLC) and the Distributed Control System (DCS) - there remain some technical differences, and these often require careful consideration in the context of the overall application. Interestingly, the philosophical loyalties that evolved around these two options in the 1980’s and 1990’s (the electrical camp vs. the instrumentation camp) appear to remain more or less solidly entrenched. A consequence of both factors is the hybrid control system, a very common architecture in plants built in the 1980’s and 1990’s. Figure 10 provides an illustration of a typical architecture for a control system designed in the ‘90’s. In this instance the PLC is generally used to manage the discrete I/O and safety interloclung, while the DCS handles the analog UO and Human Machine Interface (HMI).

-

Figure 10 Components of a plant control system (adapted from Flintoff & Mular, 1992) The evolution of smart instruments is beginning to change the role of the control hardware, as 110, signal conditioning, single loop control, etc. migrate to the field devices. Increasingly, the

emphasis of the hardware suppliers is on data management and presentation, and on advanced applications including plant management information systems and advanced controls. This is perhaps better developed in what was once the analog world, so it is the DCS suppliers who are leading the change through an increased emphasis on the functionality depicted in Figure 2. One can expect this trend to continue, as most of the more robust control and condition monitoring applications will also migrate to the plant floor. More specifically, they will be embedded in the process equipment yielding what has been termed “intelligent processes”, conceptually similar to smart instrumentation. STRATEGIES Much will be said on this topic in the companion papers. For the purposes of this introduction we can categorize strategies into essentially two levels - regulatory and advanced control. At the regulatory level the strategies are typically implemented through single-input single-output ’

2060

controllers, almost always using the Proportional Integral Derivative (PID) control law. Advanced control strategies are generally multiple input - multiple output (usually set points), and although a number of more analyfxal techniques have been applied (e.g. Decoupling Control - Hulbert and Woodburn, 1983 and Model Predictive Control - Vien et al., 1991), the industry standard appears to be a blend of artificial intelligence and analytical techniques, known as model-based fuzzy expert control. On the regulatory level, the strategies are generally implemented through PID control laws and the manipulated variables are flows, speeds, etc. on the plant floor. Despite the wealth of application history in this area, Bialkowski’s (1992) rather unsettling findings in the pulp and paper industry were found to be applicable to mineral processing. To briefly summarize: “Ifyou have been keeping score, only 20% of the loops surveyed actually decrease variability in automatic over manual mode of operation, in the short term. This prompted an interest in getting back to basics on such as issues as signal conditioning and loop tuning (e.g. Vien et al., 2000) and on performance diagnostics (e.g. Perry et al., 2000), all of which have helped to improve the quality of control at the regulatory level. In addition, effective solutions have been established for some of the more challenging regulatory loops, and these have necessarily evolved beyond pure Proportional Integral Derivative (PID) controllers. A good example of the latter is the multiple feeder control problem for large SAG milk, which is illustrated in Figure 11. In this instance, the size of the equipment generally combines to lower dynamic resilience as a result of the significant dead time between the feeders and the weigh scaIe, even though this sensor is placed as close as possible to the feeders. A further complication is that effective dead time, and controller gain is a function of the combination of feeders running (in automatic, vs, manual vs off). Another layer of complexity is added if the feeder speeds require some bias to reflect ore flow issues arising from, say, stockpile segregation. Vien et al., 2000, have described a very robust regulatory algorithm that is well suited for this problem. ”

Figure 11 The multiple feeder control problem in AG/SAG mill feed regulation One can find the first evidence of advanced control back in the 1970’s, when process engineers using control minicomputers began to supplement PID logic with ‘If-Then-Else’ heuristics to optimize circuit performance. Efforts to use analytical (i.e. mode-based) advanced control techniques have been problematic, partly because of their dependence on the reliability of sensor inputs, partly because the models themselves are simplifications of the process, and partly because the models require continuous adaptation as things change. Sandvik, 1985, provides a particularly good description of some of these issues. The introduction of the real-time expert system (with fuzzy logic) in the late 1980’s provided a platform for implementing advanced control strategies that circumvented some of the limitations of the model-based approach. However, it isn’t surprising that over the past few years the two approaches have been blended to exploit the strengths of both.

2061

Figure 12 is a schematic of a model-based fuzzy expert control structure. In this particular instance, the model can be based on process phenomenology and the adaptation achieved with an extended Kalman filter. It is also possible to employ statistical models, specifically neural networks, and frequently retrain the model routinely, or as new process conditions and/or poor predictive accuracy are encountered . Of course, within this structure both can be used. T h s author’s opinion is that whenever possible, phenomenological models should be employed, as they are more faithful to the engineering principles, physics and chemistry underlying the process. Moreover, the adaptation involves parameters with a physical significance, which makes error checlung relatively easy.

Figure 12 An advanced control structure for mineral processing Broussaud et al., 2001, have shown that the structure in Figure 12 can be applied to plants at essentially either end of the field instrumentation spectrum. Their general phlosophy is to use the phenomenological models within a soft sensor to estimate parameters such as SAG load, ore hardness, etc., which are not directly measured or measurable. Moreover, the adapted model from the soft sensor can then be used in optimization. Depending upon the optimization approach, the model can be used in its dynamic or steady state form to deduce new set points that will drive the process toward the optimum operating condition. Of course, depending upon the quality of the input data, the optimizer may make unrealistic recommendations, and one of the roles of the expert system is to filter these values prior to implementation. Not everything can be done quantitatively, and heuristics are employed to frrst ensure process stability and then to optimize performance. Ths control logic is implemented with crisp and fuzzy rules. It is of some interest to note than the major developmental focus in advanced control appears to be directed at embedded advanced measurement technologies (e.g. the image analysis systems mentioned earlier), which can enhance controller performance. To conclude, it is of some interest to note that we can distinguish a third level of control strategies, and these have been called either watchdog control or shell (or jacket) software (referred to as ad hoc functionality in the Introduction). The intent of these strategies is essentially to improve the robustness of the control applications by ensuring that the more traditional controls are protected from unusual operating conditions (e.g. start-ups & shutdowns, very large disturbances, etc.), where they would be unstable and require excessive operator intervention. Flintoff and Edwards, 1992, provide an illustration of the use of watchdog control in crushing. Astrom described this rather well (Astrom et al., 1986) and noted that given the application specificity, this has been an area largely ignored by control researchers and developed exclusively by practitioners.

2062

USERS It is essential that each operation have the necessary complement of people to support the process control system from the field instrumentation through to the users. The numbers of technical people generally increase with the size (-I/O count) and complexity (- ”analog” UO), and generally decrease with the level of sophistication of the system (e.g. diagnostic and asset management software) and training. In the constant pursuit of doing more with less, a major focus for the users of the process control systems is essentially one of training and education. Other drivers include a greater emphasis on improved control performance, and succession. Much has been written on the subject of training and education for users in process control (e.g. Vien et al., 1994, Flintoff, 1995, Rybinslu et al., 2001). Suffice it to say that this too is a process that requires management commitment to sustain the return on the process control investment. Downsizing and succession issues are combining to create an interesting dilemma for operators. As more and more sites demand improved process control performance through better regulatory and advanced control, the need for deeper subject matter expertise grows. Recruiting and retaining such people has been a challenge for the mining industry, as we compete with all of the process industries. Typically, the other sectors offer very competitive salaries and usually superior work locations. The end result is that something that could be classified as a necessary core competence for every mineral processing operation, is a responsibility that has to be increasingly outsourced. Successful outsourcing models exist in other disciplines, especially Information Technology (IT). However, given the global nature of mining, effective and timely remote communications is an essential element of such a relationshp. ThIs may be the single biggest bottleneck, but that is changing through the rapid advance of IT, and especially Internet technologies. Specialty consultants, engineering firms and equipment suppliers are all moving to establish themselves as “Application Service Providers’’ with 24x7 remote technical support. The evolution of these new business models means that the techmcal people at site must adopt a leas t e c h c a l and more managerial role toward process control. However the irony is that to be effective in the latter, they must remain quite familiar with the former.

CONCLUSIONS Process control has been an important part of mineral processing for many decades, but automatic process control has really established itself over the past 30 years. The ‘70’s can be thought of as the embryonic period of development, where the first efforts at integration involving computers were completed. The ‘80’s was a growth period, with some of the attendant growing pains, and, among other things, featured the development of better tools and techniques for implementing process control strategies. The 90’s were a also a period of some growth, but also a period of maturation as proven approaches began to emerge in both hardware architectures and especially in software structure. Exciting development in process control continues on all fronts: from field instrumentation (new sensors); to control hardware (smart instruments); and strategies (improved modeling and optimization methods); to the users (new business models, educational programs, etc.). Equally exciting is the growing profile of process control as a cornerstone of the enterprise business systems, drawing it more into the realm of production management. ThIs latter transformation will have some interesting trickle down effects as such activities as condition monitoring/predictive maintenance and real-time enterprise modeling become deeply entangled with control. The future is bright for those with an interest in this discipline. On the business side, and where it has been properly applied and supported, process control has consistently delivered on the financial projections, with paybacks of from weeks to months and returns on investments of from 20% to 200%. One can only assume that with increasing regulatory interest in resource utilization, and the increased business interest in shareholder value and sustainable development, there will be increasing demand placed on all facets of operation. ThIs is especially true for process control, since process management is effectively implemented through the elements of field instrumentation, hardware, strategies, users and maintenance

2063

programs that together comprise the process control system. The process control value chain is only strong as its weakest link, and successful operations must necessarily explore the options for continuous improvement in all elements of Figure 3. REFERENCES h r o m , K., Wittenmark, B., 1984, Computer Controlled Systems: Theory and Design, Prentice Hall, pg. 157 htrom, K., Anton, J., Arzen, K., 1986, Expert Control, Automatica, Vol. 22, No. 3, pp. 277 - 286 Bamentos R., Telias M., 1997, Nuevos Sonsores en el Circuit0 SAG, Proc. SAG Workshop, Vina de Mar, Chile Bartolacci, G., Boujila, A., 2000, Application of Multivariate Tools to Mineral Processing Data Analysis and Modeling - Flotation Case, Proc. IFAC Workshop, Automation in Mineral and Metal Processing, Finland, August, pp. 188 - 193 Bhatt, S., 1992, The Control Connection, Chemical Engineering, May, pp. 91 - 94. Bialkowski, W., 1992, Dreams vs. Realities: A View From Both Sides of the Gap, Control Systems '92: Modern Process Control in the Pulp and Paper Industry, Whistler, B.C., pp. 283 - 295 Broussaud, A., Guyot, O., McKay, J., Hope, R., 2001, Advanced Control Of SAG And FAG Mills With Comprehensive Or Limited Instrumentation, Proc. International Autogenous and Semiautogenous Grinding Technology, ed. Barratt, Allen and Mular, Vancouver, B.C., Canada, Sept., Vol. 11, pp. 358-372 Brown, N., Bourke, P., Ronkainen, S., van Olst, M., 2001, Improving Flotation Plant Performance at Cadia by Controlling and Optimizing the Rate of Froth Recovery Using Outokumpu Frothmaster, Proc. 33". Can. Min. Proc., CIM, Ottawa, pp. 25 - 38 Campbell, J., et al., 2001, SAG Mill Monitoring Using Surface Vibrations, in Proc. Int'l AG and SAG Grinding Technology 2001, eds. Barratt, Allan and Mular, Vol. 11, pp. 373 - 385 Dupont J., Vien A., 2001, Continuous SAG Volumetric Charge Measurement, Proc. 33" AGM Can. Min. Proc., CIM, pp. 52 - 67 Flintoff, B., Mular, A, 1992, A Practical Guide to Process Control in the Minerals Industry, Gastown, Vancouver Flintoff, B., Edwards, R., 1992, Process Control in Crushing, in Comminution - Theory and Practice, ed. Kawatra, SME, pp. 505 - 5 15 Flintoff, B., 1995, Control of Mineral Processing Systems, Proceedings of the XIX Int. Min. Proc. Cong., San Francisco, Vol. 1, pp. 15 - 23 Ginsberg, D., Whiten, W., 1992, The Application of Clustering to the Calibration of On-Stream Analysis Equipment, Int. J. Min. Proc., Vol. 36, pp. 63-79 Girdner, K., Handy, J., Kemeny, J., 2001, Improvements in Fragmentation Measurement Software for SAG Mill Process Control, in Proc. Int'l AG and SAG Grinding Technology 2001, eds. Barratt, Allan and Mular, Vol. 11, pp. 250 - 269 Herbst, J., Bascur, O., 1984, Mineral Processing Control in the '80's - Realities and Dreams, in Control '84, ed. Herbst, SME, pp. 197 - 215 Herbst J., Pate W., Oblad E., 1989, Experiences in the Use of Model Based Expert Control Systems in Autogenous and Semi Autogenous Grinding Circuits, Proc. Advances in AG and SAG Grinding Technology, eds. Mular and Agar, Vancouver, Vol. 2, pp. 669- 686 Herbst J., Hales L., Gabardi T., 1990, Continuous Measurement and Control of Charge Volume in Tumbling Mills, Control '90, Ed. Rajamani and Herbst, SME, pp. 163-171 Herbst, J., Blust, S., 2000, Video Sampling for Mine-to-Mill Performance Evaluation: Model Calibration and Simulation; in Control 2000, ed. Herbst, pp. 157 - 166 Herbst, J, Lo, Y.,Flintoff, B., 2002, Size Reduction and Liberation, in Principles of Mineral Processing, ed. Fuerstenau and Han, SEM, 2002 Hodouin, D., MacGregor, J, Hou, M., Franklin, M., 1993, Mulitvariate Statistical Analysis on Mineral Processing Plant Data, CIM Bull., NovDec, pp. 23 - 33

2064

Hulbert, D., Woodburn, T., 1983, Multivariable Control of a Wet-Grinding Circuit, J. AIChE, Vol. 29, No.2, pp. 186 - 191 Jones, R., Wright, A., 2001, Selecting and Configuring Load Cells for AG/SAG Mill Grinding Applications, in Proc. Int’l AG and SAG Grinding Technology 2001, eds. Barratt, Allan and Mular, Vol. 11, pp. 227 - 239 Kittel, S., Galleguillos, P., Urtubia, H., 2001, Rougher Flotation in Escondida Flotation Plant, SME Preprint 0 1-53 Koivistoinen, P., Miettunen, J., 1989, The Effect of Mill Lining on the Power Draw of a Grinding Mill and its Utilization in Control, Proc. Advances in AG and SAG Grinding Technology, eds. Mular and Agar, Vancouver, Vol. 2,pp. 687- 695 Maerz, N., 2001, Automated On-Line Optical Sizing Analysis, in Proc. Int’l AG and SAG Grinding Technology 2001, eds. Barratt, Allan and Mular, Vol. 11, pp. 250 - 269 Marklund U., Oja J., 1996, Grinding Control at Aitlk: Optimization of Autogenous Grinding Through Mill Filling Measurement and Multivariate Statistical Analysis, Proc. Intl. AG and SAG Grinding Technology, eds. Mular, Barratt and Knight, Vancouver, Vol. 2, pp. 617- 63 1 Morari, M., 1983, Design of Resilient Processing Plants - 111, Chemical Engineering Science, Vol. 38, NO. 11, pp. 1881 - 1891 Pax, R., 2001, Non Contact Acoustic Measurement on In-Mill Variables of a SAG Mill, in Proc. Int’l AG and SAG Grinding Technology 2001, eds. Barratt, Allan and Mular, Vol. 11, pp. 386 - 393 Perry, R., Supomo, A., Mular, M., Neale, A., 2000, Monitoring Control Loop Health at P.T. Freeport, Control 2000, ed. Herbst, SME, pp. 71 - 81 Pont, J., Valderama, W., Magne, L., Pozo, R., 1997, MONSAG: Un Sistema para el Mointereo On-Line de la Carga en Molinos SAG, Proc. SAG Workshop, Vina de Mar, Chle Rybinski, E., Zunich, R., Grondin, M., Flintoff, B., 2001,Operator Education: An Important Element of the Corporate Knowledge Management Effort, SME Preprint 0 1- 156 Sandvk, K., 1985, Limitations to Advanced Control in Complex Sulphde Flotation Plants, in Flotation of Sulphide Minerals, ed. Forssberg, Elsevier, pp. 433 - 446 Stephanopoulos, G., 1984, Chemical Process Control Process: AN Introduction to Theory and Practice, Dynamics and Control, Prentice Hall, New Jersey Seborg, D., Edgar, T., Mellichamp, D., 1989, Process Dynamics and Control, Wiley & Sons, New York Stowe, IS., 1992, Noranda’s Approach to Complex Ores - Present and Future, AMIRA Techmcal Meeting Valderrama, W., et al., 2000, The Impactmeter, A New Instrument for Monitoring and Avoiding Harmful High-Energy Impacts on the Mill liners in SAG Mills, Proc. IFAC Workshop, Automation in Mineral and Metal Processing, Finland, August, pp. 286 - 289 Van Nierop M., Moys M., 1995, Measurement of Load Behaviour in an Industrial Grinding Mill, IFAC Automation in Mining, Minerals and Metals Proc., Sun City, South Arfica, 1995 Vien, A., Fragomeni, D., Larsen, C.R., Fisher, D.G., 1991, MOCCA: A Grinding Circuit Control Application, SME Ann. Mtg., Denver, Colorado, February Vien, A., Edwards, R., Perry, R., Flintoff, B., 2000, Making Regulatory Control a Priority, Control 2000, ed. Herbst, SME, pp. 59 - 70 Vien, A., Palomino, J., Gonzalez, P., Perry, R., 2000, Multiple Feeder Control, Proc. An Gen. Mtg. Can. Min. Proc., CIM, Ottawa, pp. 298 - 3 12 Vien, A., Willett, D., Flintoff, B.C., Hendriks, D.H., 1994, Mill Operator Training -Where Do We GO?, Proc. 26th Annual General Meeting, Canadian Mineral Processors, Ottawa 1994, Paper no. 16, 15 pages

2065

Well Balanced Control Systems Tom Stuffco’ and Khaled Sunna2

ABSTRACT Successful control systems are characterized as reliable, relatively uncomplicated, easily expandable, and capable of supporting higher layers such as enterprise systems and advanced controls. Ths paper will explore the ingredients necessary to build a healthy control system foundation. Perhaps technically less satisfying then the specter of advanced techniques, the essential components are basic and include standards for configuration of controls, graphics, communication gateways, documentation, and personnel support. INTRODUCTION A well-balanced, successful control system has many human characteristics that are naturally desirable. Just as many humans have endearing qualities and features that make them distinct and successful, so goes the control system. A successful control system will communicate openly and easily with the rest of the world, “articulate” information clearly and concisely, demonstrate a degree of independence and be highly reliable. While it is difficult to find the right balance between reliability, scalability, and integration, these characteristics are essential to the overall success and acceptance of the system. Upon installation, the control system must successfully meet the fundamental requirement of monitoring the process while suppressing disturbances. In order to sustain this foundation, functional system maintenance practices must be well established. Establishing this maintenance function is governed by two paramount events: system design and system selection. A team with diverse talents must deliver the design essentials for both project start-up and subsequent support. End users must seek some level of assurance that the proposed system can support their immediate project needs as well as their future needs. If properly equipped, the system has good prospects for performing well throughout its useful life. If essential ingredients are missing, high aspirations for achieving world-class controls are reduced to resentful ongoing system support. DESIGN The quality with which the initial control system design w a s h performed has a significant effect on future achievements. It is imperative that the design team develops certain standards prior to embarking on a new installation project. Standards are a crucial component of the maintenance efforts that will ensue once the control systems are active. A system that lacks standards, but prefers the “flavor of the day” approach is unsupportable and unsustainable. Individuals responsible for the maintenance role under these conditions spend a significant portion of their time developing and implementing standards; which

1

2

Operations Technical Support Manager, P.T. Freeport Indonesia Company Senior Project Manager, EMA Consulting

2066

are often sub-standard because, out of necessity, they will develop them to minimize the effort required for implementation. With reference to the following diagram (Figure l), standards are most critical in the fundamental layer, including the low-level regulatory and discrete controls as well as the humanmachine interface (HMI) and data acquisition areas. The design team’s first goal must be to develop acceptable support systems for these controls. Each object within the fundamental layer should be given due consideration to ensure that resources exist to address vital issues.

External Applicatons

Enterprise

........................................................................... :undamental Layer Configuration Interface

4

Operator Interface

Data Acquisition

i

PROCESS

Figure 1: Control system Model Process A thorough understanding of the process itself needs to exist, before any of the upper levels can be effective. Expertise in the minerals processing field needs to be apparent in developing the control strategies because process controls will only acheve the level of stability allowed by the process. Understanding these process limitations and dynamics is necessary in order to define adequate performance and identify potential gains. Control engineers need to work in conjunction with operators and process engineers to ensure that control solutions can be achieved. A fm understanding of the process is necessary for the proper development of the higher control layer. Field YO The devices within the electrical and instrument (E&I) disciplines connect the process to the control layer. Obviously the system will be severely handicapped without accurate inputs or proper manipulation of outputs. Qualified E&I engineers are key to the overall project success as they provide expertise in the selection, design, configuration, and implementation of field gear. l k s E&I group ensures that field devices are placed correctly to provide adequate access, allowing for uncomplicated maintenance and calibration. In addition, the group also serves to verify that sound practices are employed during the design and installation activities associated with field cabling. Conformance to high quality standards for cable routing, cabinet wiring and terminations will have a lasting effect on ownership of the system. People tend to react very differently when faced with a cabinet that has been laid out meticulously versus one where the wires are simply “flung in” and long enough to get the door closed. Internal cabinet appearance is a solid indicator of the overall system health.

2067

During the design phase, VO assignments should contain sufficient spare capacity for future additions. After installation, these spares can be readily managed through a system that will maintainthe cabinet integrity. Because of the added complexity of sampling, on-line analyzers (e.g., density, size, assay, and moisture) present a particular challenge. Provisions for sampling are occasionally overlooked when locating these instruments in the design stage, but they are critical for proper calibration. Despite the expense, these instruments will only perform as well as their calibration allows. A group effort consisting of participants from instrumentation, process engineering, and the laboratory is necessary to correctly design sampling procedures that will minimize errors during calibration. Systems personnel are also needed if these instruments include computer interfaces, which should be incorporated with the larger control system maintenance activities. The importance of proper and up-to-date documentation is critical and the installation project must conclude with the hand-over of a proper set of "as-built" drawings. The lack of proper documentation will eventually lead to an increase in downtime and can become very costly. This documentation, as shown in the example in Figure 2, represents the final milestone marking the completion of a successful design. _4_

Field D o c u m e n t s

1

Logic Conflguratlon

I

Databases

-------

Control Systems

Figure 2: A Typical Documentation System Logic Basic standards included in the logic component are centered on documentation and control templates, using a combination of configuration application(s) and supporting databases. Control engineers must balance their desire to create "art" against maintaining consistency that ensures the fhdamental layer is sustainable. The addition of any control algorithm should be justified through an improvement in process stability that warrants the added complexity. This is not to say that elegant and sophisticated control strategies aren't necessary, but rather, that their application should be tempered with the equally valuable application of the "KISS" (Keep It Simple)

2068

principle. The less time spent on understanding inconsistent logic arrangements, the more time available for optimizing performance. Sustainability of the control logic requires proper documentation. Simple feedback controllers are easy to comprehend with loop drawings and block sheets. A control narrative that clearly explains the objectives and inner workings of the strategy, however, should accompany complex configurations like cascade loops, adaptive controls, and extensive switching arrangements.

Interface The interface level should focus on meeting the needs of those individuals who have to continually interact with it. Operators will have habit patterns where controls are expected to move in certain ways and the thoughtful application of properly designed templates that provide consistency are key. Standards should cover symbols, object sizes, text, color schemes, animation, measuring units, etc. “Poorly conceived standards have sometimes been established casually, or without meaningful consideration of operator needs. Without attention to these needs, the operator’s performance may suffer” (Considine 1993). Data Acquisition The data acquisition object in Figure 1 has been included in the fundamental control layer to cover the basic need of operational reporting. After the significant investment in a control system, the operation should not continue to rely on manual data entry for production reports. A validation method needs to be in place before automated data collection can be used, however and it is a good idea to separate “raw” from “official” data. The development of these databases is typically done with MIS support so operating statistics can be integrated with other departments. Even after the introduction of a control system, many managers prefer to have control room operators continue with manual logs as a means of verifying that key performance indicators get checked and mentally noted with a deterministic frequency. Communication Networks Essential information can flow over various assorted networks. These networks must meet the highest standards of reliability and availability through redundant, fault tolerant network design. Correct system’s support will result in compliant communication paths that are unobstructed and reliable. On contemporary networks, constrained bandwidth may be a thmg of the past, but most legacy networks are pushing the limits. Standards at this layer are aimed at preserving signal transmission times within acceptable limits and maintaining network availability. On the soft side, this is accomplished through configuration of proper phasing, update rates, and grouping of memory segments. On the hard side, th s is achieved through sound management practices including the separation of critical and non-critical communications. In addition to the absolute requirement for healthy control communications, data flow is at the heart of the system’s capability for expandability. “The need to transfer data to other systems is simply recognition that no one package can do all things equally well. In many applications, it makes sense to pass a processing task on to an expert. The ability of a package to exchange data is the main feature of so called open systems.”(Levine 1996). Miscellaneous Design Considerations Remote monitoring provisions, such as X-terminals, greatly enhance troubleshooting and reduce downtime by allowing technicians fast access to systems while off-site. Also, managers and supervisors that are “plugged in” have a much easier time directing resources to cover situations and keep abreast of operational conditions. There may be added security concerns with remote access, but sound security management policies and practices should overcome these. Adding the capability for remote access is worthwhile investment that can pay for itself many times over. A tagging convention must be established during the design phase. This convention should have both the flexibility to grow beyond the current scale, and the ability to uniquely identify

2069

process and system components. A tagging convention will typically cover the following components: . Sub-plants and process circuits . Equipment . Control system clusters or nodes . Control processors and gateways . Instruments . Input/Output points Not only do tagging conventions assign a unique identifier to process components, but they also establish a common grouping system which enables support personnel to quickly pinpoint the location and type of a given point, device, or instrument. A successful tagging system will require little or no modification when crossing system and platform boundaries. For example, the same tag name would be assigned to an UO point in the field, the control system, the advanced control system, the historian, the data acquisition and reporting system, and finally on the engineering drawings. A proper nomenclature that is established during the early stages of a plant’s life will allow great expansion flexibility and reinforce operating standards. Without this common convention, troubleshooting and design efforts become extremely complex.

Overall Design Team Considerations As noted earlier, the design team is multi-disciplined and consists of in house representatives from: management, operations, control and process engineering, computer and business systems, instrumentation, and electrical disciplines. External resources such as vendors, consultants and control engineering f m may also be on the team, depending on the size of the project and the availability of in house resources. Either way, it is imperative that the team has well-established standards to follow. Often mining companies will not have the necessary resources to undertake such a project and they need to seek outside engineering assistance. If standards for each element do not exist or cannot be given to the engineering f q then ensure that they are at least reviewed by the design team prior to the development phase. It is an awful feeling when you discover that the entire set of loop drawings have been created using a loop numbering system that you don’t agree with! If the standards aren’t reviewed, you are at the mercy of the group that has been employed to design the system. This can either be very profitable or a very costly endeavor: very profitable if the standards acquired are well thought out; very costly if the standards are ill conceived. The qualifications, track record and experience of the engineering firm along with a review of previous installations and their standards used should suffice to establish whether or not their capabilities meet your needs. SYSTEM SELECTION Without question, one of the single largest decisions impacting the control system’s future is the initial system selection. This single decision will influence (or dictate) every aspect of system development for the remainder of the mining project’s life. Once the system goes on line and is functional at some regulatory level, it is next to impossible to justify ever replacing it. If the system chosen does not continuously evolve over time through the concerted efforts of the supplier, vendor, and end user, the entire enterprise to some degree will suffer. This decision should not be made lightly or without some appreciation for the ever-evolving field of information and computer technology. When selecting a control system for the first time, you will be confronted with numerous opinions on what and whose system is best for your application. If you are on an expansion or retrofit project you are likely already locked into a supplier, which on one hand, is beneficial because the effort involved in evaluating competing supplier’s product lines is daunting. On the other hand, it can be detrimental because the slate is not clean and you will inherit any past sins (if that is the case).

2070

Define Requirements The selection process should start with clearly identhe mine’s initial and future requkments. That is, select a system that is appropriate for the purpose sought. Requirements can be as little as a few simple regulatory loops with minimal discrete controls to as large as a multiple network system with hundreds of loops spanning many miles. It is also critical to look beyond the initial needs of the department and include any corporate enterprise goals. Corporate expectations will most certainly include some form of reporting of operational results, but can expand much further. OP~OM

The choices on plat€orms include what is collectively r e f d to as “industrial systems” umtahhg any one or combinations of PCs, PLCs, and DCS architectures. Market demands and technological advancementshave and continue to blur any distinction between these architectures. For small to mid-range systems the largest remaining di-ffference lies in their historical mots and product development paths. The DCS to a large extent continues to dominate analog controls

while PLCs are the standard for discrete controls. PLC manuhcturers were enabled with the introduction of PC based SCADA (Supervisory Control and Data Acquisition) software, which allowed them to provide very cost effective means of supplying operator interfaces and data collection. These solutions were in turn made possible through the introductionof stable operating systems such as Windows NT and its derivatives. PCs are simply more microprocessors that can and are being employed at Merent layers in the architecture. The need for durability at the field and control layers, however, often prohiits the use of PCs in the field where industrially hardened equipment is a prerequisite. These are typically only available from the PLC and DCS vendors. PC based systems combined with solid field intdhces can offer an acceptable alternative, however. The advent of field bus standards and the increaSing popularity of Ethernet will continue to drive these market changes.

Open Architecture To som extent, DCS architectures are still vertically integrated (one vendor supplies every layer fiom field YO through to application solutions). This has been a necessary path of evolution because dependable, reliable methods of integrating the various elements were, in the past, none h t . These proprietary, closed architecture systems, however, are at the end of their life cycle as users are demanding the application of open standards. This transformation to horizontal structures is a welcome change for the industry. DCS vendors are supplying scalable packages for both low and high end users, while the PLC manufacturers are improving their product lines in the analog control field Gates (1999) describes the benefits of this evolution best: “Horizontal integration makes for high volume and low price. The independence of each layer mans that competition drives each layer to evolve at maximum speed... This delayering will increase competition and customer choice”. Part of the benefits that acme with the horizontal integration are the realization of ‘Ixst-ofbreed” applications from vendors who specialize in specific target areas. Data acquisition, alarm managers, loop performance managers are a few of these. Advanced Control Applications are another.

Horizontal Integration and CommunicationIssues As open standardsimprove and become more common, the tasks associated with the maintenance field should become less onerous. This is particularly true for disparate systems that have h k p m h t control configuration data structures, such as the PLC and DCS. Communications between these system are commonly funneled through serial links with limited capacity. Constraints here dictate the need to have ultra-clean communications because they are typically transferring control signals. These communications are most efficiently dealt with using contiguous memory (a grouped block of memory registers) areas. It is both annoying and

2071

dangerous when an operator or control sequence requests a motor to shutdown or startup, and the command is missed because of overloaded communication I/O buffers. lntegrated configuration software that can span the boundaries of these disparate systems wiII be welcome indeed. Currently offline relational databases that document the memory ranges for communications are the only means of supporting the signals. Consider the transmission path of a single field input in the hypothetical example in Figure 3.

OPC Middleware j

Elec. Dept. Plant A

4

PLC(BrandY)

I

4 I

t Figure 3: Hypothetical Communications Signaling The number of “hops, skips, and jumps” associated with a single input extends across multiple systems as well as multiple departments. Each of these systems has a configuration database that needs to be updated if the source is altered. The maintenance chore associated with modifying a point is huge when system boundaries are crossed. Imagine being the electrician on a routine job of reassigning an input point from reading kWh to Amps. You’d have to phone the electrical department in plant B, the control engineer, the system engineer maintaining the historian, the MIS technician, and the accountant using the signal! In addition, the variable being tracked will be out of commission for some time while everyone realigns their readings. Hopehlly as the industry becomes more horizontally integrated through open standards, other standards such as self-defining-data-formats (e.g., XML) will at the same time enable the integration of their configuration structures.

System Selection Summary In summary, regardless of the control system’s size, choose a system from a reputable supplier with a well-established track record for support and development, while insisting on some degree of openness. Computer based control systems continue to evolve rapidly and for most applications any one or combination of the major system vendors can meet your requirements. It is key to identify both your current and future needs and then find a vendor that can meet them.

2072

MATNTENANCE “Industrial automation is becoming increasingly interdisciplinary in nature. Indeed it is difficult to define what really is the hardcore of instrumentation and control technology.” (Considine 1993). Th~squote best illustrates the gray area between the root disciplines responsible for support and maintenance, and clearly accentuates the need to have demarcated lines of responsibility. Solutions for managing the control system must be designed and implemented to fit the various elements together in an efficient system. Table 1 presents a method of systematically categorizing development and support activities for the control system fundamental elements. Basic issues need to be addressed for each of the activities as they apply to each element. For instance, what resources are required? who will be responsible, where will the resources be located, and when will the activity be carried out? Table 1: Simplified System Architecture

The following lists are by no means exhaustive. They are only intended to provide examples of how to cover basic issues for an element’s associated activities. “What” hardware and software has been chosen to realize each of these elements? Each element may be viewed as having a hardware component andlor a software application that is used for the various activities of designing, configuringhuilding, documenting, monitoring, and repairinghalibrating. The size and nature of the hardware inventory depends on the location of the plant and on the availability of local vendors and suppliers. Industrial plants that operate in a remote location for example would prudently retain an adequate supply of devices and accessories that are required to maintain normal daily operation. On the other hand, the size of the inventory can be reduced to critical devices only if the plant is located nearby a reliable supplier. It is however advisable to always retain all critical items in stock. Critical items are those that have a direct impact on daily production and the lack of which would translate into a definite loss of production, Two maintenance tools that provide solid value and are worthwhile investments are alarm managers and loop performance managers. An alarm management system goes beyond the basic alarming packages that are provided by control system vendors. Generally replacing control room alarm printers, alarm managers historize all discrete alarm events on a separate computer and

2073

make them readily available for analysis, Several prepackaged tools then work collectively to provide the control engineer with the ability to analyze process alarms and system events. Searches and queries, frequency analysis, data manipulation, and even expert downtime-cause analysis through the study of preconfigured patterns are some of capabilities alarm management systems have. Automated reports from these systems aid daily maintenance functions by highlighting the more frequent alarms and distinguishing between “true” and “false” alarms. Alarm managers are also indispensable for alarm rationalization and analyzing discrete sequences during a shutdown. Support crews can retrieve an accurate picture of the events that took place prior to the shutdown and therefore better identify the contributing factors. With the proper configuration, alarm managers can provide some predictive insight on the state of control system devices. Unhealthy devices can be isolated and the problem corrected at an early stage thus preventing a total device failure and the associated loss of production. Loop Performance managers, on the other hand, use data from the data acquisition system to monitor up to hundreds of control loops and detect deteriorating performance. Since controllers are naturally error tolerant, this deterioration can extend over a long time without detection. Automatically tracking the performance of controllers is a significant aide for prioritizing tuning activities. It is also extremely useful for developing and testing alternate control strategies. “Who” or which department is responsible for carrying out each of the activities such as configuring and documenting access privileges? Both the software applications and hardware components used to maintain system security and access privileges require assignment of responsibility. Individuals will have to ensure that the security system is working, maintained, and upgraded as needed. If not done during the design stage, begin demarcating areas of responsibility by element and activity. This will ensure that scheduled maintenance practices exist for all areas. Identifying who will conduct this maintenance will in turn determine the skills and training needs.

Figure 4: Sample Skills Matrix

Adequate on-site skill inventory levels will depend on staff turn-over rates, the time required to develop the skill level, and the availability of skilled resources. An up-to-date skills matrix (as shown in Figure 4) is helpful to ensure adequate resources exist and to identify training

2074

requirements for succession planning. Critical skills require some form of redundancy and can be used to establish on-call lists and vacation scheduling constraints. The system cannot solely rely on any one individual. For highly specialized skills, cross training between departments may provide an attractive option. When deciding upon structural redundancies it is best to look at the areas most at risk. An unplanned plant outage is absolutely unacceptable and cross training, or some other form of redundant back up plan is necessary in the areas of communications, system diagnostics and hardware failures. For system support, suppliers can play a pivotal role that becomes practically compulsory at some larger system size. Formalized vendor support contracts are a common means of providing additional plant coverage. These agreements typically cover technical and logistic areas that cannot be effectively addressed by the plant’s fulltime staff. This perhaps would cover the “monitor” activity for the “control logic hardware” and “peer to peer network communication” elements identified in Table 1.A typical maintenance and support agreement offers repair services on hardware devices, dial-in services for system and process diagnostics, software version upgrades, expert advice concerning technical matters, system health checks and performance measurement. “When” or what frequency will performance monitoring (e.g., loop performance, alarm analysis, etc.) and system checks (e.g. backups) be done? Each piece of hardware and software will require inspection or evaluation at some pre-determined interval. Establishing these frequencies is based on the component’s rate of failure or rate of degradation. These basic maintenance functions could then be incorporated into a computerized maintenance planning application if one exists. “Where” will the resources be stationed for the repair or maintenance activities such as the sensor calibration area, PLC configuration terminals, etc.? The configuration work area should be separated from the operations’ area. That is, it is a good idea to have a separate configuration room where a documentation library can be established and configuration work can be carried out without distracting operations. Ideally, the locations should be close together to enable natural, easy interactions between the two groups, but they are separate functions and they need their individual space. During an emergency situation, reliable documentation at your fingertips is indispensable. AS with an airplane analogy, 99.9% of the time is carefree travel followed by a few seconds of shear terror. During these periods it is absolutely essential that you have your diagnostic information available and ready to isolate the problem.

CONCLUSIONS Successful control systems are marked by the application of sound standards and practices for managing the essential elements of the fundamental control layer. These standards are much easier to implement and adopt if they are enforced during the initial system design. Commitment to the standards by the entire team will provide a solid foundation upon which advanced techniques may be leveraged. The importance of proper and up-to-date documentation cannot be emphasized enough. The lack of proper documentation will undoubtedly lead to an increase in downtime and can become very costly. There are several disciplines that share the responsibility of supporting a control system. Without an adequate record of system and process modifications, normal support functions develop into an overwhelming task. Information technology is transforming the control industry on an almost daily basis. It is essential to select a system that will meet the current needs of the operation as well as evolve over time to meet future requirements. Control systems that are continuously cultivated and developed will provide a h g h degree of performance over their life. There is undeniably a broad range of experience and skill required to hlly develop and support a contemporary control system. The contents of this paper merely slum the surface of these topic areas. Hopellly though, the reader is left with a greater appreciation for the diverse

2075

nature of control systems in the minerals industry. If inclined to do so, readers are strongly encouraged to read the materials listed in the reference list. ACKNOWLEDGEMENTS The authors wish to thank the management of PT Freeport Indonesia for first, insisting on nothing less than the application of world-class standards during each of the several mill and control system expansions over the past decade and second, for the opportunity to write this paper. The entire control team at Freeport is also acknowledged for their effort and dedication to developing and maintaining a high level of standards. REFERENCES Considine, D.M. 1993. Process/Industrial Instruments & Controls Handbook, 4Ih Edition, McGraw-Hill Inc. Flintoff, B.C., Mular, A.L., 1992. A Practical Guide to Process Controls in the Minerals Industry, Gastown Printers Ltd. Gates W.M., 1999. Business at the Speed of Thought. Warner Books, USA, pg. 420. Levine, 1996. The Control Handbook, CRC Press, pg. 435 Mark Brewer, Kristine Chin, 1999. “Keep Advanced Control Systems Online”, Chemical Engineering, August. Morin, M.A., 2002. “The Case for Open Data Format Standards”, CIM Bulletin Vol. 95, January. Neale, A.J., Veloo, C., 1997. “Process Control at P.T. Freeport Indonesia’s Milling Operations”, Society of Mining Engineers, Littleton CO.

2076

The Selection of Control Hardware for Mineral Processing Robert A . Medower' and Robert E. Coop

ABSTRACT The selection of a cost effective control system that will meet current and future process requirements is a challenge, especially in this era of rapidly changing technology. Our objective is to provide the reader with information that w ill assist with the "control hardware selection" decision making process. We also include a section on "The future of process control technology" based upon current development trends. INTRODUCTION Minerals' processing sites are historically long lived, 20 year life of mine is not uncommon and 75 years is not unheard of. Considering that hardware technology advancements are introduced every six months, or less, it is prudent to select process automation systems that adhere to "open" industrial standards over proprietary based systems. "Open" systems are generally regarded as having the ability to embrace any new or existing technology that is perceived to have a positive impact on profitability, with minimal cost.

The selected process automation system architecture must be able to support a wide variety of communicationprotocols, hard wired and wireless, to bring data into the system, convert the data to informationand supply that information to business and regulatory agency information systems. Mineral deposits require substantial amounts of energy to transport and process into useable form and present a real processing challenge in that the deposits are never homogeneous and ore body compositional characteristics are all over the map! Extracting the desired mineral@>from the deposit at a cost that will produce sufficient profit margin, despite severe fluctuations in market price, requires process automation architecture capable of supporting an integrated enterprise operation. Such an operation encompasses the activities of individuals, process units, process areas, plant level activities, mine management and business planning and accounting systems.

In the minerals processing industry, the key to suMval is driving the cost of production per unit volume to the lowest possible level and hope that the market price will provide enough profit to sustain operations. Business information technology is rapidly becoming an important tool in the quest for productivity and achieving product quality targets. Getting key dynamic business performance information at the right time to the right person or process unit is crucial to successfully achieving corporate goals. The process automation architecture selected must therefore be able to support multiple network communication protocols.

1 Industry Consultant, Invensys Process Systems Inc., Eden Prairie, MN 2 Industry Marketing Director, Invensys Process Systems Inc., Foxboro, MA

2077

A BRIEF INTRODUCTION TO PROCESS CONTROL Some of the readers of this document may not be familiar with process control, so we present the following figures (1 and 2) as a brief introductionto the subject. Technology Trend

Application based Automation

Technologybased Automation Real-time Activity-based

Economic Trend

Accounting

Figure 1 Automatiodeconomic focus trends Until the mid 198O's, minerals processing companies employed large numbers of technicians and engineers to proactively maintain and improve operations to achieve production and quality goals. The focus was on increasing production and improving product quality to meet expanding market demands. Since then, advances in technology have improved productivity, enabling companies to substantially reduce the number of people required to achieve production and quality targets. These companies no longer have sufficient personnel to proactively devise control strategies and applications to optimize the use of assets. Increasingly, minerals processing companies are relying more heavily on their vendors to provide the expertise, products and services required to remain competitive in their market place. From the 1940's through the 198O's, purchasing decisions for process measurement devices, actuators and controllers was strongly influenced by technicians and engineers. Selection of measurement devices and controllers was heavily influenced by the "latest technology" forcing vendors to embrace and develop new technologies. The control vendors pitched leading edge technology, accuracy and reliability in their efforts to win projects. The low cost of microprocessors has resulted in accuracy and reliability of field devices to improve dramatically. The focus has shifted to the amount of process and self diagnostic information a field device can provide over a variety of communication protocols. In the near future, self validation capability will also be an important factor. Economics has been primarily a "transactional"effort. Quarterly results were compiled from data derived manually. These results were used to devise strategies for improving performance in the next quarter. With the introduction of Local and Wide Area Communications Networks and advances in business software, companies began to move from simple cost accounting towards activity based costing. In the mid 199O's, technological advancements have permitted automation and economics to merge, allowing Real-time Activity-based Accounting. Figure 2 is intended to provide a review of control product development evolution from the 1950's through 2003.

2078

Technolorn Focus

ARRlication Focus I

Economic Focus I

Expert System Rulebased NeuralNet

I I

.

I I

Distributed Control Systems I I

i

DirectDigital Control

I I I I

j

;

On-line Analvsis

Maintenance Management System

I

Multivariable Predictive control

Object-Based Automation

Programmable LogicControl

!

!

statistical Process Control

Fuzzy Logic

Dynamic Performance Management Enterprise Application Integration

!

Advanced Process Control

Electronic Digital

Supply Chain Management

I

Predictive Maintenance

I

Electronic Analog

Human Machine Interface I

I I

Pneumatic

j j I

Mechanical

Enterprise Integrated Operations

Computer Integrated Manufacturing

1 Enterprise Resource Planning

Asset Optimization

1

Communication Networks

E-Commerce

I

1950

1960

1970

Figure 2 Evolution of Process Automation

1980

1990

2000

2001

2002

2003

The j u s ~ c a t i o nfor investing in process automation has changed dramatically. Economics and production have always been the underlying factors, however, until computer technology became powerfd enough to meet processing requirements and cost effective enough to implement in field devices, users relied on human labor to meet their objectives. Current and future control system selections will be more heavily based on economics and "computer"labor. Selecting the right control hardware for minerals processing projects should be based on an anticipated return on investment. Typically, the reason proposed for investment in control technology is to improve productivity, quality and/or meet regulatory requirements at the lowest possible cost. Two relatively new justifications include asset management and predictive maintenance.

ASSET MANAGEMENT AND PREDICTIVE MAINTENANCE In the year 2002 and beyond, predictive maintenance and asset management programs will have an impact on the level of integration between machine protection and advanced process control strategies. Machine manufacturers will imbed microprocessors into their products that will communicate via an information network. The available information will be accessible for use in advanced control strategies, asset management, predictive maintenance and by the manufacturers from remote locations for fault diagnosis and performance monitoring. Asset management is increasingly important to the survivability of minerals processing companies. In the 1920's through 1970's high grade natural resources were in abundance, energy prices were controlled, environmental regulations were weak, manual labor was the primary vehicle to achieve productivity, wages were relatively low and the market could absorb all that could be produced. Since then, available ore grades are lower, product quality specifications are becoming more stringent, energy prices are being de-regulated, environmental regulations are increasingly restrictive, productivity per work hour is an important financial indicator, hourly wages in developed countries have increased substantially, and market demand is volatile. Minerals processing companies are realizing that the path to profitability lies in controlling costs by minimizing the impact on the environment, reducing "head count," eliminating waste and increasing efficiency throughout the production and distributionprocess. Process automation systems must have the ability to integrate rapidly changing technology, be a vehicle for real time bi-directional information between the process and the enterprise, run highly sophisticated process applications and become more predictive in controlling the process and maintaining equipment.

EMERGING TECHNOLOGIES There are many emerging technologies in 2002 that will have a positive impact on process automation, asset management and the availability and accuracy of real time information. MEMS, SEVAm and .NET are three of those technologies. MEMS. This technology will increasingly be used to produce sensors to be embedded in all types of process equipment to provide real time status information. The automotive industry and machinery manufacturers are already employing MEMS for this purpose. MicroElectroMechanical systems are microscopic electricaVmechanica1devices with threedimensional moving parts acting as sensors and actuators. They are relatively inexpensive, robust and easily manufactured using current integrated circuit manufacturing technology. They are widely used in a number of industries but most pertinent to minerals processing is their use in ore haul vehicles.

2080

To see the "future" (beyond year 2000) one only needs to look at the level of sophistication applied to mine haul trucks in the late 1990's. On-board receivers collect data inputs from microprocessors embedded in tires, engines and the truck frame and send that information Via wireless communications to operations management. In addition, mine dispatch systems, using information from GPS units installed on trucks, drills and trains, efficiently monitor vehicle location and permit dispatchers to control the ore characteristics presented to the primary crusher. The benefits include improved safety, increased vehicle availability, improved maintenance scheduling and lower operating costs. Technological advances will allow other process machinery such as crushers and grinding mills to become more intelligent. SEVATMA self validation technology that will improve process data quality by expanding the scope of valid data available from process sensors. This, in turn, will allow the recipient of the data to make better decisions regarding control of the process.

SEVAm , self validating intelligent sensors (the term "sensor" in this case, includes the transducer and transmitter as a unit) have the ability to not only digitally transmit information about the health of the "sensor," as intelligent transmitters do, but process information as well. The benefits include higher process uptime, higher quality products, lower maintenance and operating costs and improved safety.

.NET. One of the "Instant Messaging" products introduced to the market in 2002 as a "Web services" application. It provides instant, validated, information transfer between microprocessor based systems with very little human intervention. Two "Instant messaging" applications pertinent to mineral processing are alarm handling and predictive maintenance. The control system host would be able to communicate alarm information to selected wireless handheld displays. Process units would be able to provide a "health" report to a maintenance host computer including a list of current faults as well as wear parts that have neared their service life.

THE SELECTION PROCESS How do we begin the control hardware selection process? The first steps include generation of a bid document by; determining the scope of the project, the affect of the addition on up and down stream process units, the impact on staffing, division of labor responsibilities, return on investment and how to measure the anticipated results. Accountability and "head count" are increasingly important factors in the process automation selection process. The next step is to determine who will choose the control system. Typically, it is the project manager or purchasing agent whose primary incentive is to be under budget. Alternatively, an engineeringldesign firm is hired and given authorization to make the decision. The engineeringldesign firm's primary incentive is to supply the minimum technology required to meet the project objectives within budget and schedule constraints. Neither of these scenarios will necessarily be in the best interest of the owner, from either a short or long term profitability point of view. The budget is the overriding factor and is usually established by the owner without the influence of a qualified control engineer who could provide input on the best level of control to implement to obtain the fastest return on investment. It is interesting to note that the cost of an installed process control system, as a percentage of the total project, is typically the smallest number but is the first to be negatively affected by a budget shortfall. This happens, in spite of the fact that the control system has substantial influence on short and long term profitability.

2081

The Investment Structure Figure 3 illustrates a typical process automation investment structure. Beginning with the "Process" as the foundation, each subsequent level builds on the one below it. Without a firm foundation, advanced and Expert control applications are of very limited benefit and are typically detrimental to profitability.

The first level above the Process employs intelligent field devices to provide accurate process information and process variable manipulation. The second level, basic regulatory control, provides process stabilization. The third level utilizes intelligent process analyzers to provide quality control. The fourth level utilizes advanced regulatory, multivariable predictive and Expert control software strategies to provide process and profit optimization. The fifth level provides an integrated, bi-directional information path between the process and appropriate enterprise individuals for timely decision making.

Optimization: Expert Systems, Multi-variable Predictive Control, Advanced Regulatory Analytical Measurements Basic Regulatory Control: PLC, PC and DCS Single Station Controllersand Recorders Measurement and Actuation Devices: Pneumatic, Analog & Digital Pressure, Temperature, Flow, pH, ORP & Conductivity Valves, Actuators and Variable Speed Drives The Process

Figure 3 Process Automation Investment Structure The Impact of Variability Processing mineds is difficult due to the wide variability in ore body characteristics. Compositional analysis of the ore body is used to design the physical plant. To optimize mine life, the physical plant design is a compromise that requires blending the various ore grades to fit within the constraints of available process machinery. Mine life can be shortened if the easier to

2082

process high grade ore is presented to the primary crusher without blending the lower grade ores. If the mine operator presents a large volume of low grade ore to the primary crusher, mine life may increase, however, the cost per ton of product substantially increases. The compromises stated above, plus severe environmental conditions associated with mineral processing, present a challenge to process units, automation strategies, process measurements and actuation devices. Technological advances in products utilized by the process control industry are providing solutionsto those problems.

Measurement and Actuation Devices Process measurement and actuation devices are the "eyes and hands'' of the control system. In the 1920's and 1930's the "eyes and hands" were close coupled with the "controller" by necessity. These devices were based primarily on mechanical technology, i.e. filled systems, levers, bellows, etc. Control was accomplished by manual adjustment. The full range of the physical process sensor was constrained by mechanical and physical limitations. The accuracy of the process information available from these sensors was dependent upon the transducer technology (pneumatic, dc current or digital), process measurement type (pressure, temperature or flow), desired operating range, repeatability, turndown ratio and process constraints. Turndown is the range over which the process measurement will be acceptably accurate. For example, a standard d/p cell connected to an orifice plate has a flow measurement turndown ratio of 3:l. Rangeability (the range over which the instrument meets the stated linearity of uncertainty requirements) and uncertainty (the range of values within which the true value lies) become much more important when self evaluating technology is employed. Sensor rangeability has, in the past, been substantiallygreater than the transducer attached to it. Transducer rangeability used to be fixed at a 5:1 ratio, i.e. 3 to 15 psi or 4 to 20 milliamps, and had to be bench calibrated to the desired operating range. Intelligent transmitters speak digitally (i.e. 32 bit floating point), can easily match the range of the sensor to which they are attached and can be remotely re-ranged to accommodate process changes without negatively affecting measurement accuracy. The majority of field measurement and actuation devices purchased through the 1990's were

of the 5 :1 rangeability type.Intelligent devices have gained popularity because they can be easily re-ranged from remote terminals without negatively affecting the accuracy of the information conveyed, can communicate self diagnosed faults to process control systems utilizing communication bus technology and substantially reduce installatiodstartup costs. In the future, sensor based microprocessors will contain control algorithms permitting implementation of local control strategies. The same types of sensors used to monitor and control the process are used for process unit protection. Each process unit has physical constraints such as; bearing temperatures, lubrication, vibration, speed, power, etc., which must be monitored and controlled. Minerals processing units are large pieces of machinery that are costly to purchase, operate and maintain. Typically, the machine vendor supplies or specifies the protective devices required along with instructions on how the unit is to be protected. The owner's personnel or a systems integrator programs the protective supervisory system based on manufacturerk recommendations. In the future, the machine vendor will program intelligent sensors with embedded microprocessors with a preprogrammed protection strategy for the specific process unit. Device accuracy (the effects of non-linearity, hysteresis and non-repeatability at reference conditions) is only one of the factors in the level of uncertainty in the measurement. Installation and process variability also have an effect on uncertainty.

2083

Communications Bus Technology A modem control system may have four distinct communications levels:

. =

The first level includes communications between process measurement devices, actuators, other devices and the control system Inputloutput modules (Sensorbus, Fig. 4). The second level includes communicationsbetween Input/Output modules, other devices and the control system controllers/integrators/gateways(Controlbus). The third level includes communications between the control system controllers and the host processor, operator and engineering workstations (Nodebus). The fourth level includes communications between the host processor/workstation processor and enterprise information networks, Local Area Networks, Wide Area Networks, the Internet and other devices (IT bus).

Digital value

0

0

s

Intelligent process measurement/actuationdevices, point to point

3

ul

Field mounted devices

Digital values

3 3

Intelligent process measurement/actuationdevices, multi-drop

Figure 4 Typical level 1 Sensorbus Communications Level 1 communications. The minerals processing industry primarily employs single channel, point to point communications between analog/digital process devices and the control system. Most mineral processes are spread over large physical areas with relatively low, mixed function, I/O density at any given location, are subject to frequent electrical disturbances and subject all process and control equipment to harsh environments. Level 1 is the most difficult level to address, from a communications perspective, because it needs to accommodate; process measurements, process actuators/positioners, process analyzers, bar code readers and on-off devices like motor starters, on-off valves, switches, etc. Life was simpler when the only types of signals to deal with were 4 to 20 madc, standardized voltage values and relay contacts. The only choice was to run twisted pair, shielded copper wire from the control system to each device in the process. The same wires that carried the process signal were also used to power the process measurementlactuation devices. This wiring made control systems susceptible to disruption from electrical disturbances such as lightening, high voltage power cables, large motor operations, hand held communicationsdevices and ground loops.

2084

Fortunately, all vendors of intelligent point to point and multi-drop field devices designed their products to be compatible with existing wiring. In this case, the same twisted pair copper wire carries the power to the transmitter and the digital information provided by the field device(s). Some important issues to be aware of when specifying point to point or multidrop technology for projects include: 0 How the selected technology powers field devices 0 Interoperability 0 The number of confguration tools required (if multiple vendors products are selected) 0 The security of cables running from field devices to the control system. Interoperability is the capability to substitute a field device from one manufacturer for that of another manufacturer without loss of functionality. The primary benefit is the freedom to choose the right device for an application, irrespectiveof the chosen control system.

Level 1 Communication Protocol Comparison The most popular protocols employed through 2002 were; 4 to 20 madc, HART, FOXCOM, Foundation Fieldbus H1 and H2,and Profibus PA/DP. 4 to 20 madc. This industry standard has been in existence for many years and all major measurement, analyzer, actuation and control system vendors products adhere to this standard providing a high degree of interoperability.

Devices communicatepoint to point, must be bench calibrated, are individually powered from the control system and provide a continuous, single channel, analog signal. The maximum distance between devices under ideal conditions, using 16 Ga. shielded cable, is one mile. Realistically, in minerals processing applications, the maximum distance is substantially shorter due to the large installed quantity of high voltage equipment. Signal cabling must be isolated from high voltage cabling to minimize electrical disturbance. This requirement increases the installation costs substantially.

HART. This protocol was introduced in 1989 and has gained wide acceptance in the minerals processing industry. Currently (2002) there are approximately 150 vendors supplying products that adhere to this protocol. It uses analog 4 to 20 madc for the measurement signal on which a digital frequency is superimposed to carry information and status reporting data. Hart was developed as a transmitter protocol rather than a control system protocol, therefore, a stand alone HART M275 intelligent field device configurator is required. Digital data is transmitted at 1200 bps. Update rate is dependent upon the device and whether the signals are multiplexed through a single modem on a multiple channel I/O module or if each channel has it's own modem. Interoperability is very good, as all vendors submit their products to HART for testing and certification but there is no guarantee that all of the features of the standard are available in each vendor's product. Every device implements Universal Commands (read output, tag number, etc.) and the Common practice Commands (rerange, change tag name, etc.) but not all of the configurable parameters in the standard are necessarily available to the chosen control system. All diagnostics in the HART protocol are "device specific" commands (i.e. Command #48). Therefore, when you integrate a Foxboro or Rosemount HART device to a Honeywell control

2085

system, for example, the system would know that there was a problem due to the diagnostic message, but it may not be capable of interpreting exactly what it is. Devices communicate point to point, can be remotely reranged, are individually powered from the control system and provide a continuous, single channel, analog signal. Point to point supports up to three devices (2 masters + 1 slave). Maximum distance is 1524 m. Wiring from these devices must be isolated from high voltage cables to minimize electrical disturbance. Installation costs are as high as for 4 to 20 madc, however, startup and maintenance costs savings can be substantial.

HART protocol supports multidrop (2 masters + 15 slaves) but most major control system vendors do not support it.

FOXCOM. Foxboro introduced this protocol in 1988 for use by Foxboro manufactured intelligent field devices. Foxboro intelligent devices can be configured for full digital (4800 bps) with 10 times per second digital updates, or for analog 4 to 20 madc, with digital updates (600 bps) 2 times per second. When field devices are configured for analog mode, all information, including the measurement, is still digitally transmitted to the Foxboro control system. The digitally transmitted information is not included in the loop loading calculation, so the analog output may be wired to other devices in the control loop. Devices communicate point to point bi-directionally, can be remotely configured, modified, maintained and operated from any system console, have real time diagnostics and are individually powered from the control system. Point to point shielded, twisted pair supports up to eight devices per I/O module in a star configuration. Maximum distance is 610 m at 4800 bps, 1829 m at 600 bps. Wiring from these devices must be isolated from high voltage cables to minimize electrical disturbance. Installation costs are as high as for 4 to 20 madc, however, startup and maintenance costs savings can be substantial.

CAT5 cable Remotely located backplane Remote Location

16 Ga. Copper Twisted Pair

Multi-drop field devices

Figure 5 Foundation Fieldbus H1 Architecture

2086

Fieldbus terminator

Foundation Fieldbus H1 and H2. Fieldbus H1 and H2 are alldigital protocols (including signal and data transmissions). H1 was introduced in 1995 and II2 (Hi Speed Ethernet) was introduced in 2000. FF H1 transmits data at 31.25 Kbps over shielded twisted pair or fiber optic cables, up to 2000 m with type A cable, including spurs. Figure 5 shows a typical FT H1 architecture. Approximately eight devices share power transmitted on the twisted pair bus. The power supplies can be embedded in the I/O subsystem or external. Bus topology is multi-drop with long spurs (see figure 5). FF H2 (HSE) operates at 100 Mbps, at this speed cabling media becomes a critical factor and maximum distances for twisted pair CAT5 cable will be short, 100 m max., and fiber optic 2000 m max. Devices must be externally powered, as fiber optic cable is incapable of carrying electrical energy. Profibus PA/DP. Profibus PA/DP is an all digital protocol developed by the German Governmentand Siemensas the primary proponent. DP was introduced in 1994 and PA in 1995. Profibus PA is an extension of the remote I/O capability of Profibus DP. Over 300 vendors supply products available with profibus protocol as of 2002.

Either twisted pair or Ethernet can be used to transmit signal and data between the YO and control system. Maximum cable length for DP is dependent upon the speed of transmission; 1OOdsegment at 12 Mps to 1,200 dsegment at slower transmission speeds. Maximum cable length for PA is 1,900 &segment. The signal/data transmission cable powers devices. Profibus DP transmits data by rotating a set of values called tokens around all of the connected devices on the network, which works well with the scan cycle concept of PLC programming. A problem arises when the amount of data increases to the point where the bus becomes overloaded. When designing a Profibus network, keep in mind that there are data transmission speed and cable length limitationsto be considered. Level 2 communications. At the controlbus level are primarily between a variety of multiple vendors' control and analytical products connected to a control system. Most mineral processes employ hybrid systems that are a mix of PLC's, Analyzers and other devices that connect to the installed process automation system through a variety of protocols. Popular protocols include Modbus, Data Highway, Profibus DP, and Ethernet.

Modbus (introduced in 1978) and Modbus+ are propriem protocols designed for Modicon PLCs. Modicon published the simple, straight forward, easy to use protocol standard and offered it free of charge to anyone who wanted to use it. It became very popular with a wide variety of vendors because almost any device with a microprocessor and serial port could use it. The positive is that almost every vendor supports this protocol. The negative is that data transmission is slow and nondeterministic. The more deterministic a network is, the easier it is to predict its accuracy and the more repeatable its performance will be in actual operation. This speaks to a protocols ability to efficientlysend event or time based data between a device and a controller. Data Highway and Data Highway + are propnew protocols developed for Allen Bradley (Al3) PLCs. AB has worked with a number of DCS control system vendors to develop gateways, coprocessing modules and integrators to create bi-directional communications between Al3 PLCs and multiple vendors DCS systems.

2087

Industrial Ethernet is gaining favor as the communications network of the future. Ethernet became the prominent protocol for information technology networks and its use is expanding to include levels 1, 2 and 3 communications. It is unlikely that Ethernet will be the ''sole survivor" in the area of process control networks for future minerals processing projects. A combination of wireless, Foundation Fieldbus and Ethernet will be the best choice for future (2002 to 2007) new projects or large control system replacement projects.

Levels 3 and 4 communications. Ethernet was introduced in the early 1970's as an information tool for office and business applications.It has the ability to move large blocks of data at high speed, peer to peer between workstations, host computers and PCs. Standards for file transport, e-mail and hypertext exist for Internet applications but Ethernet application standards for automation are not as readily available. These issues are being addressed through development of standards by a number of organizations and as of 2002 there are three products that are potential survivors; EtherNetAP, Fieldbus Foundation Hi Speed Ethernet (HSE) and Interface for Distributed Automation (IDA). All three are based on transmission control protocolhnternetprotocol (TCPAP). Other Ethernet based products vying for market share include Interbus, Profinet and Modbus TCPAP. A typical Ethernet network architecture for process automation (see figure 6) integrates control stations and workstations with a variety of Ethernet compatible control products. As with any technology, there are limitations and as the volume and speed of transmission increase, collision avoidance becomes increasingly important.

Figure 6 Typical Ethernet Architecture When planning an Ethernet network, first determine the number of stations to be interconnected and their physical locations. This informationis required for step two, selecting the proper indoor and outdoor cabling for the network. CAT5 shielded twisted pair (10/100Base-T) cabling for a maximum distance of 100 m or Multimode fiber optic cable for a maximum distance of 2 km. The distance between the farthest stations on the network cannot exceed 4.2 km. The maximum number of attached devices is dependent on the amount of data throughput required.

2088

Mineral processing site communication and control system networks require a large number of connections and must process a large amount of data. Hubs are devices used to provide network layout flexibility and switches prevent network traffic on one segment from impacting performance on another segment. A key feature of an Ethernet Network is that all the devices in the network should be able to communicate with one another. To enable this feature, the "Three Hop Rule" must be adhered to. A path between any two devices in the network should have no more than three switches en route. This rule is best achieved by employing a W s t a r topology, the recommended topology for mineral processing plant bus networks. Cable Selection Selection of fiber optic cable, wherever possible, is highly recommended for all levels of communication in a mineral processing plant. Fiber optic technology is capable of sending and receiving information at high speed over great distances (up to 150 km without using a repeater) using light as the data carrier. The signal is not disrupted by outside sources like electricity, rain, humidity or radio transmissions. They are ideal for transmitting information because they are highly secure (do not induce or emit any external energy), transmit data at high speed and signal loss can be detected almost immediately if monitored.

There are two types of optical fiber cables, singlemode and multimode. Singlemode is used for data transmissions between 8 and 150 km. Lasers are used as the light source because they produce focused, parallel light to limit losses. Multimode (recommended for industrial applications) is used for data transmissions between 8 to 10 km. Light emitting diodes are used as the light source for these applications. The following fiber optic cable characteristics are recommended for harsh environments: multimode, graded-index fibers with 62.5 micron core/125 micron cladding and maximum signal losses of 1 d B h at a wavelength of 1,300 nm and 3.5 d B h at a wavelength of 850 Nh4. Two fibers are required for each communicationspath, one to transmit and the other to receive. Fiber optic systems are more cost effective than copper wire, they are lighter, less costly to maintain and do not require repeaters for distances up to 150 km. Fiber optic cables require more expertise to install, however, the benefits substantially outweigh the added one time cost. For the next few years (beyond 2002) all new mineral processing control system installations will utilize copper wire and emerging wireless technology in combination with fiber optic cables. Copper wire, along with emerging wireless technology, will be used to connect remotely mounted LO ' modules to locally mounted field devices and fiber optic technology will be the norm for longer communicationruns and to interconnect automation subsystems. An important issue that will impact the use of fiber optic technology at level 1 is powering the field devices. In most minerals processing plants, power is readily available in all process areas so cost to power the field devices from sources in close proximity should be minimal, however, most modem transmitters are two wire devices designed to accept twisted pair copper only. Twisted pair copper carries the signal as well as supplying power to the device. If the field device were designed with separate signal and power connections, an alternative would be to combine twisted pair copper wire and fiber optics in a single multiconductor cable to address the loop power issue. In the future, wireless technology may eliminate the need for level 1 twisted pair wire in some applications.

2089

Figure 7 is an example of one possible, cost effective, process automation installation architecture for most minerals processing plants. In this case, a relatively short level 1 shielded copper twisted pair (sensorbus) is used to connect intelligent field devices to a remotely mounted UO rack. A redundant level 2 multi-conductor fiber optic cable (controlbus) is used to connect the UO rack to the control system. Multidrop Remotely mounted UO Field Devices

I

-- >..l-_

I

lK'"UUleS

Fiber optic cables up to 2000 m

Level 1 Copper twisted pair run, up to 100 m. from remotely mounted YO rack

Figure 7 Level lLevel2 CommunicationsBus Architecture Example Basic Regulatory Automation System Selection The type of automation system selected for a given project is dependent upon many factors and there is no single vendor that manufactures all of the devices required to properly control a mineral process. In addition, mineral processes are considered to be large, from a control point of view, due to the large number of physical UO points installed with a high percentage of complex analog loops as well as the large processing area footprint. These criteria in 2002 still point toward selecting a DCS (distributed control system) as the primary control system. The process of extraction and concentration can present complex problems. The desired mineral is embedded in the earth's crust and must be extracted and concentrated before it can be brought to the market place. Sometimes there are multiple minerals that need to be extracted and separated which adds to the complexity. Some portions of the liberation process require advanced control andor Expert systems. This also, in 2002, points towards selecting a DCS system as the primary process automation system. Control SystemTypes Overview. There are three types of control systems competing for mineral processing projects, DCS, PLC and Hybrid [Personal Computer (PC) based control systems]. PLC's were introduced in 1969 and DCS systems in the 1980's and Hybrid systems began to gain popularity in 2000 in limited applications. Figure 8 illustrates the market sectors, as of 2002, in which the three types of control systems are positioned. Small to midsize hybrid systems are gaining market share, as all industries gravitate from hardware to software based control applications. Both DCS and PLC systems have incorporated PC's for specific functions to increase functionality and reduced costs. DCS systems utilize PC's with Windows operating systems as Operator's workstations and to connect to information networks. PLC's use PC's as application workstation's, Human Machine Interfaces (HMk) and for application configuratiodmahtenance stations.

2090

Hybrid and DCS vendors have developed reusable, prepackaged, industry specific applications to minimize engineering costs. Vendors of both system types have developed extensive libraries of process control applications that can be quickly integrated into large projects or provided with specific processing machinery to permit "plug and play." In either case, the engineering time gap between programmingPLC's vs. Hybrid and DCS systems is growing larger. 100,000

10,000

v) c)

d

.r(

1,000

2

100 10

DigiWSequential

Analog

Advanced Control

Function

Figure 8 Market Sectors by Control System Type as of 2002 PLC's. PLC's, in this comparison, are stand alone PLC's that do not incorporate a personal computer as a host. They were designed to replace relays, timers, switches and hard wired control panels. They rapidly gained favor with the automotive industry. The automotive industry is now (year 2000) moving away from hardware based systems to software based systems. Architecturally (see figure 9), all PLC's have the same basic components; Input/Output modules, a Central Processing Unit (CPU), solid state memory and a power supply. A programming device with a hard drive that is loaded with proprietary configuration software provided by the manufacturer of the PLC or by a third party vendor. PLC's are programmed using ladder logic (see figure lo), structured text (Boolean logic), ladder logic with advanced function blocks or sequential function chart (SFC). Note that each input and output is tagged with an identifier, i.e. IOOOland QO00l for digital inputs and outputs, AIOOOX and AQOOX for analog inputs and outputs, etc. Other programs that require bit status or data must reference the tag identifier to obtain the information. Process variable scaling, i.e. analog tank level, requires tag identifier register manipulation. Programming efforts can become lengthy if the applications are complex or require a large number of ladder rungs. The number of rungs required to implement an application strategy also affects control response time, troubleshooting time and startup time. PLC application

209 1

programming efforts typically require 30% or more engineering time than programming identical applications on Hybrid or DCS control systems. PLC's process inputs, logic and outputs sequentially in a "loop" or scan cycle. It runs through the loop as fast as it can and response time is determined by the worst case loop time. External events, such as interrupt driven schedulers, cannot interrupt the scan. Programming Terminal with Software

Memory

Operators Terminal with Software

External Data Network

CPU Power Supply Local Comm Link

Remote Comm. Link Figure 9 Typical PLC Architecture

The I/O modules provide a physical connection to process equipment. Inputs include analog and digital signals from pushbuttons, switches, relays, etc. Outputs include digital signals to motor starter relays, solenoid valves, etc. and analog signals to valves, variable speed drives, etc. The CPU consists of a variety of microprocessors that are programmed to perform logic and memory functions described by the application program. Program and data files reside in the CPU memory. Program files store the control application, subroutine and error files. Data files store data received from the I/O modules, status bits, counter and timer presets, accumulatedvalues and other stored constants or variables for use by the program files. Memory size is specified in kilobytes (1 KB = 1,024 words) of storage space. PLC memory capacity ranges between less than 1 Kl3 to 64 KB. The programming device normally is connected to the system only for programming, start-up or troubleshootingpurposes.

2092

The power supply converts either 240 vac or 120 vac to +5, - 15 and +15 vdc for use by the rack components. All I/O devices are externally powered. Modem PLC systems can do almost anythug that a DCS system can do but, intelligent devices, data management, fieldbus applications and integration with other vendor's products are still better handled by Hybrid/DCS systems. Hybrid systems are defined as personal computers connected to a variety of available z/O substructures,including PLC's. Electrical Ladder Diamam Pumpstart PB 1

PumpStop

-P

Tank 1 Level Low

+ CR-1-1 Pump CR-1 Run Light

CR- 1-2

=--it

n

Corresmndinp.Ladder Logic Diagram

1

I

Start I0001

stop I0002

Level I0003

Run

Pump Run Q0002

CR-1 QOOOl

Figure 10 Ladder Logic Diagram Example DCS. DCS systems were originally designed to handle large, complex control applications. Initially, the chemical, oil and gas industries benefited most from this technology. These systems become more price competitive as project scope increases and system component prices decrease. Architecturally (see figure ll), DCS systems are software oriented and the hardware, including Application Workstations, Operator Workstations, Control Processors, Gateways, Device Integratorsand Intelligent devices are designed to support the system's software. Operating software can be UNIX or Windows and both can connect to a common communicationsbus. System resources can be expanded to whatever is required for the total project. By design, power supply capacity and mounting structure capacity increase as input and output devices are added.

2093

The communication node is the backbone of a DCS system. Multiple nodes can be interconnected to provide a plantwide control network. For example, the primary crusher, concentrator, smelter, waste treatment, acid plant control systems and any other process area can be interconnectedby fiber optic network cables. Each node can handle a large number of workstations and network communications to corporate information systems. The number of devices connected to the node is scaled to project requirements and additional devices can be easily added, incrementally, at any time, to accommodate requirement changes.

Remote Enterprise Workstations Multiple Locations Information

Application/ Information Management Workstation

OperatorEngineering Workstations Multiple Locations

Redundant Nodebus High Speed Ethernet Device Integrators

Fault Tolerant Control Processors

Shielded Fiber Optic or High Speed Ethernet Cables

Multiple Protocol Input/Output Modules *4-2011~+d~ *Foxcorn *Foundation Fieldbus *Hart Profibus

Intelligent Field

pigure 11 Typical DCS Architecture

2094

Cost effectively scaling down to handle smaller projects, especially those that require a large number of discrete and small numbers of analog I/O points, can be a challenge. The "overhead of software licensing costs makes it dif€icult for DCS systems to compete with hardware based systems. Decreasing hardware costs, prepackaged application software, interoperability with third party devices, the growing importance of information technology, 24 hour, 7 days per week support availability and the end user's desire to reduce operating costs are off setting system selection cost factors. Software solutions are flexible and well suited to solving mineral processing asset management problems, see figure 12. DCS systems use an object oriented global database, permitting any system or networked hardware device to attach to the bus and have full access to all process data and control applications without having to preprogram the device. The system confguration software is always available and accessible from any local or remote workstation. Access to the configuration software is controlled by password to prevent unauthorized personnel from m-ing system or application programs. Slurry Tank 1 Inlet

9

Sluny Tank 1 Water Addition Slurry Tank Slurry Tank 1 Flow Measurement

slunyTank

Slurry Tank 1 Disch. Valve

Disch. Pwnp Figure 12 Typical Slurry Tank Control Application Note that each process device in figure 12 is identifed by name, as an object. Any program that requires status or data need only state the object name and parameter desired. For example, if sluny tank density data is required by operating personnel or by another software program, the human or software program requestor need only spec@ the object, Sluny-Tanl-1: Disch-Density.Meas, and the system will locate that information and transport it to the appropriate location. The above application is a complex control problem because the objective is to keep the slurry

tank from either overflowing or running empty while maintaining the proper density of the pumped sluny. The process variables include tank level, water addition, pump speed and density. The solution to this control problem requires a control strategy that consists of multiple cascade control loops plus a calculator block, all standard control blocks within the standard DCS software. See figure 13 for "Control Block illustration.

2095

Control blocks were developed in the 1960's to make it easier for control engineers to use computers for process control application development. AU the traditional functions, analog in, analog out, digital in, digital out, calculation, Boolean, sequencing etc. were programmed as tools called "blocks." Developing control applications was simplified because these blocks did not have to be programmed each time prior to use; they only had to be drawn from a library of preprogrammedblocks and configured to suit the application. The control block structure developers realized that programming complex control problems on computers could be intimidating. The solution to the programming dilemma was to develop tools that allowed the programmers to "configure" complex applications by grouping a number of control blocks into an entity, giving the entity a unique name and storing it in the library for future use. Each component in the control application is identified as a unique object that would be entered only once into the system database and reused as often as needed in multiple control strategies. This approach was named "Object Oriented Programming."

Figure 13 Control Block Structure

DCS systems are developed, sold and serviced by the manufacturer. The manufacturer may also offer project engineering, training and installation services. These systems are sometimes referred to as "Integrated Systems" because the DCS vendor is the single responsible source and is the primary provider of short and long term system and/or application support. Hybrid. Hybrid PC based systems are relatively new competitors in the process control industry. Pc's have been on the market for a long time, but inadequate software, poor reliability, interoperability issues and slow processing speeds kept them from being seriously considered for process control. All of these issues, for the most part, have been addressed but development is still a work in progress.

Hybrid systems are normally assembled and programmed by contractors or System Integrators. The architecture (figure 14) consists of a Personal Computer with either Windows or Linux operating systems, an HMI (Wonderware and Intellution are popular HMI vendors in 2002) and an YO substructure (either generic or PLC) with power supplies. These systems are comprised of multiple vendors' products and an issue yet to be addressed, is future interoperability as each product evolves separately. In addition, responsibility for long term support of the hardware and application engineering transfer to the end user once the control system contractual objectives are met.

2096

A primary reason that Hybrid systems are attractively priced is that all of the components, including software, are priced "ala Carte." The vendor's strategy is to provide the user with the specific functionality required for each plant area. To this end, the software licenses are offered "unbundled" and the user must decide what role each workstation will play so that the appropriate licenses are purchased. These licenses are usually defined as "Systems'*or "Suites" and are based on functionality, i.e. Development, Run Time, Viewer, Information, etc.

Figure 14 Typical Hybrid System Architecture License Types Development. Normally consists of a PC with the full complement of software installed to do system development plus a license for the number of %gs*' or points anticipated to be connected to the system. The price point for each license is usually based on a specific number of points; 500, 1000,2,500,5,000, etc. A historian can be optionally furnished with this suite. Run Time. Includes all of the software required to run the required applications but not development software. A Run Time server gets its data through its own database over the control network or over Ethernet from the development station. All process alarm detection and acknowledgment is done at the workstation level, and not at the function block level as is done in a DCS. Link The user configures the tags that each station uses for communications. "Link" software allows these stations to supply tag data to other computers running the selected HMI software, however, computers not loaded with this software will not see the tag data. In an unsecured mode,

2097

reliable data m s f e r rates of approximately5000 points per second are possible. In secured mode (extensive communication handshaking plus error checking to verify the transaction took place and the correct data was transferred) the reliable data transfer rate drops to approximately 500 points per second. Operations Viewer. Does not sit on the control system network and is used for viewing operations only. It consists of a PC with preloaded software for viewing operations in plant process areas or offices. The exchange of data is established using Ethernet to a Run Time or Development station. Information Manager. Consists of a PC connected to the network loaded with software required to create the point or tag database. The license is priced based on the number of points, i.e. 500, 2,500, 5,000, 25,000, etc. This server is a dedicated historian connected to the control network or through Ethernet to a Run Time and/or Development station. The above suites are also offered as software packages only, permitting the Systems Integrator to purchase PCs separately. The contractor or Systems Integrator is responsible for the packaging, configuration and application work as well as the design and purchasing of the appropriate cabling and communication devices, such as Ethernet switches, Hubs, etc. to permit communications between stations. DCS vendors also provide hybrid systems for entry level projects. They package these systems using unbundled elements of their DCS software and furnish low cost I/O substructures of their own, or third party manufacture to compete in this market sector. The I/O substructures employed are frequently purchased from PLC vendors. The primary advantage of a DCS vendork hybrid system is single source system responsibility, guaranteed interoperability between all hardware and software components, long term application support and 24 hour, 7daydweek service support. The control domain (Basic regulatory control) From 1908 to the mid 1980sthe control domain shifted from the process unit to the control room. As microprocessors migrated to field devices, the trend is driving the control domain back to the process unit. In 2000 the first field devices, supported by Fieldbus Foundation technology, were offered with embedded control blocks. This technological trend means that the need for traditional Input/Output module subsystems will decline as field device microprocessor power increases. Although we have focused on process measurement and actuation devices, mineral processing requires a large number of motors. As motor control system manufacturers move towards embedding powerful microprocessors into their products the need for relays, timers and switches will also diminish. As the control domain shifts to embedded systems, computers and intelligent field devices, traditional I/O subsystems will be bypassed. Hybrid and DCS systems provide more computing power, storage capacity, programming flexibility and are in better position to deal with direct connections to fieldbus and informationtechnologies due to their software orientation.

"Plug and play" is a popular theme that can apply to process units when the control domain for a given process unit is integral with the unit. For example, a crusher or grinding mill could be furnished with the protective measurementlactuation devices and controller pre-installed and

2098

configured. In place of wiring, wireless temperature and pressure devices could communicate with a local control processor that was able to communicate with the "global" community via high speed information networks. Up and down stream process units would also communicate with the network forming a process control area. When selecting process control systems for mineral processing it is important to consider system life cycle costs. Will the system selected have to be totally replaced in five years or less? Or is it flexible enough to accommodate technological changes needed for the enterprise to remain profitably in business for many years. The control domain (Analytical Devices, Advanced and Expert systems) We were once advised by a contractor's project manager that mineral processing is not "rocket science" so the simpler the control strategy the better. The statement about rocket science may be accurate, however, reality does not support the 'lsimpler the better'' portion of the statement. Every mineral processing company we've visited has either implemented, or is planning to implement, advanced regulatory control and/or Expert systems to improve operating efficiency. Grinding and flotation are two primary beneficiaries of advanced and Expert control strategies.

There are three levels of control strategy that are employed above basic regulatory control; Advanced Regulatory, Multivariable Predictive and Expert systems. Expert systems are Rule Based, use Neural Net technology or a combination of the two. These systems require robust computing power and input from process analytical devices to be optimally effective. The control system selected must have the capacity to incorporate or integrate with these tools. Analytical devices are steadily improving in their ability to provide accurate results. The primary nemesis is the sampling system. As the analyzers move from the lab to "in-line," the sampling system problems will hopefully diminish. Another problem with analyzers is the time lag between acquiring data and transmitting results. This issue has been addressed at one mine site by developing and implementing a Predictive Algorithm that provides a predicted value every minute vs. the normal analyzer cycle of every 15 to 20 minutes. Process performance improved measurably after implementation of the predictive algorithm. Analytical devices have been developed for each process area. Optical analyzers have been developed for crusher and gtinding mill feed analysis. Crossbelt analyzers are available for compositional analysis of conveyed ore. X-ray analyzers for elemental composition and fine particle size analyzers for mill discharge are in wide spread use. The control system selected must be able to cost effectivelyinteroperate with these devices. Advanced regulatory control's primary contribution to operating efficiency is reduction of deviation around desired process operating setpoints. The primary tools used at this level are; feed-forward, cascade, ratio, leaflag, etc. Multi-variable Predictive control is a matrix based, modeling, predictive optimizing process controller. Primary tools include; Adaptive Controller, Fuzzy Logic, Constraint Controller, Optimizer, Multi-model Director and Neural Net. Rule based Expert systems are based on process operating knowledge. A set of rules is developed to emulate the "best" operator.

2099

A Neural Net Expert system is software that has the capacity to learn how to react to process deviations from setpoints and provide guidance to the process controller to prevent or minimize upsets.

Evolution of the "Control Room" Until recently, control rooms were located in each process area to minimize process automation installation costs and because of technological constraints. With the introduction of remotely mounted industrial I/O subsystems, remotely controlled video systems, inexpensive wireless technology and fiber optic cables, the industry is moving toward centrally located, process wide, operations centers. It is interesting to note that the control industry began selling "distributed control" products in 1908. Pneumatic technology of the 1940's required that monitoring, control and actuation devices be in close proximity to the process unit over which they had influence. Advances in pneumatic technology in the 1950's permitted process controllers to be consolidated on control panels in close proximity to a group of process units. When electronic single station controllers, recorders and indicators were introduced in the 1960's control panels were consolidated into large control rooms so fewer operators could take responsibility for a larger portion of the process. The introduction of computers in the 1970's substantially reduced the need for control panels. Until recently (year 2000) the focus has always been on controlling the process in real time based on transactional guidance from corporate business management. With the introduction of high speed communications systems and sophisticated business and asset management software, what was once "transactional" is rapidly becoming "real-time." The primary drivers for real-time process business management include; fluctuations in market demand, increased operatinglproduction costs and market demand for higher quality, custom products. The question is; is there still a need for "control rooms?" Many mineral processing companies have already consolidated their crusher and concentrator control rooms into one location. The traditional primary function of the "control room" is expanding to include real-time process management so perhaps the "control room" is dead! Long live the **ProcessManagement Operations Center!" The process management automation center is primarily a hub that links key process information to the corporate information network. This center does not necessarily need to be located at the site. In the late 1990's,while attending a National Crushed Stone Association conference in California, we witnessed a demonstration of remotely stopping and starting a stone crushing circuit located in a quarry on the East Coast. A video system scanned the process area prior to stoppinghe-starting the circuit and the video signal was transmitted to California so we could observe the process. Neither a human nor the proverbial dog to keep the human from touching anytlung was on site. A laptop computer was used to remotely control the process equipment via the Internet. We were advised that this company operated the facility "lights out" from 4:30 PM to 7:30 AM each day. Safeguards are in place to allow the process to shut itself down if a fault occurs. Maintenance crews arrive in the morning to address any problems that occurred over night and to maintain the process equipment while operating crews replenish the stockpile. The rapid advance of technology will have a major impact on how, and from what location(s) processes are operated. Gordon Moore, physical chemist, Co-founder of Intel and author of Moore's Law: Computer processing power will double approximately every 18 months. In 2002 a personal computer with a 2+ Gigabit Pentium processor is available on the open market for approximately $1,000. In October of 200 1, Intel announced a new optical networking subsystem

2100

designed to deliver 10 Gigabit Ethernet and the worlds first complete CMOS physical medium dependent chip set for 10 Gigabit per second applications. This level of computing power will permit the operation of mineral processes from virtually any location. The primary process automation system selected must be able to rapidly handle high volumes of information, be able to be programmed using object names accessible by any workstation, easily interface with multiple vendors products and connect with a wide variety of communicationsnetworks. THE F'UTURE OF CONTROL TECHNOLOGY Throughout this document we have strived to include some idea of the future as it pertains to each element of a process control system. Perhaps a review of the mineral process from mining to shipping, and what we view the future to be in each area, is the most efficient way to summarize. Mining. For open pit and underground operations, video systems and robotics will be more heavily utilized in mineral extraction and material handling to enhance safety. The machines employed for extraction and material handling will have embedded intelligence to provide current health and predictive maintenance information. Wireless communicationswill provide web access SO that mine planning, the dispatch system, down stream process control unit and machine manufacturer can obtain remote access. On board analyzers will be able to determine the composition of the ore being transported and provide that input to down stream process units. Crushing. Video systems will monitor feed size and automatically adjust the gap to obtain the desired discharge size. Vibration analyzers, video systems, temperatures, pressure measurements, lube systems, and drive motor power monitors will wirelessly connect to an onboard microprocessor configured to optimize production and protect the crusher from damage. The microprocessor will also connect to surrounding process equipment via a fiber optic network. Concentration. Video systems and hgh speed information networks, in a trunk/configuration, will interconnect all of the process machinery in the concentrator. Each process unit will have intelligent process measurements, analyzers and actuators wirelessly connected to an embedded microprocessor designed to optimally control the unit, communicate with other process units and permit remote access. The microprocessor will also transmit current health and predictive maintenance information to the process management operations center. Process Management Operations Center. The host controller in this location will maintain all pertinent process and maintenance data and store an image of the current control strategy for automatic down load to each unit in the event that a local unit microprocessor is replaced. With this technology, process units could be easily added or removed from service. The host system could easily be reconfigured to accommodate process flow modifications and new equipment. SELECTION GUIDE For those responsible for establishing the project budget, process automation systems are functionally no longer limited to process control. They are an integral part of an Enterprise wide information network that will have a major impact on short and long term profitability. Whether the project is of limited scope, i.e. addition of a process unit,or a new process plant, it is important to consider the level of investment required to achieve the desired production and quality targets in the shortest possible time. A few pertinent questions prior to system selection and budgeting decisions. What is the scope of work? What constitutes a successful implementation and how will it be measured? How will up and downstreamprocess units be affected? What impact will contractual division of labor have on control system selection?

2101

0 0 0

0 0

What will be the impact on staffing? Who will specify, select, purchase, configure and install the system? What are their qualifications? To what degree are they accountable? Will the selection be based on the total installed price of the entire process automation system (best approach)? Or will it based on traditional methods of evaluating field devices, control system and installation as separate entities? The latter choice precludes considering installation cost savings associated with remotely mounted I/O substructures.

We've divided table 1 into two basic function types, PLC and DCS/Hybrid because Hybrid and DCS systems are rapidly merging as microprocessors become more powerful. The actual hardware selected should be based on the most cost effective system available that will meet or exceed project expectations and anticipated return on investment. The "X" indicates the best functional system type to meet the requirements of the stated variable.

Variable DCS/Hybrid Application is digital (discrete) I/O intensive, >70% Application is discrete logic intensive (simple sequencing, Booleaqetc.) Desired response speed not faster than 2ms Lowest number of engineering hours to develop applications X Global database X Common system wide database X X Interoperability with multiple vendors products Requirement for advanced and Expert control strategies X Prepackaged applications availability X X Information networking requirements Multi-tasking/ Multi-user requirements X Desired response time is 0.1 ms or less X Sophisticated alarming is required X X The process is highly interactive, process unit to process unit X Lowest life-cycle cost X Lowest Installed cost X Adheres to the open industrial standards (01s)model X Choice of operating systems Supports WEB browsers and wireless communications X Supports multiple fieldbuses and legacy I/O Accepts any Foundation Fieldbus device without proprietary resource file X X Simple graphical configurator X Ability to configure and run PID algorithms in field devices X Ease of "Change Management" X Supports object oriented naming structure X Computer aided design software for configuration and back documentation X Integration if intelligent field devices X Redundant system wide communications X Full asset management, including predictive maintenance X Workstation Configuration and diagnosticsfor system and field devices X Remote system access and diagnostics from factory support centers

2102

PLC X X X

X

An additional item for consideration is migration from legacy products to current technology. The life cycle cost of the system will increase substantially if the selected process automation system can not be cost effectively upgraded to current standards.

Table 1 provides some guidance to process automation type selection on a functional basis instead of a hardware basis. Anticipated technological advancements in communications and electronics will eliminate some of the elements currently found in today's (2002) process automation systems. In the near future, hardware based PLC and I/O substructures will be functionally replaced by embedded software.

CONCLUSION The vision of the future is, for the most part, based on technology all ready available. BMW announced the avdability of their top line model with "Drive by Wire" technology. The throttle, brakes and transmission are all controlled by computers and are disconnected mechanically from the vehicle operator. Lexis offers smart cruise control that automatically adjusts speed to accommodate varying traffic conditions. Large aircraft are being controlled from remote locations with precision, using wireless high speed communications. Video systems attached to these aircraft allow the operator to view the aircmft's surroundingsin real time. The control system of the future for minerals processing is a large capacity information system connected to individual process units by high speed fiber optic networks. The optimizing control strategy is embedded in microprocessors located in the intelligent process measurement andor actuation devices and wirelessly connected to other pertinent devices in the control loop.

Sam Walton, founder of Wal-Mart, understood technology's role in building a company. He realized that technology was not an end in and of itself, but a tool to be applied by people who knew how to use it. He invested heavily in computer technology and hired people who understood that technology was a powerful tool for boosting productivity, understanding customers' needs and allowed employees to perform their jobs better and faster. His investments in technology and people who knew how to use it, created a company valued at $220 Billion (in 2001) that consistently grew at an annual rate of 14%. When presented with the opportunity to select a process control system, choose wisely.

ACKNOLEDGEMENTS We are grateful to Peter Martin, Chief Marketing officer, Invensys Process Systems, Inc. and author of "Dynamic Performance Management" and "Bottom Line Automation," for his input to this document. In addition, we thank Kevin Fitzgerald, Director of Measurement Integration, Invensys Process Systems, Inc. for is input on the direction of communications bus technology and measurement products.

2103

BASIC FIELD INSTRUMENTATION AND CONTROL SYSTEM MAINTENANCE IN MINERAL PROCESSING CIRCUITS Joseph R. Sienkiewicz, PE’

ABSTRACT The mining and minerals processing industry presents unique challenges in the selection, application and maintenance of process measurement and control devices. Because of the typically harsh environments and difficult process conditions, special consideration must be given to the ruggedness, reliability and maintainability of these devices. This paper will explore various aspects of the device selection process, considering the degree of instrumentation, measurement types and methods, materials, and signal types. Fundamental aspects of instrument maintenance will be covered, such as complexity, accessibility and process equipment provisions. Asset management and preventative maintenance concepts will also be discussed. ENVIRONMENT Mining and minerals processing plants are often located in some of the most remote locations on earth. From high rugged mountain areas, to hostile and hot dry deserts, to wet, humid rainforest regions, these plants must operate round-the-clock. Resources such as water and power, as well as essential services like transportation and telecommunications, are usually in short supply and costly to obtain. In order to survive in today’s economy and be profitable, these facilities must often be self sufficient and extremely efficient. Personnel must also be minimized, not only for economics, but also because of the difficulty in attracting qualified people to work at these locations. To meet these challenges, modem plants must be highly instrumented so that centralized control and advanced automation capabilities are facilitated. Having a strong, well-organized plant instrumentation maintenance department, with trained and qualified personnel, is essential. Many mine sites are days away from supporting infrastructure, so extra planning is required to assure adequate availability of technical support and timely access to spare parts for repair. Arranging maintenance contracts with major instrument suppliers or independent representatives of multiple manufacturers is fairly common. These contracts often include guaranteed response times for on site availability of support personnel. Arrangements can also be made for spare parts availability on a consignment basis, either on-site, or in a nearby off-site warehouse, so those spares don’t have to be purchased. Environmental and meteorological conditions are often extreme, and their effects can adversely affect the operation, accuracy and reliability of instruments. A number of mining facilities are located between 10,000 and 14,000 feet altitude. High altitudes, due to the thin atmosphere, affect cooling efficiency. Supplemental cooling may be required in order to keep 1

Kvaemer E&C, San Ramon, California

2104

instruments within their operation ranges. Likewise, extra cooling may be required in hot desert environments for obviously different reasons. Conversely, instruments subject to freezing temperatures must be suitably insulated and heat traced. Instrument cabinets subject to freezing should include a space heater. One condition that is often overlooked is the effect of direct sunlight on instruments and enclosures. Sunlight can overheat exposed equipment in a surprisingly short time, especially if enclosures have dark exteriors and relatively large surface areas. Sun shades or pre-fabricated environmental instrument enclosures are typically used to mitigate this problem. If equipment installed outdoors could be subject to electrical storm (lightening) activity, then those associated externally wired circuits should be provided with lightening and surge protection devices. Instrumentation circuits should also be protected against electrical noise and damage fiom electrostaticdischarge, radio frequency interference, switch contact bounce, and power supply disturbances caused by load switching and lightning. Protection must be provided against windblown dust and rain, periodic washdowns, and also the corrosive effects of chemical pollutants often present at minerals processing facilities. Instruments and enclosures should always be provided with a NEMA 4X enclosure rating, which provides the necessary protection. NEMA 4X may not be required if there is no corrosive atmosphere, in which case, NEMA 4 will suffice. It should not be assumed that if stainless steel instrument tubing were used, it would withstand a corrosive atmosphere. For instance, if chloride ions are present, stainless steel instrument tubing will quickly pit and leak. Materials must therefore be selected which are suitable for the specific corrosives present. This may require an evaluation through the use of analytical instruments or corrosion coupons.

PLANT CONTROL PHILOSOPHY The quantity, type, and features of plant instrumentation and control devices depend, to a large extent, on the operating philosophy of the plant, and the degree of automation and advanced control desired. In the past, most minerals processing facility operations were labor intensive. Instrumentation was relatively simple and provided to facilitate decentralized local operating floor oriented control. With the advent of modem computerized Distributed Control Systems (DCS) the focus changed to centralized control from a main control room. Through the computer displays, operators could monitor and control all the equipment in a facility. Additional instrumentation was required to act as the eyes and ears for the operator and to provide data for process analysis. Fundamental process measurements such as temperature, pressure, flow, level, density, pH, etc., connected to the DCS through electronic signals, have to be both accurate and reliable. Sophisticated computer controlled analytical systems, such as particle size analyzers and multi-stream x-ray analyzer systems, connected to the DCS through data links, provide additional real-time analytical information to the operator, previously available only from laboratory analysis. The promise of advanced supervisory control, employing model based predictive control, expert systems, neural networks, and hzzy logic techniques, in order to increase efficiency and profitability, often require even more process measurements to be effective. Instrumentation maintenance support is of paramount importance; its value should not be underestimated. In the world of centralized control, it can easily make the difference between profit and loss if a plant is operating inefficiently due to missing or erroneous process information. MINERAL PROCESSES The processes encountered in the minerals industry present very demanding requirements of measurement and control devices. In ore crushing, conveying, stockpiling and storage operations, instruments are exposed to rocks, dust, dirt, water sprays and high vibration. In milling operations, instruments must deal with highly abrasive slurries, which cause severe abrasion and erosion of piping and related components. In flotation circuits, they must endure a variety of specialized chemical reagents and foaming fluids. In smelters, high temperatures, molten metals, corrosive gasses, and sulfuric acid are present. Autoclaving presents the additional challenges of

2105

high pressures, temperatures and highly corrosive slurries. In solvent extraction and electrowinning (SXEW) circuits, highly acidic solutions are routinely dealt with. Instruments must be rugged enough to handle all these challenges and continue to operate reliably for a reasonably long period without repair or replacement. They also must be carefully selected for their specific application to properly take the process fluid properties into account and to apply the appropriate scientific measurement method and apply the proper materials of construction.

PROCESS MEASUREMENT DEVICE SELECTION There are many issues involved in the proper selection and application of process measurement instruments. Some fundamental considerations include: environmental constraints, the nature and properties of the fluid, the condition of the fluid, geometry and orientation of pipes and vessels, installation requirements and restrictions, location and access requirements, accuracy, complexity, routine maintenance requirements, and capital and operating costs. For each type of measurement, there are usually a variety of scientific principles available; the object of the selection process is to choose the one that best matches most of the selection criteria. The majority of field mounted transmitters are two-wire 4-20 mA types. Certain transmitters, requiring more power than available from a two-wire loop (magmeters, density, sonic level, etc.), have a supplemental power source requirement (usually I20VAC) which also powers the isolated 4-20 mA output for that device. “Smart” 4-20 mA transmitters with HART (Highway Addressable Smart Transmitter) protocol have become a current de-facto standard for most applications. The HART digital data are superimposed onto the usual 4-20 mA signal. They are preferred because they are more accurate, cover a wider measurement range, are easier to calibrate, and provide a wealth of information not available with the analog signal alone. Smart Fieldbus transmitters, which communicate with the plant DCS via a digital communications network, eliminating the analog signals altogether, are becoming more common as manufacturers increasingly offer this technology in their products, and they gain wider acceptance by users. There is usually a struggle between keeping the initial cost of an instrument within budget for a given purchase, and minimizing the cost to operate and maintain that device. For example, a valve may be selected for an application based on the lowest qualified bid, but it might have to be re-built every three months in operation, while another valve, costing twice as much, may last a year in the same service. Unfortunately, many suppliers are reluctant to offer performance guarantees because of the potential variability of process properties and conditions. Sometimes only experience, and not technical compliance, will determine the better choice. The following sections are not intended to be a comprehensive tutorial or guideline on the proper selection and sizing of field instrumentation; there are many books and literature that cover this in great detail. Rather, this is a summary of common use and practice for this industry. Flow Technologies include electromagnetic, vortex shedding, coriolis mass flow, thermal mass flow, weigh scales, ultrasonic, open channel weir (with level sensing), and differential pressure (orifice plates, flow nozzles, venturi tubes, averaging pitot tubes), vaneharget, turbines, positive displacement and variable area rotameters. Flow measurements in relatively clean liquids are often made using orifice plates. Vortex shedding type of flow instruments are usually used for applications requiring measurements with turndown ratios (The ratio of the maximum measurable flow rate to the minimum measurable flow rate) of more than 1O:l. Turbine meters are used where high accuracy is required, however they are usually high maintenance items due to the wear of moving parts. Thermal mass flow transmitters are often used in column flotation cell sparging systems and other low velocity flow applications for gasses and liquids. Ultrasonic flow meters, both Doppler and time-of-flight are seldom used in this industry due to their unreliability and inaccuracy in harsh environments. They may be considered for measurement of highly corrosive fluids, since the transducers are mounted outside the pipe.

2106

Electromagnetic flowmeters are preferred for slurry service and fluids containing suspended particulate like plant process water. They have the advantage of being obstructionless and have a good turndown ratio. They must be installed in piping runs which assure a full pipe (a vertical upflowing section is ideal) and require a conductive fluid and sufficient fluid velocity. Averaging Pitot tube elements with differential pressure transmitters Figure 1 are normally used for gas flows and offer low differential pressure drops. Magneticflowmeter An air purging system is used where there are entrained solids in the gas, which can cause a plugging of the Pitot tube. Gravimetric conveyor belt scales with four idlers are usually used to weight solids flow when high accuracy is required for inventory control or custody transfer. Four idler belt scales can provide a minimum of 0.5% accuracy. Less accurate weigh scales with fewer idlers are normally used for relative measurement, for process control of mill feed and material handling systems. The belt scale electronics are normally microprocessor based and belt speed compensated. Nuclear type belt scales are used where physical constraints preclude the use of gravimetric scales. These scales are also microprocessor based and belt speed compensated. Level Technologies include bubbler, hydrostatic (pressure), capacitance, ultrasonic, radar, laser, nuclear radiation, time domain reflectometry, buoyancy, float displacement, load cells, and strain gauges. Level measurements of relatively clean liquids are usually made with pressure transmitters measuring hydrostatic head. Diaphragm seals are used if corrosive liquids or entrained solids are present. Level measurements of liquids with varying densities in vented tanks, including tanks with agitators, are normally made using sonic level sensors and transmitters. Capacitance level probes are used for level measurement in high temperature and/or pressurized tanks or vessels, or where the process fluid will tend to foam. Sonic level sensors also are used to monitor storage bin continuous level inventories, and pulp levels (using a float target in a stillwell), and froth height. Figure 2 Radar level sensors are often used in stillwell applications where high Radar level ranges are measured. Laser level gauges can be used for both liquids and solids. Although not yet widely used by this industry, this technology shows great potential for solving some of the most difficult problems encountered in bin and silo level measurement. These devices typically employ a time-of-flight measurement of near-infiared diode laser pulses. They have the advantage of high accuracy and reliability, good interference immunity, long range, and no beam divergence, which can cause false echoes. They have the ability to measure from oblique angles and are relatively unaffected by temperature, solids coning, acoustically absorbing materials (dust) or low dielectric constants. Nuclear radiation level transmitters are used for measurement in difficult process and vessel applications like autoclaves, flash vessels and gyratory crusher discharge surge hoppers. The source of radiation is usually Cesium 137 (30 year half-life) or Cobalt 60 (5.3 year half-life), and the detector is either a Geiger-Muller or scintillation type. A scintillation detector is preferred because it is more sensitive, and requires a smaller nuclear source. The owner must obtain a local or national regulatory agency license in order to use these devices on site. Feed bin loading and loss-in-weight measurement systems are usually monitored by high accuracy load cells or strain gages with transmitters for measurement of varying vessel weight which is directly proportional to inventory level.

21 07

Pressure Measurements include gauge and absolute, differential, draft, and hydrostatic.

Differental pressure transmitter

Diaphragm seals are used in pressure measurements of slurry lines, corrosive liquids, or entrained solids. Pressure measurement of lines with thick slurries are preferably made by using a rubber-lined, in-line spool pressure sensor. Differential pressure type transmitters are also used to indicate dirty air filters, including bag house dust collectors, as well as filters in liquid lines-

Temperature Technologies include thermocouples, RTDs, and infiared optical pyrometers.

Temperature measurements up to SOOOF, requiring remote transmission of the signal, are normally made by using a platinum resistance temperature detector, (RTD) calibrated to 100 ohms at O"C, with a scale factor of 0.00385 ohms/ohm/"C. Thermocouples with temperature transmitters will be used for temperatures above 800°F. The following thermocouple types are typically used for the maximum operating temperature indicated:

element with

1. TypeK 2. TypeR 3. Type N -

U

2400°F 3000°F 2400°F for Oxygen rich environment

All temperature sensors on pipes, tanks, and vessels should be installed in suitable thermowells.

Density Technologies include nuclear radiation, weight, and hydrostatic head (differential pressure).

Figure Nuclear gauge

Nuclear radiation type density gauges are almost always used for density measurement. These devices are clamped to the outside of a pipe and require a full pipe for proper measurement. Like magmeters, installation in a vertical upflowing section is ideal. The source of radiation is usually Cesium 137 (30 year half-life) or Cobalt 60 (5.3 year half-life). The detector is either a Geiger-Muller or scintillation type. A scintillation detector is preferred because it is more sensitive, and requires a smaller nuclear source. The owner must obtain a local or national regulatory agency license in order to use these devices on site.

Analysis Among the vast array of real-time analytical measurements available, the most common types used by this industry include pWORP, conductivity, oxygen content and combustion efficiency, flammable andor toxic gas and vapor detectors, multi-stream x-ray florescence, pulp particle size, and flue and stack gas emissions analyzers. Most of these complex devices require constant service, maintenance, and calibration, often in association with laboratory analysis comparisons. It is not unusual for them to require a full time attendant, whose salary must be factored into the overall operating cost for these systems.

2108

PROCESS CONTROL DEVICE SELECTION Final control devices are very important elements of fluid and material handling systems because they regulate the process flows, in order to maintain throughput and material balances, under constantly changing conditions. Three fundamental methods are used: throttling control valves, variable pulse modulation, and variable speed prime movers (pumps, feeders and conveyers). Valves Types include globe, gate, plug, ball, butterfly, angle, choke, dart, diaphragm, pinch, annular orifice, needle, regulator, and relief. Valves work on the principle of dissipating a portion of the process fluid energy. When selecting control valves, a number of factors must be considered, such as required flow capacity, body material, body pressure rating, necessary seat leakage class, body size and style, pipe connection method, trim material, (the internal parts of a valve which are in contact with the flowing process fluid) and desired flow characteristics.

Valves are sized to handle the requirements of minimum and maximum flow rate, inlet and outlet pressure, and fluid properties. Based on the process fluid properties and conditions, a required flow capacity factor, Cv is calculated, which then serves as a basis for valve size and type selection. Control valves are normally selected to absorb 33% of the total system fiiction head at design flow. This places them at a control point in the range of variability, balancing between pressure drop and subsequent energy loss, and control rangeability. If a valve is too large, it will tend to operate in a narrow range near closed. In almost all cases, a properly sized control valve will be one or two line sizes smaller than the pipe, or have a reduced port. In selecting the materials of construction, for both the valve's body and trim, special considered must be given to factors which could reduce the useful life. Among those are high fluid differential pressure and velocity, potentially causing cavitation, flashing, pitting, erosion, noise, and vibration. Entrained particles and slurries, as well as corrosive fluids also need special consideration. Expensive special alloys may be required, like titanium or super duplex stainless steel for severe corrosive applications, while harder materials, like ceramics and coatings, might be used for slurries and erosive fluids. Elastomer type valves like diaphragm or pinch valves are typically used for slurry throttling service. All valves for oxygen service exposed to fluid velocities greater than 200 feet per second should be constructed of copper base alloys, preferably Monel. 3 16 stainless steel can be used where velocities are below 200 feet per second. Instruments for oxygen service should be properly cleaned and prepared prior to installation. The geometry of the valve trim determines the valve flow characteristic curve. Three primary characteristics are linear, quick opening and equal percentage. The flow through the valve will thus follow a distinctive curve in relation to stem position. The desired valve characteristic usually is determined by examining the overall process gain and response conditions. The most common valve characteristic is equal percentage, which minimizes large flow changes when the valve is near closed, and mitigates

2109

E 50

'

0

25

50

75

1M)

X Travel Figure 7 Valve characteristic curves

subsequent wear and controllability problems. As general rules of thumb, quick opening valves are used in on-off and pressure relief applications, linear characteristics are often used in slow processes for level, low flow, and temperature control. Equal percentage characteristics are applied in fast processes for pressure control or where highly variable pressure drops in flow applications are encountered. Valve actuators are used to provide the mechanical force necessary to stroke an automatic valve. Typical types are pneumatic (spring diaphragm, spring cylinder, and double acting cylinder), air motor, electro-hydraulic, and motorized. Some considerations affecting selection and sizing are maximum differential pressure across the valve and its port size, stroke speed, stem stiffhess, performance characteristics, seat leakage class requirements, and failure action. By far, the majority of actuators in use are pneumatic. It should be noted that large valves and / or high differential process pressures require very large actuators, and special consideration must be given for their space allowance, mechanical supports, and maintenance access. Motorized valves are often used where plant air is not readily available, but as they typically use gear reducers for torque multiplication, they usually have a comparatively slow speed of response. Positioners are used on throttling valves to precisely position the valve stem in relation to the control signal, by using a stem position feedback to correct the error between desired and actual. Positioners are used, especially on larger valves, to overcome the fiiction and subsequent hysteresis in the stem mechanism, in order to maintain proper position under dynamic process conditions,to provide better accuracy for control, and to increase overall speed of response. Each type of valve has an effective range of pressure drop they can handle, with some overlap. For example, for low pressures, butterfly valves can be used. For increased pressure, a plug or ball valve would be appropriate. Globe valves would be a good choice for high-pressure applications. Special anti-cavitation trim would be required to prevent seat damage at high fluid velocities. Cost is often a major factor in selecting a control valve type. Globe valve designs are popular for relatively small lines up to 3-4”, but become very costly in larger sizes. Ball and plug valves are normally used in lines up to about 6”. Butterfly valves would be the choice for lines larger than 6”. For sluny throttling service, diaphragm valves are typically used in up to 3” lines, while pinch valves are common for 3” and larger sizes. Of course, these rules of thumb only apply if the selected valve satisfies the process conditions.

Variable Speed Variable speed (DC) drives and Variable Frequency (AC) drives are commonly connected to mechanical equipment such as conveyors, rotary valves, screw feeders, etc., to effect flow modulation of prime movers for solids handling. They can also be used in fluid control on pumps and fans instead of modulating valves. The use of these on fluid flow applications is usually more energy efficient, but usually at a higher initial capital cost. A variable speed drive in slurry applications is often preferred over the use of valves because of the high abrasive wear effects on throttling valves, and subsequent high maintenance costs, especially in larger line sizes. Pulsed Variable pulse width and pulse duration solenoids, either direct acting, or pilot duty (a small device that controls a larger device, usually with signal conversion and force multiplication),

2110

in combination with pneumatically actuated diaphragm valves or pumps, are often used to modulate slurry flows. Milk of lime addition for pH control is a common application of this. When the slurry fluid velocity is too low, the lime settles out of solution and cakes readily, so static piping and standard throttling valves plug quickly. A re-circulation pipe loop is used to keep the lime in suspension and deliver it to taps at the use points. A diaphragm valve/solenoid combination is then used at the tap, and pulsed at a frequency proportion to a pH control demand. MAINTENANCE CONSIDERATIONS Proper maintenance and calibration of instruments and control devices has a direct affect on product quality and throughput, and can have a significant impact on profitability. Reducing downtime is also a critical issue, which dramatically affects profitability. Safety Safety has become one of the most important issues in plant operations today. Improperly operating or failed process systems can cause personnel injury and even death. Careless or improper maintenance can be the root cause and must be avoided at all cost. In addition to the potential for injury or loss of life, negligent operating or maintenance practices can result in heavy fines and costly lawsuits.

Good Housekeeping Good housekeeping practices are important for preventing damage to instrumentation devices, wiring, and control cabinets. This may seem obvious, but is often overlooked and contributes to failures and subsequent costly shutdowns. Make sure all cabinet doors and instrument enclosures are securely closed. Even a small crack exposes these devices to the weather, intrusion of dust and dirt, potential corrosive environments, and even rodent damage. Locks should be used sparingly because they prevent access, especially during critical events, and keys may be difficult to find. It is far better to adhere to procedures. Rooms containing electrical and instrumentation equipment should be equipped with automatic door closures and should be regularly checked to make sure they are not blocked open. Make sure that any new room wall penetrations are properly sealed and that any HVAC equipment or positive pressure systems are operating properly. If chemical filters are used in the ventilation system because of a corrosive environment, be sure they are checked and serviced regularly. Instrument Installation Accessibility of field instruments is extremely important, both for observation of process conditions and proper device operation, and to safely facilitate their maintenance and servicing. All instruments and control valves should be located so as to be easily accessible from grade or from elevated platforms that already form part of the required access routes of the plant, with instrument indicators visible to process operators. As an example, the use of remote diaphragm seals with capillary connections to conveniently located pressure transmitters can be used. Process connections to field mounted instruments generally use 1/2 in. National Pipe Thread (NPT) pipe taps, and should include a ball or gate isolation root valve for servicing the instrument without de-pressurizing the line. High temperature and pressure process connections must be provided with double block and bleed valves for safety reasons. Siphons or pigtails are used for pressure gauges in high temperature applications for protection from temperature damage. Differential pressure transmitters are normally provided with a factory installed 3-valve manifold to allow isolation and calibration of the device on line. Control valves are often provided with a bypass manifold, including inlet and outlet isolation valves. This allows the valve to be safely repaired without shutting down the line. The manual bypass valve is normally sized the same as the control valve to allow manual throttling of the fluid while the control valve is out

2111

of service. Handwheels for manual operation are sometimes included on the control valve for temporary operation if the actuator mechanisms fai1. Process connections for slurries or fluids which are corrosive or contain suspended solids, even when employing a diaphragm seal, are often purged or flushed with a compatible fluid, utilizing a suitable back-pressure regulator, to prevent line plugging and diaphragm seizing.

CONTROL VALVE

BYPASS

Figure 9 3-valve bypass manifold

Spare Parts and Consurnables The plant warehouse usually manages the storage, retrieval and replacement of spare parts. The quantity and types of parts kept on hand is determined, to a large extent, on the effect on plant production if they are not readily available when failures occur, and how quickly they can be obtained off-site. Good design practice dictates the standardization of instrument manufacturers and minimization of variations in measurement and control techniques. Using smart transmitters, having wider measurement ranges, will cover more applications. This then allows maximum interchangeability of parts and reduces the required inventory to cover the same risk level. Instrument Shop The instrument shop is a key element in any successhl maintenance program. The desire to increase product quality and yields fuels the demand for improved accuracy and reliability. The overall goal is to maximize maintenance cost effectiveness. The primary tasks for the shop are cleaning, inspecting and testing, calibration, repair, and rebuilding of instrumentation devices per manufacturer’s instructions and calibration data sheets. The extent of a shop’s capabilities is usually in relation to the remoteness and size of the site. If the shop cannot repair an instrument due to lack of specialized equipment, or if is a warranty issue, then they will handle the arrangements for off-site servicing. The shops must be well stocked with test equipment and machinery to effectively fulfil their duty. Process calibrators must be significantly more accurate than the instruments they calibrate and should be re-calibrated themselves on a regular basis, against standards traceable to the National Bureau of Standards. As a minimum, instruments are periodically calibrated at 0%, 50%, and 100% of span using appropriate test instruments to simulate inputs and to read outputs. Smart transmitters are calibrated either from a hand-held field communicator or the DCS, per manufacturer’s instructions. Training Proper personnel training cannot be over-emphasized. Technical training programs, conducted either on or off-site, keep maintenance personnel up to date with the latest equipment and repair techniques. Having a person trained on equipment before conffonting its failure will save valuable time in not having to take a crash course from the manual while attempting to repair that device. Conducting regular training on proper procedures and safety is a must. Documentation A good maintenance program is dependent on access to complete, accurate and up-to-date reference material. An organized and well-run document control department will pay for itself many times over in timesaving alone. Maintaining equipment on erroneous or out of date information is sometimes worse than no maintenance at all. If reference material is missing, it must be obtained from off-site sources (assuming it is still available). Originals should never be removed from the record storage; only copies, preferably with a date stamp, should be allowed. Electronic record management systems are becoming an essential tool of document control departments. These systems facilitate the document identification, logging, organizing, storage,

2112

and retrieval of critical process systems and equipment reference material. These documents include vendor manuals and instructions, engineering drawings and data, operating and maintenance procedures, operating manuals and safety programs. Personnel should always be sure that the documents they use are the latest, up-to-date version. If there is any doubt, they should be checked against the document control’s master copy. When authorized changes are made to the process, equipment, or instrumentation (like adjustments to calibration ranges or control strategy modifications), the changes should be reflected in as-built information markups, on the appropriate master documents, in an organized and controlled manner. This is all too often not done, because each incremental change seems insignificant or is forgotten, until someone realizes that nothing matches the documentation anymore. Then a costly as-built remediation program must be undertaken to bring reference material up to date. Computerized Maintenance Management Tools Instrument management software and systems provide powerful tools and utilities to manage the thousands of instruments typically installed in a large plant. The “Don’t fix it if it ain’t broke” philosophy is no longer acceptable. Predictive and preventative maintenance, and regular calibration intervals can help avoid unplanned shutdowns or unsafe process conditions. HART and Fieldbus smart devices can deliver multiple parameter capability and make those data available for automatic logging and computer analysis. Data such as tag, make, model and serial number, installation / calibration date, process parameter data, process values and alarms, as well as out of tolerance and fault and failure data can be obtained by uploading iiom the instruments. Asset management and instrumentation management support software can automatically gather, store, and analyze these data in real-time to keep track of an individual instrument’s history, provide alerts of faults and failures, evaluate relative performance, determine valve performance and wear, and schedule preventative maintenance intervals. Even without smart instruments, preventative and predictive maintenance routines can be provided through the statistical analysis of process event and alarm records, equipment running hours, production totals, energy use, fkequency of failures, etc. CONCLUSION These are exciting times for plant instrumentation and control technology. New digital capabilities afford unprecedented opportunities for maximizing plant profitability, not only iiom the aspect of more accurate and reliable instruments, but in plant availability through the application of advanced computerized maintenance management tools. The potential for increased maintenance and repair efficiency and reduced plant downtime is great. Embracing this technology does not come without cost. Personnel must be trained or recruited to work with and be comfortable with this advanced technology. Management must also recognize the importance of it and be supportive, not only organizationally but also financially. The potential benefits will surely outweigh the investment.

21 13

Strategies For Instrumentation and Control of Crushing Circuits Stephen D. Parsons', Susan J. Parke8, John W. Craven3 and Robert P. Sloan4

ABSTRACT In recent years instrumentation and control advancements have imparted valuable on-line information for crusher control. These advancements have facilitated enhanced decision making from both a production and a maintenance perspective. This paper provides an overview of the instrumentation and process control strategies used within the mineral industry for primary crushing and multi-stage crushing plants. Crusher instrumentation as well as regulatory and advanced control strategies will be examined focusing specifically on strategies to improve product quality and plant throughput.

INTRODUCTION The continuous demand for high quality crushed product and subsequent high downstream costs associated with below specification product highlights the need for effective crusher control and system efficiency. As such, new crushing plants as well as older plants have embraced control and instrumentation initiatives at both the regulatory control and advanced control levels. Irrespective of whether we are referring to primary crushing or multi-stage crushing, the main objective of a crushing plant is to maintain operating conditions that result in an optimum throughput-product size relationship; this may be either maximum throughput at a constant product size or the finest size possible for a given throughput. Additionally, constraints such as continuity of flow between circuits must be maintained, and disturbances such as ore feed rate and hardness must be compensated for (Herbst and Oblad, 1985). The specific control strategies used to accomplish this depends on the downstream requirements and the circuit configuration. In terms of the mechanics of crushing, maximum throughput can be realised by ensuring that instrumentation and control strategies are implemented with the goal of maximising crusher energy utilisation. Operationally, maximum throughput can be achieved through the optimisation of plant availability. Irregular feed rates to primary crushing plants render optimisation, specifically the maximisation of crusher energy, problematic. As such, instrumentation and control strategies for primary crushing tend to focus more on fault detection and to a lesser extent on optimising the crusher mechanics. With the inclusion of recirculating loads, surge bins and subsequent feedrate control, multi-stage crushing plants employ a broader process control functionality than that of primary crushing plants. Consequently, this paper has a stronger focus on the control of multi-stage crushing plants. In general, the focus of secondary and tertiary crushing circuits is to maximise plant throughput while crushing to a specified crushed product size. Where the crusher produces a saleable product (e.g. road-stone quarries), the control objective is usually to maximise the production of certain size fractions from each tonne of feed (Wills 1997). I

MinnovEX Technologies, Toronto, Ontario, Canada MinnovEX Technologies, Toronto, Ontario, Canada MinnovEX Technologies, Toronto, Ontario, Canada 4 MinnovEX Technologies, Toronto, Ontario, Canada

21 14

System Optlmizattion kbrltecture DlaF)ram

The process control infrastructure consists of crusher control equipment (i.e. plant instrumentation), regulatory control and advanced control. Refer to Figure 1. Regulatory control provides four main functions: 1. 2. 3. 4.

Allows the process to operate at a chosen target; Minimizes the effects of disturbances; Reduces the effect of ore variability; and Provides for safe and efficient start-up, operation, and shutdown of the process.

Hence, the regulatory control system’s function is to facilitate the consistent execution of control actions under dynamic conditions. Regulatory control can be provided by either independent (local) controllers or a centralised control system with the latter being adopted as the standard for new crushing plants. The control system uses either Programmable Logic Controllers (PLC’s) or a Distributed Control System (DCS). Advanced Process Control Systems can be incorporated to provide added optimisation. To date, expert systems have been the most common means of deploying advanced control solutions for crushing plants. This chapter is organised into two main sections. The first focuses on instrumentation requirements and applications for crushing plants. The second reviews process control strategies, including fault detection, stability control and optimisation control. Instrumentation and control strategies are documented for both primary and multi-stage crushing plants.

INSTRUMENTATION Accurate and robust instrumentation is critical to the long-term sustainability and success of a control strategy, particularly given the harsh operating environment of a crushing plant. The following conditions complicate the continuous control of crushing circuits: Variability in feed (i.e. size distribution, hardness, feedrate); Abrasive nature of feed material; Noisy crusher power signals; and Unmeasured disturbances (i.e. tramp metal, chute plugging etc.).

21 15

'able 1: Process control instrumentation- standard equipment for crusher design Manufacturer's Instrumentation - generally included with crusher packages 1. Crusher unit Vibration Transmitters

RTD's Pressure transmitter Proximity Switches Flow switch

2. Crusher Drive unit Power Transducers Amp transmitter RTD's

- positioned on the adjustment ring to detect excessive

forces in the crushing chamber monitors crusher bearing temperatures - provides guidance on the crusher clamping pressure and/or the air dust seal pressure - positioned close to the adjustment drive pinions on crushers to help gauge the crusher closed side setting - positioned on water lines where water is used as the medium for the crusher dust seal -

-

measures the motor power draw measures the current draw on the crusher motor monitors crusher motor and bearing temperatures

3. Crusher Lube Package - positioned on the crusher lube lines to detect oil flow Flow switch - monitors lube oil temperature (reservoir, return line) RTD Differential Pressure transmitter - detects plugging across lube filters filter - provides guidance on the oil tank level Level switch Standard Plant Instrumentation -generally not included in crusher package

Level Transmitters Belt Weightometers Tilt Switches Amp Transmitter Metal Detectors Video Monitoring Equipment

Power Transducers Flow Switch RTD's

- measures the level of material in the crushing cavity

and/or in surge bins measures the mass of material on conveyor belts - positioned in feed chutes to detect plugging events - measures the current draw on auxiliary equipment motors including crushers, conveyors, screens and feeders - positioned before the crusher to detect the presence of tramp steel on the crusher feed conveyor - cameras positioned above conveyors and crusher cavities to provide visual feedback for operating personnel - measures the motor power draw for feeders and vibrating screens - detects restriction or valve failure in water or oil lines - continuously monitors motor, bearing and oil temperatures -

Table 2: Process control instrumentation - non-standard instrumentation and technological advancements On-line Image Analysis - positioned in primary and multi-stage crushing plants for particle size monitoring Internal equipment sensors

- equipment manufacturers are placing smart sensors

(with programmable microprocessors) within equipment to impart more data for fault detection

2116

Table 1 provides a list of commonly used crusher instruments, or what is considered the standard instrumentation for control purposes. The standard instrumentation table is broken down further into what is typically provided with the manufacturer’s crusher package and what is typically added to the design by either the operation or engineering company. Table 1 presents only the main instrumentation that should be considered for direct crusher control and does not extend to peripheral process functions. Non-standard instrumentation (application specific equipment) is listed in Table 2. This presents the higher-end or technologically advanced equipment that is not considered standard but has proven to add control value for specific applications. The instrumentation included in both the standard and non-standard tables was selected from the perspective of throughput maximization, product quality control and crusher protection.

Level Sensing Level sensors (usually ultrasonic) are used throughout crushing plants in surge bins, crusher dump pockets, crusher cavities and slurry pump boxes. Level sensors impart valuable on-line data for throughput control via mass balancing and for ensuring that adequate operating levels are maintained to mitigate equipment damage from excessive impact. Ultrasonic transmitters can be problematic in the presence of high dust concentrations. Dust clouds will reduce the transmittance of sound signals, and dust fouling or build-up on the transmitters can severely impact the reliability and/or accuracy of the instrument. As a result, radar based level sensors are gaining popularity. Proximity and nuclear density switches are often used for high level detection in dusty bins. Level sensors provide an on-line measurement of the ore level in the crushing cavity, thereby providing an indication of whether the crusher is being ‘choke fed’ (a crusher is considered ‘choked‘ when the crusher cavity is full). It is generally accepted that choke feeding secondary and tertiary crushers produces an increase in fines production and a higher overall throughput. Figure 2 illustrates the typical positioning of a level sensor above the cavity for crushers that are operated under choke fed conditions.

Figure 2 :Positioning of an ultrasonic level sensor (courtesy of Nordberg) In consideration of water flush crushers, two level sensors are typically used, one to monitor the ore level and the other to monitor the water level in the crusher cavity. Maintaining an adequate water level ensures that sufficient water is present to flush the fines from the crusher. The water level is typically measured in a protected stillwell. With respect to ore bed levels in crusher dump pockets and surge bins, the bin level and often the rate of change of the bin level, is used to regulate upstream feedrates via mass balancing, with

21 17

the step change to the feeder commensurate with the severity of the condition. This approach is more specific to advanced process control algorithms. The step changes could either cascade back sequentially to the primary plant feeder or create a simultaneous cut to all upstream feedrates. The objective in both cases is to balance the plant feedrate and maximise throughput, subsequently adding stability to all downstream processes. The control strategy employed is a function of the plant configuration and circuit intricacies. Bin level sensors coupled with plant interlocks provide an equipment protection function. Level sensors provide an indication of when the bed level is dangerously low in the bin, to the point where equipment is exposed and susceptible to excessive impact from the feed material. Plant interlocks are configured within the regulatory control system to trip the feeder well before the bin empties, thus ensuring the equipment is always under a protective layer of ore. Low-level interlocks are especially important on crusher dump pockets where unreduced ore drops directly onto an apron feeder or vibrating feeder.

Conveyor weightometer Conveyor weightometers are positioned strategically across the plant to weigh and totalise tonnage at “key” locations, such as: 1.

2. 3.

the crusher product conveyor belt, i.e. screen undersize; the plant feed belt; and the crusher discharge (jaw discharge, cone crusher discharge).

The weightometers should be positioned such that the crusher feedrate, crusher discharge rate and plant circulating loads can be directly measured or easily calculated.

Figure 3 :Positioning of weightometers in a 2-stage crushing circuit Figure 3 illustrates the typical placement for conveyor weightometers in a two-stage crushing circuit. Weightometer-2 weighs the tonnage of the crushed product and Weightometer- 1 weighs the sum of the discharge from the jaw crusher and cone crusher. The circulating load is then determined using the two values (as per Figure 3).

Conveyor Cameras Cameras are used as an operations tool to monitor problematic transfer points and/or plant equipment in an effort to detect problems prior to equipment damage and subsequent plant or equipment downtime. Common camera locations are as follows:

21 18

Conveyor head pulleys; Belt magnets; Above jaw crusher cavities; Transfer chutes; and Screening equipment such as double deck screens and grizzlies.

Proximity Switches (Crusher Setting) The crusher setting can be inferred through the implementation of proximity switches positioned on the crusher pinion teeth (as shown in Figure 4). The proximity switches count the number of pinion teeth during a bowl rotation - adjustment and the direction of rotation. Since each incremental change translates to a known “gap” increase, the crusher setting can be monitored automatically. The calculated crusher setting is re-calibrated when the bowl is tightened to the point where the mantle and bowl are touching.

-

Figure 4 :Crusher gap automation proximity switches (courtesy of Nordberg) Metal Detectors Some of the tramp metal entering a crusher may be manganese steel, which is non-magnetic and does not get picked up by the conveyor magnet. Consequently, metal detectors are positioned after the conveyor magnets to flag the non-magnetic tramp metal before it enters the crusher cavity. Power Measurement Diligently monitoring the crusher power draw via a power transducer is the most common approach used to ensure that the crusher is operating as close to and within a realistic tolerance of the maximum operating power. It is generally accepted that crushers should operate at approximately 85% of the full load amps. Vibration Sensors Vibration sensors can be mounted on the crusher adjustment ring (see Figure 5) to continuously measure adjustment ring movement and provide an alarm signal when the crushing force design limit has been exceeded due to the presence of uncrushables (such as tramp steel or wood chips) or because of changes in the feed material. Via the signal trending either the operator or the advanced control system can infer the nature of the problem and subsequently execute the appropriate control action. For example, instances where the vibration is reoccurring at equal intervals and at a relatively high frequency suggests the presence of a recirculating load of uncrushables in the circuit.

21 19

Figure 5 :Crusher vibration sensors (courtesy of Nordberg) Vibration sensors can also be used for fault diagnosis by measuring the amplitude and frequency of the bowl vibration. Figure 6 is an example of a vibration reading trend; the peaks indicate overload conditions. Historical data, including signal magnitude and trending, can aid the detection andor diagnosis of the problem, whether this be an overload condition or a crusher maintenance issue.

Figure 6 : Vibration sensing readout (courtesy of Nordberg) Currently, vibration sensors are only sensitive enough to differentiate movement in the adjustment ring from normal operating vibration.

Image Analysis Digital image analysis is continuing to gain acceptance within the industry as a means to calculate the size distribution of material on conveyor belts. There are various packages on the market, and most systems perform the following functions: acquire the digital image, perform pre-processing

2120

on the image, delineate the individual fragments in the image using digital image processing techniques and then apply statistical algorithms to determine the particle size distribution. Figure 7 provides an example of the delineated image in comparison with a direct photograph.

Figure 7 :photograph of a primary crusher product (top) and associated delineated image with fines identification (courtesy of Split Engineering) The majority of applications to date have been used to calculate the particle size distribution of the primary crusher product and feed to autogenous grinding and semi-autogenous grinding circuits. In recent years image analysis has also extended to secondary crushing plants. Primary crusher applications: Located on the feed and discharge conveyors for in-pit crushers, the feed and discharge size distributions are used together to assess the relative hardness of the fragmented rock and to provide data for the monitoring of the Crusher performance and crusher wear. It is also possible to integrate the Dispatch data to identify the location of the ore on each truck as they feed to the primary crusher and correlate the run of mine (ROM) fragmentation information to the blasting parameters. These techniques are currently being used at the Phelps Dodge Sierrita in-pit crushers (Keremy, J. et a1 2001). Secondary crusher applications: Measurement of the crusher feed ore size distribution. Located on the crusher feed belt. to provide additional information for inferencing the dynamics of the crushing plant. The measured size distribution can be used together with the feed tonnage and recirculating load measurements to classify the ore hardness (as discussed in the conveyor weightometer section).

2121

Internal equipment sensors Equipment manufacturers are placing internal equipment sensors in key strategic locations. For example, new crushers from Metso Minerals are being fitted with internal RTD’s to continuously monitor the countershaft bearing temperature. It is expected that the countershaft bearing temperature sensor will provide critical data to help flag problematic conditions before a catastrophic failure occurs. CONTROL STRATEGIES Fault detection, stability control and optimisation control are three key elements of crusher control strategies. An effective plant-wide control strategy should be an appropriate blend of the three. Typically, the regulatory control system performs the fault detection, stability control and lower level optimisation control, with the advanced control system performing the higher-level optimisation control and fault diagnosis. The control strategies are broken down into two sections, those conducted by regulatory control systems and those conducted by advanced control systems. A. Regulatory Control

Fault Detection. Fault detection, such as plant interlocks, is used as a means to mitigate damage to plant equipment. Crushing plants are susceptible to downtime since crushing equipment, by nature, is very expensive and standby units or capital spares are frequently not included in the plant design. Therefore, process control, specifically fault detection, is critical for the maximisation of plant availability and the protection of both plant equipment and plant personnel. Table 3 provides a list of conditions requiring a plant interlock. Disturbance Event High crusher oil temperature Tramp metal trip Dust seal water - no flow Crusher amp overload

Control Action Crusher trip Crusher feed conveyor trip Crusher trip Crusher trip

Stability Logic. Stability process control strategies, designed to add stability to the process with respect to throughput and a consistent product size, are first introduced as part of regulatory control, with further optimisation conducted via either higher level regulatory control or advanced control. Stability logic must be designed to effectively manage disturbance events, such as crusher peak overloads and chute or crusher plugging. Crusher disturbance logic generally employs large step changes, with the objective to quickly stabilise the process. Table 4 provides a list of operational events that promote circuit instability and necessitate disturbance logic. Disturbance Event Control Action Change in ore characteristics (size Assess the criticality of the situation. Adjust the tuning distribution, ore hardness, moisture sets by moving to a regime with either more aggressive content) or more conservative step changes to the feed rate. If the ore is harder, for example, the tuning sets should shift automatically to a regime with smaller, more conservative, step changes Irregular feedrates to the crushing Employ feedrate control logic utilising bin levels and plant the rate of change of bin levels to regulate feedrate throughout the plant.

2122

Feedrate Control. Changes in ore characteristics, including variations in the crushing work index, size distribution and feed moisture, cause fluctuations in the crusher power draw. The variations in ore characteristics are often sufficient to necessitate the manipulation of either the crusher feedrate or crusher setting, generally the only two manipulated variables available in a traditional crushing plant. While the crusher feedrate is often controlled in consideration the bin level trending, specifically imbalances between crushing stages, the determination of the maximudoptimal feedrate is typically a function of the crusher power draw, assuming that the crusher is already being choke fed. The ability to maintain the crusher power draw within a tight band of the power setting is a function of the event response time and the selection of an appropriate control action. Surge bins with variable rate feeders or conveyors should be positioned in close proximity to the crusher to diminish lag times in the system, and appropriate control logic must be formulated in consideration of the following:

0 0

the nature of the power excursions - cycling, spikes; the level of ore in the crusher cavity; an ore change inferred via process modelling or using on-line imaging; and the crusher setting.

Figure 8 provides an illustration of actual real-time data for a tertiary crusher with poor feedrate control. The amp draw fluctuations are a result of feed rate variations. Note that while the average amp draw is far below full-load, the peak levels still exceed the crusher full load power rating.

Figure 8 :Crusher trend chart, where heavy line is crusher amps, illustrates poor feedrate control (courtesy of Nordberg) Figure 9 illustrates a tertiary crusher with continuously high power draw attributed to good feedrate control. In this case a surge bin is located directly upstream of the crusher and the belt feeder is directly controlled to maintain the crusher power draw setpoint. This configuration and the control strategy utilised facilitate choke feeding and the elimination of overload peaks.

2123

Figure 9 : Trend Chart - good feedrate control (courtesy of Nordberg) Most vendor-provided crusher control packages are equipped with 3 types of control modes: 1. Auto power setpoint - crusher feedrate is manipulated to run to a power setpoint; 2. Auto level setpoint - crusher feedrate is manipulated to run to a cavity level setpoint; and 3. Manual - crusher feedrate is operated to a percent output.

Water flush crushers utilise an auto water setpoint as well, where the water flowrate is a volumetric addition based on the ore feedrate to the crusher. Typical crusher control strategies use the auto power mode to maintain a power setpoint up to the point where the level in the crusher cavity exceeds a predefined limit. The crusher is subsequently operated in auto level mode until there is an adequate level in the crusher cavity, which triggers the switch back to auto power mode. In both the auto power and the auto level modes the manipulated variable is the crusher feedrate. This approach should maximise the feedrate to the crusher, but it is the plant-wide stability and optimisation logic which regulates the feedrates throughout the plant while not exceeding the circuit constraints, such as transfer chute limitations and conveyor limits. Despite adequate choke feeding, it may still be difficult to track the power setpoint over the duration of a shift if the ore is very abrasive and contributes to high liner wear rates. It is not uncommon for plants to regap their crushers every 12 hours to maintain the product quality and crushing efficiencies.

Optimisation control. Optimisation logic can be executed via the regulatory control system, but it is generally considered “lower level” control relative to that provided by an advanced control system. The following section provides examples of optimisation control strategies implemented via a regulatory control system. “Higher level” optimisation logic implemented via advanced process control systems is discussed in the next section. This level forms a “grey area” between regulatory control (DCSRLC) and the common understanding of Advanced Process Control (Am.

In the following examples the optimisation logic considered the impact of each phase on the downstream process. The impact is quantified and considered in the overall control strategy. Attempts are being made to optimise the whole process by controlling each step, not only the restraints of the single process but each successive process. For example, the Phelps Dodge Sierrita Mine is using imaging systems on the primary crusher feed and product to monitor the crusher performance and crusher wear. The information is also being used to gauge the secondary

2124

crusher feed size and estimate the work index. In addition, the truck-by-truck basis of the feed imaging allows the possibility to trace the size information back to the bench position for fragmentation models and hole-by-hole work indices. The impact of crusher control on downstream processes has been noted at both Mount Isa Mines (Anon. 1973) and the Majdanpek Copper Mine (GrujiC 1996). At Mount Isa Mines improved crusher control strategies have led to increased grinding throughput. The crusher motor current measurements have been filtered to remove spiking caused by variable ore size and hardness, to allow the installed crusher control system to stabilize power. Reducing the magnitude of fluctuations allows the crusher to operate at higher current setpoints without increasing the probability of an overload trip. The control system on this multi-stage process balances the primary and secondary crushing stages and ensures that at least one is operating at maximum capacity. The secondary bin level is used to determine whether the primary or secondary circuit is the constraint. By ensuring balanced, near-choke-fed crusher operation, 10 to 20 percent of the total ore previously reporting above 9.5 mm now reports below 9.5 mm (but predominantly above 4.7 mm). Reduction in the 9.5 mm material is of significance in maximizing grinding throughput without loss of flotation feed sizing. The goal of crusher control at Majdanpek Copper Mine was to minimize particle size and optimise ore flow using motor stress regulation. The data processing is integrated, i.e. automatically controls and stabilises at setpoints within individual loops. The crusher feed control is based on the level in the crusher feed bin and the closed-side setting is based on motor power. Crusher throughput was increased while the average crushed product size decreased after implementing the optimisation control (refer to Table 5). In addition, liner wear decreased. The effect on the downstream grinding is indicated in Table 6.

Table 5 : Effect of Automatic Process Control on the Crushing Circuit at Majdanpek Q(t/h) I P80(mm) W(kWh/t) Steel No.of I Linings(g/t) I Months I

1 Crushing Process

I

control With auto. Process Control

Grinding and Float Process Without auto. Process Control With auto. Process Control

I

2740

6.8

W(kWh/t) 20.8 19.3

I

7.30

+208pm(%) 21.2 14.1

I

7.30

I

10

-74pm(%) 54.0 58.1

These studies indicate the positive impact of optimisation crusher control on downstream processes.

B. Advanced Process Control The group of solutions collectively referred to as Advanced Process Control (APC) constitute a wide range of applications from the more traditional Internal Model Control (IMC) and Statistical Process Control (SPC) to new age algorithms including expert systems and non-linear adaptive multi-variable controllers. Whatever the form, APC algorithms share a common trait - they are strategic as opposed to tactical in nature. That is, as opposed to making sub-second manipulations and changes at the controller level, such as maintaining a setpoint, these applications sit on a higher plane, determining and then sending setpoints to the regulatory ,system with the goal being the “optimisation” of the process. Each of the algorithms operate in a slightly different manner but share the similar concept of developing a “global” model, whether that is fundamental, phenomenological, empirical, stochastic or heuristic. This model can then be drawn upon to provide insight into the process and, with that knowledge, facilitate decision-making.

2125

While there are many secondary benefits derived from advanced process control, it is generally accepted that there are three demonstrable results: stabilisation, optimisation and education. Through a combination of increased vigilance, consistency of application and (with some of the algorithms) the ability to consider information and relationships not readily available to an operator, an advanced control system can both stabilise and push the process to the operational limits. Additionally, APC algorithms result in an intensive and comprehensive educational process. As the ability to control the plant at its operational limits is demonstrated, the people within that operation gain a better appreciation of where those limits exist and why. In many cases the process of developing the APC strategies provides fundamental insight into the operation. The benefits of such solutions appear obvious, but there were and are a number of factors that work against successful implementation. This includes the difficulty in motivating projects as crushing is seen as a “basic” unit operation with little room for improvement (and therefore limited economic benefit). There are generally three main components to an advanced control system. They are as follows: 1. Sensorddata : i. measured (automatic input); ii. measured (manual input); i. inferred 2. The execution platform : i. hardware; ii. software 3. Integration: i. design; ii. execution of the application.

Sensorddata is the most critical component of any solution. The key to this aspect, other than the obvious GIGO (garbage in - garbage out) is that much of the critical data is actually inferred as opposed to measured. As a result, data is even more critical to the outcome and success of the solution due to propagation of error. Without the ability to accurately, or at a minimum precisely, measure the variables having the most profound impact on crushing performance (i.e. ore hardness), the ability to successfully execute an APC solution had been compromised. Suffice it to say that the more empirically based the model, the higher the “domain” accuracy and the less extensible it is beyond its calibrated or understood range of application. For this reason nonheuristic-based systems have not been widely embraced to date. Heuristic expert systems, which strive to emulate operator action, have proven successful in numerous applications. They are more reliant on trends and the precision of a measurement than on the accuracy of a reading and a complete understanding of the process. In the future, as more of the key parameters within the crushing plant become available on-line, the authors can envision a model-based solution enhancing the heuristic foundation now being established. In expert systems control logic has been applied to manipulate: 1. The rate of feed to the crusher; 2. The crusher closed-side discharge setting; 3. The water addition rate for water flush crushers; and 4. Future blast pattern designs.

An appropriate example of APC are control strategies that leverage the identification of operating regimes thereby facilitating a multi state (or variable tuned) solution that responds in a manner that is consistent with the characteristics of that operational point. This is particularly valid within the crushing plant where the difference in ore hardness has a direct impact on the controllability of the circuit, with harder ore requiring a more conservative approach and the recognition of softer ore permitting the exploitation of the circuits potential.

Control Options Within a crushing circuit, ore storage coupled with an effective control scheme greatly facilitates material flow between stages. In a 3-stage crushing circuit, for example, imbalances in capacity show up in the secondary and tertiary bin levels. A sustained high level in a tertiary bin suggests

2126

excess capacity in the secondary crusher whereas a sustained low level suggests a secondary crusher capacity limitation. In an advanced control system the imbalance would be quantified and then considered in the determination of the appropriate control strategy, one where the overall process economics is maximised. Bin levels and power draw may be used to balance the sections while ensuring that at least one of the stages is operating at maximum capacity. Table 7 provides an overview of typical conditions amenable to advanced process control with the control action for each.

Figure 11 illustrates the feedrate control logic implemented as part of the Placer Dome Zaldivar Mine secondary crusher expert system (Craven 2000). Figure 10 provides an overview of the Zaldivar production process.

Figure 10 :Zaldivar Mine production process The expert system logic presented in Figure 11 represents only one of many logic sets utilised within the framework of the expert system. The rectangular blocks represent the status of the condition while the rounded blocks represent control actions. The logic proceeds sequentially starting with an evaluation of the status of the secondary bin. If the bin is not empty, or contains material, then the logic evaluates the status of the crusher power. If the power is extremely high, or high and increasing, the logic will unequivocally execute step change decreases to the crusher feedrate commensurate with the severity of the condition. The feedrate control relies preferentially

2127

on the crusher power and then on the level in the crusher cavity. Only when the crusher power is not increasing does the logic include an evaluation of the level in the crusher cavity. Lower levels in the cavity utilise larger step changes in an effort to produce “choke fed” conditions. Secondary Bln Empty

Power Extremely Hlgh

Yes

Feeder Speed Large Decrease

Power Hlgh and lncreaslng Fast

Yes

Feeder Speed Large Decrease

Y0S

Feeder Speed Small Decrease

Power Hlgh and lncreaslng

Power Hlgh and Decrearlng

Yes

Power OK and not lncrearlng

Y0S

Power Low and not lncrearlng

Yes

BowlLevel not Hlgh

yOS

c

Feeder Speed Small Increase

Bowl Level not Hlgh

yes

c

Feeder Speed Small Increase

Bowl Level not High

yes

c

Feeder Speed Large Increase

)

Figure 11 : Secondary Crushing Plant Advanced Control Logic (courtesy of Placer Dome Zaldivar Mine)

-

I n this case fuzzy logic was used to quantify the severity of the condition and to determine the magnitude of the step change. On site tuning is required to determine the actual magnitude of the step change and the required timing. The logic utilised is always site specific. For example the following advanced control logic differs slightly from the Zaldivar logic, although the ultimate objectives are the same - to maximize crusher power and consistently choke feed the crusher. The control logic for an advanced control system at Brenda Mines (Flintoff and Edwards, 1992) was based on the tertiary bin level and the crusher power, as follows:

2128

1. Low tertiary bin level a. b. c.

increase secondary crusher power setpoints; temporarily decrease tertiary throughput as long as the condition holds; if the secondary crusher is operating at maximum power or condition is prolonged, then request an increase in the secondary closed-side setting; and d. if the condition is prolonged further request that the tertiary crusher be stopped. 2. High tertiary bin level a. decrease the power setpoints on the secondary crushers; b. if the condition prolonged and any of the tertiary crushers are down, then request that a tertiary crusher be started; and c. if the condition is prolonged further request a decrease to the secondary crusher closed-side setting. In summary, advanced crusher control systems can be used to: 1. Automate crusher closed-side setting; 2. Implement adaptive control modes for a) constant setting and b) maximum power; 3. Maintain a balance between crushing sections; 4. Diagnose faults online; 5. Provide online operation manual; and 6. Trend and analyse performance indicators.

CONCLUSION This paper has provided the main instrumentation and control strategies for crusher circuit design. Although all circuit configurations and peripheral instrumentation were not considered, the information should provide insight into the main instrumentation and control strategies needed to satisfy the crushing objective - to promote and maintain operating conditions that result in the optimal throughput-product size relationship. The ability to detect faults and provide feedrate control has been the primary determinant for instrumentation and control strategy selection. It is expected that the continued development of new and smarter sensors coupled with the shift towards more of a mine/mill operating philosophy will continue to expand the control possibilities for crushing plants.

ACKNOWLEDGEMENTS The authors would like to thank Tom Bobo, Split Engineering; Jennifer Abols, Metso Minerals; and Placer Dome’s Zaldivar Mine, in particular Jim Whittaker, for their contribution to this chapter. In addition, we thank various personnel within MinnovEX Technologies Inc. for both their input and valuable discussion, in particular Glenn Dobby and Michael Schaffer. REFERENCES Anon. 1973. Crushing Control Systems at Mount Isa Mines Limited, Case Study 7 Craven, J.W. 2000. MET (MinnovEX Expert Technology) Application Manual for the Zaldivar Crushing Plant Expert System Flintoff, B.C. and Edwards, R.P. 1992. SME Proc. Phoenix, Az. GrujiC, M.M. 1996. Technology improvements of crushing process in Majdanpek Copper Mine, Int.J.Miner.Process, 1996, p. 44 Herbst, J.A.. and Oblad, A.E. 1985. Modern Control Theory Applied to Crushing Part 1, IFAC Automation, p. 301 Kemeny, J., Mofya, E., Kaunda, R., Perry, G., Morin, B. 2001. Improvements in Blast fragmentation models using digital image processing, Proc.38* Rock Mechanics Symposium, Washington, D.C. Wills, B.A. 1997. Mineral Processing Technology, Sixth Edition, Butterworth-Heinemann,Oxford

2129

STRATEGIES FOR THE INSTRUMENTATION AND CONTROL OF GRINDING CIRCUITS Robert Edwards,Andrk Vien and Rob Perv'

ABSTRACT Process control is critical to the optimized operation of all gnnding circuits. Being one of the major cost centres in a mineral processing plant, and often one of the production limiting stages, it is essential that the grinding circuit run not only smoothly, but also as close to its theoretical optimum as possible. Advances in instrumentation, the continued practical application of regulatory controls, and the maturing of the advanced control tools, such as model-based expert systems, fuzzy logic, and neural networks, have provided the right conditions for optimum grinding circuit operation. Effective and practical advanced controls, well-tuned and robust regulatory controls, and a solid layer of instrumentation are the keys to optimum circuit control. This paper discusses the techniques being employed in controlling today's mineral processing grinding circuits. INTRODUCTION Grinding circuits are designed to reduce material to a sue suitable for treatment in subsequent separation and recovery processes. Their performance dictates the efficiency with which breakage energy is imparted, thereby determining the overall efficiency of this comminution stage. Judiciously implemented process controls are an important part of ensuring cost effective operations. The grinding process is the most energy intensive, and thereby usually the most expensive, operation in a typical mineral processing concentrator. Circuit measurements and the knowledge of their relationship to the processes occurring indicate how this energy is being applied (Lynch 1977, Napier-Munn et al. 1996, Stanley 1987, and Wills 1988). The application of the energy can be considered as a function of two types of variables: - Equipment variables that are essentially fixed (mill size, ball charge, cyclone geometry) - Manipulated variables that can be changed continuously (feed rate, dilution water, mill speed, pump speed) Grinding circuit design is the selection of the equipment variables while grinding circuit control is selection of values for the manipulated variables. Grinding circuit control in its many forms basically comprises the following four steps repeated continuously: - A grinding circuit production objective is defined - Measurements from the circuit are used to deduce if the circuit is meeting its objective

' Metso Minerals, Business Line Mineral Processing - Process Technology, Kelowna, B.C., Canada

2130

-

Values of measurements required to meet objectives are defined Variables are manipulated to drive measurements towards the desired values In order to best understand grinding circuit control, and certainly prior to assembling an integrated control strategy, it is important to understand the process and the control tools available. In t hs paper, and in an attempt to better relate the processes and relationships to the measurements and manipulated variables, we will first look at each component of a typical mineral processing circuit. Important details on the structure of control loops are given before examining the typical basic circuit controls. Pertinent comments on the application of instrumentation to grinding control are given, but detailed instrument discussions can be found elsewhere in this volume. We conclude the paper by examining the interactions of the various circuit components and the application of control to find the optimum process operating range.

-

GRINDING CIRCUIT CONTROL OBJECTIVES A clear objective is an essential component of a grinding circuit control strategy. Without clear direction, even the best controls cannot help a grinding circuit achieve optimal performance and may even negatively impact production. That being said, control objectives are dynamic targets based on overall plant economics (e.g. net smelter returns) and downstream process requirements (e.g. grind versus recovery), and require continuous re-evaluation in the face of changing downstream equipment performance (e.g. flotation circuit capacity) and metal prices. The control objectives need to provide unique, realistic and attainable targets. There must be at least one degree of fieedom to accommodate ore type variations; such as changes in feed size, hardness, etc. For example, it is unrealistic to expect a grinding circuit to achieve maximum fineness AND maximum throughput - either throughput OR fineness is achieved at the expense of the other. Examples of grinding circuit control objectives are: - Maximum throughput whilst keeping product density above 42% solids and cyclone overflow size below 80% passing 75 microns - Finest 80% passing size possible, at a minimum of 30% passing 20 pm, while maintaining a fEed 2000 tph feed rate. The objectives define boundaries w i t h which the circuit can safely operate both physically and economically. Accordingly, these boundaries become the limits within which control strategies such as “constraint based optimization” can work. BASIC GRINDING CIRCUIT CONTROL ELEMENTS Grinding circuits are applied in a wide variety of applications and take on numerous configurations to best suit the overall downstream process needs. Focusing on the most widely used components in the metallic mineral processing industry, this discussion is restricted to common single and multi-stage milling circuits, both in open or closed circuit with product size classifiers. The control strategies will vary depending on the circuit configuration, but are always comprised of a combination of commonprocess and control elements. Process Elements The scope of a grinding circuit control strategy typically includes all equipment between the fine ore bins (or stockpiles) and the flotation feed (or other downstream process). The circuits are comprised of ore feeders, conveyors, grinding mill(s), classifiers, and classifier feed pump(s) and depending on ore and breakage characteristics may or may not incorporate (pebble) crushing. Common circuit configurations include primary mills in closed circuit with classifiers (Figure l), or primary mills followed by secondary mills closed by classifiers (Figure 2). In either case the process elements are similar - grinding mill, classification, and slurry transport (e.g. pumping). An understanding of the behaviour and interaction of these components is key to effectively implementing grinding circuit controls. [Earlier works by Carriere 1982, Chang 1982 and Brown 1982, though dated in terms of control hardware, contain good discussions on primary ball milling

2131

and AGISAG grinding circuit control.] Following is a summary of grinding circuit components and their control requirements.

Figure 1 - Single Stage Closed Circuit Milling

Figure 2 - Two Stage Closed Circuit Milling Primary Mills. The primary grinding mill is usually the biggest single consumer of power in the concentrator and plays an important role in conditioning the concentrator feed for downstream processing. Primary mills are typically autogenous (AG), semi-autogenous (SAG), ball or rodmills with the latter being relegated to lower throughput, or multiple circuit plants. Rod and ball mill circuits are typically preceded by secondary and tertiary crushing circuits and considered relatively stable to operate. On the other hand, AG and SAG mills are often only preceded by a primary crusher and, being inherently sensitive to feed size and hardness, are much less stable. The efficiency of primary grinding mills in general is a function of mill filling and charge motion and in AG and SAG milling, unlike primary rod and ball mills, these parameters vary considerably and need much closer attention. Intermediate (Pebble) Crushers. Crushers are used ahead of or in closed circuit with primary mills to accelerate the breakage of intermediate or ‘critical’ sized material. Crushing is more energy efficient than grinding but the capital and operating costs per ton for installing a crushing circuit are generally higher. Judicious use of crushing can increase circuit capacity up to 15% or more. Standard or small gyratory crushers have been used to pre-crush primary mill feed whereas

2132

shorthead crushers are often used to crush the oversize of primary mill discharges (Needham and Folland 1994).

Secondary Mill. It is common to require a second stage of grinding to achieve the final product specification for a downstream process. Rod mills and primary ball mills have typically been followed by a second (and sometimes third) stage of grinding. And whereas some AG mills have been able to acheve the final product size in a single stage, most AGISAG circuits have two or more stages to increase efficiency. Secondary mills are almost always in closed circuit with a classifier, so that the control of a secondary mill requires the optimization of the mill and classifier circuit. Secondary mill and classifier interactions play a critical part in the overall optimization of the grinding circuit. Classification. Classifiers are use to separate the fme material already at or below the product size specification fi-om that requiring further breakage. Classifier fmes typically report to downstream processes, while classifier coarse typically reports back to the mill for further comminution. Particles can be selected for classification based on size andor density. For example, s u e classification in a hydrocyclone can be complicated by the presence of secondary high specific gravity sulphide particles. Typical classifiers are screens, hydrocyclones and spiral classifiers, though the latter are becoming less common. Auxiliary Equipment. The significance of auxiliary equipment such as pumps, pump boxes, piping, conveyors and chutes is often overlooked in grinding circuit control strategies. Pumps and pump boxes play an integral role in delivering slurry between mills and classifiers. Unstable performance at this intermediate stage of processing can lead to flow surges and spills (with their associated uncontrolled circuit water addition) that have significant negative impacts on overall circuit performance. Chutes, pipes or conveyors that are undersized, damaged or blocked can constrict material flow, exacerbate spillage problems and further negatively impact circuit performance. The cost and effort to maintain these simple, but crucial process components are small compared to production losses they can cause. It is vital that throughput or performance limits are not due to the auxiliary equipment. Control Elements There are three basic elements to a control loop; the measurement (sensor based process measurement), the controller (computer applied algorithm), and the manipulated variable (final control element). In a simple flow loop (Figure 3) t h s would be a flow meter, a PID controller and a control valve. In a more complex mill control loop these elements might include a dynamic SAG mill model, an expert system “controller”, and a tonnage feed rate controller.

a +............

Cabling, Cable Tray Junction Boxes :

Orifice Meter for Liquid Flow Measurement

Documentation

........ :Cabling, Cable Tray Junction Boxes .%

Final Control

:

V-Ball Valve for

Flow Modulation

Figure 3 - Example Instrument - Control Loop

2133

A detailed quality analysis of the measurements should be the first step in any grinding circuit work, whether it be commissioning, development, or analysis. Following is a brief discussion on signal conditioning followed by guidelines on the application and use of process measurements. Measurements. Reliable, accurate and precise measurements are the foundation of good control. In all cases the quality of the measurement must be known to ensure it is used appropriately within the control strategy; i.e. imprecise, unreliable measurements have to treated cautiously. Additionally, the conditioning of the input signal must be appropriate for the application. For example, in a simple flow loop the digital filtering of the flow meter reading must ensure the controller does not respond to high frequency noise and only the underlying “real” process measurement. In a more complex model-based loop the relative reliability of the inputs to the mill model will be given more or less weight to preferentially bias the result towards that of the most reliable measurements. Signal conditioning is therefore an integral step in achieving reliable process measurements. Signal Conditioning. Process sampling time, digital filtering and anti-aliasing are three important techniques for signal conditioning (Vien, Edwards and Flintoff 1998). Selection of the proper process sampling time is key to effectively determining accurate dynamic process performance information. With feed rates, and thereby volumetric flow rates, varying considerably in AGISAG circuits it important to select a digital sampling time suitable for all operating conditions. For example, as feed rates increase the process time constants in pump boxes (e.g. retention time) will decrease potentially requiring a higher sampling frequency for adequate level control. A sample time that is too short will place unnecessary demands on the data acquisition subsystem, and when dealing with noisy signals may cause the controller to react to high frequency noise instead of the underlying process signal (as previously mentioned). Sampling at a rate too slow for the process will inaccurately represent the process dynamics resulting in poor controller performance. It is recommended that a sufficiently h g h frequency be selected to handle all foreseeable circumstances. Filtering is a techmque used to reduce the effects of process and measurement noise while retaining the important information contained in the signal. Filtering can be found at the instrument level where analog filters are typically used to remove high frequency noise, and at the controller level where digital filters are used to remove mid-frequency noise (though the move to “Smart” instruments is moving the digital filter into the field instrumentation level as well). Common forms are: the “exponential” filter, which places more weight on the most recent measurement; and, the “moving average” filter, whch treats all past values with the same weight. Both techniques effectively remove high frequency noise, but can over-damp the process when too much filtering is applied (i.e. “over-damping” makes the process more sluggish by increasing the apparent process time constant). Signal “aliasing” is a problem that arises from the discrete sampling nature of digital controllers. A signal can be sampled too slowly for the frequency of the process variation and can make the process appear to have a much slower variation than the true underlying process. This is the basis of aliasing and if not detected will cause the improper tuning of digital filters and PID controllers resulting in poor controller performance. Moreover, the PID controller will react to compensate for the aliased signal instead of the true signal. Aliasing is a potential problem in using moving average filters with incorrect sampling times. Texts and papers addressing controller basics provide implementation guidelines for signal Conditioning. Field Measurements. Following is a discussion of field measurements focusing not on the specific instrument, but on the issues influencing the control signal. Detailed discussions of

2134

suitable instrumentation may be found elsewhere in this text and in previous publications (e.g. Flintoff and Mular 1992, Hathaway 1982.) Weightometers - Tonnage measurements from belt scales always require some form of filtering. Highly variable raw measurements may contain tonnage spikes exceeding the scale measurement limit (called saturation) leading to under-reading of the feed rate. Feeders generally have a response time in the range 4 to 10 seconds so variations with frequencies more than 0.25 cycles per second should be removed with a suitable filter. Regular calibration checks are key to reliable belt scale measurement. Flowmeters - Magnetic flowmeters are widely used for measuring flow rates of water and slurry additions. The instruments are usually reliable and require little maintenance if installed in the correct location. Flow control valves are usually fast acting so flow measurements do not usually require much filtering. Small changes in flow can be immediately corrected by small changes in valve position. However, on large lines eliminating these many small position changes will extend the life of the larger control valves without affecting the relatively robust grinding process. The same is true for flow signals used in mass flow calculations where rapid variations in valve position can and should be avoided. Densitv - Nuclear density gauges are the most common devices used to measure slurry density in grinding circuits though differential pressure cells are sometimes used in sumps, spiral classifiers and flotation columns. The slurry density is converted to percent solids by mass using an assumed average solids density. Cyclone feed density is commonly measured, and though a useful signal, measuring cyclone overflow density can be difficult. Successful cyclone overflow measurements have been achieved however, by taking a thief stream through an on-line device such as a particle size monitor, etc. Calculated steady-state cyclone overflow density is also used in some control schemes. Power - Accurately monitoring (and controlling) mill power is important for virtually all mill control schemes, and critical in most AG/SAG applications. Electric or electro-mechanical meters are typically used to measure mill and crusher power. The mill power signals are often somewhat noisy and require some degree of filtering, while crusher power signals are typically very noisy and must be filtered to be usable. Bearing Pressure - Mill bearing pressure (i.e. hydrostatic lube system back pressure) is a critical reading for AG and SAG applications and requires special attention, as early detection of changes in an AGISAG mill load is vital, Bearing pressure readings are affected by both feed and mechanical condition of the mill and can show variations due .to the mill rotation (Perry and Anderson 1996). If the feed and mechanical variations are filtered correctly any true change in bearing pressure is revealed. If too much filtering is applied the changes in bearing pressure are dampened resulting in an unacceptable and unnecessary apparent time delay. The main noise frequency on a bearing pressure signal will likely be twice the rotational speed of the mill. If t h s signal is sampled too slowly there will be aliasing and subsequent control problems. Successful filtering techmques are fiequency (FFT) and noise cancellation based. As bearing pressure is sensitive to temperature and oil quality (viscosity effects) logic involving bearing pressure should compensate for drift. Mill weight - Mill weight is a key measurement in classical AG/SAG control strategies (Mular and Burkert 1989). Mill weight is now commonly measured with load cells strategically placed under the mill bearings and is proving a reliable replacement for the bearing pressure signal, though often both signals are used. The advantage of load cells over bearing pressure is their measurement of absolute weight and the reduction in signal noise, independent of lube viscosity, temperature and other effects (Evans 2001). They are most common on newer mills as the retrofit cost and effort on older mills can be high. The disadvantage is their relative cost compared to the readily available pressure signal, and replacement expense should a sensor fail. The installation must be done carehlly as incorrect installation of items such as the load cell cables will cause significant measurement error (e.g. keep cables away from high tension power cables and use consistent wire lengths for equal signal attenuation) (Jones and Wright 2001).

2135

In large SAG mills load cells are installed under the non-drive end of the mill (Marshall 2000). This “end weight” is used to calibrate the mill load. Prior to startup the mill is filled with water and the cells calibrated to the known mass. During this initial calibration the proportion of weight on the cells (i.e. at the non-drive end) is assumed constant throughout the mill life and used to determine the mill contents under the various loading conditions encountered during operation. The data from load cells also contains information about the condition and performance of lifters. The extraction, analysis and use of this information for maintenance and control is an interesting area of development (Marshall 2000). Sound - The wide spectrum of sound emanating from an operating mill contains a wealth of information about the internal processes and is a promising area for future mill control developments. In the past the sound intensity over a certain frequency band was used to control mill feed or mill speed, but not without some problems (e.g. the sound signal could be similar when a mill is running empty, as compared to running well, and is a function of the frequency monitored, ore type, etc.). Recent research and commercial products analyse the frequency spectrum more intensely (e.g. Pax 2001). Modem computing equipment allows for continuous digitizing and decomposition of the sound signals and allows the determination of both the position and nature of particle impacts on the liners. In one case, the characteristic spectrum of each type of event in a SAG mill (rock on rock, ball on liner, etc.) is used to determine the relative amount of each event. In another, product intensity peaks in certain frequency bands are counted to estimate the number of ball on liner impacts per second. Suitable microphones and conditioning of the signals are therefore required to isolate the important frequencies. Image - Vision systems are being implemented in AGISAG and crushmg circuits to qualitatively characterize the feed materials (Norbert 2001, Girdner, Handy and Kemeny 2001, Edwards, Flintoff and Perry 1997). The size distribution of material on a conveyor belt can be measured with one of several commercially available image based systems. The systems vary in the techniques used to manipulate the two-dimensional video image to produce a three-dimensional size distribution. The conversion must account for differing orientation of particles, stratification on the conveyor belt and the problems associated with overlapping particles. In all cases the 2-D image is first processed to delineate the particle boundaries, where; in one system an equivalent ellipse is fit to each particle outline and an empirical correction used to convert this to the size distribution, and in another the chord lengths across particles are used to determine their size and distribution. The discussion continues on the merits of each, and their usefulness for control, but in both cases several images need to be analyzed to produce a statistically significant number of measurements. Models (Dhenomenologicall- Process models comprise another form of process “measurement”. Mathematical models, so called “soft sensors”, can be used to provide estimates of process values either difficult to measure, such as ore grindability or mill filling, or key physical measurements requiring modeled redundancy (Broussaud, Guyot, and McKay 2001). The former is based on phenomenological models, which are developed from deep process understanding. In the case of AGISAG grinding they can describe the relationship between power, pressure, mill speed and grinding rates. T h s techmque also provides a model structure that can be used to describe the entire circuit, Such models are currently in use (e.g. Samsog et al. 1996) and have been used by optimization algorithms to determine the input values needed to best meet a predefined performance criteria. The models can also be used to predict near-future circuit behavior. The Kalman filtering technique provides a mechanism to adjust the model parameters according the latest actual process measurement (Herbst 1989). The model predicted values and measured process values are weighted according to their relative reliability and used to update the model coefficients. The newly calibrated model is used until another new process reading is available, the cycle then repeats. The updating process is continuous and is particularly useful if the time between process readings is long (e.g. a multiplexed particle size analyzer might provide a particle size reading every 30 minutes with estimated values calculated every minute between readings). The Kalman filter, in combination with a phenomenological model, also provides

2136

“measurement” redundancy should field instruments fail. Again, using the particle size analyzer example, should the analyzer feed lines plug causing a loss of signal then the model estimates can be used in the interim until the analyzer is repaired. Models (neural networks) - A neural network (NN) is a mathematical model that tries to match the functionality of the brain in a very simplified manner. It consists of processing elements analogous to biological neurons. The neurons each have a number of internal parameters called weights. Altering the weights will change the network response to a stimulus (e.g. inputs, measurements, etc.). The processing elements are usually organized into groups called layers with a typical network consisting of one or more layers. The network interacts with the outside world through interconnections of some of the layers with input and output buffers. The input buffer holds the data presented to the network and the output buffer holds the corresponding response of the network. The goal is thus to choose the weights for all the neurons in the network to achieve the desired inpuVoutput relationship. This process is done automatically by the NN and is known as training or learning. For control applications two networks may be required. The first network is trained to model the plant. It is typically used to infer the state of the process. A second network may be trained to control by using the first network to back-calculate the required control action. T h s control action then becomes the desired output for the back-propagation training of the “controller” network (Nguyen and Widrow 1990). Flament et al. (1990) in their conclusion, summarize well the problems and expectations of neural networks: Due to their ability to approximate any relationship between variables, artificial neural networks (ANN) carry on the expectations of many people. In mineral processing, ANN could be applied in modeling, simulation and automatic control of processing plants. However, due to their non-linear features on one hand and to the dificulties encountered in designing a network on the other hand, this new technology should be preferred only when traditional techniques fail. ANN are at their best when non-linearities or complexity are the main features of the problem to be solved. Intelligent sensors, when real devices are unavailable, sensor recordings filtering, when traditionalfilters do not work, and process modeling, when simple transferfunctions are not good enough, are some examples of such applications. Rates of change - A calculated rate of change, though strictly speaking not a field measurement, can be a very useful control signal. For example, ball mill viscous overloads are characterized by rapid drops in mill power that need to be differentiated from acceptable long term trends and inherent process noise (Austin and Flintoff 1987). Monitoring the rate of change of the mill power draw provides a key indicator that can be used to trigger corrective action before the physical manifestations of the overload are seen (e.g. excessive c h p and steel rejection from the mill discharge). Similarly, monitoring the rate of change of AGISAG power and load is the key to implementing the classical power/load mill feed controller. Rates of change may be calculated using the simple difference between values over time, the difference between current value and a heavily filtered value, a linear regression through a set of points versus time, and the cumulative sum of differences with respect to a reference.

Actuators There are a limited number of manipulated variables in a grinding circuit. Whdst there are many possible measurements and inferred values, control schemes can only manipulate one of the following values: Tonnage - The feed to a grinding circuit is one of the primary control mechanisms. Depending on the grinding circuit objectives, fixed tonnage or a maximum tonnage strategy may be required. In AGISAG mills tonnage may be the only effective control mechanism. Typically, production considerations dictate that the supervisory control strategies increase tonnage as one of the first remedial steps, but decrease tonnage as one of the last. Usually, feed tonnage is manipulated by adjusting the speed of reclaim feeders under frne ore bins or coarse ore stockpiles. Feeder dynamics and conveyor belt length may dictate the use of more sophisticated control algorithms

2137

should the time lag be significant and/or if multiple feeders are present (Anderson, Perry and Neale 1996). Water -Water additions to the grinding circuit are used to maintain primary mill feed density (e.g. water ratio to feed rate), maintain secondary mill rheology, manipulate ball mill recirculating load (e.g. to hold the cyclone feed density setpoint), and/or control cyclone overflow density. Water is often controlled to a flow setpoint ratioed or cascaded from a primary controller, e.g. feed water flow rate from the feed solids tonnage controller, or cyclone feed water flow from a cyclone feed density controller. Pump speed - Pump speed is often used to control pump box level in the face of changing feed flow rates and/or recirculating loads (e.g. cyclone feed pumps and pump boxes). They are particularly useful in AG/SAG circuits where feed rates will necessarily change to accommodate mill load conditions. Mill speed - Variable speed mills are viewed as a necessity by some, but an indulgence by others. The scale is often tipped by the capital cost of the variable speed motor weighed against its initially perceived benefits. Variable speed is a tremendous advantage where ore hardness vanes, as it will affect both the mill’s grinding rate and discharge rate. It may be the only alternative when significant changes in ore hardness or feed s u e cannot be accommodated by changes in tonnage and/or water; i.e. to maintain either mill throughput or a safe mill charge. The most flexible and expensive variable speed drive is the wrap-around motor which allows speed to be changing continuously from 0 rpm upwards. The ring motor is favored for large mills where mechanical constraints preclude the use of gear drives. A ring motor can also serve as an inching drive. More limited but less expensive drives, such as the load-commutated inverter (LCI), allow speed changes over a much narrower range such as 65% to 80% critical. Large ball mills have also been fitted with ring motors. In situations where the friability and hardness of the ore is expected to vary considerably over the mine life, variable speed drives have been used to accommodate the differing balance in grinding load between SAG mill and ball mill circuits.

Controllers The controller executes algorithms to manipulate the actuator based on process measurements, and attempts to hold the process at the prescribed setpoint (ratioed, cascaded, remotely set by a supervisory routine, or operator entered). The most common type is the Proportional-IntegralDerivative (PID) controller, though other types are seeing more use; e.g. the dead-time compensatingroutines. PID Control Theory. The PID controller is the standard control algorithm used in virtually all regulatory control systems. It is implemented in many, slightly differing forms which can give rise to some confusion. The PID is so ubiquitous that a good understanding of its use and limitations is essential for optimum control. The reader should refer to a practical process control textbook for a detailed analysis (Stephanopoulos 1984), but the important principles are described below. Given a setpoint and a measurement, the PID calculates the value of its output as a function of any combination of the following three values: - Proportional: size of the difference between setpoint and measurement, called the error. - Integral: the integral of error over time. - Derivative: the rate of change of either the error or measurement. It is instructive to look at one form of the PID controller, which is given by the equation:

-

-

where: ut K, e,

= PID

controller output at time t

= Controller gain

= Error (setpoint - measurement)

2138

T,

= Sampling time

TI

= Controller integral time

= Controller derivative time A = difference operator e.g. Aet = e, - e,l The derivative action is sensitive to small variations in the measurement so it is infrequently used in practice unless correct filtering is employed. Td

PID Sampling Interval. All measurements will be contaminated by variations of one kind or another and it is the frequency of these variations that needs to be considered. It is important to understand whether the controller andor process can remove variations of a particular frequency range (e.g. 20 seconds per cycle or slower), or whether the variation is too fast to remove and thus can be ignored. Selecting the appropriate measurement filtering and tuning constants is therefore critical and is dictated by three factors (Vien et al. 1998 and 2000): - The actual dynamics of the process to changes in manipulated variable. - The desired dynamics of the process to changes in manipulated variable. - The amount of fast and slow variations in the raw measurement compared to the dynamics of the process. One factor that is commonly overlooked is the execution frequency of the PID loop or sampling interval of the measurement (Ts). Most control systems allow this time to be set for each loop, but it is most often left at the default value or set according to processor loading rather than process control considerations. Setting T, too short makes the controller react mainly to high frequency noise rather than low frequency process changes. Setting T, too long and the controller may m i s s important dynamic information. Selection of T, remains subjective but general guidelines are: - Based on physical variables of flow, T, = 1 second or less; and level, T, = 5 seconds; - Based on open loop process response then: 0.1 < TJz- < 0.2 where ,z = dominant process time constant 0.2 < Ts/zd < 1.O where Td = process deadtime - Based on controller settings: T, > 0.01%where ~i = controller integral time

GRINDING CIRCUIT PROCESS CONTROL The level of process control; i.e. the degree of sophistication and complexity, can be loosely tied to process performance.

11

Regulatory Control

~

1

Time -W Figure 4 - Increasing Performance with Level of Control (after Rogers 1985)

2139

As suggested by Rogers (1985) and shown in Figure 4, increasing sophistication generally means an increase in process performance. Regulatory control provides a large proportion of the benefits to be derived from improved control. Successively higher levels of control contribute, but perhaps to a lesser extent, and may be required to attain the highest circuit performance. Tae ultimate theoretical optimum can understandably never be acheved (e.g. 100% energy efficiency).

BASIC REGULATORY CONTROL LOOPS The minimum acceptable level of process control is the regulatory control level. At this level the control system simply maintains setpoints to compensate for process disturbances and setpoint changes. Typical regulatory control loops in a grinding circuit are: Mill feed tonnage control (to reclaim feeder speed control) Mill feed water control (ratioed from mill feedrate control) SAG Mill sound control 0 Crusher cavity level control Pump box level control (often with a variable speed pump) Classifier feed water flow rate control (to maintain a cyclone feed density) Cyclone feed pressure control (discretely opening and closing cyclones) Ball mill feed water flow rate control (to maintain internal mill rheology) Reagent flow control Mill Feed Tonnage Control Mill feed tonnage control is usually a simple PI control loop, though because of time lags and noise can be difficult to tune. Where there is a recycle stream many operators prefer to control fresh feed rather than total mill feed, allowing a supervisory loop to make changes based on recycle tonnage. The deadtime between the feeders and the weightometer is a common problem A PID controller must be de-tuned (e.g. tuned to react slowly) if there is significant deadtime. In theory, a different type of control should be used if the deadtime exceeds the process time constant. In practice, if the deadtime is more than 5 times the response time of the feeders it is better to use a deadtime compensation controller such as a Dahlin Algorithm or Smith Predictor (e.g. if it takes 10 seconds for a feeder to complete a speed change then a deadtime of more than 50 seconds should be handled by the Dahlin algorithm). The deadtime between the feeders and the mill may appear quite long, sometimes up to 2 minutes. However, the significance of this time must be compared with the response time of the mill (i.e. to changes in feed tonnage) before deciding if any special control is justified. Variable speed SAG mill feed conveyors have been installed on several large mills in recent years with the hope of using belt speed and feeder speed changes together to accelerate changes in actual mill feed. These do not necessarily improve control because for control stability the scheme must be detuned such that the response of the feeders and conveyor is similar to that achievable with a single control loop on a fned speed belt. For example, with both feeder and conveyor being speed controlled there would be a controller with a process time constant of 100 seconds and 0 seconds of deadtime; whereas with a fixed speed conveyor there would be a controller with a -15 second time constant and 75 seconds of deadtime - it is a tradeoff between the process time constant and deadtime, but deadtime can be handled effectively with special algorithms previously mentioned, a long process time constant requires a slower control response.

Mill Feed Water Control Mill feed water is used to control the grinding conditions in the mill. Adding water can be used to increase the slurry discharge rate from the mill. The sensitivity of an AG or SAG mill to changes in water addition is ore dependent, with softer ores typically being less responsive. Water is also usually added at the mill discharge to give the correct conditions for downstream processes such as screening, classification or further grinding.

2140

Mill feed water is added to the primary mill feed chute in ratio to the feed tonnage. To ensure consistent mill density it is important to note that in this case the “feed tonnage” is the actual mill feed; e.g. fresh feed plus recycle (as pebbles or classifier underflow, etc.). Additionally, the weightometer may be a significant distance from the mill so the tonnage measurement may need to be delayed in the calculation. The feed water control can be configured to act as either a simple ratio or as a calculated mill discharge density. SAG Mill SoundBpeed A sound controller can be used to regulate SAG mill speed (Perry and Anderson 1996). The sound control loop cascades the speed setpoint to the drive controller. In some installations the sound controller cascades a setpoint to a speed controller in the mill control system, which, in turn,sends a setpoint to a controller in the mill drive. In h s latter triple cascade confguration however, the “extra” loop unnecessarily serves as an additional filtering stage and slows overall control performance. It should be excluded to simplify tuning of the sound controller. Sound to speed control can be used as mill shell protection and as a mill load adjustment (e.g. adjust speed before resorting to reducing feed tonnage). Increasing mill speed will typically increase mill sound and vise versa for decreasing mill speed. Crusher Level Control Shorthead crushers are often used as recycle crushers in AG/SAG grinding circuits. Crushing efficiency is highest when the crusher feed cavity level can be maintained for choke feed. Being able to regulate recycle crusher federate, though more expensive, is always justified. Power monitoring and control of crusher feed is used as a safety override. Mechanical devices to bypass the crusher feed are needed to divert tramp metal (that has made it passed the belt magnet and other safe guards) and to provide a release should the crusher cavity begin to overfill. A surge bin with feeder controls provides a buffer and temporary relief from high recycle rates, but cannot maintain a lower feed rate when the bin fills. A feed bypass should always be available. Manual or automatic adjustment of the crusher gap provides a further adjustment if choke feeding cannot be maintained, or the cavity level gets too full.

Pump Box Level Control Pump box level control is required to maintain balanced grinding circuit operation. It is not necessarily important to control the level at a precise setpoint, but rather it is important not to let the pump box overflow or run dry. An overflowing pump box creates spillage that needs to be cleaned up and clean-up introduces unmeasured, uncontrolled water into the circuit while at the same time increasing the possibility of chemical contamination (e.g. oil into flotation circuits). A very low pump box level allows the pump to cavitate, disrupting the slurry flow and negatively effecting cyclone performance. It is typical to allow the level to “drift” anywhere in the upper portion of the pump box (e.g. 70% level setpoint plus or minus -25%). The level is often detected using an ultrasonic device and controlled with a variable speed pump. A lower limit must be placed on the pump speed to stop it from slowing too much and allowing the slurry to settle in the lines. In cases where a fmed speed pump is used, water addition is manipulated to maintain level. This option is satisfactoryonly if downstream units can readily compensate for changes in density. If the pump is in closed circuit with a hydrocyclone then manipulation of water for level control complicates circuit operations because of its impact on cyclone feed density and the circulating load (i.e. classifier efficiency changes with changes in feed density). It is preferable to control level with pump speed in this case as it provides more direct “measurement” to “actuator” response. Classifier Feed Water Flow Rate Control Water is added to the classifier feed to manipulate/maintain the classifier feed density (especially with hydrocyclones) and the circulating load. It is often added to an operator entered setpoint, but

2141

can be cascaded from a (cyclone) feed density or mass flow controller. As a cautionary note, a great deal of water may be required to maintain the density setpoint if the circuit is unstable or the density setpoint is even slightly too high or too low. Control logic should be included to protect the water flow rate setpoint from ramping up too high or too low. In some instances, the density control is tied directly to the control valve (e.g. a direct acting loop and not a cascade loop). This practice can cause a breakdown in control if the density gauge becomes unreliable or the circuit is too unstable. Should the density controller need to be disabled for any reason, the water flow control reverts back to manual manipulation of the valve position with no feedback for the flow rate. Clearly, a flow rate setpoint is preferred over a valve position setting, especially if water header pressures are know to vary. Where classifier product (e.g. cyclone overflow) percent solids must be maintained, the classifier feed water setpoint is typically adjusted based on an online circuit solids mass balance. The classifier product is not often easily instrumented and it is usually not possible to obtain a reliable slurry percent solids reading due to piping and instrumentation considerations for direct density control.

Cyclone Feed Pressure Control Cyclone feed pressure control is necessary if wide swings in circulating load are expected (as can be anticipated with AGISAG circuits). Cyclones operate most effectively with a steady feed flow rate and steady operating pressure. When the pressure gets too high an additional cyclone needs to be opened and when the pressure gets too low a cyclone needs to be closed. This process can be automated through the use of air-actuated knife gates tied to a cyclone header pressure controller. To affect this control, the cyclone header must have enough available cyclones so that the addition or removal of one cyclone does not create too much disturbance. Typically five or six operating cyclones are the minimum. A controller can be configured to open a cyclone above a pressure setpoint plus a deadband and close a cyclone below the setpoint minus a deadband. This deadband is important in order to ensure that the cyclones do not open and close too often. Cyclone pressure control may also be llnked to cyclone feed pump speed to provide a constant flow per cyclone. Ball Mill Feed Water Plow Rate Control Sometimes secondary ball mill feed water is necessary to maintain a suitable mill rheology; e.g. if direct classifier underflow contains too little water causing viscous overloads. This can happen when treating ores with significant clay content. Ball mill water is typically added to an operator entered setpoint at a level just enough to stop the onset of overload. Too much water can decrease the mill density to a point that grinding efficiency is lost. Water addition to the classifier feed includes any mill feed water and a total water feed controller is used to control water addition in the classifier feed sump. Reagent Flow Control In many installations reagents are added to the grinding circuit. A typical example is lime addition to maintain a certain flotation feed pH. Reagents can be added dry on the mill feed conveyor or in liquid form at the mill feed water addition point. Reagent can be ratioed to a feed tonnage or to a feed metal content; e.g. frother to feed tonnage and collector to feed metal. A pH probe can used to provide a feedback signal to the pH controller. The dynamics of pH response and each lime addition point must be measured. In many operations lime is added in ratio to SAG mill feed and then to ball mill feed by the pH controller. The dynamic response of pH to lime from SAG feed to cyclone overflow is very similar to that for ball mill feed to cyclone overflow. In this case there are advantages to controlling lime valves in parallel rather than in series.

2142

SUPERVISORY CONTROL Multiple feeder control The feed to a primary mill is controlled by changing the speed of one or more feeders. The tonnage control loop must provide both disturbance rejection (regulatory control) and response to setpoint changes (servo control). In larger operations ore is typically fed by two or more feeders at a time. The ratio of speeds between these feeders is manipulated to acheve a desired ore blend or stockpile profile. It is important to ensure that changes in the feeder ratio do not disturb the mill feed tonnage control. This includes starting and stopping feeders, which is equivalent to changing a feeder’s ratio to and fiom zero. Multiple feeder control (Vien et al. 2000) is used to make the tuning of the tonnage control independent of the number of feeders running and their ratio. As an example of multiple feeder control consider a system of three feeders where a 1% change of speed of a single feeder was found to give an 11.2 tph change in tonnage. If each feeder’s speed were simply a tonnage controller (WIC) output multiplied by its ratio (e.g. the proportion of feed to be supplied by the given feeder), the gain of the process would appear differently to the controller depending on the feeder ratios used, as shown in Table 1. This would require changing the tuning of the tonnage controller depending on the feeder ratios, a clearly unacceptable proposition that would result in all feeders being conservatively tuned and likely unnecessarily slow. Table 1: Example of feeder ratio control without compensation Change in

Ratio of Feeder #

Change in tonnage

Change of speed of Feeder #

I

WIC output 1

2

3

1

2

+1%

2

3

4

+2%

+3%

+1%

0.5

0.8

1.2

+0.5%

+OX%

3

1

78.4 tph +1.2%

In order to make WIC tuning independent of feeder configuration, the speed of a feeder is calculated using the following equation: Speed, = (WCCO x N ) x

Ratio,

2

(Ratio, x ~ u n ,

,=I

where: Speedi WIC.CO N Ratioi Runi

= speed output to feeder ‘i’ = Output of tonnage controller (“A) = Number of feeders installed = Ratio of feeder ‘i’ = Running status of feeder ‘i’ (l=running,

O=stopped)

Using t h s equation, the example in Table 1 can be recalculated as shown in Table 2 SO the tonnage controller tuning does not need to change. Note that this scheme relies on a constant feeder gain (tonnage per % speed). The feeder gain may vary with speed so linearization can be added to make the feeder gain appear constant to the controller. In addition, a delay and filter can be added so that the response of the 2nd and 3rd feeders appears to be the same as for the 1st feeder. For example, consider the installation shown in Figure 5 where the time taken for a section of belt to travel between feeders (tl-2and t2.3) is significant compared to the time taken for a feeder to change speeds.

2143

Table 2: Example of feeder ratio control with compensation Change of speed of Feeder #

Ratio of Feeder ##

Change in WIC output

Change in tonnage

1

2

3

1

2

3

+I %

2

3

4

+0.7%

+1%

+1.3%

33.6 tph

+1%

0.5

0.8

1.2

+0.6%

+1%

+1.4%

33.6 tph

If the ratio of feeder 1 is decreased, the speed of feeders 2 and 3 must increase to deliver the same tonnage, which is shown as the height of material on the belt. Feeders 2 and 3 should wait until the section of belt that was under feeder 1, when it was slowed down, reaches them before changing their speeds.

I

t

t

Figure 5 - Multiple feeder discharge onto belt SAG Mill PowerE'ressure Tonnage Control SAG mill performance is a function of the mill charge. The volume and mass of the charge are seldom measured in industry, but are typically inferred from the power draw and weight of the mill either directly, by using a soft sensor, or indirectly in the control logic. The amount of charge in the mill is controlled by changing the feed tonnage (material in) or the discharge rate (material out). The discharge rate can be manipulated by changing feed water or mill speed. Characterization of any given mill is required to determine which control structure is suitable. The most common method of controlling mill charge is by using a weight (or pressure) and/or power controller to manipulate mill feed tonnage. The choice between power and weighdpressure depends on the process and mill conditions. Various schemes have been tried with combinations of power and weight/pressure. Mular and Burkert (1989) discuss triple cascade loops of power/load/tonnage or power controls with rate of change of power and pressure monitoring. All of which have been successful to a certain degree, but are restrained in their effectiveness by the limitations lnherent in the PID control structure employed. Direct PID control of power is typically problematic due to the classical AGISAG overload relationship (i.e. a decrease in AGISAG power when in an overload condition) and to the fact that maximum power draw does not always correspond to maximum throughput, especially for softer ores. In thls case it may be preferable to control load (at the optimum level). On variable speed mills power control is used as protection. For either type of control, the response is usually slow SO an appropriate control type and execution frequency must be used. PID control is often not suitable.

SAG Overload Detection and prevention of SAG mill overload is one of the most common supervisory control routines. A classic SAG mill overload is indicated by falling power draw with a simultaneous increase in mill load (e.g. increasing weight or bearing pressure). Under overload conditions so

2144

much material has accumulated in the mill that the motion in the charge is inhibited causing a drop in grinding efficiency and further accumulation of material. As the charge volume increases, its center of gravity moves towards the mill centerline, producing the drop in power draw. The larger charge volume, however, gives a rise in bearing pressure (and mill mass) at the same time. In large SAG mills treating softer/finer ores the drop in power may be not be evident as it occurs only at very high loads (possibly beyond the mill operating range) and the only indication of overload is the net gain in mill weight. If an overload condition persists the charge volume may grow until it spills out of or pushes back into the feed chute, resulting in severe mechanical damage. The exact conditions that identify an overload for a given mill must be determined by observation. The actions required to correct an overload condition will also be specific for each mill. Typically control logic examines the rate of change of power and pressure (or weight) to detect an impending overload and automatically triggers corrective action; i.e. cutting feed tonnage by half. The method of calculation of rate-of-change is important. SAG Tonnage and Speed Control The load in a SAG mill will change with varying ore conditions and may develop into a serious underload or overload condition if corrections are not made. The relative effect of tonnage, mill speed and feed water on load for any mill must be determined fiom mill observations prior to designing a suitable load control strategy. A common control strategy in SAG mills, where the objective is maximum tonnage, is to try and add tonnage as the first reaction to decreasing load. Similarly, cutting tonnage is the last reaction to increasing load.

Crusher Gap Crushers are used in grinding circuits to accelerate size reduction of fractions that are ground inefficiently in a mill (e.g. the proverbial “critical size”). The crusher throughput is controlled by the crusher gap setting. A supervisory loop evaluates the accumulation of material in the grinding circuit to determine if problems can be rectified by changing crusher gap. Modem crushers have the capacity to automatically adjust the closed side setting. These crushers usually have logic from the manufacturer to open the gap based on excessive vibration or power draw and close the gap based on power. Optimizing crusher use can have a significant effect on throughput. As an example, for a semiautogenous-ball mill-crusher (SABC) circuit the control strategy will determine if the gap should be adjusted to accommodate all recycle tonnage or if the crusher is held at the minimum setting with any tonnage that cannot be crushed simply bypassing the crusher. Ball Mill Circulating Load The circulating load (CL) is the amount of material returned to the mill from the classifier expressed as a percentage of the circuit feed. (In many circuits the classifier is a cyclone and because of this will be used as an illustration in the following discussion.) The CL cannot easily be measured directly, but can be inferred from a measurement of cyclone feed rate (density and flow rate measurement) and/or changes in the pump box level or cyclone feed pump speed and pressure (e.g. increasing CL as indicated by increased cyclone pressure, increased pump speed and/or increase pump box level). Cyclone pressure and cyclone feed pump speed control are used to essentially change the capacity of the classifier to accommodate different circulating loads. There are two independent mechanisms for changing circulating load - Changing circuit feed rate. An increase in feed rate causes a coarser product size and also a higher circulating load. The response to feed rate changes is moderately slow. The circuit moves to a new equilibrium with the maximum change occurring once the new equilibrium is reached. - Changing cvclone feed water addition. An increase in cyclone feed water addition causes an immediate drop in the cut size of the cyclone, giving a finer product in the

2145

short term. This causes a net accumulation of material in the circuit shown by an increase in circulating load. As the circuit reaches steady state, the cyclone product becomes coarser eventually bringing the product fineness down to a level close to where it started. This process can take between 20 minutes and 3 hours depending on the size of mill and the circulating load. The use of each control mechanism depends on the control objective. Ball mill circuit control is complicated by the interaction between classification and grinding and is an area where carefully executed simulation and plant tests are recommended to determine appropriate control solutions. One solution is to use a circulating load controller to control water addition and let product size control feed tonnage (Wills 1988, Pg. 298). Balancing Primary and Secondary Circuits In multiple-stage circuits the balance between the grinding load of the primary and secondary (and tertiary) circuits can be controlled to a moderate degree to increase performance. If a secondary ball mill circuit is at maximum capacity, a supervisory control can determine if it is possible to shift part of the load to the primary circuit. The converse is also possible; that is, if the primary circuit dictates capacity it may be possible to shift part of the load to the secondary circuit. The idea is to balance the circuits such that both are utilized to their maximum capacity. EXPERT SYSTEMS AND ADVANCED CONTROLLERS Expert Systems Expert systems (ES) have become widely used to monitor and control the interacting mechanisms in grinding circuits and are particularly popular in large circuits where the economic benefit of even marginal increases in performance are significant. Herbst, Pate and Oblad (1989), Hales, Colby and Ynchausti (1996), Broussaud and Guyot (1999), Sloan et al. (2001) have documented successful installations with “typical” performance improvements in the range of 4% - 8% increases in throughput. The challenge for a concentrator’s management team is to qualify these “small” benefits in light of the various process factors that complicate the calculations. Gritton (2001) and McKay and Broussaud (2001) have provided techniques using random period odoff tests to accurately perform such a “statistically-defensible”benefits analysis. That being said, it is important that the circuit first be characterized to find the appropriate structure and control parameters to ensure that maximum benefit is derived from the ES. That is, the control strategy should be determined first and only then should the implementation strategy be contemplated. Whilst the term “expert system” is now used to describe a range of advanced control solutions there are important differences between the actual control implemented. The term “expert system” strictly refers to systems that use crisp or fuzzy logic with heuristic “rules”, but has become synonymous with a more general range of products that include ES’s at their core in support of models, neural nets, etc. Broadly categorizing, the distinguishing features of the systems are: - Rule-based control (fuzzy or crisp) - Rule-based control + models (soft sensors) - Rule-based control + models (soft sensors) + model based optimization

Grinding circuit ES’s rely on encapsulating the type of analysis and decision making used by an impeccably trained and experienced operator. This knowledge is written as a set of interrelated rules to derive actions (e.g. setpoint changes) based on prevailing circuit conditions. The rules are applied consistently and continuously and if well written will allow all operating shifts to perform, as a minimum, at the level of the best operating crew. The setpoints are manipulated to achieve the predefrned objectives within safe operating limits as set by the technical staff, operations, management, etc. This lets the operator direct the operation of the circuit whilst the expert system deals with interactions and identifies and eliminates any developing problems.

2146

Augmenting field measurements with “soft sensor” estimates may further enhance control. AS mentioned previously, models can be used to predict unmeasurable values, such as grindability, or estimate measured values (e.g. between measurements) based on current conditions. Current modeling techniques being employed provide on-line updating of the model parameters, such that the models are continuously tested with new process data and recalibrated when necessary. At the highest level, these models are used to calculate circuit setpoint changes that will maximize grinding performance.

Rule-based control The structure of a typical rule-based grinding controller can be broken down into sequentially executed components that: - Check the validity of each measurement used - Identify any alarm conditions and decide on corrective action - Analyse the circuit to identify any process problems and decide on corrective action - Analyse the circuit to identify whether objectives are being reached and decide on action. - Evaluate the relative importance of the actions and execute the most important.

This cycle is repeated at each control interval; e.g. between 20 seconds and 1 minute, depending on the dynamic response time of the circuit. Optimizing control: Model based When a model is available as a soft sensor, it is a natural extension to use this model to perform optimizing control. Model Predictive Control (MPC) is t-ically multivariable in nature and often involves predicting mill behavior for the near future (Hales, Colby and Ynchausti 1996, Herbst, Pate and Oblad 1989, Broussaud and Guyot 1999). One of the keys to a successful MPC implementation is to establish the appropriate optimization criterion for the process variables. The criterion will often involve weighting certain variables more heavily than others, or penalizing utilization of certain manipulated variables more heavily than others. For example, grinding circuit product size is more important than pump box level and manipulating water should be preferred to manipulating feed tonnage for short-term variations. Though not used extensively in the mineral processing industry the optimization function can also include model predictions. The optimization function can also include constraints (i.e. equipment capacity limitations) either within the grinding circuit itself, or within upstream andor downstream processes. Ultimately, the optimization fbnction could directly express economic considerations. Part of the hesitation in using MPC in mineral processing is the need to provide a shell around the main MPC algorithm to ensure a robust application. This shell may consist of an expert system that validates sensors, checks the model predictive capabilities and ensures that the MPC results are sensible. It may also include a component that controls the model parameter adaptation so that the model does not try to adapt to abnormal situations such as a sensor failure. EXAMPLE CONTROL STRATEGIES To summarize the discussion let us briefly consider some examples of grinding circuit control. The first example is a circuit under simple regulatory controls, and the second is a circuit including model based controls.

SAG mill / Ball mill circuit basic control A simple PID control structure, as shown in Figure 6 , is comprised of: - SAG mill bearing pressure controller cascading a setpoint to the mill feed tonnage control - SAG mill feed water ratioed to mill fiesh feed + recycle tonnage - SAG mill speed controlled by sound

2147

-

SAG mill overload detection or high motor power alarm causing a forty percent cut in feed tonnage Recycle Cpebble) crusher level controlling feed to the crusher Cyclone feed density controlling water addition to the sump Cyclone feed pressure controlling number of operating cyclones (e.g. opening and closing cyclones on pressure) Cyclone feed sump level controlling pump speed Particle size analyzer monitoring cyclone overflow size

-

AIT Size Indicating Transmitter DIT - Density Indicating Transmitter FIT - Flow Indicating Transmitter FIC - Flow Indicating Controller IIT Current Indicating Transmitter JIT - Power Indicating Transmitter LIT - Level Indicating Transmitter LIC - Level Indicating Controller PIT - Pressure Indicating Transmitter PIC - Pressure Indicating Controller SIT - Speed IndicatingTksmitter WIT - Weight Indicating Transmitter WIC - Weight Indicating Controller XIT Sound Indicating Transmitter

-

'

(size)

FIT

DIT

~

XIT (sound) JIT

L1c SIT IIT

Figure 6 - Simplified Regulatory Control Loops for a SAGBall Mill-Crusher Circuit SAG Mill / Ball Mill Circuit Advanced Control The same regulatory control loops are employed as in the basic circuit of Figure 6, with the addition of a model-based component. The model-based optimizing control uses a Kalman filter to continuously adapt the parameters of a SAG mill model, ball mill model and cyclone classification model. An optimizer with the objective of maximizing throughput subject to a constraint of 80% finer than 120 microns uses the model inputs and finds the tonnage, SAG feed water and cyclone feed density setpoints that move the circuit toward higher throughput. An expert system verifies these new setpoints are reasonable, that the modeling is valid, and that the physical circuit is capable of accepting the recommended changes. The new setpoints are then applied subject to limits set by the operator (e.g. operator limited tonnage increases because of low stockpile levels, limited flotation capacity, etc.). CONCLUSIONS Strategies for the instrumentation and control of grinding circuits have matured over the past few decades and now encompass regulatory through optimizing controls. Where once regulatory PID controllers were the only available tools the practitioner may now select from a number of more advanced techniques. The fundamentals of circuit control, however, remain unchanged. Field instruments and PID controllers must not be ignored in the drive towards more sophsticated applications. Correct instrument installation and maintenance, suitable signal conditioning, good loop tuning and the appropriate controller implementation remain critical to sustain stable operations from a solid foundation of control. The control hierarchy builds on the regulatory base to include supervisory and Optimizing strategies. Regulatory controls provide process stability, supervisory controls ensure setpoint adaptation to changing process conditions, and optimizing controls maximize the fineness or circuit throughput objectives. A strategy based on a set of focused and obtainable objectives, supported by the requisite field instrumentation, will help ensure its success.

2148

The hture will see the incorporation of enterprise-wide models to optimize not only the grinding circuit, but also the mine and haulage systems (feeding the circuit), the flotation plant (upgrading the product), the dewatering plant and transportation system, and the smelter I refinery producing the metal. When taken as a whole, the inefficiencies that exist between the various production stages can be minimized and the overall enterprise optimized for maximum economic returns. REFERENCES Anderson L., R. Perry, A. Neale 1996. Application of Dead-time and Gain Compensation to SAG Feeder Control at P.T. Freeport Indonesia. Proceedings of the 28Ih Annual Meeting of the Canadian Mineral Processors, Ottawa, Ont. January 1996. Pg. 360. Austin J.W., B.C. Flintoff 1987. Production Improvements Through Computer Control at Brenda Mines Ltd. Paper presented at the American Mining Congress Meeting in San Francisco, September, 1987. Brown C.M. 1982. The Selection of Instrumentation and Control Systems for Semi-Autogenous Grinding Circuits. In Design and Installation of Comminution Circuits, ed. A.L. Mular and G.V. Jergensen 11, Chapter 41. American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc. Baltimore, MA: Port City Press, Inc. Broussaud A., 0. Guyot, 1999, Factors Influencing the Profitability of Optimizing Control, Symposium Proceedings, Control & Optimization in Mineral, Metallurgical, and Materials Processing, ed. Hodouin, Bazin and Desbiens. Quebec City, Canada, CIM, pp. 393-403 Broussaud A., 0. Guyot, J. McKay 2001. Advanced Control of SAG and FAG Mills with Comprehensive or Limited Instrumentation. Proceedings of the International SAG Conference held in Vancouver, B. C., September, 2001. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 358. Camere K.C. 1982. Computer Control of Grinding Circuits. In Design and Installation of Comminution Circuits, ed. A.L. Mular and G.V. Jergensen 11, Chapter 38. American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc. Baltimore, MA: Port City Press, Inc. Chang J.W. 1982. Primary Ball Mill Circuits. In Design and Installation of Comminution Circuits, ed. A.L. Mular and G.V. Jergensen 11, Chapter 40. American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc. Baltimore, MA: Port City Press, Inc. Edwards R.P., B.C. Flintoff, R. Perry, 1997. Developments in Distributed Control and Expert Systems for Process Control. In Proceedings from the SAG '97 Workshop, Vina del Mar, Chile. Evans G. 2001. A New Method for Determining Charge Mass in AGISAG Mills. Proceedings of the International SAG Conference held in Vancouver, B.C., September, 2001. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 33 1. Flament F., J. Thibault and D. Hodouin 1990. Potential Applications of Neural Networks for the Control of Mineral Processing Plants. CIM Con$, Hamilton. Flintoff, B.C., A.L. Mular Eds. 1992, A Practical Guide to Process Controls in the Minerals Industry. Chapter 7 . MITEC. Published in Vancouver: Gastown Printers Ltd. Girdner K., J. Handy, J. Kemeny 2001. Improvements in Fragmentation Measurement Software for SAG Mill Process Control. Proceedings of the International SAG Conference held in

2149

Vancouver, B. C., September, 2001. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 270.

Gritton K., 2001, Methods to Document the Benefits of Advanced Control Systems, SME Annual Meeting, Preprint 01-020. Hales L.B., R.W. Colby, R.A. Ynchausti, 1996. Optimization of AG and SAG Mills Using Intelligent Process Control Software. Proceedings of the International SAG Conference held in Vancouver, B. C., September, 1996. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphcs. Pg. 632. Hathaway R.E. 1982. Selection and Sizing of Instrumentation and Control Systems; Size Controlled Grinding Circuits. In Design and Installation of Comminution Circuits, ed. A.L. Mular and G.V. Jergensen 11, Chapter 39. American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc. Baltimore, MA: Port City Press, Inc. Herbst J.A., W.T. Pate, A.E. Oblad 1989. Experiences in the Use of Model Based Expert Control Systems in Autogenous and Semi Autogenous Grinding Circuits. Proceedings of the International SAG Conference held in Vancouver, B.C. September 1989. Vol. II. Vancouver B.C. Canada: First Folio Printing Co., Ltd. Pg. 669. Jones R., A. Wright, 2001. Selecting and Configuring Load Cells for AG/SAG Grinding Mill Applications. Proceedings of the International SAG Conference held in Vancouver, B. C., September, 2001. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 227. Lynch A.J. 1977. Mineral Crushing and Grinding Circuits; Their Simulation, Optimization, Design and Control. New York Elsevier Scientific Publishing Co. Maerz N.H. 2001. Automated On-Line Optical Sizing Analysis. Proceedings of the International SAG Conference held in Vancouver, B.C., September, 2001. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 250. Marshall C., 2000. Successful Load Cell Utilisation to Measure AG/SAG Mill Charge Mass, IIR Crushing & Grinding 2000, Perth WA, 18-19 May 2000. McKay J., A. Broussaud, 2001, Benefits Analysis of Expert Control Systems, SME Annual Meeting, Preprint 01-171. Mular A.L., A. Burkert 1989. Automatic Control of Semiautogenous Grinding (SAG) Circuits. Proceedings of the International SAG Conference held in Vancouver, B.C. September 1989. Vol. II. Vancouver B. C. Canada: First Folio Printing Co., Ltd. Pg. 651. Napier-Mum T.J., S. Morrell, R.D. Morrison, T. Kojovic. 1996. Mineral Comminution Circuits; Their Operation and Optimization, Julius Kruttschnitt Mineral Research Centre, The University of Queensland, Australia. Needham T.M., G.V. Folland, 1994. Grinding Circuit Expansion at Kidston Gold Mine. Annual Meeting of the Society of Mining Engineers, Littleton, CO. Preprint 94-104. Nguyen D.H., and B. Widrow 1990. Neural Networks for Self-Learning Control Systems. IEEE Control System Magazine, Vol. 10, No. 3, pg. 18.

2150

Pax R.A. 2001. Non-Contact Acoustic Measurement of In-Mill Variables of a SAG Mill. Proceedings of the International SAG Conference held in Vancouver, B.C., September, 2001. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphcs. Pg. 386. Perry R., L.Anderson 1996. Development of Grinding Circuit Control at P.T. Freeport Indonesia’s New SAG Concentrator. Proceedings of the International SAG Conference held in Vancouver, B.C., September, 1996. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 671. Rogers J.A. 1985. Optimizing Process and Economic Gains. Chemical Engineering. Vol. 92, No. 25, pg. 95. Samog P.O., P. Soderman, U. Storeng, J. Bjorkman, 0. Guyot, A. Broussaud, 1996. Model-Based Control of Autogenous and Pebble Mills at LKAB Kiruna KA2 Concentrator (Sweden). Proceedings of the International SAG Conference held in Vancouver,B.C., September, 1996. Vol. 11. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 599. Sloan R., S. Parker, J. Craven, M. Schaffer 2001. Expert System on SAG Circuits: Three Comparative Case Studies. Proceedings of the International SAG Conference held in Vancouver, B.C., September, 2001. Vol. II. Vancouver, B.C. Canada: Pacific Advertising Printing & Graphics. Pg. 346. Stanley G.G. 1987. The Extractive Metallurgy of Gold in South Africa, Chapter 3, The South Afiican Institute of Mining and Metallurgy, Monograph Series M7, Johannesburg,RSA. Stephanopoulos G. 1984. Chemical Process Control; an introduction to theory and practice. New Jersey: Prentice-Hall, Inc. Wen A., R.P. Edwards, R. Perry, B.C. Flintoff 1998. Back to Basics in Process Control. Proceedings of the 281h Annual Meeting of the Canadian Mineral Processors, Ottawa, Ont. January 1998. Pg. 337. Vien A., R.P. Edwards, R. Perry, B.C. Flintoff 2000. Malung Regulatory Control a Priority. In Control 2000, ed. J.A. Herbst. Society of Mining Engineers. Littleton, CO. Pg. 59-70. Vien A., J. Palomino, P. Gonzalez, R. Perry 2000, Multiple Feeder Control, Proceedings of the 32nd Annual General. Meeting of the Canadian Mineral Processors, CIM, Ottawa, pp. 295-3 12. Will B.A. 1988. Mineral Processing Technology;An Introduction to the Practical Aspects of Ore Treatment and Mineral Recovery. Chapter 7. Camborne School of Mines. Printed by Pergamon Press.

2151

Strategies for the Instrumentation and Control of SolidSolid Separation Processes Gerald H. Luttrell' and Michael J. Mankosd

ABSTRACT

Solid-solid separation processes are used to upgrade a variety of materials such as coal, tin, iron ore, and heavy mineral sands. These processes, which include density, magnetic and electrostatic, are often operated without online instrumentation or automatic controls. In many cases, this mode of operation is dictated by complexities associated with the control of large numbers of individual separators. This article discusses the problems associated with the real-time adjustment of process setpoints based on online measurements of concentrate grade and suggests alternative modes of operation that are better suited for plant control and optimization. Case studies that illustrate the potential impact of these different approaches are presented. INTRODUCTION Solid-solid separation processes are routinely used to separate minerals based on differences in intrinsic properties such as specific gravity, magnetic susceptibility, or conductivity. Common examples of solid-solid separation processes are listed in Table 1. These unit operations are often incorporated into parallel and/or multistage processing circuits at industrial sites. Parallel circuits are required because of particle size limitations and throughput restrictions associated with these processes. For example, coal preparation plants are typically forced to include three or more parallel circuits as part of their basic flowsheet because no unit operation currently exists that can upgrade a very wide range of particle sizes. Likewise, mineral sands plants are forced to distribute feed slurry to hundreds of spiral separators or dozens of electrostatic separators because of throughput restrictions associated with these processes. Multistage circuits are commonly used in many of these operations to sequentially recover a variety of saleable minerals from the same feed stream. Several stages of cleaning and scavenging in series are also frequently needed to overcome inherent inefficiencies in these processes created by the random misplacement (imperfection) and bypass (short circuiting) of particles. Table 1. Examples of common solid-solid separation processes. Separator Type Coarse Particles Intermediate Particles Dense Media Spirals Density Pneumatic Jigs Shaking Tables Drum Magnet Drum Magnet Magnetic Roll Magnet Roll Magnet Electrostatic Screen High Tension Electrodynamic Roll Electrostatic Plate Miscellaneous

___

Ore Sorters

Fine Particles Centrifugal Separators Tilting Frames High-Gradient Matrix WHIMS Carousel

-----

I Department of Mining & Minerals Engineering, Virginia Polytechnic Institute & State University, Blacksburg, Virginia. * Eriez, Erie, Pennsylvania.

2152

The inherent complexities associated with the use of multistage unit operations andor parallel circuits makes solid-solid separation processes less amenable to traditional approaches for automatic control. In some cases, this difficulty is created by a lack of reliable and affordable online analyzers that are needed to monitor large numbers of process streams. In other cases, the online manipulation of individual process setpoints to automatically control concentrate quality is simply impractical. For example, the development of a control system that could simultaneously monitor and adjust the setpoints of hundreds of spirals would be a formidable task. Some processes, such as wet drum magnets, are essentially impossible to control since they do not have any user adjustable variables. As a result, industrial instrumentation is typically limited to conventional sensors for monitoring volumetric flow rate, mass flow rate, slurry density, or pulp level. These parameters are commonly maintained at predefined setpoints using simple feedback loops for stabilizing control. Many of the key process variables that impact process performance are visually monitored and manually adjusted by plant operators. Common examples include the selection of specific gravity setpoints for dense media separators or the setting of splitter/cutter positions for spirals, magnetic separators, and electrostatic separators. Control loops are rarely used in industrial plants to optimize performance based on online measurements of concentrate grade or recovery. The installation of online analyzers and controls for the realtime adjustment of individual process variables has been attempted in some cases, but these endeavors have been largely unsuccessful in terms of improving productivity or profitability. Despite the problems described above, most industrial plants that utilize solid-solid separation processes have the potential to significantly improve metallurgical performance using online instrumentation and automatic controls. This improvement can only be realized, however, by properly implementing a control strategy that incorporates a fundamental optimization principle known as the incremental grade concept. This concept points out the shortcomings associated with the realtime manipulation of process setpoints for plant optimization. More importantly, the concept also suggests alternative methods for optimizing plant yield that involve the supervisory control of plant blending practices. This article provides a general review of the concept of incremental grade control and describes industrial control schemes that currently make use of this optimization strategy.

INCREMENTAL GRADE CONCEPT The optimization of solid-solid separation processes based on the concept of constant incremental grade has long been recognized in the technical literature (Mayer, 1950; Dell; 1956). The basis for this important conceDt can be best illustrated using the simple circuit diagram shown in Feed Figure 1. In this example, the circuit incorporates two different separators that are configured to upgrade the oversize and undersize streams from a classifier. The combined mass yield (Y) for the overall circuit can be s, Classifier s2 calculated using:

fl

t-l

Y = S,Y, +S,Y, Separator # 1 (Yl)

in which SI and S2 are the respective percentages of feed material reporting to separators 1 and 2 and YI and Y2 are the respective yields of concentrate generated by separators 1 and 2. Likewise, the grade of the combined concentrate from the circuit can be calculated using:

Separator #2

'T'

(YZ)

Concentrate I

where GIand G2 are the concentrate grades produced

2153

Figure 1. Simple two-unit circuit.

by separators 1 and 2, respectively. The optimization of circuit performance requires that the combined yield (given by Equation 1) be maximized subject to a constraint imposed on concentrate grade (given by Equation 2). For the case of Equation 1, this objective can be achieved mathematically by taking the derivative of Y with respect to Y and setting the result equal to zero. This gives:

Equation 2 can also be easily rearranged to provide a second governing expression for Y. In this case, an expression for maximum yield is obtained by taking the derivative of Y with respect to Y2. After setting the result equal to zero, this gives: - ay = - (s,, l ~ + G l ~ ) + ~ [ Y 2 ~ + G 2 ) = 0 ay, G ay, ay,

-$(Yl

s,

-+Gl XI ay,

(EQ 5 )

au2

Combining Equations 4 and 6 gives:

The optimization criteria expressed by Equation 7 is not readily apparent. This dilemma can be resolved by considering how a separation process is impacted when the concentrate yield is changed by an infinitesimally small amount (AY). In such a case, the resulting yield and grade can be computed using the following mass balance expressions: 'new

= 'old

Ay

The term G* is simply the grade of the last increment of mass added to the concentrate when the yield is increased by an infinitesimal amount. Equation 9 can be rearranged and simplified to show that:

G* = YnewGnew - YoldGold = (Yold Ay)(Gold AY

-IAG)

AY

- YoldGold

= y-*'+G

AY

(EQ 10)

Thus, the left hand and right hand sides of Equation 7 are simply the incremental grades produced by separation processes 1 and 2, respectively. Thus, the maximum yield of concentrate can only be obtained when each separator is operated at an identical incremental grade. This concept is valid for any number of parallel circuits and is independent of the characteristics of the feed streams. It can also be shown that plant profitability is maximized when constant incremental grade is maintained regardless of the operating costs of the different separation processes (Abbott, 1982).

2154

The fundamental basis for the incremental grade concept can be demonstrated using the simple illustrations provided in Figure 2. Each of the two illustrations show two feed streams comprised of various amounts of valuable particles (dark colored material), gangue particles (light colored material), and middlings. The makeup of the first feed stream makes it relatively easy to upgrade since most of the valuable particles and gangue particles are well liberated. In contrast, the second feed stream is difficult to upgrade due to the large percentage of middlings particles. Figure 2(a) shows the result that is obtained when the separation is conducted to provide a target grade of 75% for each feed stream. For the first feed, all but the pure gangue particles must be recovered to reach the target grade of 75%. This operating point produces a concentrate yield of 75%. For the second feed, a yield of only 53.8% can be realized before the target grade of 75% is exceeded due to the poorer quality of this feed. The concentrate from these two feeds produces a combined yield of 64% at the desired target grade of 75%. At first inspection, this appears to be an acceptable result. However, a closer inspection shows that particles containing up to 75% gangue reported to concentrate when treating the first feed stream, while at the same time many particles containingjust 50% gangue were discarded when treating the second feed stream. The only way to prevent this problem is to treat the feed streams at the same incremental grade. As shown in Figure 2(b), operation under this condition increases the concentrate grade for the first feed from 75.0% to 89.3% and decreases the grade for the second from 75.0% to 65.9%. However, when the two streams are blended together, the combined concentrate still meets the required target grade of 75% while providing a concentrate yield of 72%. Thus, the yield obtained by operating at constant incremental grade is significantly higher than that obtained by operating at constant cumulative grade (i.e., 72% versus 64%). In practice, it is easy to be fooled into accepting the lower yield because of the presence of multiple feed streams. The optimum result is intuitive, however, when the feed streams are blended together prior to treatment. In this case, the pure particles would be recovered first, then the 75% pure particles, then the 50% pure particles, and so on until no additional particles could be taken without exceeding the target grade. The incremental grade concept was originally developed for separations involving mineral systems that included true middlings particles (such as those shown in Figure 2). However, the Feed #1

j

fi

0 Separator

I

H

Y = 9/12 = 75.0% G=6.75/9

Separator

Y = 7/13 =53.8% G=5.25/7 = 75.0%

BEE (a)

Feed #1

Feed#2

Combined Yield = 16/25 = 64.0% Combined Grade = 12/16= 75.0%

Separator

I 1 1

I

Y = 7/12 =58.3% G = 6.2517 = 89.3%

Imm

(b)

Feed #2

0 Separator

Y = 11/13

!= 84.6%

G = 7.25/1 I

=65.9%

mu

HE!

Combined Yield = 18/25 = 72.0% Combined Grade = 13.5/18= 75.0%

F Ire 2. Comparison of concentrate yiel and grades obtained by operating at (a) constant cumulative grade and (b) constant incremental grade.

2155

basic concept can be extended to include even those particle systems that are completely liberated. A separation that involves only two pure minerals would still generate products with different “effective” incremental grades because of inefficiencies in the separation process. In cases such as this, the separation is optimized when the effective incremental grade is maintained at the same value in all producing circuits. As such, the incremental quality concept can be universally applied to nearly all types of solid-solid separations. The question that arises at this point is how can incremental grade be monitored? There is currently no technique for the online measurement of incremental grade. Fortunately, this important parameter can be related to a measurable parameter for many mineral systems. For example, consider the separation of a high-density valuable mineral from a low-density host gangue. For this two-component system, the grade of valuable mineral in an individual partible is directly proportional to the reciprocal of the particle density (p) according to the expression: Incremental Grade (%) = where pI and p2 are the densities of the light and dense components, respectively (Anon., 1966). The incremental grade concept states that maximum plant yield is obtained by operating all circuits at the same incremental grade. Since Equation 11 implies that incremental grade is linearly related to the inverse of specific gravity, this concept can now be modified to show that maximum yield is obtained by operating all circuits at the same specific gravity setpoint. This statement is true regardless of the size distribution or liberation characteristics of the feed, provided that ideal separations are maintained in each circuit. If the separation is less than ideal, minor corrections are necessary to determine the actual specific gravity setpoints required to provide a given incremental grade (Armstrong and Whitmore, 1982; Rong and Lyman, 1985; Clarkson, 1992; Lyman, 1993; King, 1999). Similar types of expressions can be derived for other mineral systems that relate incremental grade to other physical parameters. The importance of this type of relationship is that incremental grade can generally be maintained indirectly in industrial operations by holding the separation parameter of interest at a constant value. Common examples include maintaining specific gravity setpoints for density separations and maintaining field strength and splitter positions for magnetic separations. PLANT CONTROL STRATEGIES The incremental grade concept is an important consideration in the control of solid-solid separation processes. As stated earlier, this optimization principle requires that all parallel circuits that contribute to a plant concentrate must be operated at the same incremental grade to achieve maximum yield. What is less obvious is that this optimization concept also requires that the same incremental grade be maintained at all points in time throughout the entire duration of a production cycle. For example, a poorly designed control system may recover individual particles containing a low amount of valuable mineral at one point in time (when the feed contains an abundance of high-grade particles), and then later discard individual particles containing a larger amount of valuable mineral (when the feed contains few high-grade particles). The lower yield from a low-grade production period will never be compensated by the increase in yield realized during a period of high-grade production. As a result, a plant that continuously raises and lowers incremental grade for grade control purposes will always produce less concentrate than a plant that maintains the same incremental grade. Obviously, this realization has tremendous implications in the design of a plant control strategy. The incremental grade concept simply does not support the traditional control strategy that utilizes feedback from online analyzers to make real-time adjustments to circuit setpoints. This strategy can improve the consistency of the concentrate grade, but cannot optimize plant yield. On the other hand, circuits operated under constant incremental quality optimize total plant yield. The only downside is that the concentrate grade may vary considerably throughout the production cycle in response to fluctuations in the feed

2156

quality. For some period of time, these natural variations may cause the concentrate to exceed the target grade and to be unacceptable to downstream customers. An attractive solution to the variability problem associated with incremental grade control is blending. Blending may be conducted before or after the separation. There are two alternatives for the former scenario. The first involves complete homogenization of the feed ore over extended periods of time (e.g., several days) using stacking and reclaim systems. This type of blending system eliminates short-term variations in feed quality and makes it possible to maintain a stable and constant setpoint for the plant separators. In fact, a control system with a rapid response time would not be needed in this case since any change in the physical properties of the homogenized feed ore would occur very slowly. In many cases, adjustments to setpoints can be made manually in response to data received from standard sampling programs. The turnaround time for the analytical data would simply need to be shorter than any gradual change in overall stockpile quality. This approach would also provide the plant with a relatively constant feed in terms of size and grade. The stable feed would improve plant performance by eliminating overloaded operations and would permit the throughput capacity of the plant to be maximized. Unfortunately, blending systems capable of handling the full feed tonnage treated by most modem processing facilities would require very large stockpile areas and would be expensive to install, operate and maintain.

A second alternative for feed blending is to mix different quality feed streams in appropriate proportions just prior to separation to provide a consistent concentrate grade. Figure 3 provides a flowchart that illustrates the logic behind this type of control strategy. In this particular example, the feed stream to the processing plant is segregated into two stockpiles according to how each responds to upgrading (e.g., difficult and easy). In practice, these two feeds could be segregated based on the fact that they were extracted from different high grade or low grade areas of the same mine or occurred as different geologic splits in the same mined area. In any case, the ore from both stockpiles is fed to the plant while an online analyzer is used to monitor the overall grade of the final concentrate. Based on feedback from the analyzer, the ratio of feed material from the piles is adjusted as required to maintain a constant concentrate grade. This strategy allows the quality of the concentrate to be adjusted online without changing the predetermined setpoints that optimize plant performance. Furthermore, precise determination of the cleaning potential of the different feeds is not required since the control scheme determines the required mix ratios in real time. The feed streams simply need to be sorted into piles with “better” and “worse” qualities. Since complete homogenization of the feed ore is not required in advance, these stockpiles can have relatively small volumes that are easy to manage and more cost effective. Unfortunately, the

I

Figure 3. Feedback control strategy designed to manipulate the ratio of two feed streams (A and B) to maintain a constant concentrate grade.

2157

potential exists for one of the stockpiles to be depleted if long-term changes in the feed quality occur. In this case, a supervisory control scheme must be used to adjust the plant setpoint. For example, if the easily treated ore is in danger of being completely consumed, the setpoint may need to be lowered and yield sacrificed to maintain the desired concentrate grade. Alternatively, the system could be designed to issue an alarm to an operator who could choose to maintain the same setpoint and ship lower quality ore to a different market. In any case, the goal of this particular control scheme is to absorb short-term changes in ore quality through variations in stockpile levels without requiring a change in circuit setpoints. If feed blending is not practical, another scenario for plant yield optimization is to blend products on the concentrate side of the plant. As with feed blending, two alternatives are possible. One option would be to homogenize the entire concentrate from the separation processes over a sufficiently long period (e.g., several days) to ensure that a consistent product is generated. This system has the advantage that the tonnage of concentrate to be blended would be substantially smaller than the tonnage of feed that would need to be homogenized using a feed blending system. Second, the blend system could be configured to combine different proportions of material from different concentrate stockpiles that have been sorted according to grade. This could be conducted automatically using an online analyzer or manually if the turnaround time for sampling and analysis it not too great. The variations in product quality are handled by shifts in the quantities of material stored in the stockpiles. As with the feed based system, a supervisory control loop would be used to monitor the stockpile volumes so that adjustments to setpoints could be made if one or more of the stockpiles starts to be depleted. The major downside to a control system based on concentrate blending is that it does not provide the plant with a consistent feed. Variations in the feed quality can make it very difficult to maintain stable setpoints for many solid-solid separation processes. For example, changes in feed quality that alter the feed tonnage to a spiral circuit can have a dramatic impact on the specific gravity setpoint (Mikhail et al., 1988). CASE STUDIES Control of a Coal Preparation Plant Modem coal preparation plants commonly use density-based separators to remove highdensity inorganic matter (rock) from low-density carbonaceous matter (coal). The primary market specification for the concentrate is usually ash content. Unfortunately, the feed coals to preparation plants are typically subject to significant variations in terms of particle size and quality. Factors responsible for these variations include natural fluctuations in the physical properties of the coals and routine changes in production rates from multiple sections or mines. These disturbances make it difficult for plant operations to maintain a consistent coal quality and to maximize clean coal production. Therefore, control systems are needed to help alleviate these difficulties.

Consider the simple case of a 500 tph preparation facility that operates only with heavy media circuits. The plant currently receives run-of-mine feed from two different coal seams. The primary seam, which is mined during three 8 hr shifts, is capable of providing a high yield at the target grade of 7.5% ash. In contrast, the second seam is very difficult to upgrade and, as such, is mined during only one 8 hr production shift. In order to select an appropriate control system for this plant, two sets of partition simulations were conducted over a 24 hr production period. The first set of computations were conducted to simulate the performance of a traditional control strategy involving the realtime adjustment of specific gravity setpoints based on feedback from an online ash analyzer (Figure 4a). As discussed previously, this type of control system does not optimize yield. For comparison, another set of simulations were conducted in which the specific gravity setpoint was held constant and the feed blends were adjusted in response to feedback from the online analyzer to ensure that an acceptable product grade was maintained (Figure 4b). Previous studies have demonstrated that the incremental ash of coal particles can be directly related to the specific gravity of any particular density class (Abbot and Miles, 1990; Luttrell et al., 2000). This relationship was presented earlier in Equation 11, Therefore, constant incremental ash (and

2158

.- - -

&)

I

SGSepoint

Figure 4. Online control systems for a coal preparation plant based on (a) manipulation of specific gravity setpoints and (b) manipulation of feed rate ratios.

maximum yield) can be maintained by holding specific gravity setpoints in the heavy media circuits at a constant value. Figure 5 compares the operating data from the two sets of control simulations. When the circuit is configured to automatically adjust the specific gravity setpoints, the ash content produced at any point in time remained relatively constant as a result of the online feedback from the analyzer. However, this requirement forced the specific gravity setpoint to vary greatly over the production period to compensate for (i) natural variations in the quality of the two feed coals and

1.9 1

1.8

0

I

40

- 30

. I

+ E

c9 0

5 0 m

1.7 -

-

s E 2 W

.c,

1.6 -

-

20

1.5 -

1.4

1.3

- 10

I 0

s

5

I

I

,

10

15

20

lo 25

Time (Hours) Figure 5. Simulation data comparing two online control strategies (open circles media setpoint control, filled circles - feed blend ratio control).

2159

- heavy

80 70

-

2 b

z

P)

60 50

$

40

-

30

c,

20

10

-€+

SG Control

-El-

Blend Control

0 0

5

15

10

20

25

Time (Hours) Figure 6. Comparison of plant yields obtained via manipulation of specific gravity setpoints and manipulation of feed blend ratios. (ii) the large disturbance created by the addition of the poor quality coal during the middle third of the 24 hr production period. In contrast, the control system based on feed blending provided a clean coal of consistent quality while keeping the same specific gravity setpoint. In this case, the large surge of poor quality feed coal that would normally have to be dealt with during one shift is now spread out over the entire duration of the production period by the control system. The fluctuations in quality are handled by increases and decreases in the stockpile volumes of the two coals. More importantly, the blend control is based on the incremental grade concept that maximizes total plant yield. As a result, the blend based control system produced a total of 8133 tons of clean coal over the 24 hr production period, compared to just 7878 tons for the traditional control approach. This improvement provides 255 tons of additional saleable coal from the same tonnage of feed. As shown in Figure 6, the lower production associated with the traditional control strategy is due to sharp drops in yield during brief production periods that can never be made up. In today’s market, the production improvement represents an increase in revenue of more than $1.6 million annually for this particular example (i.e., 255 tons/day x $25/ton x 250 days/yr = $1,625,625). Thus, the financial gains afforded by a properly designed control system can be very substantial. Control of a Magnetic Separation Circuit for Mineral Sands The incremental grade concept can also be applied to control circuits that involve the blending of concentrates. Consider the simplified magnetic and electrostatic separation circuit shown in Figure 7. This multistage circuit is designed to separate heavy mineral sands, primarily rutile and zircon. Both minerals are nonmagnetic and are separated from other magnetic sands in the first stage of the circuit by drum-type magnetic separators. In the electrostatic circuit, rutile is a conductor (C) and is thrown from the drum of the electrostatic separator, while zircon is a nonconductor (N) and is pinned to the separator drum. Multiple stages of concentrate cleaning are required to generate products of an acceptable grade. As a result, material that reports as middlings (nonconductors from the rutile cleaners or conductors from the zircon cleaners) represents an off-spec product that cannot be sold without hrther processing. As a standard operating practice, the off-spec material is normally recycled back through the circuit for recleaning through the electrostatic section of the separation circuit

2160

Magnetic Product C I I

I

C

X

- A

Actuafor

Analyzer

N

C

-1

Off-Spec Concentrate (a) Figure 7. Online control systems for a mineral sands concentration circuit based on (a) adjustment of splitter setpoints and (b) adjustment of middlings back blending.

For control purposes, an online analyzer was installed to monitor the TiOz content of the rutile concentrate. As shown in Figure 7a, one possible control strategy is to use the analyzer to provide a feedback signal to pneumatic actuators that in turn vary the position of the product splitters on the electrostatic separators. Several series of mathematical simulations were conducted to evaluate the performance of this control strategy. These simulations were carried out using dynamic partition models based on empirical characterization data. The response of the control system to variations in feed stream quality is summarized in Figure 8. The simulation data show that this

-

- Concentrate TiO, >n

80 - C 0 2 2 3 u u u C !Z Z 2 0 2 Z 0 2 E

2161

ZC, 3 0 3 -

65

n

system provides a stable grade by varying the positions of the splitters. However, this configuration does not necessarily optimize circuit performance since the large changes in the splitter setpoints throughout the production period are likely to create large variations in the incremental grade. As discussed earlier, this will not permit the yield of rutile concentrate to be maximized. A better approach would be to design a control system that makes use of the incremental grade concept to optimize circuit performance. Unlike the previous case study involving coal, a control system based on feed stream blending is difficult in this particular situation due to limitations associated with upstream processes. Fortunately, a control system based on material blending is possible via the recycling of off-spec concentrate. As shown in Figure 7b, a temporary stockpile of off-spec material can be created for blend control purposes. The signal from the analyzer can be used to control the rate at which off-spec concentrate is recycled back through the circuit. Mathematical simulations conducted using this control strategy show that this approach allows a relatively constant concentrate grade to be produced without varying the setpoints of the splitters (Figure 8). Normal variations in the quality of the feed to the circuit are absorbed by changes in the active volume of the off-spec stockpile. The simulation data summarized in Figure 9 shows that the control system based on concentrate blending provided a higher overall yield than the system based on splitter setpoint control. In this case, the average yield was increased from 45.0% to 45.8% with no reduction in concentrate grade. This improvement, which represents a net increase in concentrate tonnage of about 1.76%, would provide over 3,000 tons per year of additional concentrate. Thus, the implementation of this type of control system would be expected to provide a significant financial return. It is important to note that this particular control strategy does not necessarily ensure that the incremental grade is maintained. This shortcoming is due to the fact that the effective incremental grade does not depend entirely on splitter position for this particular type of separation. For example, it is widely recognized that the effectiveness of electrostatic separators can be influenced by environmental factors such as temperature and humidity. Researchers at the Julius Kruttschnitt Mineral Research Centre in Australia have recently developed a variety of online sensors that may make it possible to address some of these problems. Nevertheless, it is certainly more likely that the effective incremental grade is more consistent when blend control is used than when the splitter positions are varied over large ranges.

.I

h 30

-

.c,

. I

3

g

5

20 -8- Splitter Control +Blend Control

10

0 1

0

I

I

I

I

5

10

15

20

25

Time (Hours) Figure 9. Comparison of plant yields obtained via manipulation of splitter setpoints and manipulation of middlings back blend ratios.

2162

SUMMARY A theoretical principle known as the incremental grade concept provides useful insight concerning the optimization and control of solid-solid separation circuits. According to this concept, a plant limited by a constraint on concentrate grade will produce maximum total yield when all circuits are operated at the same incremental grade. This concept applies not only for a fixed point in time, but also for the entire duration of a given production cycle. Therefore, the incremental grade concept supports (i) operation at fixed process setpoints and blending of feed ore before separation or concentrates after separation to maintain product consistency and (ii) the online measurement of concentrate grade to adjust feed or concentrate blends. This approach dictates that plant engineers think outside the box (figuratively and literally) since blend-based control systems may influence plant operations as well as mining extraction, materials handling, and saledmarketing activities. In most cases, the incremental grade concept does not support the real-time adjustment of separator setpoints based on online measurements of overall concentrate quality. Data obtained from case studies for two different solid-solid separation systems indicate that large financial returns are possible through the implementation of supervisory control systems that utilize the incremental grade concept to optimize the performance of solid-solid separation processes.

REFERENCES Abbott, J., 1982. The Optimisation of Process Parameters to Maximise the Profitability from a Three-Component Blend, Is‘ Australian Coal Preparation Conf., April 6- 10, Newcastle, Australia, 87-105. Abbott, J. and Miles, N.J., 1990. Smoothing and Interpolation of Float-Sink Data for Coals, Inter. Symp. on Gravity Separation, Sept. 12-14, Cornwall, England. Anonymous, 1966. Plotting Instantaneous Ash Versus Density, Coal Preparation, Jan.-Feb., Vol. 2, No. 1, p. 35. Armstrong, M. and Whitmore, R.L, 1982. The Mathematical Modeling of Coal Washability, lst Australia Coal Preparation Conf., April 6-10, Newcastle, Australia, 220-239. Clarkson, C.J., 1992. Optimisation of Coal Production from Mine Face to Customer, 3rd Large Open Pit Mining Conference, Aug. 30 - Sept. 3, Makcay, Australia, 433-440. Dell, C.C., 1956. The Mayer Curve, Colliery Guardian, Vol. 33, pp. 412-414. King, R.P., 1999. Practical Optimization Strategies for Coal-Washing Plants, Coal Preparation, Vol. 20, pp. 13-34. Luttrell, G.H., Catarious, D.M., Miller, J.D. and Stanley, F.L., 2000. “An Evaluation of Plantwide Control Strategies for Coal Preparation Plants,” Control 2000, Mineral and Metallurgical Processing, (J.A. Herbst, Ed.), Society for Mining, Metallurgy, and Exploration, Inc., (SME), Littleton, Colorado, pp. 175-184. Lyman, G.J., 1993. Computational Procedures in Optimization of Beneficiation Circuits Based on Incremental Grade or Ash Content, Trans. Inst. Mining and Metallurgv, Section C, 102: C159-C 162. Mayer, F.W., 1950. A New Washing Curve. Gluckaufl Vol. 86, pp. 498-509. Mikhail, M.W., Salama, A.I.A., Parsons, I.S., and Humeniuk, O.E., 1988. “Evaluation and Application of Spirals and Water-Only Cyclones in Cleaning Fine Coal,” Coal Preparation, Vol. 6, pp. 53-78. Rong, R.X. and Lyman, G.J., 1985. Computational Techniques for Coal Washery Optimization Parallel Gravity and Flotation Separation, Coal Preparation, 2: 5 1-67.

2163

Strategies for Instrumentation and Control of Thickeners and Other Solid-Liquid Separation Circuits Fred Schoenbrunn, Lynn Hales, and Dan Bedell

ABSTRACT

Some of the process variables that are commonly monitored on a thlckener are torque, rake height, bed level, bed pressure, feed rate and density, underflow rate and density, settling rate, and overflow turbidity. Many of these are easily measured, while some can be difficult. Combining these signals into a coherent control strategy requires forethought and an understanding of the fundamentals of thickener operation. A wide variety of control strategies have been implemented on thickeners, using various combinations of sensors. In recent years improved flocculants, higher throughput rates per unit area, and desired higher density underflow concentrations have required the development of better control strategies to successfully operate sedimentation equipment. This has been complicated by plant expansions that have placed increased loads on existing sedimentation equipment. Successful control strategies consider the process goals, plant fluctuations, sensor reliability, and system response times.

A historical review will be discussed followed by discussion of the latest developments in sensors, control equipment, and control strategies. INTRODUCTION Thickeners are one of the workhorses for industry and are involved in numerous steps in flowsheet design. They have been instrumental to the mining industry growth and in the processing of minerals. The thickener over the years has received little attention and has generally been treated as a wide spot in the process line with little or no instrumentation. Concentrate thickeners and some clarifiers were the first units to have instruments applied to them due to their impact on production and profitability of a plant. Before the use of polymers thickeners were normally sized with excess capacity so that they could “respond or handle flow and feed variations” and not adversely impact production. The advent of natural flocculants followed in recent years by synthetic flocculants has increased the efficiency of sedimentation rates allowing increased production through a given size of thickener. This has resulted in the time variable being dramatically shortened. Thickeners may use a “single” flocculant system or a “dual” flocculant system. The single component system may use a cationic, anionic, or non-ionic flocculant to achieve underflow density. The dual flocculant system will use a cationic or coagulant (to obtain acceptable clarity) followed by the addition of an anionic or an-ionic flocculant to achieve good underflow density. The use of flocculants had led to the discovery of a relationshp of solids concentration to flocculant addition concentration that gives what is called “Settling Flux” (See Figure 1). It now understood that there is an optimal feed concentration at which the solids will settle at a given flocculant dosage. T h s discovery has helped advance sedimentation technology but introduced yet another variable into the equation.

2164

Optimum

Feed Solids Concentration Figure 1 Settling flux Thickener controls have been generally based upon the actual underflow or overflow solids with some controls based upon polymer addition rates. The biggest challenge for conventional thickeners has been the long lag time between variable changes and results. High rate thickeners have reduced the residence time but are still relatively slow in cause and effect relationships. The emergence of ultra high rate and ultra high density thickeners has further pushed the sedimentation envelope. With each of these developments and advances in sedimentation technology there has been the need for more and better instrumentation and controls. Modem thickeners are becoming better instrumented and industry is discovering that unit operations are more predictable and profitable as a result. The modem generation of thickeners will not successfully function without the aid of quality instrumentation. Because of instrumentation the mining industry is now beginning to control the thickener, inventories, and manage the process flows and improve the overall efficiency of the operations. The following sections will take some of these variables, types of instruments to measure them, and discuss the types of controls used to manage them. The general objectives for thickeners and clarifiers are to produce clean overflow and maximum solids concentration in the underflow. Flocculants are typically used to agglomerate the solids to increase the settling rate and improve the overflow clarity. Tluckeners generally operate continuously and are used in a wide variety of industries, and in numerous applications. The term “thickener” will apply to both thickener and clarifier unless noted otherwise throughout this section. Thickener control involves a number of complexities and variables such as varying feed characteristics, changes in feed concentration, solids specific gravity, particle size distribution, pH, temperature, and reaction to flocculant which can all contribute to variations in performance. Accurate information about what’s happening inside the thickener is often difficult to obtain. In addition to these variables, various phenomena such as “sanding” and “island formation” may occur whch can be difficult to predict and interpret from the data. The two independent variables that are typically used for control of thlckeners are underflow rate and flocculant addition rate. A t h d variable, the feed rate, is generally used only in an emergency to avoid impacting plant production. The dependent variables include underflow density, overflow turbidity, rake torque, solids interface level (bed depth), solids inventory (bed mass), solids settling rate and underflow viscosity.

2165

Historically, most control schemes have used one or two of the dependent variables to control the independent variables. For example, using the underflow density to control the underflow rate, which can be by variable speed pump or if the underflow is by gravity, using a controi valve or orifice. Another possible scheme is to use the bed pressure to control the underflow rate and bed level to control the flocculant rate. The range of conditions that it can recognize and to which it can respond will limit any control scheme. None of them so far have been able to resolve all of the possible inputs for specific conditions and react to them. For example, the two control schemes described above do not consider the rake torque and both can have problems from high torque if the feed particle sue distribution suddenly becomes coarser. Various algorithms have been used to control thickeners with varying degrees of success. Rule based expert systems have been developed for use on thickeners since the early days of computerized control systems, but have been cumbersome for implementation, troubleshooting, modification, and tuning. With the recent developments in expert control software, these issues have been greatly simplified. FLOCCULANT

RECYCLE

RB1

CLEAR Off

CLEAR Olf

Figure 2 Thickener with minimal instrumentation

2166

Figure 3 Thickener with additional instrumentation

INSTRUMENTATION Torque Rake torque is an indication of the force necessary to rotate the rakes. Higher rake torque is an indication of higher underflow density or viscosity, deeper mud bed, hgher fiaction of coarse material, island formation, or heavy scale build up on the rake arms. Rake torque measurement is usually provided by the thickener manufacturer as part of the rake drive mechanism. Typical methods involve load cells, motor power measurement, hydraulic pressure, or mechanical displacement against a spring. They are all generally reliable and reasonably accurate if set up correctly. The’type supplied depends on the manufacturer and the type of drive supplied. For example, if a hydraulic drive is used, then hydraulic pressure is the best method to use for torque measurement. Torque measuring devices are designed to produce a signal that may be utilized for alarming or control.

Rake height Rake lifting devices are frequently used to minimize the torque on the rake arms by lifting them out of the heavy solids and enable the rake to continue running during upset conditions. It is desirable to prevent the rake drives from running extended periods at torques above 50-60%, to prevent accelerated wear on the drive. Lifting the rakes a small distance is usually effective at relieving the pressure on the rakes and thus reducing the torque. Because of this, using the torque indication in a control strategy must also consider the rake height in order to effectively control the ~ckener. Rake height indicators are typically supplied by the thickener manufacturer. The two most common methods are ultrasonic or a potentiometer type with a reeling cable. Both are reliable and accurate. In many plants a bar or rod on the end of the rakes serves as a “visual” indicator of height and arm position in the tank. Lifting of the rakes allows a short period of time to make corrections before being forced to shut down the thickener.

21 67

Bed level There are several general types of bed level detection instruments; ultrasonic, nuclear, float and rod, and reeling (with various sensors). Each has advantages and disadvantages, which are discussed below. There is not a standard bed level sensor that is recommended for all applications. Ultrasonic bed level sensors work by sending a pulse down from just under the surface, which in theory bounces off the bed surface back to the receiver. Elapsed time is used to calculate the distance. Advantages are non-interfering location, measures over a large span, and relatively inexpensive. The downside is that they do not work on all applications. If the overflow is cloudy, it can interfere with the transmission or causes too much reflection to give a reliable signal. Scaling affects accuracy and can cause drifbng or loss of signal. Using them on concentrate thickeners has proved to be particularly troublesome. Nuclear bed level sensors work by either sensing background radiation level or attenuation between a source and detector, depending on whether the solids have a natural background radiation level. The sensor is comprised of a long rod that extends down into the bed with radiation detectors spaced along the length. If the ore changes from not having radiation to having it, there will be problems. The advantages are that it is relatively reliable when properly applied. The downside is that it measures over a limited range, may interfere with the rakes (a hinged version that will swing out of the way when the rakes pass by is available), and is relatively expensive. Float and rod types work with a ball with a hollow sleeve that slides up and down on a rod that extends down into the bed. The ball weight can be adjusted to float on top of the bed of solids. These are subject to fouling and sticking, and can be installed and measure only in the area above the rakes, however, they are relatively inexpensive. Reeling devices work by dropping a sensor down on a cable, and sensing the bed level by optical or conductivity sensors. In theory they are non-fouling and get out of the way of the rakes, but in practice, stories abound of sensors wrapped in the rakes. A number of plants have found them reliable and they can cover a large range. The price is midrange. Freezing wind and cold temperatures can cause icing problems. Vibrating or Tuning fork sensor. These are designed to sense a difference in the vibrating frequency in different masses of solids. These are used in Europe and Africa with some success. Bubble tube or differential pressure. This is an old but tried and true method of bed level detection. There may be some plugging or fouling of the tube over time. External density through sample ports. Slurry samples are taken from nozzles on the side of the tank and pass through a density meter to determine the presence of solids. T h s system can be set up with automated valves to measure several different sample points, typically sampling once every five or ten minutes. Requires external piping and disposal of the sample stream. Access can be problematic on occasion.

Bed pressure Because thickeners maintain a constant liquid level, the pressure at the bottom of the thickener is an indication of the overall specific gravity in the tank. If the liquor specific gravity is constant, the overall specific gravity is an indication of the amount of solids in the tank and can be converted into a rough solids inventory. This can be a very effective tool for thickener control. Because of relative height to diameter ratios, it is considered somewhat less useful for very large diameter thickeners. Differential pressure sensors are used to measure the bed pressure, leaving one leg open to the atmosphere to compensate for barometric pressure variations. Care must be taken in the

2168

installation to minimize plugging with solids. This is frequently done by tilting the tank nozzle on which the DP cell is mounted downwards from the sensor so that solids tend to settle away fIom the sensor surface. A shutoff valve and a water flush tap are recommended to allow easy maintenance. Flow rate Flow rates for feed and underflow lines are useful, particularly when combined with density measurements in order to generate solids mass flow rates. Since flocculant is usually dosed on a solids mass basis, knowing the mass flow rate is very useful for flocculant control, providing a fast response system. Flow rate measurement is an absolute necessity for the newer generation of the ultra high rate and ultra high density thickeners. Since the streams being measured are usually slurries, the flow rate is usually measured by either magnetic flow meters or Doppler type flow meters. As long as these instnunents are properly installed in suitable full straight pipe sections, avoiding air if possible, they are accurate and reliable. If the feed stream is in an open launder, flow measurement is more difficult but can be accomplished using ultrasonic devices. Density Nuclear gauges are the norm for density measurement. Nuclear density instruments require nuclear handling permits in most countries. It should be noted that there are now some types that use very low level sources that do not need nuclear licensing, reducing the hassle of using these. Density gauges should be recalibrated regularly, roughly every 6 months, as they are subject to drift. As with flow meters, proper installation is important for reliable operation. Small flow applications may be able to use a coroilis meter to measure both mass flow and percent solids with one instrument. Settling rate The settling rate in the feedwell is a good indication of the degree of flocculation, and can be used to maintain consistent flocculation over widely varying feed conditions. A settleometer is a device that automatically pulls a sample from the feedwell and measures the settling rate. The flocculant can then be adjusted to maintain a consistent settling rate. In most applications they require regular maintenance to maintain consistent operation. Overflow turbidity Overflow turbidity can be used to control flocculant or coagulant. There may be some significant lag time between the actual flocculation process and when the clarified liquor reaches the overflow discharge point where the sensor is typically positioned. These sensors and meters are generally used as alarms or for trim only. In most applications they require regular maintenance to maintain consistent operation.

CONTROL ARCHITECTURE AND EQUIPMENT Normally thickeners are part of an integrated control system where the objective is to remotely start and stop the equipment, monitor operating conditions and performance, stabilize operations based on operating and feed conditions and lastly, optimize performance based on economics andor operational goals. Tools to accomplish these goals include programmable logic controllers (PLC’s), distributed control systems (DCS’s) and expert control systems.

21 69

Programmable Logic Controllers are normally used to perform starting and stopping functions in processing plants. They do have the ability to also implement continuous control loops but this use is only used minimally. Distributed Control Systems are the workhorse of plant control systems and normally are used to coordinate the monitoring of process data and the subsequent stabilizing control of important process parameters. Specifically, for thickeners, monitored process parameters might include feed flow rate, overflow flow rate, overflow turbidity, underflow density, underflow flow rate, rake position, rake power and torque, flocculent dosage rate, bed level, and bed mass. Stabilizing control loops might include flocculent dosage rate, underflow density, or underflow flow rate. Bed pressure is sometimes used as an indication of solids inventory. This can be used to help the system to determine whether a h g h bed level is the result of decreased settling rate or increased solids inventory. In some cases, a drive torque target is used as an indication of acceptable underflow rheology. A typical P&ID incorporating the instrumentation required by stabilizing control loops previously discussed is shown in Figure 4.

CONTROL STRATEGIES Process goals vary widely throughout the industry and are necessarily aligned to some degree with the basic plant control hardware. For plants with distributed control systems it is common to have a number of stabilizing control loops such as, flocculant dosage rate that is ratioed to the solids mass flow rate of the feed and an underflow density control loop that is maintained by varying the discharge pumping rate.

2170

Where expert control systems exist in grinding and flotation it is natural to extend them to include thickening. The basic concept behind expert control is the continuous monitoring of process conditions along with process and economic objectives and then using advanced expert logic and or on-line models to determine what the best process set points should be to optimize the performance of the target unit operation or plant area. A generic thickener expert control strategy that monitors the performance of the thickener in a tailings dewatering circuit and specifically controls polymer dosage and discharge rates is now described. Table 1 shows a list of process measurements and control actuators employed by the generic thickener expert strategy.

It is common for expert strategies to run the monitoring and control rules at a fixed time interval. Often times this frequency is a function of the retention time of the unit operation being controlled. For example every five minutes the generic thickener expert strategy determines the appropriate control actions through the following decision tree: Evaluate for emergency response situations: These are situations that require immediate attention. It is assumed that these conditions, if not addressed, would lead to a process emergency. 1. Check the current drawn by the underflow pump. If the current is too high, decrease

the load on the pump by reducing flow through the thickener; or reducing flocculant rate. 2. Check the underflow density. If the density is too high, decrease the polymer dosage. 3. Evaluate the torque on the rake. If the torque is high, increase the pump speed. 4. Check the sludge level in the thickener. If the level is dropping rapidly, decrease the pumping speed and decrease polymer dosage. Optimize thickener performance: If no emergency conditions exist then the generic thickener expert strategy seeks to optimize performance. To obtain optimum performance, two targets are used. 1. An underflow density target is used to ensure optimum solids content in the tailings impoundment, and optimum water reclaim for the mill. 2. A bed level target is used to obtain optimum loading in the thickener without overloading the drive mechanism. An important detail that the expert system designers have to deal with is how simultaneous emergencies are dealt with. If it is decided that every five minutes only the first emergency determined would be dealt with then a condition where the rake torque is high and the sludge level

2171

is dropping quickly, the generic thckener expert strategy will only respond to check the rapidly rising torque by increasing the discharge rate. While the increased pumping rate will also affect the bed level situation, the generic thickener expert strategy described here only responds to the highest priority situation. It is easy to see however that there may be a more expert way in taking into account all process conditions simultaneously so that it is possible to deal with multiple emergency conditions while taking into account optimization goals. This additional level of control logic prevents multiplying (or masking) the effects of control actions by responding to multiple emergency conditions with similar or conflicting remedies. Table 2 shows possible combinations of the underflow density, bed pressure, and bed level process variables and what an expert system might do to optimize the thckener as conditions vary.

The amount of the set point changes that are made for the optimization strategy depends upon a number of factors which is beyond the scope of this discussion, however, the very idea of the complications of this issue introduces the concept of the art of control.

CONCLUSIONS Reliable instruments are a prerequisite for a functional control system. The control system must be designed based on the variables that can be consistently measured. Stabilizing control systems can be designed to use these signals to control the operation of the equipment. Expert control

2172

systems can be implemented on top of functional stabilizing systems to optimize the operation of the solidliquid separation equipment. Successful control strategies consider the process goals, plant fluctuations, sensor reliability, and system response times.

REFERENCES V. Dooley, “Mud Level Gauges - A Comparison of Techniques”, Fourth International Alumina Quality Workshop, Darwin, Australia, June 1996. Nelson M. G., R. P. Klepper, and K. S. Gritton, “Design of an Expert Control System for Thickeners”, Presented at SME ’97, Denver, Colorado, February 1997. Allen J. P., M. G. Nelson, and K. S. Gritton, “Automated Thickener Control; Instrumentation and Strategy”, Presented at SME ’97, Denver, Colorado, February 1997. Johnson G., S. Jackson, I. Arbuthnot, “Control of a High Rate Thickener on Gold Plant Tailings”, The AusIMM Annual Conference, Rotorua, New Zealand, 1990.

2173

Strategies for Instrumentation and Control of Flotation Circuits Heikki Laurila', Jarkko Karesvuori' and Otso Tiili'

ABSTRACT Flotation cells have increased in size dramatically over the past ten years, making it more viable to integrate greater levels of instrumentation on each cell, given the value of the material. Advances in instrumentation also have allowed better measurement of specific flotation parameters. Combined, these factors have created an opportunity for the development of more effective strategies for flotation control. Over the past decade, strategies for flotation control have been delivered using advanced control techniques such as, model based controls, expert systems, and neural networks, all of which have been employed with varying degrees of success. The most recent applications of expert and neural controls look promising and may also lead to more robust model based controls in the future. The ideas, strategies and instrumentation behind these concepts are discussed in this paper.

INTRODUCTION Flotation has a long history and nowadays it is one of the most broadly used processes in the mineral separation industry. However, it remains quite an inefficient process. Although a great deal of research and development has been conducted in the field of flotation process, over many years, there are still economic benefits to be achieved through improved operations. Flotation can be used in most mineral separation applications, as it is suitable for a large variety of minerals and a wide range of particle sizes and densities. Flotation is performed in cells ranging in size from laboratory scale up to 160 m3 tanks. The newest tanks available on the market have volume of 200 m3. This allows use of flotation in processes of very large as well as very small capacities. Flotation is the most difficult stage of ore benefication and the performance of the flotation circuit has a very significant effect on the total performance of the concentration path. Typically, the recovery rates in flotation are between 85 and 95 %, the difficulties in improving performance derive mainly from the complexity and the nonlinearity of the flotation process itself. Earlier the lack of instrumentation and proper technologies for instruments also complicated the achievement of acceptable efficiency. A large number of different control strategies for flotation have been presented in literature and implemented in flotation plants over the past 30 years and development of new instruments continues. Flotation is facing a new era in terms of automation and process control. There are three main reasons for this. Flotation circuit design is moving away from multiple recycle streams and towards simpler circuits. This removes a degree of safety and stability, as poor operation will send material directly to the tailings rather than recirculating it. The advantage is that the selfcompensating nature of the circuit is reduced and it becomes easier to regulate and optimize the process (Henning, Schubert and Atasoy 1998). The size of the flotation tanks in large-scale flotation plants has been increasing in recent years (Figure 1). While earlier plants had a large number of relatively small tanks in series, future 1

Outokumpu Mintec Oy., Riihitontuntie7 C, P.O. Box 84, FIN-02201 Espoo, Finland

2174

plants will have a smaller number of larger cells. This trend leads to decreasing number of instruments but along with simple circuit design, the demands of reliability and accuracy for instrumentation have increased.

Figure 1 The recent growth of the cell volume Recent developments in instrumentation have provided new devices such as image analysis based devices for froth characteristics measurement and new digital fieldbus technology has enabled the production of smart instruments. These instruments can provide more information than traditional analog instruments in using self-diagnostics to provide information about the quality of measurements and status of the device. FLOTATION PROCESS Separation of valuable mineral from the ore in flotation process is based on the different surface properties of minerals. Some minerals have a tendency to attach stronger to air bubbles than others in water treated with special chemicals. Minerals having this tendency are called hydrophobic, however, most of the minerals are not naturally hydrophobic. Therefore, several chemicals must be introduced in water-ore slurry in order to convert valuable mineral particles into hydrophobic and to make other particles as hydrophilic as possible. Air bubbles formed by airflow to which mineral particles attach, rise to the surface of the slurry, forming the froth, which can be collected. Flotation is very complex process having a large number of affecting variables, nonlinearities, interaction between variables and randomness. Estimates have been presented that there are about 100 affecting variables in the flotation process (Arbiter and Harris 1962). The raw material and grinding process prior to flotation contribute a relatively large part of the variables. From the viewpoint of process control the most important variables are:

.. .. .. .

Slurry properties (density, solids content) and slurry flow rate (retention time) Electrochemical potentials (pH, Eh, conductivity) Chemical reagents and their addition rate (frothers, collectors, depressants, activators) Slurry levels and aeration rates in the cells Froth properties (speed, bubble size distribution, stability) Particle properties (size distribution, shape, degree of mineral liberation) Mineralogical composition of the ore

2175

Mineral concentrations in feed, concentrate and tailings (recovery, grade)

' Froth wash water rate (specially in flotation columns)

In industrial scale, flotation takes place in interconnected cells, which compose the different sections of flotation circuit. Each section has its function in the total flotation process, as high recovery and high grade cannot be achieved in a single stage of the process. Whilst the design of a flotation circuits vary significantly, the basic operation is similar and some basic rules can be stated. Each section is composed of flotation cells. Earlier sections could contain up to dozens of relatively small cells but nowadays the total number of cells has reduced as cell sizes have increased. The tailing of each cell becomes the input of the next cell in the circuit. From the last cell of the section, tailing is led to the first cell of other section, except the cells where the tailing is sent to final tailing. Usually concentrates of the single cells in a section are combined to one concentrate flow, which is then directed to next phase of process. Sometimes regrinding or thickening is situated between the flotation sections. Regrinding may be necessary when floating two minerals with very different optimal particle size distribution or when final concentrate grade cannot be achieved because of gangue contamination in non-liberated particles. Thickening may be needed to increase solids content of slurry. Figure 2 presents the outline of two flotation circuits. The circuits are generalized examples but describe some basic structures and sections of flotation circuits. Both circuits are divided into three sections - rougher, scavenger and cleaner but connections between sections are not similar. a.) Closed flotation circuit Feed

Tai Li ngs

I

I

Cleaner Cleaner Concentrate I

I

I

b ) Open flotation circuit

Tailings

Feed

Figure 2 Flotation circuits (a. Closed, b. Open) Feed coming from the grinding circuit is led to the rougher. Before the actual flotation, slurry may be conditioned in a conditioner, where some of the reagents are added to slurry. Some measurements are often performed in conditioner, pH, for instance. In rougher flotation, most of the fast floating valuable minerals are separated from the slurry directly to the concentrate. In

2176

other words, the recovery is held high at the expense of the grade. This is performed in conditions where fairly thin froth beds and high aeration rates are used in the flotation cells. The tailings of the rougher are introduced to the scavenger, where slowly floating fine and coarse mineral particles are floated in conditions where smaller aeration rate and even thinner froth beds than in rougher section are used. The tailings of the scavenger are sent to the final tailings of the circuit. The concentrates of the rougher and scavenger are refloated in cleaner section in order to increase the grade of the final concentrate. The froth beds in cells of the cleaner flotation are thicker than in other sections, which leads to the decrease of the water recovery and increased rejection of hydrophilic gangue. The connection of the tailings of the cleaner flotation makes the difference between the circuits presented in Figure 2. As in the open circuit, the cleaner tailings form the final tailings of the circuit, with scavenger tailings closed circuit, re-circulates the cleaner tailings back to the rougher. In a closed flotation circuit, the control of circulating loads is essential. Material can accumulate in the circuit as circulating loads increase. This may lead to situations where recovery suddenly drops far below a satisfactory level after being on target and stable for long periods. This occurs when the accumulation has reached certain point where circuit conditions become unstable. The means to correct the situation after it has occurred are minimal. Almost inevitably, the recovery stays low for certain period and may even reduce the average recovery below long-time target values. Figure 3 illustrates the effect of poorly controlled circulating loads on tailings grade.

Figure 3 The effect of poorly controlled circulating loads CONTROL STRATEGIES FOR FLOTATION Flotation process control is a challenging and important task in the ore benefication chain. The efficiency of the flotation process largely controls the economics of the overall mineral processing plant (Hodouin et al. 2000). Flotation plants are difficult to operate. Non-linear dynamics, coupling among control loops, large and variable dead times, strong and continuous unmeasured input disturbances, imperfect knowledge of the phenomenology of flotation, impede process control. Frequent lack of appropriate and precise instrumentation makes supervision and control even more difficult. (Osorio, P6rez-Cornea and Cipriano 1999). A universal way to control a flotation plant cannot be given. Each plant has its special features in terms of cell configuration, instrumentation, ore and chemistry, which have led to a large number of different control strategies and methods used and reported in literature. The only

2177

undoubtedly common feature between flotation plants is to maximize profit. This objective is pursued in many ways. Some typical aspects and methods are presented in this paper. Plant-wide flotation control strategy can be divided in layers, which are presented in Figure 4.

OPTIMIZATION Maximize profit

Process disturbance rejection, Control of grade and recovery

/

BASE LEVEL CONTROLS Control of slurry levels, aeration rates and reagents dosing

\

INSTRUMENTATION Valves, Measurements, XRF-analyses, Image analyses

...

Figure 4 Control system hierarchy of the flotation plant This starts with instrumentation, which is the basis of all control and must function well. The whole plant optimization is strongly dependent on both well performing and reliable instruments and properly tuned single control loops. For each measurement or control task, a large selection of devices is offered in the market. Attention must be paid in choosing the right ones for the process. As in every instrumentation case, intimate knowledge of the process, for it’s special features and demands, is vital. As a process of many affecting variables and high complexity, flotation has a large number of variables to be measured and to be manipulated. This leads to extensive variety of different instruments used in a flotation plant. In spite of the new digital fieldbus technology, the majority of instruments are still traditional devices, using 4-20 mA analog signal technique. Instruments are connected to plant automation system via I/O-units, where analog signals are converted to digital signals. The digital fieldbus technology will replace the analog technique and provide totally digital communication between field instruments and the automation system. Base level control consists of traditional PID (Proportional Integral Derivative) controls of slurry levels and aeration rates and ratio control of reagent additions. The derivative term is usually excluded in tuning. The plant-wide flotation control strategies use slurry levels, aeration rates and reagent additions as control variables. Instrumentation and Base Level Controls Slurry Flow Measurement. Slurry flow measurement is mainly performed by magnetic flow meter. Measured fluid must be at least weakly conductive (more than 5 microsiemens per centimeter). The measurement is based on Faraday’s principle of induction, which states that tension is inducted to conductor when moving in the magnetic field. Magnetic flow meter consists of an electro-magnet wrapped around a length of process pipe, which is lined with an insulating material. Electrodes are installed in the wall of the pipe on opposite sides and these enable an electrical circuit to be formed through the liquid and measuring device. The material selection for the electrode and linings depends on the properties of slurry. Common lining materials are; hard rubber, Teflon, polyurethane and aluminum oxide. Electrode materials are; stainless steel, platinum and tantalum. The AC-magnetization technique used in earlier times has been replaced by DC-magnetization, with high measurement frequency (more

2178

than 30 measurements per minute). Magnetic flow meters do not impose energy loss, as there is no obstruction to flow. In general, slurry flow measurement is problematic, due to solid particles of slurry and suspended air bubbles, particularly the latter, which will decrease the performance of magnetic flow meter. If slurry contains magnetic material (e.g., magnetite) special demagnetization is needed. Concentrate flow measurement in open channel can be done by a dam arrangement, using a V-shaped cutout and ultrasonic level transmitter. Schematic illustration of arrangement is presented in Figure 5. Known geometry of the channel and level measurement can be converted to flow rate. This method does not provide flow rate accurately but gives a rough estimate, which can be used in cases where other flow measurements are not possible,

Figure 5 Arrangement for flow measurement in open channel Slurry flow measurements have an important role in determination of circulating loads. They are also used in circuit mass balance calculations. Although frother addition rate is mostly adjusted according to the tons of ore in feed, some flotation plants have found it reasonable to use feed volume rate instead. Elemental Assaying. On-stream XRF analyzers are very important instruments in flotation, as they provide elemental assays from the process flows. Manual elemental assays in the laboratory are not as useful for real time process control because of the long delays. On-stream XRF analyzers have decreased the delay to reasonable scale. A primary sampler acquires a primary sample flow from the process stream. It is directed to the secondary sampler, which reduces the size of the sample suitable for the X-ray analysis equipment, then returned to the process. Modern XRF analyzers can report the assay of several elements and solids content from the sample. Usually the analyzer system has many sampling points from different stages of the flotation circuit and one XRF-analyzer can accommodate up to 24 sample lines, measured in sequence. The measurement time for one sample is from fifteen seconds to one minute, the cycle time 5-15 minutes, depending on the number of streams to be assayed. An on-stream XRF analyzer schematic is presented in Figure 6.

2179

Figure 6 The flowchart of the on-stream XRF analyzer Elemental assays by on-stream analysis is the only way to obtain on-line information about the performance of flotation in real-time, which enables remedial action for concentrate and tailings grade control as well as recovery control. Based on grades reported by an on-stream analyser, advanced controls adjust the setpoints of base level controllers. Also feedforward action for proper feed type treatment is made possible by real time assays from on-stream analyzers. Density Measurement. Some on-stream XRF analyzers and particle size analyzers provide density measurement but specific density meters are also commonly used. Nuclear density meters are suitable for slurry density measurement, although suspended air bubbles in the slurry often make measuring impossible. Therefore the location of the instrument must be carefully chosen. A suitable place for density measurement is a process pipe, which in normal conditions is always full of slurry. Nuclear density meters are based on the attenuation of the nuclear radiation by the media. Density measurements, along with slurry flow measurements, enable the calculation of total mass flows, which are required in mass balance calculations. Slurry Level Measurement. Accurate level measurement in flotation cell is troublesome, due to usually thick froth bed and variations in slurry density, which complicate methods using direct ultrasonic measurement of level or hydrostatic pressure. Also, the concept of level is not definitive, since the transition from slurry with bubbles to froth with slurry, is not sharp. The most typical instruments used for measuring the slurry level in cells are a float with target plate and ultrasonic level transmitter, a float with angle arms and capacitive angle transmitter and reflex radar. Instruments are presented in Figure 7. Ultrasonic level transmitters emit a series of ultrasonic pulses, which echo from the target plate. The transmitter receives echoed pulses and measures the travel time. In order to separate the real echo from false echoes, acoustic and electronic noise, the signal must be filtered. Similarly, before converting the travel time to distance and output, it must be temperature compensated. Reflex radar, using a plastic or Teflon plated metal rod, can sense various boundary layers and gives the possibility of measuring both slurry level and froth thickness. Problems associated with this technology are due to severe changes in slurry and froth conductivity. Froth bed thickness measurement by microwave radar or certain special ultrasonic transmitters has also been under development, although problems occur because echoes are influenced by froth properties.

2180

Figure 7 Instruments for slurry level measurement Slurry Control Valves Process features such as large flowrate capacity changes and abrasive mineral slurries, limit the number of possible valve solutions for slurry flow control. Control valves used in flotation must be robust and durable, specific selection of outflow valve depending on the structure of the cell. Pinch valves and dart valves are suitable solutions for flotation duty. Traditionally, manually operated overflow weirs have been used on tails streams. In older flotation plants, the mechanical design of the cell bank has forced the automation of overflow weirs instead of the installation of control valves. The pinch valve is simple and economical slurry valve solution. It consists of three main components, valve body, sleeve and actuator. The pinch valve actuator manipulates the cell outflow by flattening the sleeve with jaws. The only component in contact with slurry is the sleeve, which, with the correct material selection for the application, makes the pinch valve relatively easy to maintain. However, sleeve material may lose elasticity over time. This can cause stickiness in the valve operation, so that the sleeve does not follow the jaws correctly. This problem can be eliminated with positive opening tags, which connect the sleeve to the jaws. The construction of the dart valve is well known. The slurry flow through the valve is manipulated by a vertically moving cone, controlled by an actuator, which varies the area of the cell outlet. The dart valve is located in additional box connected to cell. Along with choosing the right valve type, the sizing is also important. The control valve should be sized to operate in normal process conditions between 30 and 60% open, optimal for control performance and allows some disturbances to occur before saturating. Dart valves have wider opening range for reasonable control performance than pinch valves. Pinch valves have highly nonlinear characteristic curves near extreme positions of the opening. Depending on the size of the valve, a response time of 10-20 seconds from fully open to fully close is desirable. Particularly large pinch valve actuators require special attention to ensure sufficient actuator speed and in large cells, the use of two valves in parallel is practical, since the control properties of big valves are not adequate. In a dual arrangement, one valve may be operated manually whilst the other is automatically controlled or both valves may be automatically controlled by a slurry level control loop. Slurry Level Control. Slurry levels are controlled by manipulating the outflow of the cell. The present value of the level is measured and compared to the setpoint and the controller calculates the signal used to manipulate the opening of the control valve in the cell. The majority

2181

of flotation plants use the PID-algorithm for this task. In Figure 8 the instrumentation for slurry level control in the flotation cell is shown. PROCESS

fl

INSTRUMENT _. . . . .AIR. SUPPLY ...

,__,_ .

'I

I

IPTRANSDUCER PNEUMATIC,Cx'.LINUER ANU,

'

I

CONCENTRATE UART-VALVE

i

,J,

J

c 1

I'

"

Figure 8 Instrumentation for level control in the flotation cell In a series of flotation cells, PID-control of slurry levels is troublesome. Feed disturbances travel slowly through the cell bank and as each cell independently compensates for the disturbance separately, they concurrently drive the level of the following cell off the setpoint. Therefore more sophisticated methods have been developed to deal with level control. These strategies consider the problem as a multivariable task, where the whole series of cells is monitored and compensations between cells are calculated. Jamsa-Jounela et al. (Jamsa-Jounela et al. 2001) presented feedforward control combined with traditional feedback control in order to diminish disturbances caused by the inflow of the cell bank. Stenlund and Medvedev (Stenlund and Medvedev 2000) presented a model-based decoupling control for levels in a series of flotation cells. Single level control loops are strongly interconnected and the aim of the control strategy was to decouple the consecutive level control loops, Perfect decoupling in this context means that, a change in either the slurry level or the control signal will have no influence on the level in the cell before or after the cell where the changes occur and are measured. A block diagram of feedforward control and decoupling control is presented in Figure 9.

2182

Uicomp

PID-controller

Compensation Control Valve

hi

,

Figure 9 Level control loop with either decoupling or feedforward compensation (i" cell, h cell level, u control signal, e error, F flow) Airflow Measurement and Control. Airflow can be measured in many ways. Common measuring instruments for flotation airflow are thermal gas mass flow sensor or differential pressure transmitters with venturi tube, Pitot tube or annubar element. Each instrument type has its advantages and limitations. The functionality of thermal gas mass flow sensors is based on the tendency of the flowing air to cool the sensor. The instruments using the sensor of this type are considered accurate but they are relatively expensive compared to other devices and measured air must be clean. These instruments are also factory calibrated, which complicates any later changes. Differential pressure flow meters are popular in industry, including flotation, due to their low price, simple principle and fairly low requirement of maintenance. There is a large selection of differential pressure flow meters available and for flotation airflow measurement, the transmitter combined with the venturi tube or Pitot tube is the most common. An orifice plate is not suitable solution due to the significant pressure loss it causes. A Venturi tube (see Figure 10) consists of converging conical inlet, a cylindrical throat, and a diverging recovery cone. Pressure is measured before the conical inlet and in the center of cylindrical throat. The pressure difference is related to flow and is calculated from the measurements by the transmitter. The Venturi tube is simple, reliable, and accurate if well calibrated and causes tolerable pressure loss. However, it is an expensive device and demands lot of space. The transmitter with Pitot tube or annubar element (see Figure 10) provides airflow in the pipe by measuring the static and the total pressure. A Pitot tube has only one measuring point, while annubar element has several across the pipe, providing the average velocity in the pipeline. Hence the result provided by annubar element is not as dependent on the velocity profile as by Pitot tube. Both techniques are quite accurate and the observed pressure drop is small. Problems associated with differential pressure flow meters are installation related, that is, large pipe sizes and flotation plant layouts where it is difficult to locate sufficient straight pipe sections. One solution is to use smaller pipe diameter in the section where instrument is to be installed, since straight pipe section length requirement is a function of the diameter.

2183

Figure 10 Schematic illustration of Pitot tube, venturi tube and annubar element Butterfly valves are used for airflow control in the cells, as they are low-priced and their performance is satisfactory for the task. Airflow control loop contains a measuring device and control valve, where the measuring dcvice is situated before the control valve in the pipe. Airflow is measured and that value is compared to the setpoint. Flotation airflow control is not as problematic as slurry level control; therefore, a properly tuned PID-controller has been determined adequate for this control task. However, correct sizing is important for airflow valves because many problems in airflow control are due to oversized valves and poor airflow control can disturb slurry level control. Unfortunately, oversized airflow valves are very common in flotation plants, often a result of the incorrect assumption that bigger is better. Advanced control strategies and operators use aeration rate as control variable for grade control and circuit balancing. Therefore the setpoints of the airflow control loop are frequently manipulated. Flotation cells with self-aspirating aeration mechanisms often do not have automatic airflow control. The available range of airflow control is anyhow limited. This problem is pronounced at high altitude. Reagent Control. Reagent feed rates are often set to worst-case values that contain considerable safety margin under average conditions. In addition to increased chemical costs, this may result in loss of selectivity of the flotation. A large variety of different instrumentation solutions for reagent dosing are available, due mainly to two special features of this task. Firstly, the reagent flows are quite small and therefore difficult to control and to measure. For instance common term of frother addition is milliliters per minute. Secondly, the reagents have different physical and chemical properties. A simple but inaccurate way of reagent dosing is the use of onloff valves with estimated liquid flow. Estimates are regularly corrected by manual checks. If accurate reagent addition is needed and flow volumes are large enough for individual control loops, the best results are achieved by using inductive flow meters and control valves. Metering pumps are also suitable, even though they are high-priced and have high maintenance requirements. If flow volumes are too small for earlier mentioned methods or if economy is a priority, the flowing systems are possible solutions. A dosing system presented in Figure 11 is suitable for flotation chemicals with reasonable conductivity (e.g., Xhantates, Cyanides, H2S04). Slow process response makes the use of this partly On-Off system possible. Levels of- the reservoir vessels are measured by ultrasonic level meters or by flange model pressure transmitters. Instead of actual measurement, simple level switches can be used to indicate high and low level. The idea is to control the main flow with inductive flow meter and control valve or variable speed drive of the pump, then afterwards, split the main flow to adequately small flaws for addition points. It is also possible to have a storage tank so high that gravity drives the flow, which is adjusted by a control valve.

2184

Individual reagent dosing control for each addition point is performed by On-Off ball valves with pneumatic actuator and sequence logic.

Figure 11 Dosing system for reagents with reasonable conductivity Reagents having no conductivity (oils, frother) cannot be measured by a magnetic flow meter and a solution could be the use of a Coriolis flow meter. A metering pump is another possibility but often the volume of reagent involved is too low for satisfactory accuracy in operation. One useful way to control reagents having no conductivity is described in Figure 12. The system consists of constant level tank (overflow) and 3-way solenoid valves having constant volume pipes above and process lines below. As the solenoid opens, it fills the constant volume pipe and when the solenoid closes, it lets the filled liquid to go to process line. The control system takes care of the On-Off pulses to solenoid valves.

Figure 12 Dosing system for reagents with no conductivity Most flotation plants use a ratio type control (grams of reagent per ton of ore) for the addition of reagents to the flotation circuit. The feedback action being carried out by plant operators, who adjust the ratio according to measurements provided by the on-stream analyzer (Hodouin et al. 2000).

2185

Figure 13 shows a typical feedforward ratio control of xanthate dosing in copper-lead bulk flotation. In this case collector addition is already made in grinding circuit. On-stream analyzer assays provides lead and copper grades in feed and mill feed is measured with a belt scale. Predicted collector addition is determined from regression calculation, which is multiplied by an operator factor giving operator possibility to make percentage changes to predicted addition in either direction. Regression constants are determined with history data analysis (Jones et al.).

I

I

I

Predicted Reagent Requirement (P) c, + c, x JMFX (%Pb + %Cu)

I

I

I

+ Xanthate Setpoint cc/min

PxF

Figure 13 Feedforward ratio control of xanthate (Jones et al.) pH, Eh and Conductivity Measurements. During the last decades attention has been paid to the effect of electrochemical potentials in slurry on the performance of flotation. Electrochemical measurements can give important information about the surface chemistry of valuable and gangue minerals in the process. Electrochemical potential measurements are almost the only method for detection of the chemical situation directly from the process (Ruonala 1995). pH is the most commonly measured electrochemical potential. pH measurement is a special case of ion selective measurement, whereby pH measurement is sensitive to the hydrogen ion concentration. By using ion selective electrodes the concentration of certain ion can be measured. Ion selective electrodes are electrochemical cells where potential difference between the membrane of the electrode and the examined solution is developed. The magnitude of the potential difference is proportional to the logarithmic activity of the selected ion in the solution. Unfortunately, the potential difference cannot be reliably measured, directly. In addition, a reference electrode is required to measure the potential of solution. Measuring equipment are electrodes and pH-transmitter. Sometimes pH measurement can be replaced or complemented by conductivity measurement, which can give approximately the same information about the flotation process and the instruments are less expensive. In general the conductivity measurement may be better than pH measurement in highly alkaline conditions, where reliable and accurate pH measurement is difficult to achieve. Conductivity measurement does not work well if large amount of air in the slurry or ore properties cause variations in the conductivity. Eh measurement with platinum electrode (redox) can be useful in some special cases, such as Cu-Mo flotation, where additional nitrogen is added to flotation air.

2186

pH measurement is troublesome because the electrodes are easily contaminated by active substances in the slurry. Therefore pH measurement often needs a sampler system or automatic washing system along with regular checks and maintenance. Slurry pH is controlled by PID control, which adjusts lime or acid addition rate to the slurry. pH control is also problematic due to very slow response of the system to control actions. Therefore, any control interval must be set long; to allow the process to react to the control actions before the controller executes the next action. Other electrochemical Potential measurements. Recently other electrochemical potential measurements have been under study. The use of minerals as working electrodes makes it possible to detect the oxidation state of different minerals and handle their flotability. The idea and equipment are similar to pH measurement, apart from the electrode materials, which are chosen according to the mineral. Stability of the electrodes has been a significant problem in on-line use although some encouraging results have been achieved using different mineral electrodes for process studies (Ruonala, Heimala and Jamsa-Jounela 1997). Some minerals float well within certain limits of electrochemical potential. Improved recovery by the control based on electrochemical potential measurements may be possible if the correlations between assays of feed, tails, concentrate, recovery and electrochemical potentials can be found in data analysis. Ruonala (Ruonala 1995) developed and implemented the control of sulfuric acid by cascade control, where the master loop potential of NiS-electrode gave the setpoint changes to pH of the slave loops in conditioner and flotation cells. Data collection was made before the implementation of the control. In the daily study, statistical and local optimum potentials were found. About 0-2 % increases in nickel recovery was reported, depending on the ore type. Froth Image Analysis. Operators have always used their eyes to characterize froth. Based on their visual perception they have determined proper control actions for the particular process. However, the weak repeatability over time and the uncertainty due to the varying judgement of each individual have made this information unreliable. In automatic control systems, the use of information obtained by human visual perception has been difficult as well. Therefore new machine vision based instruments have been developed to obtain information about froth characteristics and some commercial products are already available on the market. These instruments build their functionality on the methods of image processing. Images provided by integrated digital camera of the instrument are analyzed with different models and algorithms. There are structurally two types of froth image analysis systems. A centralized type gathers images from several cameras installed on flotation cells, to one central computer, which performs the image analysis for images collected from all units. A distributed type device has camera and computer integrated together, so that all calculations are performed in the instruments. Therefore distributed system is very flexible, consisting of single instruments, that one by one can be connected directly to the plant automation system. Conversely, a centralized system forms a distinct froth image analysis system, which then may be connected to plant automation system. Along with the froth measurements, both types provide a visual image for the operator to view froth in the cells. In general the difficulty with froth imaging has been relating the information extracted from the images to the flotation performance. Different indices and measurements describing properties of the froth such as bubble size distribution, bubble shape distribution, color, number, density, speed and stability, are provided by these instruments, yet there are only few real applications reported (Cipriano et al. 1998; Sadr-Kazemi and Cilliers 1997). The use of froth speed in control tasks has, however, proven its usefulness (Brown et al. 2001). To summarize the issues discussed above, an example of a flotation circuit flowsheet with typical instrumentation is presented in Figure 14. Conductivity and pH are measured only in the conditioner. On-stream analyses are taken from feed, tailings and concentrate of the circuit, also from several flows between the flotation sections. Slurry flows, pulp levels and airflows are measured at several critical points. Most of the reagents are already added in grinding circuit,

2187

except frother, which is introduced to slurry in the conditioner and the re-addition of sodium cyanide in the cleaner. Monitoring of the overall process, as well as each section of the process, is enabled by the extensive instrumentation. Level and aeration rate controls perform the fine adjustment of process condition, as flow measurements and on-stream analyses supervise critical circulating loads and performance of various sections. Conductivity measurements provide information about the electrochemical state of the feed and pH-measurement is needed for pH-controlled lime addition.

F = Fliiw

L =Pulp LCWl A =On-Lmc analysis (Cu.2s. Fc)

c = Co!ld"cu"ily

Figure 14 Typical flowsheet of flotation process ,Cu-circuit of Pyhasalmi concentrator (Koivistoinen and Miettunen 1985) Advanced Controls Advanced controls are mainly controlling recovery and concentrate grade. The setpoints for an advanced controller are determined by a plant optimization system or by an operator. Based on setpoints, elemental assays provided by on-stream analyzers and other measurements, the advanced controller determines the control action, which often means the adjustment of the setpoints of base level controllers. Elemental assays from the tailings and the concentrate enable the feedback control of flotation as the performance of the circuit is monitored and corrective actions are made by the advanced controller. Elemental assay from the feed provides the possibility of the feedforward control actions. Reagent additions, aeration rates and slurry levels are adjusted, based on information of the incoming ore. Many advanced controls and particularly the feedforward type controllers, are relying either on process models or rule-based advanced model. Aeration rates, slurry levels or reagent additions rates are increased or decreased by IfThen rules, if recovery or concentrate grade is deviating from its setpoint. Several deterministic process models to estimate metallurgical responses in flotation have been developed. Niemi, Maijanen and Nihtila (Niemi, Maijanen and Nihtila 1974) presented a fairly simple model for flotation, derived from mass balances. The flotation cell is considered as a

2188

tank with ideal mixing and first order flotation kinetics are used to describe floating of the mineral. Several affecting variables were ignored. Bascur (Bascur 1982) chose a population model approach for development of the flotation process model. The idea was to divide mineral particles into several populations depending on their size, mineralogical composition and state in the slurry (e.g., free in slurry, in slurry attached to bubble). Transfer functions between populations are defined and the hydraulic model is included in the population model. Flotation models have considerable weaknesses due to the complexity of process. Generally all the models suffer from the influence of unmeasured variables on the results of the model. Very complicated models make their use in control purposes difficult, as they demand many measurements and parameters, with inevitable problems in robustness of the model. On the other hand, simple models have problems in accuracy. Various advanced control techniques such as adaptive control, multivariable control, optimal control, predictive control and fuzzy control, have been developed and explored for flotation control. All methods have generally the same control objective and the same operational principles but the approaches are different. Model-based control strategies are often based on models presented above, however the population model, developed by Bascur, is usually substantially simplified. The control objective is to drive the recovery and the grade to their setpoints by manipulating the control variables, which are reagent addition, aeration rate, and slurry level. The suitable values for control variables are obtained by advanced control techniques and the values obtained are then used as setpoints for respective base level controls. The tendency is towards the use of aeration rate as a control variable since the slurry level control is more complicated and response of the reagent addition suffers from with considerable delay. Control objectives are normally combined to one criterion, containing weighted terms for both recovery error (deviation from setpoint) and concentrate grade error. Often control variables are as well included to the criterion, which is then minimized with one of the techniques mentioned above. Brown et al. (Brown et al. 2001) stated that the speed at which the froth is recovered over the lip of the cell has a very direct and consistent influence on the grade and the recovery of the flotation circuit. Exploitation of this relation has been difficult due to the absence of a suitable device for measuring the froth speed. As mentioned earlier, new machine vision based instruments can now provide this function. Therefore, control of the froth speed is possible to execute by using slurry level, aeration rate and frother dosage as manipulated variables. A hierarchical control for the froth speed has been developed (Brown et al. 2001). The aeration rate was used for fast and fine froth speed adjustment, while for relatively big step speed adjustments, level and frother changes were applied. The aeration rate was chosen as the primary adjustment variable, as air could be controlled more tightly to a setpoint than the level, level measurement being noisier than air measurements. The froth speed control was successfully extended to control of concentrate grade. It was found that the concentrate grade is more strongly correlated to froth speed than to level or to aeration rate. For this reason, the concentrate grade controller was developed. Based on the grade information obtained from on-stream analyzer, a fuzzy controller adjusted the setpoint of the froth speed. The recovery from the cells under the new control strategy was found to be significantly better than the recovery from the corresponding cells in other bank under manual control. Authors are confident that controls of this type will become common in many flotation plants.

Optimization When considering the optimization of the flotation process, there are some requirements for the automation system and the knowledge of the process. Appropriate instrumentation and well-tuned base level controls are essential. A quality and extensive data bank with process data including onstream and laboratory analysis data is also required.

2189

The data bank enables the creation of grade-recovery curves for different ore types. Graderecovery curves are the basis of most optimization methods as it describes the relationship between the mineral or metal recovery and the grade of the concentrate for a certain ore type. Typical relationship between these factors is presented in Figure 15.

Grade-Recovery Curve 96

1

91 90

25

24

27

26 Grade [Oh]

Figure 15 Grade-recovery curve

As seen in Figure 15 an increase in recovery decreases grade and vice versa. Therefore the optimal point on the curve is somewhere between the maximum recovery and maximum grade. There are some ways to determine the optimal point. Formulas to calculate the value of products and operating costs of production are desirable as some strategies approach the problem from the economical point of view. One idea of how to find the optimal point on certain grade-recovery curve is the principle of isoeconomic contours (Flintoff 1992).

Grade-Recovery Curve with lsoeconomic contours

24

26

25 Grade [%]

Figure 16 Grade-recovery curve with isoeconomic contours

2190

27

In Figure 16 dotted lines are describing the equal economical recovery of the process as a function of recovery and concentrate grade. The maximal profit will be achieved by finding the maximum point of grade-recovery curve in proportion to the isoeconomic contours. Calculated isoeconomic contours are based on net smelter return, which sets the price for produced concentrate. It consists of market price of the produced metal, concentrate grade, refining and smelting fee, quality-based penalties depending on the contents of certain impurities and other costs (freight rate and dewatering costs). The following equations show an example of net smelter return calculation.

T=

T, Recovery

where

MI% TI MI% MZ% Mi%GDF PMi CSMELTMG CFEFMMG,Mi IMP ~PLIMIT CIMPURITY Mo MOLIMIT CMO COTHER

T Recovery

is total price for concentrate ton is primary valuable metal grade in concentrate, (e.g., copper) is secondary valuable metal grade in concentrate, (e.g., silver) is deduction factor for valuable metal i (i=l ...2) grade in concentrate is price factor of valuable metal i (i=l.. .2), USD/(% x concentrate ton) is smelting fee, USDkoncentrate ton is valuable metal i (i=l.. .2) refining fee, USD/(% x concentrate ton) is impurity grade in concentrate (e.g. zinc in copper flotation) is impurity grade limit is impurity fee, USD/(% x concentrate ton) is concentrate moisture is concentrate moisture limit is moisture fee, USD/(% x copper concentrate ton) is fixed fee €or dewatering and freight, (USDkoncentrate ton) is total price for primary valuable metal ton in ore is primary valuable metal recovery

In spite of control strategies and expert systems developed, there still are a large number of flotation plants operating purely on the experience of the operators. In these cases, operators examine the curves above, the recent state of process, the available measurements and adjust setpoints of the controllers on the strength of their knowledge. Sometimes cells are even operated totally manually. An experienced operator can often find right corrections to make a given process perform well but inconsistency and lack of continuity provide the need for optimization systems. New expert systems are concentrating to solve the issue of the feed type classification, which is a challenging and important task. Each feed type should be specially treated, as their graderecovery curves are feed type-specific as well as finding the optimal reagent dosing. The performance of the expert system is founded on the success of the feed type classification (Laine 1995). The success of the classification is, in turn, based on the on-line information that the system receives and on the algorithm the system uses. On-line measurements input to the system must include adequate information to distinguish feed types. The algorithm also must be able to classify the information into different feed types. The knowledge on the type of the feed of the process can be utilized in feedforward control, which sets the suitable process conditions for each feed type.

2191

type

Know.base A

Validation A

w Operator Measurements Infof Measurement si nals

Feed I

Set lpoints

Inputs

v

Set points

Control si nals Instrumentation

Process

Product I

,

Figure 17 The structure of the expert system in Hitura concentrator (Jamsa-Jounela, Karesvuori and Laurila 2000) The expert system knowledge base consists of rules to which the knowledge of process is converted. Normally If-Then rules use measurements, classified feed type and process models to define optimal control actions. In Figure 18 an example of optimization of reagent dosing is shown (Koivistoinen and Miettunen 1985).

2192

,---_--____-----____----------------------------------------------------------

Niirmal -recover Roughing debricxating

CuSO, set pint: stcp Upwuds

STABILIZING

NO I

I

Rlbring lhe input variahles

I

vanations of the input variahlcs

IX)

lhc rrutpul

OPTIMIZING Identificatiim: Parameters descrihlng thc effects of ZnSv'Zn-circuit feed and CuSO, on the output variahles are estimated by lincnr regressim.

Rcliahility of the ~cgressi"~ modcl is tesbd

Thc new CuSO, scl point is dcbrmincd by lhc ~iplimizing ulgorithm

Figure 18 Optimization of CuS04 dosing in Pyhasalmi concentrator (Koivistoinen and Miettunen 1985)

CONCLUSIONS There has been a veritable plethora of strategies, techniques, technologies, and instrumentation applied over many years by many people, scientists, researchers, suppliers and the industry itself, pursuing the elusive goal of optimizing flotation performance over prolonged periods. The pursuit has met with some encouraging progress in the past few years. A combination of extensive accrued knowledge of the process and control regimes, modern instrumentation and data processing all contribute to the possibility of optimum performance being achievable through plant automation. Nevertheless, the highest and most successful level of optimization will always require that the most basic levels of control are in place, operating well and maintained so. REFERENCES Andersen, R.W., Gronli, B., Olsen T.O., Kaggerud I., Romslo K., and K.L. Sandvik 1981. An optimal control system of the rougher flotation at the Folldal Verk concentrator, Norway. Proceedings of 13'" International Mineral Processing Congress vol. 2, ed. J. Laskowski, Varsova. 1517 - 1537. Arbiter, N., and C.C. Harris 1962. Flotation kinetics. In Froth Flotation, ed. D. W. Fuerstenau, New York, Edwards Brothers Inc. 215-246.

2193

Aumala, 0. 1996. Teollisuusprosessien mittaukset. Tampere: Pressus. B ascur, O.A. 1982. Modelling and computer control of aflotation cell. Ph.D. thesis, University of Utah, Utah. Borer J. 1985. Instrumentation and Controlfor the Process Industries. New York: Elsevier. Brown, N., Bourke, P., Ronkainen, S., and M. van Olst 2001. Improving flotation plant performance at Cadia by controlling and optimizing the rate of froth recovery using Outokumpu Frothmaster, Proceedings of the 33“’ Annual Meeting of the Canadian Mineral Processors, ed. M. Smith. 25-37. Cipriano, A., Guarini, M., Vidal, R.,Soto, A., Sepdveda. C., Mery, D., and H. Briseiio 1998. A real time visual sensor for supervision of flotation cells. Minerals Engineering. 11(6):489. Flintoff, B.C. 1992. Measurement issues in quality “control”. Presented at the 1992 Toronto CMP Branch Meeting. Herbst, J.A., Pate, W.T., and A.E. Oblad 1992. Model-based control of mineral processing operations. Powder Technology 69:21. Hodouin, D., Bazin, C., Gagnon, E., and F. Flament 2000. Feedforward-feedback predictive control of a simulated flotation bank. Powder Technology. 108:173. Henning, R.G.D., Schubert, J.H., and Y. Atasoy 1998. Improved flotation performance at Fimiston plant through better level control. Presented at International Symposium on Gold Recovery, Montreal. Hulbert, D.G. 1995. Multivariable control of pulp levels in flotation circuits. Preprints of the 8”* IFAC International Symposium on Automation in Mining, Mineral and Metal Processing, ed. I. J. Baker, Sun City. 71 - 76. Jones, J.A., Deister 11, R.D., Hill, C.W., Barker, D.R. and P.B. Crummie. Process control at the Doe Run company. Jamsii-Jounela, S.-L., Dietrich, M., Halmevaara, K., and 0. Tiili 2001. Control of pulp levels in flotation cells. Preprints of the 10”’ IFAC International Symposium on Automation in Mining, Mineral and Metal Processing, ed. M. Araki, Tokyo, 8 1-86. Jamsa-Jounela, S-L., Karesvuori, J. and H. Laurila 2000. Flotation process neural data analysis and on-line monitoring, Proceedings of the 32nd Annual Operator’s Conference of the Canadian Mineral Processors, ed. M. Tagami, Ottawa. 441-457. Koivistoinen, P., and J. Miettunen 1985. Flotation control at Pyhasalmi. In Developments in Mineral Processing 6, Flotation of Sulphide Minerals, ed. K.S.E. Forssberg, Amsterdam: Elsevier. 447-472. Koivo, H.N., and R. Cojocariu 1977. An optimal control for a flotation circuit. Automatica 13:37. Laine, S. 1995. Ore type based expert system for Hitura concentrator. Preprints of the 81h IFAC International Symposium on Automation in Mining, Mineral and Metal Processing, ed. I. J. Barker, Sun City. 321-327. Niemi, A.J., Maijanen, J.S., and M.T. Nihtila 1974. Singular optimal feedforward control of flotation. Preprints of the IFAC/IFORS Symposium on Optimization Methods Applied Aspects, ed. I. Tomov, I. Popchev, G. Gatev, M. Kitov, N. Naplatanov, Varna. 277-283. Osorio, D., PBrez-Correa, J.R., and A. Cipriano 1999. Assessment of expert fuzzy controllers for conventional flotation plants. Minerals Engineering. 12(11):1327. PCrez-Correa, R., Gonzilez, G., Casali, A., Cipriano, A., Barrera, R., and E. Zavala 1998. Dynamic modelling and advanced multivariable control of conventional flotation circuits. Minerals Engineering. 11(4):333. Ruonala, M. 1995. The use of electrochemical mixed potential measurements for the process control and expert system development at the Hitura mine. Proceedings of Automation in Mining, Mineral and Metal Processing. Ruonala, M., Heimala, S., and S. Jounela 1997. Different aspects of using electrochemical potential measurements in mineral processing. Int. J. Miner. Process. 5 1:97. Sadr-Kazemi, N., and J.J. Cilliers 1997. An image processing algorithm for measurement of flotation froth bubble size and shape distribution. Mineral Engineering. 10(10):1075.

2194

Schubert, J.H., Valenta, M., Henning, R.G.D., and I.R. Gebbie 1999. Improved flotation performance at Karee platinum mine through better level control. Journal of the SAIMM JadFeb (1999). Stenlund, B., and A. Medvedev 2000. Level control of cascade coupled flotation tanks. Preprints of the IFAC workshop on Future Trends in Automation in Mineral and Metal Processing, ed. S-L. Jamsa-Jounela. Helsinki. 194-199.

2195

Pressure Oxidation Control Strategies John Cole and John Rust'

ABSTRACT In mineral processing operations there are more items one wants to control than can control. Gold pressure oxidation facilities are the same way. It would be nice to see, touch, taste and feel what is happening inside an autoclave, as with other unit operations. To measure pH online inside the autoclave could be critical to enhancing recovery or lowering costs. The only sense left to an operator is smell, which is very useful to discover a leak in the circuit, but not for process control. This paper will address current and future control strategies utilized in gold pressure oxidation operations.

INTRODUCTION Newmont Mining Corporation owns and operates the Twin Creeks and Lone Tree Mines in Humboldt County, Nevada. Twin Creeks is located 85 kilometers northeast and Lone Tree 55 kilometers east of Winnemucca, Nevada. Both mines utilize pressure oxidation to treat high-grade refractory ores. Control strategies for these plants and others are similar. This paper will discuss control strategies for gold pressure oxidation circuits for whole ore feed. Control of concentrate pressure oxidation circuits are similar, but not exactly the same. The greatest difference being the lack of slurry preheating. The Twin Creeks refractory process facility is typical for all whole ore pressure oxidation plants. The process flowsheets are shown on Figures 1 and 2.

#'

John Cole is Chief Process Engineer and John Rust is Senior Metallurgist for Newmont in Nevada

2196

Figure 2 Process Flowsheet Pressure oxidation is a pre-treatment step to gold recovery. Ore is delivered from the mine to stockpiles. All facilities utilize size reduction to start the ore processing. Specific ore body needs dictates the configuration and size of grinding equipment. Twin Creeks utilizes a SAG mill, primary ball mill and secondary ball mills for size reduction. Target grind size at Twin Creeks is very fine. Post grinding most facilities de-water the slurry to the maximum extent. Excess water consumes too much heat in the autoclaves affecting the circuit heat balance. Depending on the geochemistry of ore, the thickened slurry may be acidified. Slurry is conditioned with sulfuric acid as required to reduce the carbonate content of the feed. The slurry is then fed to the pressure oxidation circuit. Generally there are several stages of slurry heating utilizing steam from the let down flash vessels. The pre-heating step is critical in whole ore circuits. Heated slurry is pumped into the autoclave via positive displacement diaphragm pumps. The autoclaves are horizontal cylindncal pressure vessels with multiple compartments. There is at least one agitator in each compartment. Oxygen is generated onsite by a cryogenic air separation plant. Oxygen required for the oxidation reaction is added to each compartment. Steam and water may also be added to each compartment to control the temperature. Autoclaves operate anywhere from 160°C to 23OOC at pressures ranging from 1,225 W a to 3,150 Wa. These pressures reflect the saturated vapor pressure plus the non-condensable gas pressure. Vessel retention times are generally 45 to 60 minutes. Oxidized slurry exits the autoclave through a ceramic choke valve. The pressure let down system consists of flash vessels, same number as slurry heaters, which reduce the slurry pressure and subsequently the slurry temperature to atmospheric levels. The slurry is further cooled either through a series of tube and shell heat exchangers or counter-current

2197

decantation thickeners. The cooled, oxidized slurry then reports to a neutralization circuit where limestone and/or milk-of-lime is added to neutralize the acid and provide protective alkalinity for the gold leaching and recovery circuit. Neutralized slurry may then be processed through any number of conventional gold recovery means. PROCESS CONTROL Process control for pressure oxidation circuits can be categorized as physical and metallurgical controls. It is the same as any mineral processing plant. Physical controls include; level, flow, mass flow, pressure and temperature. Metallurgical controls include; ore blending, slurry density, slurry viscosity, eWpH, free acid level, level of sulfur oxidation, rate of sulfur oxidation, oxygen utilization, and ferrous iron titration. Without proper physical controls, metallurgical control of the process would be extremely difficult. Physical control of a pressure oxidation circuit is relatively straightforward and not too complex. Though proper physical control is not complex, the outcome of that control may not be cost effective. An example would be over using steam and water for temperature control. Figure 3 shows a simplified pressure oxidation flow sheet with only slurry flow shown.

Hlgh Pressure Flssh

s

I" r r y Hsaler

Physical Controls Level Control. There are many types of instruments to measure level such as ultra-sonic, differential pressure, nuclear devices and even strain gages. Because all of the vessels operate at elevated temperatures and some at elevated pressures, measuring the level in these vessels is difficult. Through years of experience nuclear devices have proved to be the most reliable means of pressure oxidation vessels level measurement. Nuclear sources are used in the heaters, autoclaves, flash vessels and other minor vessels for level measurement. Vessel level is controlled in the slurry heaters and the autoclave, measured only in the other vessels. In the case of the heaters a nuclear source is mounted on the outside of the heater vessel with a strip detector mounted on the opposite side. Level is measured by the amount of radiation reaching the detector. Level set points are set at a maximum comfortable level, about 80% of range. This provides the maximum suction head on the discharge pump and maximum surge capacity in case of upset conditions. The actual level range monitored is a small part of the vessel. The measured level range starts near the bottom or outlet of the vessel and ends below the flash steam inlet. Inside the vessel above the steam inlet pipe are sets of trays designed to cascade the slurry contacting it with the steam. If the slurry level rises above the steam inlet, the steam will collapse causing very violent shaking of the heaters. To prevent this a separate point source nuclear device is installed as a high level interlock. The heater level controls the speed of it's feed pump, either directly or cascading through flow control. The goal of the control scheme is to have steady levels in the heaters.

2198

At Twin Creeks the first stage heater operates at atmospheric pressures. The second stage heater operates at elevated pressures. There is a pressure control valve on the vent outlet of the second stage heater. Steam to heat the slurry is supplied by the respective first or second stage flash vessels on the autoclave discharge. Level control for two vessels in series utilizing pumps to control the levels is easy. The added complexity of the autoclave circuit is the temperature, steam flow and source of steam. Since the source of steam is the flash vessels any change in throughput or autoclave level will impact the amount of steam reporting to the heaters. In the second stage heater this change in steam flow will change the vessel pressure which in turn will either increase or decrease the head on the feed pump. This will cause a change in flow, which leads to an upset in heater vessel level control. As stated earlier the goal of the control scheme is steady heater level. With the added complexity of varying pressures, a preferred control method utilizes a level signal cascading through flow control. Figure 4 is an example of slurry heater level control.

From

Grinding

+ Figure 4

I G e h o P u m p

1

For the autoclave, the nuclear source is mounted inside the autoclave in a double-lined well. The well is located in the vapor zone near the top of the autoclave. A strip detector is mounted on the outside of the autoclave in a position that the slurry level in the autoclave is measured. Actual level measured is generally only at the top of the vessel. The level range is from the horizontal centerline to the top of the last compartment wall in the autoclave. This represents a 0% to 100% level. The autoclave slurry outlet is set near the horizontal centerline of the vessel therefore it is desirable to have the level higher than the outlet. Otherwise both gas and slurry would exit the autoclave through this pipe. A slurry level above the compartment wall could allow shortcircuiting. The desired slurry level is fifty millimeters below the last compartment wall. The level detector sends a signal to a right angle choke valve with a modulating plug. As the autoclave level increases the plug opens allowing more flow to exit the autoclave. The reverse is true for decreasing autoclave level. Tuning and proper operation of this valve is critical to autoclave operation. If this valve operates too quickly or is tuned to hold a tight level it can cause several other problems in the circuit. These will be described in detail later. Figure 5 shows the autoclave vessel level control scheme.

2199

Pressure Flash

Law Pressure Flash

Figure 5 Flow Control. There are two major items where flow is controlled, slurry and oxygen. Both are mass flow, but controlled as volumetric flow. There are other flow controllers such as water or steam but these are utilized for temperature control. Flow set points are set based on throughput requirements. Slurry flow is measured by a magnetic flow meter. Slurry density is measured by a nuclear device. Combination of flow and density provides mass flow. Slurry flow into the autoclave is controlled by the autoclave feed pump speed. These pumps are positive displacement diaphragm pumps. The slurry flow in these pumps is linear to the speed. The speed controller is a manual set constant. The operator increases or decreases flow based on operating conditions, generally the faster the better. As the operator increases the speed of the feed pump the heater feed pumps will also increase in speed based on the level of the heater vessels. An increase autoclave feed pump speed will lower the level in the second stage heater. The lower level will send a signal to the second stage heater feed pump to increase the speed. This in turn will lower the level of the first stage heater and cause the first stage heater to send a signal to the first stage heater feed pump increasing its speed. Figure 6 shows this control method.

2200

L o w

Tern p s I" r r y H e a l e r

f Figure 6

IG e h o

P urn p

I

T

Oxygen flow is measured on the main oxygen header from the oxygen plant by a differential pressure orifice plate. This flow is pressure and temperature compensated to standard conditions. This gives the total oxygen flow to the autoclave. The operator utilizing a manual set constant controls total oxygen flow. There are also individual compartment flow meters. These are vortex shudder types. They are not pressure and temperature compensated. The compartment flow meters are used to distribute oxygen for metallurgical purposes therefore the level of accuracy required is not as high as total oxygen flow. Oxygen flow changes and distribution of the flow to the compartments are made based on the metallurgical performance of the plant. Oxygen flow control is shown on Figure 7.

Figure 7

220 1

Pressure Control. Pressure control is utilized in the second stage slurry heater, the autoclave and the autoclave agitator seal water system. In all cases the control device is a valve. Set points vary with operating conditions. Autoclave pressure set points are generally 500 to 700 kPa above the boiling point at that temperature. This is known as the non-condensable over-pressure. The majority of the non-condensable gas is oxygen. A right angle choke valve with a modulating plug controls pressure in the autoclave. The autoclave control valve has to be quick reacting due to the rapid changes in pressure. Autoclave pressure can swing as much as 350 kPa with the loss of a feed pump, oxygen flow or an agitator. Most of the oxygen is added to the first and second compartments in the autoclave. If one of these agitators stops all of the oxygen in those compartments reports to the vapor zone rapidly thereby increasing pressure rapidly. Oxygen is added to the autoclave at a fixed rate based on the tomes of sulfur fed. Failure in one of the two feed pumps would halve the sulfur feed though the oxygen flow would remain constant. The result would be a rapid increase in autoclave pressure until the pressure control valve compensates for the change. A loss of oxygen flow would cause a rapid decrease in autoclave pressure. Autoclave pressure control is illustrated on Figure 8.

I

9

Figure 8

The seal water system is designed to provide a water pressure slightly higher than autoclave pressure for the agitator double mechanical seals. The set point is generally 350 kPa higher than the autoclave. Since the set point pressure is tied to the autoclave pressure, any changes in the autoclave pressure affects the seal water system. Therefore the pressure control valve must be very quick acting. Rapid decrease in autoclave pressure does not cause a problem for the seal water system. A rapid increase in autoclave pressure will encroach on the seal water system pressure causing the system to shut down and go into emergency back up. Once pressure is stabilized the seal water system would have to be started. Temperature Control. There is a relatively narrow band the temperature in the autoclave is controlled at. For most autoclaves that band is 15OC. Too low of temperature and oxidation is inhibited or ceases. Too high of temperature and the pressure limits on the vessel are approached. Temperature control in the autoclave is simple providing the ore is within design specifications. There is the ability to add steam for heat and water for cooling to all compartments. The goal in operation is not to add either steam or water for temperature control. Proper ore blending and adjusting the incoming slurry temperature and density is the preferred control method. Feed rate is also important in temperature control. Too fast of feed rate causes low temperatures in the first compartments, high temperatures in last compartments. Too slow of feed rate causes high

2202

temperatures in the first compartments, low temperatures in last compartments. Temperature control for the autoclave is shown on Figure 9.

Steam

P-

i

I

Water

Figure 9 Slurry temperature control in the heaters is essentially pressure control for the second stage. The first stage temperature is not controlled but merely a function of steam flow. Since the first stage heater is operated at atmospheric pressures the highest temperature obtainable is 100°C. Steam from the first stage flash vessel is directed through the second stage heater. The steam comes in direct contact with the slurry. A control valve on the vent outlet regulates the vessel pressure for both the flash and heater vessels. The slurry temperature will approach the steam temperature in steady state operation. Therefore to increase the slurry temperature increase the vessel pressure and decrease pressure to decrease slurry temperature. Changes in operating heater temperatures are made based on autoclave performance. For ores low in sulfur a higher slurry heater temperature would be appropriate. For high sulfur ores lower heater temperatures are would be correct. There is a maximum slurry temperature to autoclave feed pumps can be exposed to, but no minimum. The maximum is dictated by the rubber components in the feed pump. Slurry heater temperature control is shown on Figure 10.

2203

Low Temp Slurry Healer

I

HighTemp Slurry

Scrubber Blast Tube

I

.

%r

Figure10

Metallurgical Controls Without proper control of any of the above-mentioned physical controls, metallurgical performance would suffer. There are metallurgical control aspects of level, pressure, and temperature and flow control. These will be discussed also.

Ore Blending Proper feed to the autoclave is the first and most important aspect of successful operations. An autoclave’s function is to pre-treat an ore so that valuable elements may be recovered. The autoclave must bum the sulfur that occludes the gold to allow gold recovery. An autoclave operator is not concerned with the gold grade feeding the plant. Gold will enter and exit the autoclave in the same form. An operator is extremely concerned with the sulfur and carbonate assays, and other ore constituents that oxide or change form. The sulfur is the fuel for the process and the carbonate can be beneficial or detrimental depending on the amount. There is a minimum and maximum amount of sulfur or carbonate an autoclave can process. Too low of sulfur and not enough heat to sustain the reaction, too high of sulfur and not enough oxygen to maintain throughput. Carbonate behaves the same way, too low is usually not an autoclave problem but can be a neutralization problem. Too high of carbonate and the reaction will be quenched due to lack of acid. Other elements such as copper can affect down stream circuit performance with increased cyanide consumption. As with any circuit the autoclave operates best with the least amount of upset. Having a consistent feed is the best way to prevent upsets. An unanticipated high carbonate feed can be very expensive in terms of unwanted down time and the cost acid to destruct the carbonate. An unanticipated low sulfur feed can also be expensive due to the steam required to keep the reaction going. In ore blending absolute assay values are not as critical as anticipating assay values. It does not matter as much if the ore has a 3% sulfur assay or 5% sulfur assay, as long as the operator knows the sulfur value and can adjust the circuit for them. Slurry DensityNiscosity Slurry density is important for circuit throughput. Slurry viscosity plays a role in oxidation. The higher slurry percent solids, the more circuit throughput at a constant pumping rates. From strictly a production standpoint the thicker the better. Usually, though not always, the more dense the slurry the greater the viscosity. A typical autoclave feed solids target would be 50 percent solids. Too low of density, less than 45 percent solids, throughput is lowered due to pumping limitations and heat balance is affected. Lower density means more water to pump and heat and cool. Water takes five times more energy to heat than solids. Therefore it is critical to have the minimal amount of water in the feed as possible. Even

2204

with high sulfur in the feed, a low density feed will require steam addition to maintain proper temperature. Slurry density is controlled upstream of the autoclave in the grinding thickener prior or by water addition to the autoclave feed. The latter works only if the slurry is thicker than desired. The measure of slurry viscosity is relative to previous viscosity measurements. The control of slurry viscosity is more art than science. There are several ways to increase viscosity and only one cost-effective way to reduce viscosity. Ore types, high clay, generally produce high viscosity slurries. Other operator-induced methods of viscosity enhancement could be over flocculating a thickener or excessive use of acid in the acidulation circuit. The one true viscosity reduced is dilution with water. An operator desires the highest feed percent solids possible thereby maximizing throughput. If the slurry becomes too viscous, sulfur oxidation is ultimately hampered. The autoclave agitators can not mix and distribute the oxygen effectively in too viscous of slurry. The oxygen added to the autoclave must be physically dispersed and intimately contacted with sulfur in order to oxidize it. The more viscous the slurry the smaller area of influence the agitator will have. The oxygen will form large bubbles and vent through the slurry to the vapor zone in the autoclave. Therefore only part of the slurry will be oxidized. Excessively high viscosity can cause similar temperature control problems as too low of density. With limited oxidation in the first compartment due to high viscosity there will not be enough heat generated to maintain the proper temperature. Steam addition would be required to maintain the temperature. The oxidation reaction would move towards the later compartments where typically less oxygen is added. The result would be to starve the slurry for oxygen caused by higher than normal sulfur assays in the later compartments without commensurate levels of oxygen. Though the only control for high viscosity is dilution, knowing when to dilute can be as difficult. Generally if the first compartment temperature is not at anticipated levels and the reason is not apparent, dilute the feed. This will accomplish two things, the solids feed rate will be reduced along with the viscosity. Both of which should help oxidation rate and temperature profile. Oxidation. To achieve sulfur oxidation a few items are required, heat, acidic environment, and oxygen. To a large extent the amount of each of the items determines the rate of oxidation; more heat, more acid, and excess oxygen equals faster rates of oxidation. On the surface one would argue that complete oxidation at the fastest rate possible is always desired. This is not always true. The extent of oxidation required is ore specific and the minimum required oxidation should be targeted. In the case of the Lone Tree Mine, the original design was for partial oxidation of the sulfides. The desire was to oxidize the entire fine-grained pyrite, which was gold bearing, and none of the coarse pyrite, which was barren. The target was set at 75% overall sulfur oxidation. Though it was possible to achieve more oxidation, it was undesirable from a cost standpoint. The more sulfur oxidation meant more acid generation, which in turn meant more lime consumption. The gold recovery did not improve with increased oxidation. To start an oxidation circuit the autoclave is filled preferably with acidic slurry. The acidic slurry allows for easier start ups since the slurry is already acidic. When starting without acidic slurry the first addition of oxygen will go to creating an acidic environment. The autoclave is then heated at a slow rate with steam. Slurry flow through the circuit is started when the autoclave is at operating temperatures. As soon as slurry flow is established oxygen flow to each compartment is established. As the oxidation reaction starts the temperature will increase. As the temperature increases in the autoclave the slurry flow is increased. Both oxygen and slurry flows are increased simultaneous until the desired flow rates are achieved. It is easy to overcome the reaction by increasing the slurry feed too fast. The start up and operation of an autoclave is analogous to a campfire. To start a fire one must have kindling and as the fire is burning larger pieces of wood are added. Too much wood too early and the fire will go out. Not enough air and the fire will go out. Not enough wood or too much air and the fire will burn out quickly. The operator must strike a balance with the fuel (sulfur) and air (oxygen). Once the circuit is operating normally decisions must be made as to the targeted throughput, oxygen addition rate, oxygen distribution, temperature, pressure and acid addition to the feed. Most of these decisions are made in advance and not always by the operator. Throughput is dictated by management, oxygen addition is a

2205

function of sulfur assay and throughput, temperature and pressure are usually at design levels, leaving acid addition and oxygen distribution for the operator to manipulate. Ultimate controlling parameters for daily operations vary from facility to facility. Different plants use ferrous iron titrations on the autoclave discharge, free acid titrations on the autoclave discharge, eH of autoclave discharge or sulfur assays. Sulfur assays are the desired control parameter though possible the most difficult to obtain quickly and accurately. Circuit feed rate is adjusted based on the assays. Oxygen addition rate is determined by assays from the autoclave feed. The pounds per hour of oxygen fed to each autoclave is determined by the pound per hour of sulfide sulfur fed and a set oxygen utilization factor. Oxygen utilization is the amount oxygen consumed in the reaction. The design oxygen utilization ranges from 60% for whole ore facilities to more than 80% for concentrate autoclaves. Adjustments to the preset oxygen flow rates are made based on compartment temperatures or other control parameters. Generally if parameters are less than expected oxygen flow is increased. Oxygen distribution is also preset, generally most of the oxygen reports to the first half of the autoclave, since the amount of sulfur is greater \in the first half. The carbonate assays of the feed determine acid addition to the autoclave feed. The amount of acid addition is tempered against the autoclave discharge free acid titration. If free acid titrations are higher than desired, feed acid addition is reduced or stopped. There is a minimum desirable free acid level. This is maintained to insure an acidic, oxidizing environment inside the autoclave. Temperature and pressure usually remain at design levels. A decrease in either temperature or pressure usually reduces the rate of oxidation. Since most if not all facilities operate at near maximum temperature and pressure there is little upside by increasing either one. The use of eWemf readings in the neutralization circuit has been used as an online measure of oxidation at the Twin Creeks and Lone Tree Mines. The eWemf varies with ferrous iron levels. It is also an excellent cyanide consumption predictor. As the emf becomes more negative cyanide consumption increases. This is an indication of less than optimal oxidation. Throughput or any of the other controllable parameters are adjusted accordingly. Figure 11 illustrates the complete control strategy for a pressure oxidation circuit. Individually the control loops are simple. When all of the loops are put together control of the circuit becomes more difficult. As pressures in the heater vessels change so does the flows and heater levels. This affects the temperatures in the heaters and subsequently the autoclave temperature, pressure and level.

2206

CONTROL IMPROVEMENTS This list of desired control improvements is short. To improve pressure oxidation control would require the development of a few key instruments. In some cases the instruments exist only are not reliable, such as oxygen analyzers. With such instruments other improvements in pressure oxidation equipment or circuit configuration could be implemented with confidence of success. The single highest operating cost in a pressure oxidation plant is oxygen, followed closely by lime for neutralization. Therefore improvements in the measure and control of oxygen and acidity will lower the operating cost of a pressure oxidation plant. Oxygen control would involve both online sulfide sulfur analysis on the slurry feed and discharge streams, and gas analysis of the autoclave vent. Online sulfide sulfur analysis would provide the level of oxidation. Precise oxygen flow could then be added to the autoclave based on the oxidation. The vent gas analysis would be used as a trim to the oxygen addition and provide online oxygen utilization. Operating conditions such as individual compartment temperatures and autoclave pressure, along with oxygen distribution could be changed to improve performance. Reliable vent flow measurements would also be required to complete a gas balance. Online measurement of pH or eH inside the autoclave could be used to measure the rate of oxidation. The goal being minimizing acid generation without sacrificing the level of oxidation. An ore feed change or the addition of a neutralizing agent, such as limestone or trona, to the autoclave feed could be made to control the acidity of the autoclave without fear of over addition and quenching the reaction. With some or all of these control improvements a supervisory control system could be incorporated. An expert system would be the next step in pressure oxidation control. Such a system could currently be utilized, but could only control portions of the plant. Expert systems

2207

could be used control acid addition based on feed carbonate assays and free acid levels on the autoclave discharge. Such a system could minimize the acid used thereby reducing operating costs. An expert system on oxygen addition may be beneficial for the autoclave operation, but would have to also control the oxygen plant to realize full benefit. To simply adjust the oxygen addition rate to the autoclave without adjusting the oxygen production rate would serve no purpose. CONCLUSIONS Pressure oxidation control has advanced greatly through time. Though the control can be difficult it is not extraordinarily complex. For the most part all pressure oxidation facilities utilize similar control strategies and equipment. Therefore through the innovation of a few key instruments and control strategies the operation of all pressure oxidation facilities would be easier and less expensive. ACKNOWLEDGEMENTS The authors wish to acknowledge all those responsible for the successll operation of the pressure oxidation circuits throughout the world. We also would like to thank the management of Lone Tree Mine, Twin Creeks Mine, and Newmont Mining Corporation for their support and allowing presenting this paper.

2208

Development of a Mineral Processing Flowsheet - Case History, Batu Hijau Tom de Mull, Stuart Saich and Karen Sobel

ABSTRACT The Batu Hijau project was the largest mining project executed up to the time of completion. The concentrator was the largest ever built from “grassroots”. Initial ore discovery was made in May 1990 followed by several years of work investigating the geology and mineralogy, which culminated in the preparation of a final feasibility study in July 1996. An EPC contract was awarded to Fluor Daniel for the basic design, detailed engineering and development of the project in August 1996. During the detailed engineering phase the original design basis was thoroughly scrutinised and updated as a result of further flotation testwork carried out using sea water. This testwork allowed for a second review of the flotation kinetics leading to significant mid stream changes in circuit design which were incorporated into the final installation at a net cost savings. A flotation circuit model was developed using pilot plant data obtained from testwork carried out by AMMTEC in Perth, Australia and subsequently used as the basis for flotation circuit mass balances and equipment sizing. This flotation model together with a well defined process design criteria were used to size flotation circuit equipment and associated hydraulic systems. The initial ramp up rate of the concentrator achieved the aggressive targets set by the operations personnel. Analysis of early operating data during startup indicated that overall plant performance was well within original design expectations, but that internal circulating loads where greater than expected. Debottlenecking studies were subsequently carried out which confirmed visual observation as to certain equipment modifications that could be made to enhance recoveries. The flotation kinetic model has been updated to reflect actual operating conditions, and to develop a greater understanding of scale up issues for the large flotation circuit equipment involved. INTRODUCTION The $1.83 billion Batu Hijau project is located in Sumbawa, Indonesia and is owned and operated by PT Newmont Nusa Tenggara (F’TNNT). FTNNT is an Indonesian company owned 80 percent by a partnership between Newmont Mining Corporation (Newmont) and Sumitomo Corporation. A local mining company, PT. Pukuafu Indah, holds the remaining 20 percent. Fluor Corporation (Fluor) was responsible for engineering, procurement and construction of Batu Hijau, the world’s largest greenfield startup mining project ever constructed. The copper concentrator is designed to treat 120,000 tonnes of ore per day. The open-pit mine uses electric shovels and haul truck to transport the ore to the primary crushers. A 5.6-kilometer long overland conveyor carries the ore to the concentrator, which consists of a coarse ore stockpile, two-train SAG and ball mills, primary and scavenger flotation cells, vertical regrind mills, cleaning flotation cells and counter-current decantation thickeners to wash the sea water from the concentrate. The thickened concentrate slurry is stored in two tanks, then pumped to the port site where the concentrate is filtered and stored for shipment. Tailings produced by the concentrator flow by gravity from the process plant to the ocean, where they are disposed of via submarine tailings placement. The tailings are deposited three kilometers from the coast, at a depth of approximately 108 meters below the surface. The tailings migrate towards the Java Trench and are ultimately deposited at depths of several thousands meters. Early metallurgical test work on Batu Hijau ore was carried out at Newmont’s research facilities. This involved extensive metallurgical testing involving impact crushing, grindability tests, abrasion, flotation, filtration, flocculation etc to enable the definition of initial process

221 1

configuration. Newmont also enlisted the services of Lakefield Research (Ontario, Canada), and other testing facilities, during this phase to support their effort. Fluor first became involved in the Batu Hijau project in 1994 to carry out the Optimization Study. The purpose of this work was to optimize major aspects of the project such as plant throughput, process plant location, mode of concentrate transport and location of the port facilities. In 1995, Newmont and Fluor began work on the feasibility study. Core samples from the ore body were initially sent to Lakefield for preliminary laboratory locked cycle testing. Results of this work were used to define a proposed flotation circuit which was subsequently tested using composite samples of the expected mine plan feed in a batch pilot plant. This work was completed in late 1995, with an addendum issued in July 1996. The main deliverables from this stage of work were the definition of a proposed flotation circuit configuration and preliminary equipment sizing for the final feasibility study report. The feasibility report was issued in February 1996. This paper focuses on development of the process from this point forward.

PILOT PLANT TESTWORK During the preliminary design phase the need for a source of fresh water, suitable tailings location and subsequent water recovery was investigated. Due to the adverse seismic conditions in the region an alternative source of water (sea water) was recommended, along with the use of sub-sea tailings disposal. The proposal to use sea water in the circuit prompted the need to carry out further pilot plant testwork on suitable composites to confirm flotation kinetics and overall recovery. This was carried out at AMMTEC in Perth, Australia during the early part of 1996. The pilot plant used was set up as a scaled version of the proposed circuit developed during the feasibility study. Results from the revised flotation testwork using sea water instead of fresh water were made available to the design team in late 1996. Samples of feed, concentrate and tailings material from the AMMTEC sea water pilot plant testwork were sent to the Council for Scientific Research Organisation (CSIRO) in Australia for Scanning Electron Microscopy (SEM) analysis. Results from this work proved invaluable in interpreting the results of the flotation testwork. RAW DATA ANALYSIS Several pilot plant testwork campaigns were carried out at AMMTEC, using sea water and further composite samples. This work resulted in the generation of a significant amount of raw data that required further analysis before valid conclusions could be reached. Of the various tesc campaigns “Trial 12” was run with the desired intent to provide metallurgical information for scale up Raw data obtained from ‘Trial 12’ included the following typical metallurgical information. Solids flow rates. Pulp densities for each stream. Particle size distributions for individual streams. Metallurgical assays (Copper, Gold, Sulphur and Iron) for individual streams and as a function of particle size within each stream. To facilitate meaningful conclusions, analysis of the raw data was carried out using the following methodology: Mass balancing of raw data using SysCADMassBal. Visual data smoothing using EXCEL. Generation of recovery curves for both particles and individual elements within each stream using EXCEL.

Mass Balancing As is typical for metallurgical sampling the data obtained from the Trial 12 test runs did not balance, eg. total mass of copper reporting to concentrate and tailings streams did not equal total 2212

mass of copper in feed stream. Some form of statistical mass balancing of raw data is required prior to subsequent analysis or modelling. In order to achieve a suitable mass balance for solids, liquids, particle sizes and individual elements a statistical mass balancing package, SysCADMassbal was used. With this tool a reasonable mass balance of particle sizes across multiple unit operations and grades within each size fraction was developed. The raw metallurgical data for each process stream (i.e. solids flow rate, particle size distribution and assays) were used together with typical sample standard deviations to statistically manipulate the raw data into a mathematically balanced data set. This ‘massaged stream data’, was subsequently used in the data smoothing exercise, detailed below, prior to use in process modeling. The use of unbalanced data, or only using feed and concentrate stream data to develop a process mass balance can easily lead to incorrect conclusions.

Visual Data Smoothing After mass balancing the raw data it became apparent that a discontinuity existed in the particle size and grade around the 38-micron size fraction. In reviewing the methodology used to analyze particle sizes and prepare sufficient material for subsequent assaying, it was revealed that the samples were wet-sieved at 38 microns and that all material finer than 38 microns was subsequently sized in a “cyclosizer.” This equipment uses a series of very small cyclones to separate the minus 38 micron material into size fractions. Because the mineral specific gravity affects the split, any gold or gold bearing material reported to coarser size fractions than their size warranted. This discontinuity was removed via visual data smoothing and further mass balancing using EXCEL.

Particle Size (microns)

-+

0

RawtestwakdalaFeed%bjmass

-Smoothed

RawtestwakdataCons%bjymass

-Smwtkddata Feed%bjmass -SmwtheddataTailssbty mass

Raw testwork dala Tails % bj mass

data Cons %by mass

2213

RESULTS OBTAINED FROM RAW DATA ANALYSIS Analysis of the mass balanced and smoothed data focused on the recovery of material to the concentrate streams as a function of particle size. This analysis was completed both globally across the overall circuit, and also for specific flotation banks within the circuit. By plotting the mass of material within each size fraction for both the feed and concentrate streams, the size range over which preferential flotation occurs could be identified. This method of analysis allowed for an ‘easy to visualize’ representation of the effect of either ‘over’ or ‘under grinding’ of both the primary grinding circuit product and the regrind circuit product. The main objective of the exercise was to minimise (and equalize) flotation losses in either the coarse or fine size fractions through optimisation of grinding design criteria. In addition to the above analysis an investigation into the optimum rougher concentrate polishing mill retention time was carried out. Results are indicated in the following sections: Optimum Primary Grind size Analysis of the material mass recovery to the concentrate stream for the overall flotation circuit was used to identify the optimum primary grind size. Figure 2 indicates the particle size distribution for both the feed and concentrate streams.

Figure 2 indicates that the flotation circuit does not effectively recover very coarse material (i.e. > 210 microns), and that recovery of particles finer than 38 microns is diminished. This is typical of flotation circuits, but of importance was the easily identifiable size range across which primary grinding should be carried out. In addition to the mass recovery distribution for individual particle sizes, the recovery of copper as a function of particle size was also examined. Figure 3 presents the copper distribution as a function of particle size for both the feed and concentrate streams.

2214

Figure 3

- Overall Copper distribution in Flotation circuit

The above figure once again highlights a similar trend in recovery as a function of particle size with a slightly narrower recovery range indicated.

Optimum Regrind Particle Size A similar analysis of particle size distributions for the feed and concentrate streams around the cleaner circuit was also carried out. Figure 4 highlights the results.

Interpretation of Figure 4 indicates that the optimum regrind size lies within the 38-145 micron size range. The relatively large amount of minus 38 micron material present should be avoided due to the high losses (low recovery) of material within this size fraction. The pilot plant was operated at the proposed regrind specification of 80.0 percent passing 25 microns, which resulted in the generation of a significant quantity of ultra fines material.

2215

One of the significant conclusions reached from the raw data analysis was the requirement to increase the regrind size from 80.0 percent passing 25 microns to 80.0 percent passing 80 microns. Significant cost savings were realized as the design team grinding experts were able to reduce the regrind milling installation to three mills from the original configuration of six. The recommendation to increase the regrind size was subsequently confirmed in further locked cycle flotation tests.

Rougher Concentrate Grinding Optimisation The concentrate collected from the first cells in the pilot plant ran at a grade just slightly lower than desired final grade. A review of the SEM photographs of this rougher concentrate identified that several large (>180 microns), high grade (>30.0% Cu) particles were present. The recommendation was made that this material be collected separately (split flotation circuit) and passed through a polishing mill to enable some minor particle breakage without over grinding the concentrates. In order to determine the effect of polishing on recovery a concentrate sample was subjected to several batch grind-flotation tests to optimise design basis. Figure 5 indicates the findings.

-No Regind -30

secs Regrind -45

sec Regrind -360

sec Regrind

The above figure indicates the improved kinetics as a result of polishing the product for up to 45 seconds but with longer grinding times kinetics drop off and losses occur as a result of overgrinding. The optimum polishing mill product size found was 80 percent passing 82 microns.

FLOTATION CIRCUIT MODELLING A flotation kinetic model was developed in EXCEL to reflect the proposed circuit design. The kinetic model used the following basic parameters: Simple first order rate equation. Rate kinetics for each significant copper bearing mineral. All non-valuable material lumped together as gangue. Mineralogy obtained from Scanning Electron Microscopy (SEM) work. Rate kinetics as a function of particle size developed for copper for each unit operation. Pulp densities from testwork used.

2216

Rate Equation Used The form of the rate constant applied within the flotation model was as follows:

Where Ri Ri(max) Ki “i”

-

=

Recovery of mineral “i” at time t minutes Maximum recovery of mineral “i“ at time t = infinity Rate constant for mineral “i” Each mineral and gangue modelled

The mass balanced and smoothed data obtained from the sea water pilot plant testwork was then used as the basis to tune relevant rate constants to equate to actual pilot plant performance. For the copper balance, recovery data as a function of particle size for each section of the circuit allowed for the development of flotation rate constants for each size fraction. By developing the rate constants on a size basis the effect of a change in regrind size criteria could be analysed. The rate constants for individual particle sizes were then summated to represent overall copper mineral flotation rate constants. Revised rate constants for the increased regrind size were used in the final model Rate constants for gangue and gold were tuned to achieve overall grades and recoveries as experienced in the pilot plant.

Minerals used in flotation model The following minerals or elements were included within the flotation model: Chalcopyrite Bornite Covellite Gold Gangue Results from Scanning Electron Microscope (SEM) analysis were used to define the mineralogy for the original model. The ratio between valuable copper bearing minerals was changed to suit expected mine plans for final modelling but the rate constants for individual minerals assumed to be the same.

Pulp Densities Results from both the Lakefield and AMMTEC pilot plant testwork indicated that the pulp densities for scavenger concentrate streams were of the order of 510% solids by mass. Review of similar operations and experience of project consultants (Dr R Klipmel) indicated that operating pulp densities should be higher at around 10.0- 15.0 % solids by mass. Lower pilot plant pulp densities were expected as a result of the objective to maximise recovery at the expense of grade. However when the recommended pulp densities were used within the flotation model, poor copper recoveries as a result of short pulp residence times indicated the need to re-evaluate sizing of the cleaner flotation circuit. Flotation Modelling Results The kinetic flotation model as based on the sea water pilot plant testwork was then used to predict performance of the proposed flotation circuit that was, at that time, in the detailed design phase. Conclusions from the flotation modelling were as follows: The volume of scavenger concentrate, at the relatively low pulp density, was significantly higher than expected. Pilot plant pulp densities of the order of 2 4 % were achieved, whereas the simulated value of 12.0-14.0 % still resulted in excessive volume of scavenger concentrate. The original regrind size specified for the scavenger concentrate would result in excessive losses in the fine size fraction. 2217

Overgrinding of concentrates should be avoided due to high losses. An optimum rougher concentrate polishing mill retention was developed. Based on these findings several modifications to the original flotation circuit configuration were recommended and subsequently tested using the flotation model. These were as follows: Install a dewatering circuit on the scavenger concentrate stream. Install regrind screens (as used in the iron ore industry) to avoid overgrinding of high density, fine particles that would report to cyclone underflows. Resize cleaner flotation circuit to suit revised mass flows. Reduce the number of regrind mills from six to three as a result of reduced regrind duty.

PROCESS DESIGN Process Design Criteria The basic design criteria for the Batu Hijau project was as follows:

Nominal throughput Nominal throughput Plant availability Plant surge factor Average plant throughput Maximum plant throughput

43,800,000 120,000 92.0 +I- 15.0 5,435 6,250

tpa tpd % %

tph tph

In addition to the above listed mass throughputs, a range of expected feed grades (from the mine plan) was imposed upon the system to enable calculation of concentrate production rates and grade. The flotation model developed and based on the AMMTEC “Trial 12” data was updated to reflect the proposed dewatering circuit, the proposed circuit configuration and the process design criteria to verify equipment selection.

Process Engineering At the same time of completion of the flotation modelling development the final vendor bids for flotation equipment were received. This proved opportune as the proposed vendor equipment sizes were not exactly the same as the equipment listed in the feasibility study. Typical issues identified were such as one vendor offering six rows of 100.0m3 Rougher scavenger cells versus a competitor offering five rows of 127.0 m3 cells for the same duty. Once the preferred vendor (commercial and mechanical) was selected the relevant equipment sizes were entered into the flotation model and equipment selection optimised. At this time a detailed review of th? pr.oposed flotation circuit configuration and equipment sizing was carried out by owner representatives, engineering company personnel and outside consultants (Dr R Klimpel). During this phase the recommended operating parameters for the dewatering circuit were set within the flotation model such that the regrind circuit would be fed with a constant pulp density independent of rougher scavenger circuit operation. The flotation model was then tested using the ranges specified in the design criteria and overall recovery and grade recorded. Then number and configuration of rougher/scavenger flotation cells was fixed but he configuration and number of cleaner cells varied to optimise operating flexibility and maximise recovery. The following table lists the feasibility study circuit flotation equipment versus final selected equipment.

2218

Rougher cells Scavenger cells First cleaner cells Cleaner scavenger cells Second cleaner cells Final cleaner cells

Rows # 6 6 1 1 1 1

Cells/row 2 8 4 4 5 3

Cell Size m3 100 100 100 100 20 20

Total volume m3 1200 4800 400 400 100 60

Rougher cells 5 1 127 635 Scavenger cells 127 5715 5 9 47 5 i7n 1 A First cleaner .. - cells .- . . _.Cleaner scavenger cells 1 4 42.5 170 Second cleaner cells 1 10 17.0 170 Final cleaner cells' 1 4 14.5 58 * The vendor recommended a final installation of five flotation cells due to insufficient concentrate collection lip length. ._.I

~

Once the final circuit sizing and configuration had been completed the flotation model was run at three design tonnages and associated grades to generate expected mass balances for subsequent slurry material handling equipment. Slurry pump and sump design and sizing were now redone using the new mass balance.

CIRCUIT PERFORMANCE AND DEBOTTLENECKING Startup of the Batu Hijau copper concentrator and subsequent ramp up to design tonnage went extremely well. Within ten days of starting the second primary grinding circuit the hourly average mill throughput had achieved the design 5,435 t/hr. The design maximum throughput of 6,250 t/hr was achieved four weeks later. Performance testing was included within the EPC contract. In order to verify the successful completion of each required performance test, operating data from key process points was recorded into daily data sheets and reconciled against production reports. These were then collated over time to verify successful completion of each performance test. In addition to being a useful method of capturing successful performance, the data was also used to verify the original design basis and update the flotation kinetic model to identify any differences in original design expectations versus actual plant performance. Detailed analysis of the operating data, flotation model updates and equipment sizing has been carried out in the form of de-bottlenecking studies. Three of these have been completed out to date. The following are some of the major issues highlighted as a result of reviewing operating data and updating the flotation model to actual operating conditions. Initial full scale plant flotation kinetics were significantly lower than pilot plant results. Rougher and scavenger concentrate pulp densities were significantly higher than expected. Circulating load around cleaner circuit higher than expected. The debottlenecking studies have been used to assist operations personnel in confirming site conclusions and making changes to the circuits to enhance overall performance. Further details as to findings are as follows:

Flotation Kinetics The initial ore fed to the concentrator consisted of slightly oxidised material, which resulted in a reduced pH of pulp reporting to the rougher-scavenger flotation circuit. Primary lime addition was re-directed from the regrind circuits to the primary grinding circuits to increase pH and improve

2219

flotation performance. The types of frother and collector, and addition points were also changed during initial startup to optimize flotation recovery. The first two cells in each row of rougher-scavenger cells were designed such that concentrate could be collected from both of them, and sent to the polishing mill. The expected operating procedure was that only concentrate from the first cell would normally be sent to the polishing mill. On original startup however, it was found that the collection rate of concentrate from the first cells in each row was significantly lower than expected. The grade was slightly higher than expected but due to the low flow rate, concentrate was collected from the first two cells in each row of cells and sent to the polishing mill. Discussions with the flotation cell vendor and review of similar operations lead to the conclusion that internal concentrate collection launders were required. Once installed the kinetics improved significantly, and were comparable to original pilot plant testwork. One of the key parameters not investigated within the flotation model was lip loading factors. Back calculation of expected lip loading from the pilot plant testwork indicated a requirement of approximately 1.05 t/m/hr concentrate whereas the physical maximum obtained was 0.55 t/m/hr. Figure 6 on the following page indicates the original rate constants used for design and typical operating values.

-AMTEC

Trial 1 2 Data.

k = 0.8

:4- . May 2000 Avsrage. k = 0.19

-Onginel --+-July

Design Value. k = 0.442 2001 Aue k= 0.393

Apnl2000Avsraga. k -Novsmbsr2000

I

0.179

w e . h = 0.374

Operation of the rougher-scavenger cells improved back towards original design expectations once the internal launders were installed in late 2000 and early 2002.

Rougher-Scavenger Concentrate Pulp Densities The initial expectation of the rougher concentrate pulp density as developed from the pilot plant testwork was 20.0 to 25.0 percent solids by mass. Once the plant was started and began to operate at design tonnage it quickly became apparent that the actual pulp densities were significantly higher. These have been found to run as high as 45.0 percent solids by mass and at the expected copper grade. Of note is that the flotation feed pulp density ranges from 32.0 to 38.0 percent solids by mass. Cleaner Circuit Feed And Circulating Loads A dewatering cone settler was installed to enable the recycle of excess water back to the scavenger circuit due to the expected low concentrate pulp densities from the scavenger flotation cells. Whilst this primary function has become of lesser importance due to higher than expected

2220

concentrate pulp densities, the cone settler served another critical function. This is in the decoupling of operation of the cleaner circuit from the rougher-scavenger cells as a result of the large residence time of the settler itself. Surges from the primary grinding and scavenger flotation circuits do not affect the downstream processes. The feed rate and pulp density of material to the cleaner circuit was thus very stable and assisted operators in not having a continuously fluctuating feed to this part of the circuit. Initially the recirculating load around the first cleaner and first cleaner scavenger flotation cells was higher than expected. This was found to be due to the reduced flotation rate within the rougherhcavenger circuit with higher scavenger concentrates reporting to the first cleaner cells. Once the internal launders in the rougher cells were installed and the load shifted back to rougher cells and subsequently the second cleaner cells the circulating load improved significantly.

Regrind Screens and Cyclones The regrind screens as installed became a significant maintenance effort due to continual blocking and scaling up as a result of the presence of lime. This resulted in a dilute screen product to the regrind mills. The regrind screens were replaced with cyclones (after startup) which have performed well. The required pulp density is now being fed to the regrind mills. Throughput Studies Three debottlenecking studies have been completed using hourly average operating data over a minimum of one month operating time frame. As a result of this a valuable database of plant performance has been used to evaluate flotation kinetics as a function of throughput. This was analysed on a shift basis and the model tuned to reflect overall performance. An ‘effective cleaning performance’ of the rougher cells has been developed using the ratio of rate constants for valuable mineral versus gangue. As the total SAG feed rate increases the grind becomes coarser and the flotation circuit residence time decreases. By plotting the ratio of rate constants vs SAG throughput an indication of plant performance can be gained. Figure 7 illustrates this.

Juv/Oct 2001 Data

+

Operation pria to internal launders in rwghers

2221

---

July/Ocl2W1 Data

CONCLUSIONS The successful startup of the Batu Hijau copper concentrator and subsequent ramp up to design tonnages was the result of extensive metallurgical studies into flotation performance coupled with simple methods of analysis. The evolution of the flotation circuit design in tandem with detailed engineering, and subsequent incorporation of resulting recommendations, resulted in the construction of a flexible and robust installation. The scale up factor from pilot plant to full scale installation was of the order of 20,000. Notwithstanding this large scale up factor, initial plant recoveries were within expectations and individual streams within 30.0 % of initial mass balances. Subsequent equipment and circuit changes have reduced this to approximately 10.0 %. By “closing the loop” the original kinetic parameters have been updated using actual plant data within the original flotation model and equipment performance factors defined. This has enabled the flotation model to become more accurate resulting in an excellent de-bottlenecking tool for future plant operations. ACKNOWLEDGEMENTS The authors wish to thank the management of PTNNT for their assistance and opportunity to present this paper. In addition a special note of mention of the invaluable input from Dr R Klimpel in guiding the team in the flotation circuit sizing and selection.

2222

Specification and Purchase of Equipment for Mineral Processing Plants Carey Hunker and Sam Maldonado

ABSTRACT Effective specification, bidding and purchase of equipment can have a significant impact on the capital cost and quality of new facilities. It is essential for the equipment purchaser to understand the competitive bid process and the types of commercial agreements which govern. The use of industry standards has become a predominant method of defining equipment design and manufacture. In specifying and selecting equipment suppliers from the global marketplace, the role of quality standards such as I S 0 9000 is increasingly important. THE COMPETITIVE BID PROCESS Considering that suppliers of equipment are frequently willing to provide quotations based on a brief verbal request, why go to the time and expense of preparing detailed specifications? The competitive bid is the reason. There is perhaps no greater tool in the competitive bidding process than a clearly written specification which clearly summarizes the purchaser’s requirements using generic, industry standard terminology. Unlike a telephone call, the specification provides sufficient detail to allow each bidder to understand all the criteria against which the bids will be compared. Use of a specification provides identical information to each bidder, promoting a fair and ethical bidding process. TYPES OF COMMERCIAL AGREEMENTS Two types of commercial agreements are most common in the purchase of equipment. Purchase Orders are applicable when equipment is installederected by the purchaser. Typical examples might include valves and pumps. Contracts are applicable when equipment is installederected by the supplier. Typical examples might include field-erected tanks and conveying systems. Other commercial arrangements may include rentals and loan arrangements (such as field trials of new prototypes). It is important for the purchaser to understand which commercial approach is appropriate and use appropriate terminology in the specification.

THE SPECIFICATION Specifications are intended to ultimately form part of a purchase order or contract. The specification contains the technical and engineering details that explain exactly what is being purchased as well as defining the responsibilities of each party to the purchase order (or contract). Types of Specifications Two distinct approaches are typically followed in preparation of specifications. The fabrication specification is applicable when the equipment is designed by the purchaser. Within this type of specification, the purchaser essentially requests a quotation on fabricating the purchaser’s design. The specification typically contains specifics on

2223

materials, welding, tolerances and coatings. A fabrication specification is typically accompanied by detailed drawings outlining the purchaser’s design. A typical example might include a chute or storage bin. The pe@ormance specijication (sometimes referred to as a duty specification) is applicable when equipment is designed by the supplier. With this type of specification, the expected performance of the equipment is specified, but the detailed design of the equipment is left to the supplier. In reality, there are numerous gray areas between these two approaches. For example, when specifying atmospheric tanks, it is common for the purchaser to dimensionally define the tanks, while the determination of wall thickness and location of weld seams is typically left to the tank fabricator. A challenge for the purchaser is therefore to determine what aspects of the equipment design are best performed by the purchaser, and what by the supplier. Luckily, for most common types of equipment, guidelines are typically provided through the use of industry standards.

Industry Standards Prior to the advent of industry standards, equipment specifications often described every component, nut and bolt of specified equipment in voluminous detail. Suppliers struggled to decide which standard model to offer in response to long, wordy and often unclear specifications. As a response, many industries introduced standards which not only standardized the design of equipment between manufacturers, but also clarified the relationship between the purchaser and the supplier. An industry standard typically classifies equipment into types, groups, classes, categories or services that are easily identified by the purchaser and recognized by the supplier, simplifying and clarifying the bid process significantly. The purchaser therefore need only identify the applicable industry standard, and provide the data recommended by the standard. An example of an organization preparing industry standards is the Crane Manufacturer’s Association of America (CMAA). A purchaser wishing to specify a “top running single girder electric overhead traveling crane” needs only to refer to CMAA Standard 74, which fully defines the design for this equipment. As part of this standard, a brief, three page “Crane Inquiry Data Sheet“ is provided which summarizes the data requested by CMAA manufacturers to provide a quotation. Within this standard, four classes of cranes are defined, from Class A (standby or infrequent service) to Class D (heavy service). Use of this standard greatly simplifies the purchaser’s job, minimizes the data needed to be communicated while allowing the purchaser to fully define the type of equipment requested in a manner that will be clearly understood by the supplier. Many industry standards in the United States now fall under the auspices of the American National Standards Institute. Founded in 1918, ANSI is a private, non-profit organization that administers and coordinates the U.S. voluntary standardizationand conformity assessment system. Similar organizations in other parts of the world include International Organization for Standardization (ISO) in Europe and the Japanese Industrial Standards Committee (JISC) in Japan. Further information on these organizations may be accessed at http://www.ansi.org (ANSI), http://www.iso.ch (ISO) or http://www.jisc.org (JISC).

Format While the format of specifications regrettably varies widely, the master specification format defined by the Construction Standards Institute (CSI) has gained widespread acceptance in the United States as a common format for specifications used for construction and procurement. Founded in 1948, the Alexandria, VA, based organization’s goal is to enhance communication by providing a common system of organizing and presenting construction information. CSI’s members include architects, engineers, constructors, specifiers of construction products, suppliers of construction products, building owners, and facilities managers.

2224

The CSI Manual of Practice, Section 11, provides techniques that are useful in preparing specifications. Further information on CSI formats can be accessed at http://www.csinet.org. Some general guidelines for preparing specifications follow: Tenses: Use the imperative mood, the command form, to explain actions that the reader must perform, or is responsible for. In this style of writing, the subject, "you" is understood, and the verb is at or near the beginning of the sentence or the main idea, e.g., "Grind all welds to 1/8" and clean thoroughly". Use the indicative mood to define, identify, or describe. For these functions, the verb, shall be, is like the equals sign in an equation, and the verb links the subject of the sentence to the discussion of it, e.g. "The completed weldments shall be free of defects and slag inclusion." Wording: The purchaser should select and use words carefully. Each should be used in its precise meaning. Once a word and its meaning are selected for use, the same word should be used throughout the specification whenever that particular meaning is intended. Spelling used in specifications should be consistent. Whatever standard is established the most important thing is to be consistent throughout the documents that are being prepared, Some common problem areas: Insure, Assure, and Ensure.

To insure is to issue or procure an insurance policy. Assure is to give confidence to or convince a person of something. Ensure is to make certain in a way that eliminates the possibility of error. Shall and Will. Shall is used with reference to the work required to be done by a contractor. Will is used in connection with acts and actions required of the owner or the architect / engineer. The words "must" and "is to" should be avoided. Install, Furnish, and Provide. Install means to place in position for service or use. Furnish means to provide or supply, and provide means to furnish, supply, or make available. Definite article "the" and indefinite articles "a" and "an" need not be used in most instances. Poor: Apply an oil paint with a brush to the wall. Correct: Apply oil paint with brush to walls. All. The use of the word all is usually unnecessary. Poor: Store all millwork under shelter. Correct: Store millwork under shelter. Contractor. Avoid using Contractor as the subject of the sentence. Poor: Contractor shall lay brick in common bond. Correct: Brick shall be laid in common bond. Preferred: Lay brick in common bond.

Content As discussed earlier, specifications are intended to ultimately form part of a purchase order or contract. The typical specification therefore defines both the responsibilities of each party,

2225

explains exactly what is being purchased, and provides all relevant technical and engineering details. A typical specification may contain the following: Scope of supply: a summary of what is to be supplied, typically clearly defining the purchaser’s and supplier’s responsibilities. References: a summary of applicable industry standards and codes. Performance criteria: For performance specs, a description of the specified performance. This is often provided in tabular fashion on data sheets. Materials: Materials of construction. Tolerances: Tolerances of dimensions or performance. Design life: The required life of the equipment, or specific components. Dimensional parameters: Physical size of equipment. This is often completely defined in fabrication specification by the purchaser’s drawing or sketch. Protective Coating: Requirements for painting or protective coatings. Noise Levels: Acceptable noise limits. Site Conditions: A summary of climate, seismic zone, geographic location and environmental conditions. Utilities Available: Utilities may include steam, air, water. Power: Voltages and power availability. Spare Parts: Requirements for capital and maintenance spares. Welding: Specifications for welding, particularly with fabrication specifications. Acceptable Sub-Component Manufacturers: The purchaser may only accept subcomponents from a specific supplier. Quality Requirements: Purchaser’s requirements for quality control, non-destructive examination and testing. Data Requirements: Purchaser’s information requirements, both for bid and after purchase, including timeframes. Special requirements for package systems, which may include emissions limits, modularization and requirements for hazardous operation reviews.

LOCATING AND QUALIFYING BIDDERS Bidder Selection Mineral processing is one of the most geographically diverse industries in the world. Locating quality suppliers and contractors in some locales can be challenging. In some countries comprehensive trade directories are available to help locate suppliers of specific types of equipment. Examples of these include: The Thomas Register (http://www.thomasregister.com/) offers both North American and European Directories. The Fraser’s Trade Directory (http://www.frasers.com/) provides a listing of Canadian suppliers. The South African Bureau of Mines provides a listing of suppliers at (http://www.bullion.org.za/bulza/chaor~wmdir/wmdsrvc 1.htm#equip)

Bidder Qualification Ensuring all bidders are qualified prior to commencing the bid process will benefit both the purchaser and supplier. For this reason, bidders are usually pre-qualified based on a selection process that may use any of the following criteria as a basis: A demonstrable track record, experience and references

2226

Degree of shop/factory loading Stability, financial strength Geographic location Innovation, technology Maintainability (availability of spare parts and service) Reliability and complexity Safety record/experience modifier Quality assurances and/or IS0 9001 compliance Pre-manufactured equipment or stockpiled materials Qualifications of tradesmen (machining, welding, electrical)

THE BID PROCESS Ethical Bidding The need to maintain a legal and ethical bid process needs no justification. Suppliers can expend considerable money and effort in preparing bids. Significant effort should be made to ensure that bidders are afforded an equal opportunity to compete on the same terms as their competitors. In bids involving significant sums of money, sealed bids are often received, and opened only after all are received in a witnessed environment. Some aspects of an ethical bid process: Technical discussions, bid clarification responses and meetings, and discussions of alternative proposals conducted in a manner fair to all bidders. Bid information considered confidential as it may contain proprietary information. Distribution limited to those performing evaluation. No indication given to a supplier regarding the competitive nature of his bid during the evaluation period or negotiations with bidders. The following practices are not considered part of an ethical bid process Bid shopping - playing one bidder against another as a negotiation strategy. Requesting quotations without intent to buy, without advising the supplier that the request is for information purposes only and only for the purpose of providing leverage against a competing firm. Threatening suppliers with loss of future business unless they comply with unreasonable requests. Requesting suppliers to bid on a larger volume of business in order to secure lower pricing for a known lesser quantity. Deliberately introducing confusion into the negotiations with unusual issues, terms or false figures.

Evaluation of Bids A Bid Evaluation typically includes two components, a technical and commercial evaluation. The technical evaluation may include an evaluation and comparison of: Conformance/exceptionsto specifications Performance and efficiency Safety and operability Utility requirements Materials of construction Lifecycle Physical size

2227

Supplier’s quality plan The commercial evaluation may include and evaluation and comparison of Approval drawing cycle Price adjustment (escalation) Service representative cost Spare parts cost Storage Costs Return conditions & restocking charges Taxes Terms of payment, discounts, early payment, progress payments, incentive, and penalty Customs clearance costs Export packing charge and stowge Freight (inland, air, ocean, & rail) Import duties Insurance costs Warranty & performance guarantee Export credit financing Nearest supplier service Shipping point Liquidated damages acceptance Price basis Currency and exchange rates Schedule Supplier’s shopload Previous experience with supplier Governing items and conditions of purchase Bid validity period Union labels Cancellation charges

ENSURING QUALITY The topic of quality control and quality assurance is too broad to be addressed fully here. However it should be noted that a specification does need to define the purchaser’s expectation of the quality of the finished product. In many cases, but not always, industry standards can assist in defining quality requirements. Where this is not the case, other standards may be utilized to define to define one aspect of the manufacture. For example it can be specified that all welds must meet the requirements of a specific standard of the American Welding Society. The standards of the International Organization for Standardization (ISO), in particular I S 0 9001 (Quality management systems - Requirements) are gaining in popularity worldwide. Further information on the I S 0 9000 series of quality standards is available at http://www.iso.ch/iso/en/iso900014000/iso9000/selection~use/iso9000family.htrnl. Whether based on I S 0 9001 or not, it is common practice for the purchaser to review and ensure the supplier’s quality plan is adequate, preferably prior to award. It is also common practice for the purchaser to visit or inspect the purchaser’s shop before or during the fabrication of the equipment. The purchaser may wish to impose “hold points” in the fabrication process, where fabrication must cease until the purchaser is satisfied that all quality requirements have been met. If this is the purchaser’s intent, it should be identified clearly in the specification so the supplier’s schedule can incorporate these work stoppages

2228

OTHER CONSIDERATIONS Used Equipment When used equipment is desired, it is important to remember that the supplier’s technical resources may not be available to assist in selection. It is not uncommon for used equipment to be sold by clearing houses and agents who may have little technical knowledge of the equipment itself. For this reason, the purchaser may wish to specify a specific size or model number, rather than a performance requirement. Where possible, involving the original manufacturer will likely prove highly valuable in determine if equipment can be re-used and the degree of refurbishment required. Languagdnternational Considerations. Due to the geographically diverse nature of the mineral processing industry, equipment may be procured from any corner of the globe. When working in any region, the purchaser must be prepared to work in the local language, and in accordance with local laws and customs, which may vary widely. Impact of the Internet. The advent of the internet has changed, and will no doubt continue to change the manner in which commercial transactions occur throughout the world. At the time of this writing, the mineral processing industry awaits the debut of Quadrem, http:\\www.quadrem.com. Founded by 14 companies (Alcan Aluminium Limited, Alcoa Inc., Anglo American plc, Barrick Gold Corp, The Broken Hill Proprietary Company Limited (BHP), Corporacion Nacional del Cobre de Chile (CODELCO), Companhia Vale do Rio Doce (CVRD), De Beers Consolidated Mines Ltd., Inco Ltd., Newmont Mining Corporation, Noranda Inc., Phelps Dodge Corporation, Rio Tinto, and WMC Limited) the venture intends to create a platform to bring together mining, minerals, and metals producers and suppliers in more than 100 countries utilizing a common catalogue of products in multiple languages. The intent is to allow participants, regardless of size and location, to access and trade with a large pool of suppliers both locally and around the world.

2229

The Management and Control of Costs of Capital Mineral Processing Plants David W. Stewart

ABSTRACT The management and control of costs of capital mineral processing plants is dependent on the actions and understanding of all those involved in the development and execution of the project. The following process describes how project costs are monitored against an established budget and how the process varies through the different phases of the project from basic engineering, through detailed engineering, procurement and construction.

INTRODUCTION The basics of cost management and cost control are straightforward. They are: Develop the budget Monitor against the budget Report deviations early for corrective action Revise the budgevforecast (as appropriate) Monitor against revised budgevforecast Repeat the cycle. This cost control cycle is illustrated in Figure 1.

Figure 1 - The Cost Control Cycle

2230

a

For cost management and cost control to be effective this cycle must: 0

Occur continuously Bediligent Involve all project team members Result in all project team members being kept informed.

PROJECT SCOPE A clear and concise definition of the project scope is an essential ingredient in establishing and controlling the costs of a project. All participants in the project should be kept fully informed on the defined project scope. The project owner must make clear what his expectations are. The physical boundaries of the project should be defined and should address: Infrastructure - such as access roads, power supply, communications facilities, housing, port facilities and air strips

Mine facilities - such as mine development, maintenance and offices Process facilities - such as crushing, grinding, flotation, etc. Operating materials - spare parts, initial fills of reagents and initial grinding mill ball charge. The operating requirements of the plant should address: Plant throughput Plant operating and maintenance schedules Plant reliability expectations Manual versus automatic control requirements. Each party with project budget responsibility must have a defined scope on which to base their budget development. Any deviations from that scope would then result in a related change in budget, making it important that the budget based scope definition be as precise as possible to minimize any cost “surprises” in the future.

COST ESTIMATE After successful completion of a feasibility study, basic engineering typically ensues to refine and confirm the project scope. At the end of basic engineering, a preliminary estimate is prepared. As the project develops, estimates of the total cost of the project are updated. The project scope details improve as the project progresses, and the further the development of the project, the greater should be the accuracy of the cost estimate. Cost estimates can be categorized by the accuracy of the estimate and may be referred to as order-of-magnitude, preliminary and definitive as illustrated in Figure 2. These terms for classes of estimates are sometimes used during the feasibility study phase also. The order-of-magnitude estimate being very early in the study with minimal engineering completed and the preliminary estimate after the study has been completed.

2231

30

Order of Magnitude

Preliminary

Definitive

E

1 20

&

0

30 0 4-%-~otto

scale + b -

10 20 30 40 50 60 70 80 90 100

ENGINEERING PERCENT COMPLETE

Figure 2 - Classes of Estimates Cost estimates/forecasts are a combination of direct costs and indirect costs. Direct costs (equipment, material and labor) are generally specifically priced, whereas indirect costs tend to be a function of the project duration and schedule. For this reason and to estimate escalation costs, it is essential that any cost estimate or forecast is based on a carefully prepared project schedule. Cost estimates are generally assembled by assessing the costs o f : Process equipment (crushers, mills, pumps, etc.) Construction materials (concrete, steel, pipe, cable, etc.) Construction labor Temporary construction facilities (offices, warehousing, power generation, etc.) Temporary construction services (survey, inspection, safety, quality control, etc.) Engineering (design), procurement and construction supervision services Design and growth allowances Contingency and escalation Owner’s costs (these may include the cost of the owner’s organization, land acquisition, interest costs during construction.) To develop these estimates of costs, each of these components have to be quantified and pricing determined. Quantifying these components depends largely on progress of the design. Sizes and quantities of pieces of equipment will normally be obtained off the process flow diagrams or the equipment list. Quantities of bulk materials (concrete, steel, cable, etc.) will be extracted from the 3-D design model or taken off detailed drawings as those are developed or, prior to modelldrawing development, can be based on historical ratios. Services will generally be based on staffing plans and expected hours to perform those services. The pricing of materials is obtained by getting quotes from manufacturers for equipment and materials. Labor and services pricing is obtained by getting quotations from contractors who will perform the work or provide the services.

2232

If specific current data is not available, such as during early stage feasibility studies, all pricing data can be developed from historical data. This data would have to be adjusted for escalation, location and any other special circumstances. Even if current data is used, comparison with historical data is helpful in verifying the validity of the data. A design (quantity) allowance is an amount included in the estimate/forecast which is based on the degree of design completion and a comparison of the designed quantities with historical experience. The design allowance covers quantities we know from experience are not included in the designed quantity at a particular point in the engineering phase. A growth (cost) allowance is an amount included in the estimate/forecast which we know from experience is needed to cover adjustments within the defined scope that cause increases in the price of quoted or awarded materials. As the project progresses and scope and pricing are more refined, these allowances will be reduced. Contingency is an amount of money included in a budget, estimate or forecast for costs, which based on past experience, are likely to be encountered but are difficult or impossible to identify at the time the estimate or forecast is prepared. Contingency is intended to cover estimate errors and omissions, design developments, pricing variations, and the like. It is not intended to cover changes in scope, force majeure events, labor strikes, etc. There is often a tendency on the owner’s side to view contingency as available for his use to cover additional scope and owner cost over-runs. This is not the intent of contingency.

ESTABLISHING A BASELINE BUDGET Early in the project’s development, a baseline budget must be established, if costs are to be controlled effectively. The initial cost estimate for the project becomes the project’s initial budget and is the baseline against which all project costs are monitored and reported. The preparation of the project budget consists of reformatting the approved cost estimate into significant and readily identifiable components, such as purchase orders, contracts, and work packages which can be monitored in accordance with the planned and actual progression of work. In preparing the budget, the estimate details will be expanded or compressed to assure the optimal alignment for monitoring and reporting. The budget data is organized to conform to project requirements regarding reporting format, code of accounts and the project-specific conditions. As the project progresses, budget details will be adjusted, as necessary, to conform to changes in the project implementation plan or the code of accounts. This is accomplished through a system of formal budget adjustments. The cost control group maintains a log of budget adjustments, budget transfers between cost accounts, and scope changes. A continuous audit trail of the changes in budget should be maintained at all times. A series of tabulations and trend lines (curves and graphs) are developed during the budget process. These include quantities, job-hours, and costs for such items as plant equipment, bulk materials, installation labor, contracts, indirect accounts, services and construction equipment, at both summary and detail levels. These will be used during the monitoring process to comparatively depict actual quantities, job-hours, and costs and expenditures against the planned timeline. These tabulations and trend lines are further updated at each formal cost forecast development.

CODE OF ACCOUNTS Controlling costs requires a constant comparison of the three major categories of project costs. These categories are: Committed costs Paidcosts Project budget. To facilitate this process, the code of accounts for the project is established, so that these three categories can be collected into their respective code of accounts “buckets”. In simple terms,

2233

monitoring the project costs involves comparing the three categories in each of the related “buckets”. The code of accounts provides the framework for identifying the project’s physical facilities and for categorizing the quantities, job-hours and costs for plant equipment and materials, installation and services. It is a standardized system for defining, recording, monitoring, reporting and auditing cost information. The code normally consists of one code for facility/sub-facility (usually numeric) for direct cost accounts, distributable cost accounts and EPCM services accounts and another code for commodities (often an alpha code) in the direct facilitykub-facility accounts. For example: 0310.RAD 0310 would indicate the facility, such as Grinding RAD would indicate 10”to 12” carbon steel pipe - R being pipe, A being carbon steel and D being 10” to 12”. The development and maintenance of the project code of accounts should be the responsibility of the cost control group. The code of accounts must be established and agreed to by the owner and the project team and issued immediately after project award to confirm the defined scope. The code of accounts is applied to labor, material, contracts and services, and is monitored throughout the project life both in the design offices and at the construction site.

COST TRENDING AND FORECASTING A trend is a deviation from the established baseline - budget or schedule. The trend program is a formalized process for identifying and evaluating deviations as early as possible, such that timely appropriate action can be taken by the project team to alleviate the impact of detrimental trends, or to exploit the benefits of improving trends. This program is most effective during the formative design and procurement stages of the project. The essence of cost trending is timeliness rather than precision, enabling the project team to make decisions early enough that it can influence the direction of project costs. This is illustrated in Figure 3.

Figure 3 - Cost / Schedule Influence Curve

2234

Trends are generally of the following types:

0

0 9

Additions or deletions to project scope Design development resulting from improved engineering definition Price variances resulting from bid quotations Changes in installation productivity Changes in work execution methodology Estimate errors and omissions Escalation, re-evaluation, etc.

The trending program is initiated at the inception of the project and should continue throughout the project period. Each member of the project team should be encouraged to initiate trends. These will then be analyzed for validity. For this reason, it is essential that all project team members have a clear understanding of the project scope and the budget. Trends are order-of-magnitude estimates prepared in accordance with the code of accounts and include direct, indirect, and engineeringhervices costs, contingency and schedule impact. A weekly trend meeting attended by key members of the project team is held for the purpose Of:

Validating identified potential trends Discussing pending trends that have been previously identified for status and action Reviewing trend estimates prepared during the week Reviewing the project status and latest project decisions for identification of new trends Discussing owner comments. On a regular basis (weekly) trend reports and approved trends are issued formally for information and decision-making. The “current forecast” of project costs is continuously updated by adding the approved trends to the previously approved forecast. At the start of the project a plan should be established for the frequency of formal forecasts of the total project cost. These should occur every 3-6 months throughout the life of the project. A cut-off date is defined for each forecast and on that date all available data is collected, quantified and priced. A forecast establishes the actual costs through the cut-off period and adds to it the expected costs of the remaining “to-go” work. The resulting figures are the expected cost at completion determined at that period in the project’s life.

DEFINITIVE ESTIMATE At a certain stage in the development of the project, the baseline budget should be updated to reflect firm project scope, finalized plant layout, specifications for major plant equipment and materials, layout and process design for the major plant systems (essentially design for major buildings and structures), and a firm construction plan and schedule. This is commonly known as the definitive estimate. At the time a definitive estimate is prepared, design engineering is ideally at least 40% complete (and no less than 30%),all major design decisions have been made, all major equipment has been committed or firm price quotations have been received, and construction has started, assuring that actual construction experience has been obtained as the basis for the cost of construction.

SCOPE CHANGE CONTROL The scope change monitoring program establishes a method for the timely assessment of the effect of changes in scope on project cost and schedule from an established base estimate and schedule. This program records and reports the overall effect of scope changes and provides a basis for obtaining owner agreement on contract scope amendments.

2235

Changes in scope include alterations to the project‘s technical scope and/or to the contractor’s scope of services as requested, directed, and/or authorized by the owner. The importance of control of the scope of the project and identification of deviations from the base defined scope cannot be over-emphasized. The identification of these scope changes and realistic pricing and impact on the project schedule in a timely manner are a key to good project cost control. As soon as a change in scope has been identified, a rough order-of-magnitude estimate is needed for the owner to decide whether or not to proceed with the change. It is the responsibility of the project team to ensure that scope changes to the project’s physical facilities and/or the contractor’s scope of services are identified, documented and processed. For the change control procedure to work effectively, it is important that all members of the project team be encouraged to identify anything that they believe to be a change in scope. This information must be transmitted to the cost control group who will then prepare a cost estimate for the defined change. The cost control group will prepare a preliminary evaluation of the change at the total project cost level and submit the estimate and schedule impact to the appropriate members of the project team and the project manager. After their review and approval, it will be forwarded to the owner for review and approval to proceed. After approval, the cost of the change will be included in the current budget (project budget plus approved scope changes). The status of all changes will be summarized monthly and will be submitted to the project team to keep them fully informed of the changes to the budget.

TRACKING QUANTITIES, HOURS AND COSTS An essential part of cost containment is tracking the components that have a direct impact on the project’s cost. Two key components are the quantities of materials to be installed and the engineering, construction and management service hours involved. To be effective these have to be tracked as soon as they are identified and compared with the budget figures. Ideally, as the detailed design progresses, quantities of bulk materials (concrete, steel, cable, etc.) that the design represents should be compared regularly with the associated quantities in the budget. Any deviations should be noted and discussed immediately with the project team. Similarly, as materials are installed, they should be quantified and compared with the “trended” budget quantities and, again, deviations from the baseline should be recorded and brought to the attention of the project team. A further control of quantities ideally would occur at the following stages: Quantity budgeted Quantity designed Quantity purchased Quantity shipped Quantity received at site Quantity distributed to the location for installation Quantity installed. The expenditure of service hours on design, construction and management, should be tracked against the planned weekly or monthly expenditure of the budgeted hours. Material costs are monitored via a budget allocation system. The engineering department (in the design office or at the construction site) prepares material requisitions (MR’s) which provide type, specification, quantity and disposition of material. The cost control group compares MR information to the budget and provides a cost code for each item and the associated budget amount. After bids are received and analyzed, the cost control group compares the pending commitment value with the budget amount, noting deviations. Deviations will be reviewed by the project team to determine if possible action is required to stay within the budget value. The deviations will be identified as “trends” and will be incorporated into the current cost forecast update.

2236

COMMITMENTS AND COSTS A commitment is a contractual obligation to pay another party for performing services or providing materials. Cost and commitment reporting is a monthly function that keeps the project informed of the commitment and paid cost status of the project. It enables a comparison by cost code of the following items:

0

baseline budget current budget (original budget plus approved scope changes) current forecast paid costs this month paid costs to-date commitments this month commitments to-date

Preparation of the cost and commitment report consists of compiling, into a single database, the commitments from the awarded purchase orders and contracts, the paid cost data from the accounting and labor ledgers, updating the current budget, and timely transmittal of the report to the project team. The current budget data is updated by budget adjustments each month to reflect approved scope changes and budget transfers. The current forecast is revised as trends are approved or when a new project forecast is developed and approved. Cost and commitment ledgers are maintained in native currencies and therefore separate ledgers are required for each currency. For project reporting, the commitments and costs should then be converted to a common project reporting currency. The reporting currency and the method for conversion from native currencies must be established and agreed to at the start of the project.

HISTORICAL DATA At the completion of the project, all final actual costs, quantities and hours should be assembled into a project historical report. This report should include a complete definition of what the costs cover and should include flow diagrams, key project drawings, major specifications and data sheets, the project schedule, etc. This data is then available for use as the basis for studies for plant expansions or similar new plants. In addition, it can be used as a check for estimates and pricing received from third parties when reviewing other potential projects. CONCLUSION The key elements to the management and control of costs of capital mineral processing plants are having a clearly defined scope, establishing a realistic baseline budget, monitoring commitments and costs against the baseline budget and identifying deviations early enough to take corrective action and updating the baseline budget as conditions change. This process must be adhered to consistently and frequently.

2237

Schedule Development and Schedule of Control of Mineral Processing Plants Parrnod Kurnar

ABSTRACT The schedule development and schedule control of mineral processing plants is dependent on all those involved in the development and execution of the project. The following process describes how project schedules for engineering, procurement, and construction are developed through the different phases of the project and what steps should be considered in managing the project schedule.

INTRODUCTION The basic philosophy of schedule development and schedule control is simple. It is: plan the work work the plan monitor the plan 0 report deviations as early as possible take corrective action 0 repeat the cycle For the schedule to be effective this cycle must be continuous and involve all project team members The schedules are developed starting with a milestone schedule and then progressing to more levels of detail. The levels of schedule concept is called schedule Hierarchy and this is described in the following section.

LEVEL OF SCHEDULES The schedule hierarchy is shown below:

2238

Milestone Schedule

A clear and concise definition of the project scope is essential for developing project schedules. All participants in the project should be kept fully informed as to what the scope of the projects is and the project owner must make clear what his expectations are. The physical boundaries of the project should define and address: infrastructure- such as access roads, power supply, water supply, communications facilities, port facilities and air strips, mine facilities - such as mine development, maintenance, offices process facilities. temporary facilities - camp requirements, temporary shops required for construction, etc.

-

Master Summary Schedule Level I

The Master Summary Schedule covers the entire scope of the project and is the basis for client progress reporting. It is used by the project team to evaluate summary progress against established milestones and sets the overall project schedule parameters in the development of more detailed schedules. Client interface activities are also shown on this schedule. A Level 1 schedule identifies the contractual milestones for the project and the key activities to be performed to ensure those milestones are achieved. When applicable, the planner ensures that “Freeze Design” dates are identified as key milestones for design documents including design criteria, process flow diagrams, general arrangements and the main single line diagram. These milestones will typically be carried down into more detailed lower level schedules. The schedule is prepared by the planning engineer and developed in a simple bar chart format. The upper section of the schedule identifies the major contractual and/or key project milestones; the lower section shows the major activities grouped by engineering, procurement and construction. The construction activities identify the physical plant facilities. The critical path of the project will be highlighted, so that it is readily visible. The Master Summary Schedule defines the time commitments between contractor and the client in terms of notice to proceed, receipt of design criteria, major equipment lead-time, and completion dates. The summary checklist (listed below) for schedule development is offered to assist the planner in the development of the project schedule client desired milestones: notice to proceed, permits issued, plant ready to accept feed major equipment /material lead times project scope schedule assumptions schedule restraints - environmental permits - major commitment restraints; full funding, etc. - site accessibility - relocation of utilities - availability of construction powedwater - construction camp - climatic conditions - weather, altitude - manpower availability - sources of permanent powerlwater - availability of client provided items: permits, permanent power, equipment

2239

Based on the above checklist, the applicable information is collected and the Master Summary Schedule is developed. At the minimum it should reflect notice to proceed, start engineering, award and delivery of major equipment, availability of power and water, start construction of major facilities, and plant ready to accept feed. In developing the schedule, the planner takes into account when engineering can be started for critical areas, what information is required to complete certain parts of engineering so the major equipment can be ordered, and when construction can start. Normally the critical path in the mineral processing plant is installation of the grinding mills and drives. The planner will verify when the engineering has sufficient information so the mills and drives can be procured and delivered. He will also establish when the foundation design can be completed based on mill vendor information available, so the construction can start. Simultaneously, it will be established when the building steel is required and when the design will be complete to support fabricating the building steel. Based on estimated quantities and job hours required to install the concrete and steel the installation durations are established with input from experienced construction personnel. Any other critical areas are also planned based on the similar logic. The planner will also review when the permanent power and water are required to start the plant.

-

Project Facility Summary (Intermediate) Schedule Level II The next level of project scheduling is called the Project Facility Summary Schedule. The work breakdown structure (WBS) provides the framework for identifying the project’s physical facilities and for categorizing the type of work. The WBS is established in co-ordination with the cost engineering group so that the code of accounts and WBS are in line with each other. WBS establishes the levels of detail, so that project can be depicted by area, facilities within the area, sub-facilities within major facilities and then by type of work. Type of work is defined by engineering, procurement and construction. For Example: Level 1 - Project Level 2 - Area (infrastructure, mine facilities, process facilities) Level 3 - Facilities (crushing, concentrator,port facilities) Level 4 - Sub-facilities (grinding, flotation, regrind) Level 5 -Type of work (engineering,procurement, construction Level 6 - Discipline (civil, architectural, mechanical, electrical) Level 7 - Commodity (earthwork, concrete, steel, mechanical equipment) This schedule identifies the total scope of each facility of the project, its major components and milestones. It establishes the fundamental logic for performing engineering, procurement and construction activities on the project. Dates for major design releases, commitments and deliveries, durations of plant and facility installations, mechanical completion, and commissioning are clearly identified. The Project Facility Summary Schedule serves as the main link between engineering/procurement activities and construction activities during the development of the detailed engineering/procurement schedules. The construction activities at this level drive the engineering and procurement schedules. This schedule is the basis for estabilshing initial planning dates on the interface table. An Interface Table defines the construction need dates and engineering delivery dates for each facility and by type of work. As subsequent Level I11 and IV schedules are developed, impractical durations or logic may be identified which result in changes to the Level I1 schedule to ensure a workable plan. The Level I1 schedule is prepared by the planning engineer and developed in a bar chart format. The upper section identifies the major contractual and/or key project milestones. The section below shows the major engineering, procurement, and construction activities within each major project facility. Major equipmentlmaterial issue-for-bid, award, and delivery dates are all major construction contract issue for bid, award, and mobilization dates are shown. The critical path of the project will be highlighted, so that it is readily visible.

2240

-

CPM Schedule Level III This schedule covers the entire scope of engineering, procurement, construction, and pre-operational testing activities. It is prepared at a level of detail to effectively plan, schedule and coordinate the project’s work activities. The CPM Scheduleprovidesa detailed basis to monitor and evaluate progress for the identification of potential schedule impacts and possible development of workarounds. The schedule is comprised of the engineering and procurement logic schedule that is the basis for the milestone dates for each engineering deliverable in an engineering progress and performance reporting (EPPR) system (see Level IV schedules), and of the construction logic schedule which is the basis for the detailed construction schedules (also identified under Level IV schedules). Construction contractor schedules are reviewed against this schedule, and project detailed cash flow projections are based on this schedule. The construction portion is resource loaded with job hours and commodities to develop construction progress curves and commodity curves. The Level 111schedule is prepared by the planning engineer and developed as a CPM schedule using a computerized program such as Primavera. It is prepared by facility and by discipline for engineering and procurement activities and by commodity for construction activities. For the engineering and procurement section, major categories of deliverables (e.g. design criteria, flow diagrams, equipment foundations) are identified for each discipline. Key milestones for each also are identified (e.g, issue for internal co-ordination, issue for approval, issue for construction, incorporation of vendor prints). Development of major material requisitions and contracts are shown and the logic relationships between key vendor prints and design drawings are identified. Engineering Work Packages (EWP) are identified and tied to the relevant contracts, fabrication or construction activities. The EWP is a descrete package of construction activity and includes all the design drawings, specifications, and material lists necessary to allow construction effort to be complete. Key information from outside parties (e.g. topographical and geotechnical surveys) are identified and logically linked. For the construction section, all major activities are identified by commodity within each facility. The links to the engineering and procurement section will be through the EWPs, and contracts, and equipment and material deliveries. Look ahead extracts (90/120 day schedules) are run from the Level III schedule as needed and will be used to identify any deviations from the plan. At each update period the reason for any actual or forecast deviation from the target (plan) will be recorded and if related to a trend or change order the related number will also be recorded. This will result in a continuously updated record of schedule impacts on the overall project.

The schedule development sequence and updating process is shown in Figure 1 below:

2241

Schedule Development & Updating Process Project Kick-off Schedule

Project Master Summary Schedule (Level 1)

Material Procurement Report

Construction Contracts Status Report

Interface Table Contract Package Scope Summaries

Engineering Procurement Logics

Construction Logics 90 I 120 Day

Look-ahead Schedule EPPR System (Engineering Logs)

3-week Lookahead Schedules

2242

Control Logs and Short term Schedules - Level IV The engineering control logs and shortterm schedulesare referred to as the Level IVschedules. These are used by engineering and construction for planning and performing their work. The engineering control logs show drawings, specifications, material requistions, and related engineering tasks. Each deliverable is identified together with construction issue dates and intermediate milestone dates such as issued for coordination and client approval. These control logs are used for monitoring engineering progress and performance.

PROGRESS AND PERFORMANCEMONITORING Earned Value (EV) is used to determine performance and progress. This EV system is often overlooked because the term “Earned Job-hours” is used so frequently. Jobhours are, in fact, a “weighting” factor in the EV system. This point is discussed in further detaile below. Performance Performance is always measured against budget. For Engineering, it is the number of hours spent to achieve the deliverable milestone versus the hours budgeted. For Construction, it is the number of hours spent to install a unit of work versus the hours budgeted. To develop a composite performance for non-similar tasks, hours are spent and “earned” for each task, then summed and compared. The “earned” hours are calculated from the completed work item and the budgeted hours. Progress Progress is always measured against planned, which reflects the latest scope. Physical progress represents the “quantity” of work performed. Schedule or planned progress represents the time frame in which the work is performed. For engineering, scope is generally identified by deliverables. For construction, scope is generally identified by quantities. For any single item, it is a measure of work performed versus work planned. A composite progress for non-similar tasks can be calculated using a job-hour weighting, similar to the performance calculation, but in this case the “earned” hours are calculated from the completed work item and the planned hours. Some of the progress and performance reporting tools are: Progress and performance reports Progress and performance curves Engineering drawingdeliverable release curves Procurement purchase order commitment curves Commodity release and installation curves Quantity installation curves

SCHEDULE MANAGEMENT The objective of schedule monitoring and managing the schedule is to assess the current status of the project, identify deviations to the plan and implement any corrective action so that the project is completed in the most timely and cost effective manner. The planning engineer becomes the eyes and ears of the project as far as project schedule is concerned. The planning engineer, with input from the project team, (engineering, procurement, and construction) updates the schedule on a regular basis and compares the status against the approved schedule. Any deviations are reviewed and analyzed and workarounds are developed to ensure that the work is completed within the overall schedule.

2243

Curves for engineering progress and construction progress as well as procurement purchase order commitments are developed to monitor the progress. In addition manpower curves are developed for engineering and construction work. The planning engineer tracks the actual progress versus the planned on a regular basis. Also the number of people planned versus actual are compared. The monitoring process results are discussed with the project team, the project management, and the project owner on a regular basis. This provides the project a tool for corrective actions if and when required. In order for the schedule to be effective, it is essential that all project team members participate in the schedule planning and monitoring activities for the project.

CONCLUSION The key elements to the schedule development and schedule control of mineral processing plants are having a clearly defined scope, considering the realistic schedule restraints, client identified milestones, and developing a realistic schedule. The schedule control is simple, by working the plan, monitoring the plan, reporting the deviations and taking the corrective action. This process must be adhered to consistently and frequently.

2244

The Risks and Rewards Associated with Different Contractual Approaches P.J. (Jeff)Gard’

INTRODUCTION Significant changes have evolved in the mining industry toward the end of the last millennium. Mining companies have become significantly more focused on cost and return on assets. This has led to dramatic changes in the contracting industry with owners expecting contractors to accept significantly more risk. The object of this paper is to examine the risks and rewards, both from an owner and a contractor perspective, associated with differing contractual approaches. PROJECT GOALS All too often the goals for a project are stated as delivering a high quality facility, in a safe manner, within the capital budget and ahead of the agreed schedule. These aspects are all very important but the true measure of project success in any industry is normally the same - to maximise shareholder value. Herein lies a serious dichotomy. How can this goal for the mining company be achieved coincident to the identical goal for the contractor? If the correct balance between risks and rewards for both parties can be achieved, then the goals for both parties will be similarly achieved.

RISK SHARING PRINCIPLES Various authors have identified principles which should govern the allocation of project risks between the various parties involved in the project. To determine these principles the following questions must be answered: 0 0

0

0

What is the source of the particular risk? Which party can best manage the events that may lead to the occurrence of this risk? Which party can best manage this risk, should it eventuate? Will the actual cost of the risk or the premiums charged by the party accepting the risk be reasonable and appropriate? Will the occurrence of this risk lead to other possible risks for any of the various parties involved?

If at project commencement the allocation of the risks is not understood or is deemed to be inequitable, then the project goal is unlikely to be achieved. The proper identification of the risks involved will lead to the form of contract that will best suit the project.

1

President, Mining and Minerals, Fluor

2245

STANDARD FORM OF CONTRACT The following standard forms of contract will be considered:

0

0

RCPPF - Reimbursable Cost Plus Percentage Fee where all contractor’s “true” costs will be reimbursed plus a fixed percentage of these costs will be paid as a fee. RCPFF - Reimbursable Cost Plus Fixed Fee where all contractor’s “true” costs will be reimbursed and a fixed fee paid. RCPIF - Reimbursable Cost Plus Incentive Fee where all contractor’s “true” costs will be reimbursed and an incentive fee paid on the achievement of certain pre-stated goals. FP - Fixed Price where the contractor will deliver the facility for a fixed price. LSTK- A fixed price contract in which the contractor delivers a facility that is warranted to perform at a specified level.

CONTRACT TYPE VS RISK The following chart depicts the relative allocation of risk between the owner and contractor for the various contracting scenarios. Project Definition Amount of Risk Financial Uncertainty Risk Allocation

Poor High High

Reasonable Moderate Moderate

High

CONTRACT FORMS Reimbursable Cost Contracts This form of contract gives minimal risk for the contractor but the highest risk for the owner, since the majority of risks associated with errors and omissions, schedules and quality are borne by the owner. Most owners agree that the following advantages are given by a RCPFF contract: 0

the owner pays for what he wants the owner can control who the contractor assigns to project and level of effort to meet schedule and/or cost targets the contractors profit margin is virtually transparent the project is generally not adversely affected by a loss making contractor

Most contractors agree that the following advantages are given by this form of contract: a guaranteed fee or profit margin for the work performed limited exposure for errors and omissions

2246

Unfortunately this form of contract has many disadvantages for the high-risk taker the owner: 0

0 0

0

What are the “true” contractor costs to be reimbursed? Disputes all too often ensue in the areas of payroll burdens, overheads and managerial time. Some portions of these items may have to be paid from the fixed fee and therefore the contractor is almost forced to minimise such costs. It is paramount to the success of this form of contract that all “true” contractor costs are reimbursed. What is the contractor’s commitment to minimise project cost? What is the commitment to utilise the best-suited people for the project? This will be dependent on the total workload of the contractor and the mix of contract forms, within this backlog of work. What will stop the contractor overstaffing the project in times of low workload?

The most common form of reimbursable contracts are based on a fixed fee, where the contractor has frozen their profit margin irrespective of final cost of services provided. This form of contract does not, however, distribute significant profit risk to the contractor. Other variations of the conventional cost reimbursable contract have evolved to provide a more equitable sharing of risk between the owner and contractor. Contractor’s fees can be put at risk by tying attainment of project goals of cost, schedule, safety or plant performance to payment. Other variations of the cost reimbursable contract include placing caps on the total services contract value, as is the case in the Guaranteed Maximum Contract or to a lesser extent capping certain components of the services cost such as contractors overheads or expenses related to the work. Another important issue with the reimbursable cost contract is the selection criteria used to select the contractor. If the owner chooses a competitive bidding process, do they select the contractor on the basis of lowest fee? This will not necessarily guarantee the optimum overall capital cost or full life cycle cost for the project.

INCENTIVE CONTRACTS Incentive contracts are thought to more equally share the risks between the contractor and the owner. They give a middle of the road position between a RCPFF and a FT contract but are often difficult to reach agreement to give the equal risk sharing. Incentive contracts must be based on clearly defined and quantified parameters. Care must be taken not to have parameters which could in fact be in conflict with each other. The most common parameters that are used are: 0 0

0 0

capital cost schedule, possibly including ramp-up time safety quality

The difficulty in this form of contract often arises from the lack of agreement of the risk sharing by both parties. It is important that differences in risk sharing are recognised when the risks are: 0

0 0

controllable by the owner controllable by the contractor not controllable by either

2247

Fixed Price (FP) Contracts This form of contract gives less risk for the owner but greatly increases the risk for the contractor. With this contract, the owner will pay a fixed price to the contractor, irrespective of the actual cost to the contractor for the performance of the contract. The contractor is always attempting to reduce costs by improved efficiencies, value engineering and intense negotiation with vendors and subcontractors. The owner’s risk in this instance is that there may be a reduction in the quality of the facility where this has not been identified in the contract. It is common knowledge that it is extremely difficult to completely specify, without ambiguity, the requirements in a contract. This often gives a loophole for the contractor with this form of contract, or conversely, it leaves the contractor open to commercial risk by the owner in trying to claim something that the contractor did not envisage. Another disadvantage for the owner with a fixed price contract is that the contractor may require an excessive premium to fix the price. Even then, the owner has no guarantee that this will be the actual final cost of the facility - it is, in effect, a guaranteed minimum cost. If the fixed price is too low the entire project may be at risk since the contractor may not be able to meet their contractual commitments. A contract awarded to the lowest fixed price bid through a competitive tender process is not necessarily the best for the project. Choosing the fixed price contract strategy should be avoided when:

specifications are not clear and enforceable uncertainty is significant the reputation and financial security of the contractor are not beyond question A variation to the fixed price contract format is the inclusion of incentives for attainment of project goals of schedule attainment, process ramp-up, or safety. These incentives may further focus the contractor to the owner’s objectives in addition to their own fiscal targets.

There is no doubt that this form of contract should be avoided in the very early stages of any project. At this time there will be poor scope definition, high owner risk and probably financial uncertainty. A more appropriate use of this form of contract is to start the project on a RCPPF basis and convert to a fixed price when the scope is properly defined and there is more financial certainty.

Lump Sum Turn Key The extreme of passing risk from the owner to the contractor is a Lump Sum Turn Key (LSTK) contract. A LSTK contract passes virtually all of the project risk to the contractor. For an agreed, fixed price (the “lump sum”) the contractor undertakes to build a facility and hand it over to the owner when it is a fully operational unit confirming to previously agreed design criteria (the “turn key”). In addition there are usually bonus/penalty agreements on schedule for mechanical completion and the time needed for the plant to reach its design metallurgical performance. The contract has many advantages for the owner, particularly small and medium sized mining companies who do not have the financial resources of their major competitors. Because the risk is nearly all transferred to the contractor, then the risk for the project to the financial institution that is providing finance for the project is against the contractor rather than the mining company. Providing that the contractor has a strong balance sheet, then it becomes considerably easier for the owner to arrange project financing. For the contractor, as well, there are also advantages. He becomes master of his own destiny, insofar as the design of the facility is concerned. Provided that the facility conforms to the criteria laid out in the contract, he is free to build it and commission it how he chooses rather than as a compromise to the owner’s wishes.

2248

One of the few remaining risks for the owner is that he must provide an ore feed that corresponds to the conditions of the contract. The warranties for a flotation plant, for example, that has been designed to accept an ore feed of 3% copper, 1% zinc and 5% iron may well not be valid if the ore feed on start-up is 1%copper, 3% zinc and 15%iron. Lastly, because the contractor is taking nearly all of the risk, then his fee premium for this type of contract is the highest of all of the normal contracting methods.

PROJECT ALLIANCES A project alliance takes the concept of risk sharing to its highest level. In a project alliance the owner, usually multiple contractors, as well as other stakeholders form an alliance to execute the project in which they are collectively incentivised on the successful outcome of the project. As an example, an alliance might comprise: 0

0

0 0

The Owner An Engineering Contractor A Construction Contractor A Key Equipment Supplier An Environmental Consultant

Targets might be set on capital cost target and a schedule target and bonus/penalty arrangementsput in place around those targets. The project might miss its schedule target because of an environmental delay for example, but all of the parties, not just the environmental consultant, would share the penalty. The object of an alliance is to encourage teamwork in achieving the project goals. In the example used, the environmental delay may have been caused by an engineering error, or perhaps an incident caused by the construction contractor. By incentivizing the group collectively cooperation to avoid delays and overruns is rewarded. Project alliance have typically been used for large “mega projects” involving multiple contractors.

CONCLUSION Even with the inherent risks of a fixed price contract it is still a very common form of contract. The RCPFF contracts are not popular with risk-averse owners. Incentive contracts do offer the middle of the road, but they are not used as widely as they could be. This is probably due to the lack of knowledge of this form of contract and their possible motivational effects on the contractor. Negotiation of a fixed price with a trusted contractor may be the preferred option for an experienced owner. The setting of realistic goals is necessary for all relationships but is particularly true for incentivized contracts. When a contractor realizes that he cannot achieve any of his incentive, his performance rapidly changes to become a “salvage” exercise and the quality of his performance may suffer. In any contract situation, there must be trust between both parties and a definite intent to minimise adversarial situations. This will give benefits to both the contractor and the owner and allow information to be shared and decisions to be made, culminating in a successful project.

2249

Success Strategies for Building New Mining Projects Robin J. Hickson

ABSTRACT The increasing complexity of issues surrounding new mining projects, whether inside or outside North America, today stretches the technical competence of the selected Engineering & Construction firm, as well as the operational and financial capabilities of the Client Owner. In today’s knowledge-driven economy, wealth creation is both embraced and challenged from multiple quarters. A compilation of best practices, i.e. “do’s and don’ts’’ for mining projects is offered to practicing engineers for both Engineering & Construction contractors and for Owners. These strategies are globally drawn from a range of recent mineral undertakings, some successfully completed, some not. While the focus of the paper is toward assisting the young new project engineer, the strategies themselves apply to most mining projects. INTRODUCTION Successful building of a new mine project, like any business undertaking involving the interaction of human beings requires not only a clear understanding of the goal by the participants, but also a knowledge of the strategies that are best likely to achieve that goal. Before discussing the key elements of these strategies, it is first important to know what each of the two major players; the Client Owner and the Engineering & Construction contractor (E&C) is seeking when they come together at project initiation. (Note: “Client” and “Owner” terms are used interchangeably herein.) THE OWNER PERSPECTIVE ON E&C COMPANIES What is it that mining Clients expect when they select an E&C Company? The answer is simple; competence to successfully execute the project while, at the same time, providing the Owner with real project risk reduction. The bigger the Client, typically the more risk adverse he will be! The components that the Owner looks for within the E&C organization include: 1. A successful track record - of completing similar projects on time and within budget. 2. Engineering competence - instantly available, specific, above-average specialized process

and technical skills, i.e. those skills that the Client does not keep in-house. 3. Project management capability: 0 Familiarity with the particular geographic location Experience in building projects of comparable size 0 Knowledge of the selected technical process type. 4. Best team. A core of key players who have successfully worked together previously. 5 . Global interacting offices with genuine willingness to provide skills from all offices, not just the signature office. Owners seek an E&C organization with offices that can work together worldwide, and complement each other to provide the Owner with a best global solution that maximizes local in-country experience and knowledge. 6. True understanding of the Request for Proposal (W)requirements, not a mere restatement of Client’s words in the resultant bid submittal. 7. Proven advanced technology designs, and an ability to timely deliver internal know-how. 8. Up-to-date, reliable, accurate, and comprehensive, in-house cost estimation database. 9. Realistic forecasts - of cost, schedule and process technical performance.

2250

10. The cost accuracy figure provided to the Client should reflect work completed; not just restate the RFT. Any accuracy better than requested will reduce the Owner finance costs. 11. Non-alignment - with a particular technology or subcontractor. Clients mistrust an E&C’s own patent recommendation, even when it’s truly best. 12. Sourcing prowess - i.e. volume pricing, inspection capability, expediting competence, and logistics skills. The E&C contractor should have appropriate technical and global understanding of the scope of supply, as well as a sufficient network relationship with vendor organizations to both provide procurement areas of opportunity as well as to facilitate resolution of potential problems before they become major project issues. 13. Ability to control both time and cost, and to punctually and accurately report project performance and progress. 14. Competitive rates. (Note: While E&C cost is important, it is rarely the final decider.) 15. Vision to think in terms of total project, not as separate study phases for engineering (E), procurement (P), construction (C) and start-up. Segmentation of project elements can create gaps. It takes a successful execution and meshing of all project phases from initial idea to Client turnover to achieve a successful project outcome. (See Figure 1) 16. A partner that genuinely desires to align with the Client and all other key project players. 17. Exceptional safety performance track record. 18. An executive sponsor with clout and know-how to fix project issues to Client satisfaction. 19. Unswerving commitment to stay in minerals industry to be there for the Client tomorrow!

PRE.COYUISSIONING Sowcs. Proiscl Hmagsmanr Manual lor Mining

NO

STARFUP I RAMP.UQ

Figure 1 Project Stages Flow Sheet

2251

The Client’s selection of its E&C firm should be objective, not subjective. Usage of a ‘Project Bidder Selection Evaluation & Ranking’ form such as shown in Table 1 is a proven methodology to accomplish this task. The criteria and the criteria weights used on such evaluation form need to be consensually set by the Client evaluation team prior to receipt of bid packages from the E&Cs.

THE E&C COMPANY PERSPECTIVE ON OWNERS What does an E&C contractor hope for, when it vies for an Owner’s business? The answer is more than just “being chosen for the job!” Winning the work, but then being treated unfairly, can be worse than losing. Thus, along with winning the bid, an E&C seeks the following from the Owner: 1. Open and frank communication. A Client willing to do things right. No forcing of the E&C firm to cut critical corners. 3. All parties working to the same key schedule milestone dates, i.e. no hidden agendas. 4. “Total cost” focus, not transaction price. Quality disappears when lowest price is key. 5 . A belief in the merit of adequate funding and staffing for excellent project controls. 2.

6. An open-minded willingness to benchmark and incorporate “lessons learned” from others. 7 . If advocating “fast tracking” via construction-driven engineering, then a full understanding of the additional risks being incurred is required, along with a willingness to fund early involvement of construction and commissioning personnel into the up-front project design. 8. Owner personnel (with responsibility for design and engineering) located in the same office as E&C contractor. Even in today’s cyber world, human interface still is critical. 9. Allowed usage of E&C contractor’s own in-house systems, i.e., their 3-D CAD technology, their own document management, materials sourcing, project controls, scheduling systems, etc. It is a detriment to have to work within another party’s system. 10. Strong belief in partnering principles, and meaningful alignment with, and respect for all stakeholders. The Client needs to view the E&C contractor as a partner, rather than the enemy! 11. Freezing of scope, process design and process flow drawings ( P m s ) in a timely manner; and then sticking to the freeze. 12. Realism in setting Client review times for design criteria, drawings, studies, bid lists, contracts, purchase orders (POs), change orders, etc. Clients who extend their own review times, but still expect overall schedule to stay on track are being unfair. 13. Willingness to issue (and act upon) negative trend notices. Clients that insist on burying negative trends (to avoid bad news getting to their head office) eliminate their best chance of reducing the undesirable consequences. 14. A gainshare system of risk and reward; alignment of Owner, E&C and major suppliers. 15. Concurrence on definition of project completion spelled clearly out at project start. 16. Clients need to be honest and realistic. Too many facts are being spun today. Clients, particularly one-project juniors, often demand that the E&C buy-in to dubious data, as the price of project work. 17. Ability for the E&C to earn a reasonable profit margin. 18. RFP should be a bid document, not a guise for free value engineering. A billion-dollar RFP costs thousands of dollars to prepare. Asking for multiple rebids is unethical. 19. Payment of invoices on time. A number of Clients don’t pay their entire bill, and there is a disquieting trend to not pay the final invoice, or the retention. With retention often greater than the E&C profit margin, such non-payment practice can cause E&C doom.

2252

-

Table 1 Prqject Bidder Selection Evaluation and Criteria Ranking Form Weight Factor

Items PROJECT TEAM STRENGTH & QUALIFICATIONS Executive Management I Sponsor Support Project Manager Total Engineering Capability - CiviVStructuraYMatefld Handling/Mechanicai~l~~c~s~mentation Project Controls Competence: - Controls Manager - EstimatodScheduler ProcurementExpeditingAagistics Knowledge Construction Management Experience & Site Capabilities: - Construction Manager - Reld EngineeringlQAIQCIContracts/Safety Feasibility Study-FinancialAnalyst Organization Chart - CompletenessResponsiveness Key People - Commitment to Keep on the Job TECHNICAL EXPERTISE Process Technology Familiarity Process Conuols Design Capability Product Expertise, e.g. Cu; Mo; Au; Ni; Coal Familiarity in Building a Like Facility InfrastructureDesign and Tie-ins Capability RESPONSIVENESS TO PROJECT RFO Technical Understanding of Project Design InnovatiodAlternatives Mine Plan Integration Capability Constructability Operations Simplicity Maintenance Minimization Performance Warranty Operating &Maintenance Manuals LOCATION KNOWLEDGE Familiarity with Client Construction Experience at Project Site Construction Experience in an On-going Operation Special Concerns: : e.g. Host CountryBonding PROJECT IMPLEMENTATION Work Plan Approach Ability to Team with Owner - Pamering EnvironmentalI Permit Compliance Know-how Safety Commitment PROJECT CONTROLS SYSTEM SCHEDULE REASONABLENESS Ability to Completeby ?? Establishment of Key Milestone Dates OPERATING COST ESTIMATION EXPERTISE CAPITAL COST ESTIMATE REASONABLENESS Is it Lowballed? Is CF Optimized? Battery Limits? COMMERCIALTERMS ResponsivenesslAlignment with Client Requirements Is IncentiveRenalty Fee Option Proposed? Clarity/Simplicity OFFICE LOCATION PRESENTATION EFFECTIVENESS TOTAL SCORE

(37) 2

9 6 5

3

~~

~

~~~~~

2253 2253

1 1 100

ABC Company % wt

CBA Company % wt

BAC Company 70 wt

20. Both Client and E&C should only pursue litigatiodarbitration as the final resolution of differences, not a first step. Disagreements between Client and contractor too quickly go to legal resolution rather than to face-to-face settlement by the parties, and it seems that any issue is fair game today, whether or not it was in the original work scope. Lawsuits are being filed against E&C contractors today for project over-production, for underproduction, for the E&C not accepting Client data, and for the E&C not challenging the Client data sufficiently up front.

SCOPE DEFINITION Project success requires candid decision making at the project front-end leading to a clearly defined scope of work. Too often this is all too vague. The task of defining scope and design concepts starts before the contract between the E&C and the Owner is signed. The proposed E&C Project Manager (PM) and process engineers need to be involved alongside the E&C sales team in proposal evaluation and review of Client’s RFP scope. The process engineer understands the functionality of the project scope. He or she may be the sole capable judge for the E&C of how difficult the project will be to design and operate as expected, what the technical risks to costs are, and, ultimately, how the design will (or won’t) work. Caution: If internal expertise is lacking, the E&C must acquire appropriate outside specialist consultants to review the pertinent issues of concern.

CAPITAL COST ESTIMATE The control budget estimate, be it a lump sum project or reimbursable, must be priced realistically; not set arbitrarily to come under a Client target nor forced by an E&C to beat bidder competition. Once a capital estimate is set, the PM and the Project Controls Manager must stay sufficiently familiar with the components to be capable of effectively monitoring the trending program. Contingency Capital cost contingency is a specific monetary provision within the capital cost budget that is intended to cover variations in the forecast value of cost or schedule. Contingency is not a provision to cover variations in scope or quality. The contingency allowance is added into a project based on the level of engineering completed and on the deemed risk. A contingency allowance, when correctly calculated, will normally be fully spent during the course of the project. A project contingency needs to be set properly to cover risk, i.e. at the 90* percentile confidence level. A Monte Car10 simulation of accuracy risk on each line item of the capital budget to achieve a
0

Project location Degree of definition of project scope Level of undefined project risks Potential for Owner’s scope to occur Potential for project design quality change to take place.

2254

This reserve is an unusual component; it is not present in most projects. And if present, unlike normal contingency, it may or may not be spent, dependent upon whether the envisaged scope 1 quality change actually occurs. The management reserve needs careful administration, typically needing an Owner level above that of Project Director for expenditure authorization. This higher level approval is introduced to avoid the reserve being used as a “slush fund” by the project team.

Estimation Accuracy and Minimum Engineering Level For a lump sum estimate, the E&C has to establish what constitutes the minimum-adequate level of engineering upon which to safely base a lump sum price? While, theoretically, if contingency is set properly for risk, a lump sum can be established from any engineering level, experience shows mineral projects cannot portray risk well without minimally framing the project with at least some 15% of the engineering effort, i.e. a preliminary or ‘‘Type 3” estimate. A “Type 4” definitive estimate, with some 40% engineering complete is the normal minimum for a lump sum estimate however, but a “Type 5” detailed estimate, with >65% engineering, is always preferred. Capital cost estimates are categorized into Types 1 through 5 based upon the level of effort behind the quantity take-offs and the pricing knowledge. The types can be summarized as follows: Type 1 Magnitude Estimate, with 0-2% engineering complete. Used for scoping studies. Accuracy 230% or worse. Type 2 Conceptual Estimate, with some 5-10% engineering complete. Used for prefeasibility studies. Accuracy 220-25%. Type 3 Preliminary Estimate, with some 15-30% engineering complete. Used for feasibility studies and baseline budgets. Accuracy 215%. Type 4 Definitive Estimate, with some 40-60% engineering complete. Used for project control and simple lump sum bids. Accuracy 210%. Type 5 Detailed Estimate, with > 65% engineering complete. Used for project control and more complex lump sum bids. Accuracy +5% or better. Bulk material quantities, particularly steel, cabling and piping, are perennial problems for estimating engineers whenever engineering definition is lacking. Material pricing and unit cost seem to be less of an issue. Capital cost estimates rely heavily on the E&C in-house estimation database. This database thus needs continuous updating to maintain relevance. Real life variances discerned in the Home Office (HO) and field must be routinely fed back to the E&C’s estimating department. To achieve project success within the engineering phase, prior to engineering start the discipline and process engineers must buy-in to the manhour budget for their areas. If not, successful adherence to budget and schedule is unlikely. They then must be held accountable!

SCHEDULE A realistic schedule with logic tie-in milestones to procurement cycles and to subcontract activities is a necessity. Initially, a project master schedule, using a Gantt bar chart that shows overall time frames of each major project phase and the significant milestones will suffice. However, to actually execute the engineering or the construction phases a master integrated schedule with logic tie-ins for all the individual work elements (typically Primavera Level 3 or beyond) is a requisite.

2255

I

Scope Approval

!S

Project Approval

Start Construction

Englneerlng Complete

End Conslrucllon

0

?z 0 I-

Tha Impact Changes Ha v e on Project Cost

m 0

* Planning

.-

Prellmlnary Conceptual Englneerlng Englneerlng * Envlronmenlal Environmental Revlew Revlew Eallmallng * Esllmsllng Cot1 Englneerlng Scoplng Study * Prs.leaslblllty Sludy rn Feaalbllity Study

.

O~ls11Englneerlng

* Environmental

Permllllng Procurement Cost Englneerlng

* DetaII Englneerlng

-

Envlronmental Permlnlng Procurement Conslrucllon Coal Englnesrlng Deflnltlve Esllmate

* Conalrucllon Procurement Cost Englneerlng Perm11 Flnallza11on

me- comlrrlonlng Slarl-Up Turn-Over

Figure 2 Profiie of Project Management Control Influence PROJECT EXECUTION - OWNER ISSUES Today’s Owner approaches a new project with a myriad of concerns. Avoidance of the all too common historical mining project execution issues is paramount: 0 0

0

0 0

Capital overruns Completion dates missed Forecast production output achieved late Operating cash costs frequently never achieved Environmental standards compromised,particularly in the less developed countries Insufficient regard for safety.

Project overstatement could also be added to the above. Compulsive optimism seems to drive certain mining project advocates. Insufficient benchmarks are taken to check projects against reality, or if taken, they are ignored or explained away. But these pitfalls are preventable. Consistent project success lies in a rigorous execution of basic project management tenets. Project Characterization The Owner must be willing to spend sufficiently up front, where the optimal influence over project outcome resides. (See Figure 2.) This is the time to decide to build or not. Cheap estimates up-front lead to bad outcomes. Proper initial project characterization requires a formal process:

1. Multi-department, integrated initial evaluation (scoping) assessment process. This is used by Owner in the initial idealopportunity phase, when initial “golno go” business strategy is executed. This initial scoping assessment process requires timely participation by the Owner’s pertinent support disciplines, (sales, human resources, tax, legal, treasury, etc). These disciplines are brought into the decision-making process early, to introduce a quick, objective, collective fatal flaw assessment of front-end project risk.

2256

2. A Project Management “Best Practices” program. A well-defined Project Management process provides the originator of an “idea” with the tools to develop that idea into a winning project that will meet corporate Owner objectives. A formal project management program, following a regimen of proven best practices should kick in post the scoping phase, i.e. from pre-feasibility through start-up. By following a set outline of best practices, management is assured that project viability is adequately challenged, correct questions are posed, the project is properly characterized, necessary personnel are placed, and that an appropriate process is implemented. The goal is to set an optimum execution strategy to deliver the best life-cycle performance for the project. 3. Technical underpin to design criteria. Each project needs a focused set of site-specific technical plans and designs. And, as a complement, rigorous benchmarking against comparable project history is also necessary, to honestly assess the potential for success and to fully highlight the risks. 4. Fatal flaw review. Periodic reviews should be scheduled to look for fatal flaws and better technical solutions. These should be conducted with non-project personnel, experienced with the concepts involved. Project and/or process engineers are not always able to stand back at sufficient distance to take a look at the total picture. Opinions from outside the project team should thus be routinely sought and never refused consideration. 5. Feasibility Study completed to “bankable quality.” Even if external monies are not sought, a feasibility study should be of a quality that could merit bank project financing. When a project cannot persuade outsiders to put up their own money without demanding that the corporate Owner underpin the investment, it is a strong indication that the project should not be pursued!

Project Independence To achieve an unbiased project characterization, the Owner’s project group needs independence from the Owner’s operations and exploration departments. Unless the Owner’s project department is separated from an Owner’s operation’s “stay-in-business” concerns, and from Exploration’s “pet project” influences, mischaracterization by project advocates will occur. However, the large central project group, the norm of 30 years ago, is no longer tenable for this task. The solution is to keep a small central, specialized Owner’s project group, ideally less than six persons, that has the knowledge to hire the appropriate outsource entities to execute any project. The small size of this group ensures that the carry cost for the Owner is perceived as an opportunity, not a burden. A low carrying cost also minimizes the job security concerns of Owner personnel, thus avoiding any project creation tendency. Outsourcing can provide the same, or better, cadre of skills as an internal central department, but without the overhead. Project Success To arrive at a successful outcome, everyone has to have the same goal. Thus a project must adopt a Project Creed - define project success before starting out. From Day One everyone must understand the exact success goals, and the metrics that they will be measured by. Final project success should mean achievement of all six of the following parameters: 0

0

Project handed over to Operations on schedule Project completed within budget Plant’s name-plate production met after ramp-up Predicted unit cash costs achieved by the date specified Zero Lost Time Accidents Full environmental compliance.

It is vital that the first four criteria are all given equal weight. If an Owner elects to emphasize only one, two, or three of these first four success criteria, then the project team will utilize the non-

2257

selected criteria as their “safety valve.” This is bad. Ideally, the metrics for these success parameters should be set initially by the Owner and the E&C together to create the highest net present value ( W V )for the project, preferably using a gainshare program for the potential shared benefit of Owner, E&C contractor and suppliers, And, as the final success ingredient; one has to have the right corporate culture to have any chance of project success. The requisite elements of this culture are as follows: 0

0

Single Point Accountability. One cannot segment out project portions to different corporate departments and expect a successful project outcome. All areas: initial characterization, engineering and construction, owner costs, project commissioning, project controls etc., need to be under one entity, the Owner’s Project Director. He/she needs overall control of project. These project elements are inter-linked; a modification to any one element affects all the others. Honest Characterizations. The project group must operate within a corporate mandate in which it is more important to characterize a project correctly than to achieve a specified internal rate of return (IRR). If job security hinges on coming up with a positive NPV, human frailty will ensure mischaracterization ensues. One Team. Successful outsource utilization requires that external personnel be treated as insiders. Second class treatment of external people will result in a second-class outcome.

-

PROJECT EXECUTION E&C ISSUES Constant reinforcement of proper project management procedures by the E&C firm is a fundamental necessity, to ensure that a success mentality pervades. Simply handing out a project manual and hiring a world class E&C firm cannot alone ensure a successful project outcome. Project Team Senior E&C management has to ensure that appropriate human resources are supplied at project inception, and that the key project members are compatible, within the project team and with the Client. Further, the project team should be based together, in one place, wherever the project focus is. A PM cannot be effective during the field construction phase when based in the comfort of the home office. The PM has to be constantly engaged, wherever the action is. Figures 3 and 4 illustrate typical project organization charts for the home office and for the field respectively.

2258

I

CIImnt Project Dlroclor ELC Ex*cutlvr Sponsor

ELC

miac EhC

Cllsnt Controls Yanagrr

Prolect Ysnsgor I

I

ELC Contracts Yansgor

l

'

I

.

I

EhC Cornrnarclal Y g r

I

4n 7 Estlmstor

Yochanlcal

1-4

Coordlnmtor

lnstrurnrntsllon

NOTE: All positions are E6C unless olhemiss noled Eng. SpacIaIIsts

Docu msnI Control

Figure 3 Project Organization Chart: Home Office

NOTE. Alloosilions

Figure 4 Project Organization Chart: Field Office

2259

are

E6C unless olheruise noled

Cost control and planning manpower needs are, unfortunately, often under-budgeted, particularly in the field. As a generality projects need to budget, then assign, significantly more cost and control engineers than have historically been utilized; i.e. sufficient personnel capable of timely and accurately capturing project trends. A key to smooth project execution in the field is integration of construction and commissioning personnel into the project team at the front end of engineering. Proper attention to construction and start-up planning during early engineering will positively affect many aspects of project outcome. A Contracts Administrator is needed to handle the subcontractors on projects of any significant complexity or size (>$25 million); the PM cannot effectively handle these additional duties alone. For major lump sum projects (>$50 million) a Commercial Manager to interface with the Client should probably also be installed reporting to the PM, as shown in Figures 3 and 4. Partnering Project orientation needs to include formal team building. In larger projects, a formal team building exercise (partnering), led by a third party, should be included at project kick-off. Partnering builds upon the merits of aligning Owner stakeholders and outsource project support resources together to empower team success. It is important to preach and employ an integrated team approach throughout the project life.

0 0

Partnering builds consensus and eliminates the “us” versus “them” mentality Partnering creates a trusting environment where issues are discussed openly and frankly Partnering allows a “best man for the job” philosophy to exist, and removes turf battles.

Project Procedures Existing generic project procedures have to be modified to best facilitate the setting of ‘project specific’ requirements, such as mobilization, logistics, start-up, etc. Projects need policing for quality. Efforts must be extended toward uniform application of best practices during all phases of a project. E&C management and/or quality audit teams should target at least one multi-day site review per quarter. Further, in the first three months of project initiation, it is useful to introduce an external reviewer to ensure proper procedures are established. Insurance coverage (deductibles, caps, coverage limits) must be fully understood at project start. Insurance limitations should be noted as a reminder in the Project Manager’s monthly report. Formal dispute resolution tactics need to be established early in the project. This will not avoid Client and subcontractor confrontations, but it will facilitate a civil path through the issues. Project Sponsor The E&C project executive sponsor has to take hisher role seriously, and needs to be willing to commit time, particularly in cultivating Client senior management contacts and relations. While the sponsor role traditionally is mostly focused on Client concerns, it is suggested here that the E&C project sponsor also play a more active internal E&C role. The suggestion is that regular, though not necessarily frequent, internal meetings led by the PM take place, to include both the E&C sponsor and the senior project staff. If internal conflicts are present, this process can provide a mechanism to let those issues rise in a timely manner to the sponsor’s attention. Draw on Experience The smart Client recognizes the merit of bringing into the project the wisdom of others who have “trod the path before” to: 0 0 0

Benchmark critical components against external operating installations Incorporate constructability and commissioning reviews during engineering Conduct an external fatal flaw analysis

2260

Insert seasoned operations personnel into all project phases through to commissioning Remember, there are no old mistakes that can’t be repeated. Project Site Site familiarity is critical. The engineering leads and the Construction Manager (CM) should walk the site together, prior to engineering start. Site conditions, equipment availability, plant tie-ins, Client preferences, labor rates, critical equipment deliveries, and construction package preferences all need to be established and understood by the team members, up front. Project Information Field construction disciplines (Controls, Procurement, Accounting etc. as shown in Figure 4) must retain a report link to their HO lead. While project disciplines always report to the PM via a line structure, each must also retain a reporting link to their functional HO department head. Electronic transmittal of HO drawings to the field is a necessity today. The computers of all project team members on site need e-mail linkage and connection by LAN (Local Area Network) as well as WAN (Wide Area Network) to the E&C contractor’s global Intranet.

CONTRACT The discussion that follows is focused on the contract between the Owner and the E&C contractor. However, most of the points would also apply to a contract between the E&C and subcontractors. Judicial Usage of Lump Sums Lump sum bids should be used&o when the project scope is well defined. It is a misconception that a “hard money” lump sum contract will shift the cost overrun risk from the Owner on to the contractor. Lump sum is merely an agreement for the contractor to provide, for the lump sum price, all services necessary to satisfy the contract battery limits. If limits are not precisely defined, all one gets from a lump sum, are change orders to the contractor’s benefit. Because Owner and Contractor interests are not identical, no contract is truly ideal. Various combinations of the lump sum (L) and reimbursable (R) elements are shown in Table 2. Banks and junior mining companies prefer lump sum, in a mistaken belief that their financial exposure is capped, and that their risk is transferred to the E&C. Major mining companies are more pragmatic about the evolving nature of pertinent data within a grass-roots project; thus they tend to prefer cost reimbursable, at least through detail engineering. Lump sums are not generally appropriate until after the feasibility study is complete. And, if value engineering is contemplated, a Client should ask that this be undertaken prior to fixing the lump sum amount. The contract for lump sum work has to include sufficient detail such that it reads, to all parties, as an absolute fixed scope of work. Thus any “allowances” included within a lump sum submittal must be fully explained, defined, and agreed-to by the Client prior to project kick-off.

2261

Table 2 Types of Contract ~

~

CONTRACT TYPE COST ELEMENT

1

2

3

4

5

6

7

8

9

10

Design Services

L

L

L

R

R

R

R

R

R

R

Engineering Services

L

L

R

L

R

R

R

R

R

R

Management Services*

L

L

L

L

L

L

L

R

R

R

Equipment Supply

L

R

R

R

L

R

R

R

R

R

Material Supply

L

R

R

R

L

R

R

R

R

R

ConstructionManagement

L

L

L

L

L

L

L

L

R

R

Construction

L

L

L

L

L

L

R

R

R

R

Indirect Costs

L

L

L

L

L

L

R

R

R

R

Profit and Overheads

L

L

L

L

L

L

L

L

L

R

* Management Services include project management, scheduling, procurement, expediting, logistics, cost control, and financial administration L stands for Lump Sum; R stands for Cost Reimbursable Type 1: Type 2 through 6: Type 8: Type 10:

Represents the Turnkey Lump Sum contract Represent Lump Sum contracts, Type 5 being the most common lump sum contract within the minerals industry. Represents the “Fee Plus” contract. Represents the “Cost Plus” contract where all elements are fully reimbursable. Profit and overheads would be expressed as a percentage.

I

Source: Project Management Manual for Mining

Use of Bonuses / Penalties Bonuses for achievement of successful project completion are worthwhile. The bonus needs to be set up-front, when recipients have the maximum chance of influence over outcome. The bonus should not be set, however, until after project scope is set in an approved Feasibility Study. An incentive/penalty scheme should include both key E&C and Owner staff, tied to the following: Key milestone dates, including project completion Capital cost Safety Constructed facility performance Incentives and penalties preferably should be symmetrical (equal amounts around the set target). A gainshare system of risk and reward for the Owner and E&C contractodsuppliers is best - where all participants have the potential to share the commercial benefit of a successful outcome and, equally, all have the potential to share the pain of failure. i.e. everyone wins, or everyone loses. (See Figure 5, Examples of Project Gainshare, courtesy of M. Entwistle, Kvaerner E&C 2001 - see footnotes, Reference 7.)

2262

CAPEX GAINSHARE

SCHEDULE IMPROVEMENT GAINSHARE

- t t

ve Ceo

t\ I

12 weeks DELIVERY AHEAD OF TARGET DATE “STEADY OPERATION” CONTRACT DATE i n m i

Schedule Improvement gainshare

TARGET COST

I

DELIVERY AFTER TARGETDATE

12 weeks (weeks)

- ve Ce

UNDERSPEND (‘) Non-owner alliance

partner8 shares by pre-agreement dillerent for each

-

v LOSS

Capext gainshare

Figure 5 Examples of Project Gainshare Bonus / Penalty A project incentive (bonus) program, if established, has to be fully understood by all parties at project start; and if the bonus is to act as a true incentive for performance improvement, then the bonus monies need to be paid to the participating parties immediately after project completion.

PROJECT MANAGER An obvious critical component of project success is the assignment of appropriate project leadership. Odds favor a best candidate for the Owner’s Project Director coming from outside the Owner company. A project leader needs: 0 0 0 0

Familiarity with the geographic location Knowledge of the process type Experience in handling a project of this size Fluency with the “local” language.

Overcoming Client Preferences / Demands Prior similar experience for the E&C Project Manager is also good. But, while a PM’s similar experience is always helpful to others on the project team, it is not an absolute requisite (except, perhaps at a remote location site), At some level within the E&C team though, relevant experience does become valuable and even necessary. Otherwise, the Client probably would never have selected this particular E&C to perform the work. The E&C PM must therefore proactively embrace the relevant experiences of subordinates on his team. The E&C organization must ensure that its PM is someone who can serve the project’s needs. A good PM is one who cares about project success to the point of being passionate about it, and is willing to accept ownership of the effort. Moreover, a good PM is committed to the right approach and an honest and forthright manner of applying the skills necessary for success. A Client will always state that he prefers a manager with prior experience in the type of project being constructed. While “nice to have” this is not always warranted. More importantly, it may not necessarily satisfy the E&C contractor’s need for adequate project management and control. Project Manager Responsibility / Accountability Project Managers (and similarly, Owner Project Directors) need to be given clear project responsibility then held accountable for project outcome, as much for properly informing corporate management of project issues, status and performance each month, as for delivering the desired result. PM non-performance, certainly repeated non-performance, needs to result in removal of the PM. The identification of problems within a project is not, however, synonymous with negative performance. By addressing problems, solutions are sought and implemented. A constant vigil toward problem areas and their management is the hallmark of a good PM.

2263

If a PM is to be burdened with overall project accountability (which he should be!), then he must be similarly accorded overall project responsibility. Thus the PM, not the corporate office, is the person who should be accorded jurisdiction over who can be removed from the project team, and when. Personnel reassignment during a project life must be the exception, rather than the rule. The PM also needs to be sole entity drawing on project contingency and escalation. Outside monetary draws on a project will kill PM accountability.

Project Manager Performance A PM must have the all-around skill-sets to be capable of managing all aspects of a project, including procurement, controls, construction, start-up, and project closeout, i.e. not just the more frequently encountered front-end studies and engineering. The PM needs to be fully involved with establishing, negotiating and closing the Client contract. The PM and other key personal must fully understand the contract, particularly scope definition, performance guarantees, and penalty clauses. The PM and the lead process engineer should review the contract before it is signed. Advice needs to be timely sought from all relevant quarters to avoid contract provisions that may inject undesirable implications. A quality oversight process to measure the fulfillment of the PM’s responsibilities needs to be visibly in place. Regular project oversight review helps prevent problems with project performance that could remain hidden to management until too late in the schedule to correct. Pitfalls of Project Management Many things can derail a project. As W. J. Tinsley noted in his 1999 IMM Presidential address, PM success must avoid focusing on apparent shortcomings; but rather it requires an addressing of the actual failings themselves. (See footnotes, Reference 8.) These pitfalls of project management have historically included: Insufficient planning for project execution Loss of emphasis on overall project goals Inattention to safety and quality Lack of focus on value Poorly defined scope, budget, and/or schedule Inadequate project control system Improper management of change Lack of understanding and acknowledgement of the effects of change Poor communication between parties / lack of trust.

ENGINEERING Clear, full definition of technical concepts must be established prior to the start of detail engineering. This should be “a given”, but all too frequently is not! Recognize that starting on detail engineering before process concept (i.e. PFD’s, mass balance) is complete will result in rework and inefficiency. To further minimize any possible rework, the “failure mode analysis” (Hazop) review, which is typically conducted at the 70% engineering stage, should additionally be undertaken at an earlier stage of drawings, i.e. prior to drawing number Rev. 0. Engineering Checks Budget and schedule must allow the appropriate manhours and time to conduct a proper check of drawings by engineering such that complete documents can be issued in accordance with the agreed schedule. Drawings must not be sent out to meet a preset schedule knowing that the drawings are incomplete, not properly checked, or that corrections will be needed. Accountability

2264

for quality of engineering documents resides with the E&C department lead manager. These discipline leads have to enforce the protocols for checking as outlined in their own procedures. Cost and schedule constraints may lead E&Cs to sometimes try and cut corners, to do insufficient internal checking. The PM must be constantly alert to not let any such quality compromises creep into the project. Design reviews must always be held; and such reviews need to be conducted with all affected disciplines. The results of design reviews must then not be allowed to be suppressed simply because they might cause delay or cause conflict with the Client.

Freezing Design The design freeze must be a hard freeze, not a soft freeze, and have the commitment of the Client, project management and design engineering. One must avoid the endless desire to continue engineering to achieve “perfection”. An inability to freeze design concept leads to engineering overruns and then to “knock-on’’delays. A viable freeze starts with communicating expectations, and then consensually establishing appropriate budgets for each discipline at project start. A freeze will not hold if unilaterally imposed. A freeze must be based upon sound technical judgment and not be arbitrarily undertaken to merely accomplish a shortsighted goal of satisfaction of say, a schedule milestone. If sincere doubts persist, then design must be revised. Certified Vendor Drawings Detail engineering should ideally proceed on the basis of approved vendor drawings. However, the Client must recognize that, today, critical vendor certified drawings are not truly final until after fabrication is released, in spite of the certified connotation from the vendor. Without this release, certified drawings can change. Thus it is a fatal flaw to claim 100%engineering prior to award of fabrication. Vendors do not release their own subs to fabricate prior to the E&C drawing release.

3-D CAD Models Use of 3-D CAD models should be utilized to the maximum practical extent. While there is little, if any, engineering cost or time reduction from utilization of 3-D over 2-D, there is a real field construction savings from the better interference recognition of 3-D, and the better visualization of the ultimate facility layout. Exceptions need to be made, however, where there are limited 3-D CAD model skills in joint venture partners, to accommodate third party engineers, and for certain small projects. It must also be recognized that 3-D models are not always usable by a segment of the foreign fabrication shops or by all constructors in the field. Thus the E&C needs to be prepared to supply 2-D drawings when required, and not blindly insist on supplying 3-D models in all instances. Engineering Leadership / Skill Base Engineering skills serving the minerals industry are constantly aging; capability and expertise can erode over time. A project must not be allowed to suffer from a lack of guidance by experienced project engineers. While training programs and adding younger engineers into the project engineering talent pool are the best long term answers for the E&C, outsourcing often has to be embraced for the short term project good, particularly where internal expertise is lacking. The PM needs to keep aware that the E&C contractor may not proactively accept the outsourcing of project work elements. Many E&C companies struggle to recognize that they do not contain all the specialized skills for every job, or that their mature engineers may not all be usefully experienced for the particular project in hand. Remove the “Maximization of Billable Hours (cost reimbursable?) Mindset” This is more than just recognizing the difference between lump sum and cost reimbursable at project outset. Cost reimbursable contracts, by themselves, are not open doors to an endless engineering effort. Schedule constraints exist on all projects; the best method of reducing overall project cost is a reduction in schedule along with the proper management of construction labor.

2265

Timely delivery of engineering documents is key. Attempts to maximize billable manhours flow contrary to this objective and, in the eyes of many Clients, are constant points of contention. The number of manhours that a project requires is simply a tool for use in project planning, not an objective to be met. With respect to lump sum contracts, all participants must understand the philosophy of “minimum adequate” billable manhours. What isn’t spent, is additive to profit, a portion of which can be returned to the project team either through an incentive plan established at project start or through a performance reward plan distributed at project end. Client Representative Client personnel in the E&C office are to be encouraged, but must be managed. Clients must not be allowed to g o to individual engineers and change engineering concepts without going through the E&C contractor’s trend and change order process. Selection of the control DCSRLC system and configuration should always involve the Client. PROCUREMENT Procurement normally encompasses sourcing, purchasing, expediting, and then shipping and receiving. However these last two items have been separated out into a following separate Logistics section within this paper - to highlight their importance. Sourcing The first procurement step on most major minerals projects is to agree on a clear procurement plan, then identify all critical, long lead-time equipment and prioritize the sourcing of those items. The next step is to prepare a bidder list for early Client approval. Most major E&C contractors maintain a database for pre-qualifying vendorshppliers. This database needs to be referenced before the project team (PM, project engineer, process, procurement, Client) nominates their vendor list. This ensures optimal vendor selection. For schedule efficiency, the Client needs to place a person in the E&C office with authority to sign purchase orders. Specific requirements of the financing institutions (e.g. Exim, EDC) that relate to tax credits, financing assistance, etc., must be identified prior to order placement, and then actively tracked. Keep in mind when placing orders the need to maximize the standardization of components, i.e. define “smart spares.” This needs to be addressed in both engineering and in procurement. Purchasing For domestic projects it is generally more efficient to purchase goods “FOB jobsite” as ownership of purchased goods then stays with the vendor until goods are received on site. The vendor is thus responsible for packing, preparation and accuracy of packing lists, inland freight, and delivery. The terms of purchase must be determined as a strategy during the early stages of the project to account for overall insurance coverage. Expediting For all critical goods, the project must appoint project expeditors and prepare a shop expediting/surveillanceplan with early warning status reporting of potential slippage. Before critical goods are released for shipment from vendors, the material must be inspected, packing inspected and packing lists verified. Pre-shipment inspection at a supplier’s facility does not typically include verifying packing lists because the goods are open for inspection and not yet packed. A second visit would thus be required for designated “critical” goods after packing is complete. If export packing is to be performed by others, then a new packing list will be needed. “1 Lot as per Attached List” terminology is to be avoided as a line item on purchase orders. The use of “1 Lot” may be unavoidable, however, in cases of complex equipment as the buyer and seller may not know component breakdown until after the goods are produced.

2266

Field Issues The E&C contractor will normally furnish the large-quantity imported bulks (pipe, fittings, cable, etc.) to subcontractors as the E&C can generally do this at considerable cost and time savings for the project. Subcontractors (subs) should generally purchase the small bulk items, especially on overseas projects. Subs typically have better local knowledge; therefore shorts, damaged and small bulk items should remain their responsibility. Procurement continuity must be maintained from home office engineering into the field. While it is generally beneficial for the responsible procurement person to move with the project to site, it may not always be appropriate. For example: If engineering continues in home office and it is necessary that purchasing remain close Where local agents can better perform the small value, rapid response field transactions For work overseas, where it is essential to employ a field agent with language fluency.

LOGISTICS Offshore projects as well as large, logistically complex projects will require a logistics study. Such study usually results in a requirement for the services of a freight forwarder and a customs broker. To properly handle logistics, first one must understand the transit cycle. This probably depends upon the season and the state of the roads. Three days from port site may stretch to three weeks, or more, in the wet season. Freight Forwarder If a freight forwarder is utilized, one should award his contract early. A freight forwarder is chosen on basis of real (not claimed) experience and competency. Principal issues for selection are strength of representation in the country where the project is located, strength of representation where goods are being shipped from, and past successful performance. The freight forwarder’s scope of responsibility typically includes collection, inland transportation to port of export, seaworthy packing, export crating, booking of vessels, customs clearance, and inland transportation to site, as well as producing his own packing list based on actual packed contents. (Such list needs to be as accurate as the vendor’s list.) If the freight forwarder is tasked with subcontracting trucking services from vessel to jobsite, then the freight forwarder assumes responsibility for proper equipment use and rigging. Receiving Port It is vital to appoint a good, local transit agent at project outset. More important, make sure “your guy” knows how to work with government officials to move goods idout of port. Develop procedures for seamless customs clearance. If building one’s own port, hire experienced port facility personnel early (harbor master, warehouse personnel etc.) Review and improve, as necessary, existing port infrastructure. Establish a heavy equipment off-loading facility as well as a “bonded yard” for “in bond” shipments at the port of entry. To save transportation costs, try to use dedicated ships with consolidated cargoes. This will require establishment of consolidation marshalling yards, both at the port of entry and at the staging area. Materials Tracking The E&C procurement group typically provides a bulk material management procedure (usually a proven computer software module today) to track materials from MTO (materials take-off) estimation, procurement, shipping, warehousing through issuance of materials to contractors. The E&C logistics plan should provide specific transport routing instructions for each PO package, for each geographic area of supply.

2267

Consolidated packing lists are needed on site for each container. The freight forwarder (or vendor) must provide a consolidated packing list per container. Site uses these packing lists to receive all pieces, for customs clearance and for tax clarification. Be vigilant in ensuring that the vendor interprets PO specifications accurately. The logistics lifeline to a remote site can be four months or more; receiving the wrong part can ruin a schedule.

SPECIAL NEEDS OF THE REMOTE LOCATION PROJECT While the project execution strategies outlined to this point could pretty well all apply to both domestic North American as well as to remote foreign locations, it is worthwhile looking at the unique special needs of the remote location, international project. Local Knowledge Application of knowledge gained from similar undertakings in the same region is paramount. Lack of familiarity with local conditions will negatively impact progress and construction productivity. Understand the weather seasons. When are the rainy months, the droughts, the freeze-ups, and the thaws? In many locations, land (or river) transportation can be impossible in certain weeks of the year. Seasonal effects must be taken into account in the schedule. Experienced Personnel The key is to assign a resourceful, innovative, experienced site project team, preferably a team with relevant international construction experience, that has successfully worked together previously in a remote environment, i.e., one with a winning track record. Ensure that there is a commitment of these selected personnel to relocate. If they don’t want to be there, they won’t succeed! Procedures Set Early A project execution plan must be in place prior to mobilization. Establish the work breakdown structure and control budget in the initial 90 days. The control schedule (up to Primavera Level 3) and overall manpower loading also has to be established early, with a build-in of job factors (labor skills) for local conditions. It is imperative to remain schedule-driven through all project phases. Set the task force center in a civilized location, as close to site as possible, equipped with a computer and CAD network. Get agreement on design standards (e.g. US, Australian, French, Russian); then make sure that everyone is cognizant of that agreement. Pre mobilization, understand all project-permitting requirements through operations start-up. Environmental Issues Embrace Western environmental standards. Lending agencies require these anyway, so why fight them? As the 2001 Wellcome Trust Survey on public attitudes to science guided (see footnotes, Reference 9), for acceptance by society, projects need to respond positively to the public concerns of the non governmental organizations (NGOs.) 0 0

0

Select designs (and fuels and reagents) that are environmentally friendly Don’t make environmental compliance an engineering encumbrance Involve the environmental manager from Day One Force the alignment of the E&C contractor with the Owner’s commitment to the local community and environment.

An open, pro-active involvement with the key stakeholders (including local community, state and federal agencies) does work. Such a strategy facilitated the successful execution of the recent

2268

Alaska-Juneau Gold Mine closure project by the E&C firm Kvaerner, and Echo Bay Mines, along with receipt of the Year 2000 Alaska Governor’s “Reclamation Project of the Year” Award.

Establishment of Initial Site Undertake a familiarization visit to the project site by the core project team within the first four weeks. Establish a temporary facility and camp as close to the permanent site as possible. Adequate mobile equipment and transportation mechanisms need to be available from Day One. The initial focus is not on the facility that the Owner wants built; rather it is the rapid establishment of the site for habitation. Thus, appoint a knowledgeable camp manager immediately - to get the initial, basic catering and accommodation facilities in place. A realistic understanding of the true availability of existing housing is crucial. Data Gathering Assess local road and beaching capabilities. Survey the site for available materials and resources (aggregates, sand, concrete, craneage, fab shops). Gather geotechnical and survey data. Early Works Upgrade site access roads at the project front-end. Time and effort expended on ensuring access suitability for equipment will eventually pay large dividends in maintaining schedule. Know where your earth-moving equipment spare parts are sourced. Don’t wait for the first breakdown! Place satellite communications as soon as possible. Land-based telephone lines in remote locations are generally unreliable. (It is OK to use local landlines as back up, however). Today, a project web site connecting site, Client, task force center, and support offices is becoming a must. SAFETY AND HEALTH The project safety, health and environmental plans need to incorporate the Client, local and E&C contractor standards. Safety can never be an afterthought. Raise the Safety Profile: Have a Commitment to Zero Every project needs a senior management commitment to a safe project and working environment. Statistics from the Construction Industry Institute of the USA show that when the company president and senior management regularly review construction safety performance, accident incidents reduce a staggering 86%. (See footnotes, Reference 10.) It is also a good practice to make safety performance part of contractor remuneration. Mine construction safety performance is worse, on average, than mine operations safety performance. This does not have to be so; 11% of mining related construction projects completed over the past four years had zero lost time accidents (LTAs), e.g. Kvaerner’s Solvay Soda Ash Project in Wyoming (1.2 M hours), Fluor Daniel’s Cerro Verde Copper Project in Peru (3.2M hours), Henderson Molybdenum’s 2000 Project in Colorado (1.4 M hours). Every project’s goal should be zero LTAs. “Best practices” for achieving zero accidents are as follows: Demonstrated senior management commitment Proper staffing for safety (<50site workers per safety professional) Site specific pre-project I pre-task safety planning Formal safety orientation, training and education (> 4 hours per month) Worker (and family) involvement (including safety perception surveys) Formal recognition and rewards (distributed at least bi-weekly for 0 LTAs) Subcontractor management (requirement for site specific plans) Accident I incident reporting and investigation (with sanctions for non-compliance) Drug and alcohol testing.

2269

Health & Medical A reliable supply of water is crucial, both construction and potable, particularly on the remote site. Know where the water is coming from before personnel mobilize. Problems with potable water supply will ultimately result in the outbreak of illness. Have medical evacuation facilities in-place before the site team is established. Establish medical facilities onsite; at the very least, a first aid center - and, if in a remote location, preferably with a nurse andor doctor. TRAINING A training needs survey is required very early in a project; i.e. a local job skills assessment coupled with workforce requirements (for both operations and construction crafts). CONSTRUCTION Set the construction-contracting plan with its engineering deliverables early, then establish the work packages, and structure the schedule (and cost estimate) around the work packages. E&Cs typically furnish project licenses; they have to ascertain that all engineering, construction management and construction licenses necessary for the jurisdiction of the project are in hand. Materials & Subcontractor Resources In addition to standard vendor assessment criteria, construction management has to verify sources of raw materials, financial strength of suppliers and local subcontractors, as well as local fabrication shop capabilities. Local subcontractors are generally willing, but not always sufficiently competent. 0

0

0 0

Maximize usage of local and regional resources. Work packages should reflect this resource base. Maximize offsite fabrication, plant packaging and/or modularization (e.g. substations, rebar cages, pipe racks). Match to port and site craneage, as well as to assembly yard facilities. Set up a bulk materials management entity, with multi-discipline support. On-site power needs to be sufficiently under project management’s own control to be reliable.

Fast Track Approach fast-track methodology of project execution with extreme caution. While a constructiondriven project is normal, and makes sense (the majority of money is expensed in this project phase), fast-track has, in multiple hind sights, been used more as an excuse to delete candid frontend risk evaluation, and to justify insufficient planning and/or poor project controls by proponents of less-than-robust projects, than for real project execution improvement reasons. Lessons learned from fast-track projects indicate that:

0

0

For “construction driven” fast-track projects to succeed, project management procedures need bolstering, not finessing. This is not the time to miss out any crucial project execution steps. Undertake constructability reviews during both the design and procurement phases. Take construction related activities into full account prior to site mobilization. Move a cadre of key experienced design engineers from the engineering phase to the field for the duration of the construction phase, to quickly clarify design interpretation queries. Quality assurance (QA) and quality control (QC) plans are always required. Remoteness of site cannot justify substandard quality.

2270

COMMISSIONING Install a separate start-up commissioning team. A separate commissioning team best assures that the construction contractor has truly achieved “practical” completion, and that the plant is really ready for the Owner to take “care, custody, and control.” 0 0

Commissioningteam leader needs to be a “start-up” specialist, inserted for the task. A start-up and commissioning plan is needed by sequence and by task through both cold commissioning (each plant system initiated in isolation without any process product) and hot commissioning (integrated plant run-in with process product) on into production ramp-up. Such plan should identify all necessary spares, tooling and first fills. The commissioning team needs to integrate multiple participants (Owner plant operations and maintenance personnel, vendors, construction leads, and experienced start-up staff) within a defined strategy and hands-on roles. One should avoid lump sum for all the post mechanical completion services, such as start-up and commissioning activities. Scope, in these stages, is too susceptible to change.

PROJECT COMPLETION Define “completion” at project start. To facilitate a smooth and painless demobilization, the E&C needs to establish as tight a definition of project completion as possible within the contract (particularly for mechanical completion). To achieve this goal, all parties must agree upon the completion criteria (and language) Many Clients are not cognizant of the differences between mechanical completion, commissioning completion, substantial completion, practical completion, first-product completion, ramp-up completion, performance test completion, etc. It is best to set specific dates for release of design and field construction staff, to try and prevent key personnel leaving early for the next job’s paycheck. Consider offering demobilization bonuses to key team members to encourage them to stay through completion release date. Capital Savings Capital under-runs go back to the Owner’s treasury. These project savings aren’t a free “slush fund” to be spent on non-project items. The Owner’s operations team needs to be clearly told this at project outset. Lessons Learned Every project needs an honest appraisal of project outcome, i.e. a lessons-learned close-out report. This should be a written audit, conducted three to six months after the formal project turnover. The appraisal findings need to be fully circulated. Burying mistakes does no one any good. Use the circulated close-out report as a catalyst to stop repeating the same errors again in the future. Incorporate lessons learned from prior projects into future projects. PROJECT CONTROLS Project control is a vital component of the project management system, and a big part of what Clients expect when they hire an E&C firm; i.e. the ability to organize and control a concentrated expenditure of money. Every project needs a timely, clear report of true project status from the project site each month and a clear forecast characterization of where it is trending. If an E&C company cannot do this, it cannot effectively manage projects. Project Controls System Project success requires adherence to a full suite of project controls throughout the life of the project. Controls have to be implemented at project outset; not part way through the project. The controls function is poorly understood by most Clients. Clients like to believe a “good” project

227 1

manager alone makes the difference. Intelligent Clients know that good controls create good project managers! A project control system is a complete and comprehensive process, whereby all aspects of project execution are monitored and reported against the originally approved scope of work, budgeted costs and project schedule. The foundation of the project control system starts with the “project execution plan”. It then flows through the project procedures manual and project control reporting documents to the conducting of formal review meetings and the regular issuance of required management reports. The project control system provides the timely data that gives the PM the basis for decisions regarding the activities of the project. Effective project control requires: A clearly defined scope of work at the front end of the project - and agreed to by all! An understanding of what is controllable Constant attention to the details Assignment of a controls manager Freeze of the scope of work, post project definition Circulation of regular weekly and monthly reports to all key stakeholders Full, open reporting of problems and risk issues Mandate of strict adherence to the change order procedure One person, the Project Manager, to approve change orders. The goal in every project is “no surprises.” This requires that a control schedule and a base cost estimate (control budget) be set early (in the first 90 days), followed by constant analysis and forecast trending.

Controls Issues between Client and E&C A project is controlled by the E&C contractor to a set scope, not to a budget! This is frequently a very difficult concept for the Owner to grasp - Clients are used to solely thinking in terms of budgets. The degree of Client approval needed in the various project execution steps must be established in writing and communicated to the E&C Project Manager and his staff. It is best to establish the progress-reporting philosophy for both Owner and E&C contractor prior to project start. Set the types and frequency of report, agree on a trending procedure; and how “percent complete” is to be measured and combined between activities. Project Changes The field cost engineer’s role is one of proper forward trending and planning, not of historical record reporting. Commitment reporting is mandatory and must be maintained current at all times. Procurement and contracts managers must route their data through project controls. Project control is a communications issue. It is imperative that cost engineers and accountants be kept apprised of project changes. An individual needs to be assigned to handle project controls in the field with the single responsibility of accurately getting the trends into the project reports on a timely basis. All projects need a formal change notification and change control procedure established up front, set in the procedures manual, and then proactively adhered to. The trend notification form and the change log, itemizing those changes are the primary control documents. Project staff need to properly and fully utilize the trending mechanism; the PM has to have the backbone to formally write and then timely submit all resultant change orders immediately to the Client. Late or nonnotification from field E&C personnel on the basis that a “delicate” or “friendly” Client relationship will somehow better facilitate future change submittals is simply unacceptable. Typically lump sum contracts require Client change notification within a specified, short elapsed time from the event. Delayed notification is most times tantamount to the E&C conceding any recovery for the change.

2272

Contingency Management A contingency I escalation drawdown management procedure must be formalized for all projects. Tracking of contingency is a project controls tool. Drawdown logs must be instigated on all projects. Tracking contingency (including escalation) against “remaining to commit” is a key indicator toward forecasting the health of a project. Transparent contingency management is requisite. Non-contingency slush funds (escalation, growth etc.) need eliminating. All non-specific funds (escalation, growth, risk allowance etc.) need to be on one budget line, within contingency. Audits Routine audits should be conducted throughout the project life cycle. A project audit provides formal oversight that evaluates adherence to project management “best practices” and gives independent confirmatory evidence of project progress.

REPORTS Regular weekly and monthly reports must circulate in a timely manner to all key stakeholders. These reports are an integral part of the controls function.

Monthly Report For North American corporations, the “executive summary” and the “issues and concerns” section at the front of the monthly report has to be in English, as well as the host language. Each report needs an overall project status curve and table comparing “overall actual‘’ against target each month, not just individual element progress. Each report also needs a contingency drawdown log. Overall project progress is preferably reported on an earned work value basis, but if this isn’t available, then manhours are generally the next best substitute, particularly for the front-end engineering phase. It is imperative to agree up-front on the relative weighting of the project elements, i.e. for the Engineering (E), Procurement (P), and Construction (C) activities utilized in reporting overall project progress. Monthly reports have to be readable, accurate, short and concise. Summary costs, a schedule with a table of milestones, charts, graphs and photos are preferred over repetitious text. Variance reporting within the monthly report must be timely. Shortfalls being buried as long as possible do the project no good. Honest variance reporting needs to be enforced. Negative issues have to be dealt with openly. Monthly reports are not external political documents published to give the Client a “warm and fuzzy” feeling. The importance of publishing honest reports needs to be established with the Client up front; particularly the pricing of cost trends, the delays in schedule trends, and the timing of when these are reported. The monthly report due date needs to be a set (preferably the same) date each month. Accommodation, however, has to be given for incorporation of the E&C’s accounting system into the monthly report. Client thus needs to agree up-front on each month’s cut-off date. Due dates have to be met, even if gaps show in the monthly report. Flash Reports Single-page flash reports showing project KPIs (Key Performance Indicators) should be issued by the end of the first business week of each month. Project Closure Reports Close-out reports should be requisite on all major projects. Manhours should be allocated for this at project initiation. These reports are a vital road map to project performance improvement.

2273

CONCLUSION Careful attention to the following: 0 0 0 0

Defining a clear scope prior to project initiation Setting of a control budget with appropriate contingency Establishing a milestoned schedule within a detailed plan of project execution Adherence to proper control and management of change over project life,

are all vital, fundamental ingredients to project success. With these elements firmly in place, then by following the strategies outlined within this paper, having the right team of E&C and Client individuals, and working under a partnered framework wherein all stakeholders are aligned to a common set of goals, any mineral project can be successfully completed anywhere on this globe. ACKNOWLEDGMENTS Certain strategies outlined within this manuscript are borrowed from prior unpublished works of the author and Terry Owen - VP Capital Projects, Inco Limited. Most of the strategies themselves were developed from real life Client experiences gained by Owen and Hickson while working on mineral projects around the globe with Kerr McGee, Freeport McMoRan, Cyprus Amax and Phelps Dodge. The author is grateful to the management and employees of all these organizations, as well as to prior employers, New Jersey Zinc, Asarco and Gold Fields, and to his present employer, the international E&C firm Kvaerner, for the education that they all so generously endowed over the past four decades. The opinions expressed within this paper, however, are solely those of the author. REFERENCES Hawley, R. February 14,2001. Making the best of Valuable Talent. Hawley Group Report 1. to the Engineering & Technology Board. London, UK. Editorial, February 2001. Leading the Field. Engineering First Magazine - Issue #15. UK. 2. Hickson, R. J. July 3 1, 2000. E&C Companies; the Mining Client Perspective. Unpublished 3. report to the Kvaerner Board, San Ramon, CA. Hickson, R. J. May 18,2001. Mining Project “Do’sand Don’ts” for E&C Firms and 4. Owners. Proceedings of 51” Annual MPD Meeting of the SME. Colorado Springs. Hickson, R. J. March 1,2000. Project Management for Dummies; How to improve your 5. project success ratio in the new millennium. Preprint No. 00-133 in Proceedings of 2000 Annual SME Meeting. Salt Lake City. Published in SME Transactions 2000, Volume 308. Owen, T. L., Hickson, R. J. September 1997. Project Management Manual for Mining. 6. Unpublished manuscript. Entwistle, M. Spring 2001. Delivering Value - Clean Fuels Alliance in Australia. Examples 7. of Project Gainshare. Kvaerner E&C Bulletin #6, page 13. Tinsley, W. J. October 1, 1999. Change & Challenge. IMM Presidential Address to 8. Yorkshire Branch of Institution of Mining & Metallurgy. Pontefract, UK. Wellcome Trust Survey, February 2001. Science and the Public: A review of Science 9. Communication and Public attitudes to Science. Office for Science & Technology, UK. 10. Charles, R. January 24, 2002. Best Practices Zero Accidents Program; Construction Industry Institute of USA. Kvaerner Environmental, Health & Safety Champions Meeting. London, UK.

2274

Pre-Commissioning, Commissioning and Training Timothy Watson, P.Eng., AMEC

ABSTRACT The preparation for the start-up and commissioning of a new facility needs to begin at the time the project is conceptualized so that adequate time can be built into the project master schedule to allow for the ramp-up to full production. Significant planning and preparation are required by both the owner and the engineer throughout the various phases of the project to ensure the smooth and successful start-up of the facility. It is not enough to begin the preparation for start-up a few months prior to the introduction of fresh feed if the owner expects to go into the start-up with a well trained and confident operations and maintenance team. The overall schedule for the project needs to developed from back to front; how the facility will be started up must drive the sequence of construction which will in turn set the priorities for the engineering and procurement of the new facility. The start-up and commissioning of a new facility is preceded by the following stages: 0 0 0

Owner's education program Mechanical completion Pre-commissioning.

It is essential that the owner's education program start long before the new facility is mechanically complete and it should continue throughout the plant checkout and precommissioning phases. Mechanical completion of a new facility involves the construction contractors and the owner's project management team with assistance from the pre-commissioning team. As each system reaches mechanical completion the pre-commissioning team will be responsible for the final checkout without the introduction of fresh feed. The introduction of fresh feed is usually the responsibility of the owner with assistance from the pre-commissioning team. This paper will provide an overview to the procedures and the responsibilities of the various participants involved in the mechanical completion, pre-commissioning and commissioning of a new facility.

INTRODUCTION The start-up of a new facility can be significantly enhanced by the development and implementation of a commissioning plan of approach. Some of the elements a commissioning plan of approach should address are:

0

0

Commissioning sequence Mechanical completion, pre-commissioning and commissioning requirements Performance tests Roles and responsibilities during construction, pre-commissioningand commissioning Owner's education and training program Safety.

2277

The development of the commissioning plan of approach during the initial phase of detail design will allow the requirements of plant commissioning and start-up to be built into the project master schedule. When the master schedule is developed, the sequence in which the plant will be commissioned should be the basis of the schedule. This paper will discuss the various elements that are key in planning for the successful startup of a new facility.

DISCUSSION Commissioning Sequence The development of a commissioning plan should begin during the initial phase of detail design and should continue throughout the detail design and construction phases of the project. The plan as to how the facility is going to be commissioned should play an integral role in the development of the project master schedule. The development of the master schedule should be based on the sequence in which the facility is going to be commissioned. The commissioning sequence will then drive the construction completion requirements, which will then set the priorities for the procurement and engineering activities. The transfer of care, custody and control (TCCC) of a new facility to the owner will be facilitated by the use of a construction testing, pre-commissioning and commissioning plan. The basis of the commissioning plan is the identification of the systems, both process and non-process, into which the project can be sub-divided, that will assist in the orderly completion of construction and pre-commissioning ahead of plant commissioning. The identification of the systems should consider plant layout, process requirements, mechanical, electrical and control systems requirements and any specific contract and/or sub-contract requirements. Depending upon the simplicity or complexity of the plant, the size of the plant and/or the requirements of the owner, the definition of the process and non-process systems can be presented in different ways. For a small plant with a relatively simple process, marked up flowsheets and single line diagrams can define the process a4d non-process systems. For a larger, more complex plant, the definition of the process and non-brocess systems can be accomplished by assigning each piece of mechanical equipment, each pipeline, instrument and electrical item to a process or non-process system. The presentation of this data should be supplemented using a narrative system description, color-coded marked-up process and instrument diagrams (P&ID’s) and electrical single line diagrams. To assist with the timely completion of construction and the transition into precommissioning and commissioning as soon as realistically possible, system-based scheduling should be used to drive the completion of construction. For the pre-commissioning and commissioning plan to be effective, the system definitions and a systems-based schedule needs to be presented to construction in a timely manner so that construction can complete their work on a system basis rather than an area basis. Mechanical Completion, Pre-Commissioningand Commissioning Requirements In the engineering and construction business, there are a number of different terms that have meaning in relationship to project completion and responsibility for the assets of the new facility. The terms may vary from industry to industry and country to country, but what is important is to ensure there is no misunderstanding with respect to who is responsible for the facility and the necessary testing requirements. For the purposes of this paper, the following terms will be used and defined: mechanical completion, pre-commissioning and commissioning. Mechanical Completion. Mechanical completion is generally defined as the installation of the facility in accordance with the contract, drawings, specifications and vendor documentation. There are a number of activities that fall within mechanical completion, some of which are: 0

Soils and compaction testing, particle distribution testing of backfill materials

2278

Torquing of structural steel bolts Hi-pot testing of power cables Ground resistance testing Die penetrant or x-ray testing of welds Cold alignment of rotating equipment Paint thickness testing Hydro-testing, flushing and re-installment of piping systems Chemical or mechanical cleaning of specialty piping systems Removal of rust inhibitors and installation of lubricant, greases and fluids Instrument calibration where possible Functional checkout of electrical equipment and instruments both locally and through the distributed control system (DCS) and/or programmable logic controllers (PLC) Motor run-in following the functional checkout of the motor control circuit Splicing and alignment of belt conveyors, belt feeders, apron feed feeders, drag conveyors, bucket elevators, etc. Final alignment of rotating equipment following motor run-in Completion of touch-up painting, pipe and equipment thermal insulation, pipe and equipment labeling Completion and turnover of all quality assurance and quality control (QNQC) documentation.

Pre-Commissioning.Pre-commissioning is the period when all of the process systems are run on air and/or water without introducing fresh feed into the plant or facility. All the equipment is run through the PLC and/or the DCS in order to verify all personnel safety, equipment safety and process interlocks are functioning as intended. In some instances it is not possible or desirable to run some pieces of equipment or entire process trains on air and/or water. In these instances it is still possible to verify a portion or all of the control system through the use of a process simulator or simulating the individual control loops utilizing signal generators. The pre-commissioning period is a good time to introduce the Owner’s maintenance and operating personnel into the checkout team. Maintenance personnel can assist in collecting data while the equipment is running on air and/or water. The data collected during this period can form the basis of the ongoing plant condition monitoring program. The control room operators can begin to get hands-on operating experience on a system-by-system basis without the pressure of running with feed. Pre-commissioning is the time when equipment vendor representatives assist with the checkout of individual pieces of equipment or entire process trains. The use of vendor representatives ensures the equipment is checked out and started up in accordance with the manufacturer’s requirements and also ensures the equipment warranties are not violated. Some of the activities that occur during pre-commissioning include: Verification of all personnel safety, equipment safety and process interlocks Drop tests can be performed to verify the accuracy of level transmitters, flow transmitters and pump capacities on water Conveyors should be run empty, the belts trained under no load conditions and the weigh scales should be calibrated Nuclear density transmitters should be zeroed on water

2279

0

Motor power andlor current draw and equipment baseline vibrations should be recorded while the equipment is running on air or water Preliminary loop tuning should be carried out where possible. It is important to realize that the majority of the loop tuning can only occur when the facility is running on actual process fluids.

Some of the other activities that occur during pre-commissioning to get the plant ready for the introduction of fresh feed include: 0

0

0

Dry out of refractory lined fireboxes for burners or dryers Dry out and heat up of refractory lined furnaces and reactor vessels Chemical curing of brick-lined hydrometallurgical vessels Chemical cleaning of boilers Cleaning of steam piping.

Generally, the completion of pre-commissioning activities signifies the transfer of care, custody and control (TCCC) from either the engineer or the contractor to the owner. The TCCC from the engineer or contractor to the owner is usually accompanied by a number of process system turnover packages. The turnover packages should contain the QNQC documentation that was prepared by the contractor as part of their mechanical completion requirements, the precommissioning test reports, any reports prepared by specialty vendors and a copy of the contractually agreed upon as-built drawings. Commissioning. The start of the commissioning period of a new facility is usually signified by the introduction of fresh feed and extends until the new facility has reached full design production or a percentage of design production. The owner is usually responsible for both the facility and the necessary operations and maintenance activities during the commissioning period. Many times a new facility will ramp up to full production quickly without experiencing significant difficulties. At other times, the time to get a new facility to design production may be extended due to a number of different factors or a combination of factors, for example: 0

0

0

0

The bulk sample taken for the testwork program may not be representative of the entire ore body The friction factor of the flowing fine ore or concentrate may not have been incorporated into the design of fine ore or concentrate mass flow bins Screen harmonics may not have been dampened properly when supporting one or more vibrating screen within a steel structure Pumps that require their motors or impellers changed due to the solution or slurry specific gravity or froth factors being different than those used for the design Atmospheric or climatic data may not have been incorporated into the design correctly.

Performance Tests Some contracts may have specific performance test provisions that dictate the facility is run at nameplate capacity, or a percentage thereof, for a period of time. The performance test is usually run after the facility has been commissioned and started up and is conducted by the owner with assistance from the engineer or process licensor. During the performance test it will be necessary to record confirming data. To ensure the accuracy of the data critical instruments should be recalibrated prior to the commencement of the performance test. If the method of calculating and evaluating the results are not defined in the contract, they should be agreed to before the initiation of the performance test.

2280

Roles and Responsibilities During Construction, Pre-Commissioningand Commissioning The roles and responsibilities of the owner, engineer and contractor during construction, precommissioning and commissioning should be contractually defined and understood by all. Construction. Through to mechanical completion the contractor and/or contractors have prime responsibility for the assets of the new facility. The individual contractors should be expected to use the process commissioning sequence and scheduling information in planning their activities and work with other contractors to complete large multi-contractor systems. The project quality assurance/quality control ( Q N Q C ) documentation should be utilized to document the various tests or measurements conducted throughout the construction period. The main interface with the contractors is the project construction management team. As construction progresses, members of the pre-commissioning team should participate in the functional check-out of all equipment and control loops by providing assistance to the contractor through the operation of the DCS and/or PLC. Having members of the pre-commissioning team assisting the contractors with the functional checkout can reduce the overall duration of the schedule by eliminating the independent verification of the plant electrical and control system wiring.

Pre-Commissioning. The completion of the last 15 to 20% of construction should occur on a system basis rather than an area basis. This approach allows pre-commissioning activities to commence as early as possible on the project because the remaining construction activities will be completed on a prioritized system basis. Prior to the pre-commissioning of a system all the pieces of equipment and components that make up the system should be visually inspected for completeness and assurances should be provided that all the required Q N Q C testing and documentation has been completed. Utmost care needs to be exercised during the pre-commissioning period as new equipment is put into service or heated up. Many of the construction workers may still be on the job completing the last of the systems or punch list items and the operation of equipment may present many new hazards to the construction workers. The pre-commissioning of a system should continue for a period of time necessary to prove out the system. The system should be considered complete when all the mechanical, piping, electrical, instrumentation and control systems have been operated, checked and calibrated, and are in a state that commissioning and/or start-up can commence. During the pre-commissioning period, the project construction manager is usually still responsible for the site, but there is significant interface between the construction management team and the pre-commissioning team to ensure all tests are properly scheduled, coordinated and conducted in a safe manner. The pre-commissioning period is a good time to begin to get some of the Owner’s operation and maintenance personnel involved as part of their education program prior to start-up. Commissioning. The commissioning phase of a new facility usually commences with the introduction of fresh feed into the facility. During the commissioning phase the owner is usually responsible for both the operation and maintenance of the new facility. It is not unusual for members of the pre-commissioning team, construction management team and selected members from contractors to be requested by the owner to provide support during the commissioning phase. The additional manpower during the commissioning phase of a new facility is required to assist with those activities that are not usually encountered during the normal operation of an existing facility, for example;

PLC or DCS support may be required to assist with the final loop tuning once fresh feed has been introduced. For those facilities containing horizontal grinding mills, millwright support may be required to re-torque the head and shell bolts after the grinding mills have operated for a defined period of time.

2281

a

For plants containing equipment driven by V-belts, the V-belts will need to be retensioned a number of times as the belts go through the initial stretching. PLC andlor DCS support may be required to assist with troubleshooting configuration problems that were not identified during the pre-commissioning period. Variable speed drives may need to be re-ranged once actual operating conditions have been encountered during the initial commissioning period.

Over and above the engineering and craft labor support that is required to deal with the additional work load that normally occurs during the commissioning period, problems may be encountered that also require additional support. Examples of the types of problems that may arise during the commissioning period are: Replacement of screen decks on sizing screens with improperly sized apertures Replacement of pump impellers or motors that were improperly sized or that require resizing due to the actual conditions being different than those allowed for in the initial system Pipes or pipelines that were improperly sized Premature replacement of wear components because the ore or slurry is more abrasive than initially anticipated Premature replacement of refractory lining systems because they were over stressed or cycled during the initial commissioning period Replacement of instruments, electrical or mechanical components that failed prematurely during plant commissioning. Despite the various problems that may be encountered during the commissioning period, a well-trained operations and maintenance staff provided with some additional support can deal with most situations quickly and efficiently.

Owner’s Education and Training Program The start-up of a new facility that has been designed and constructed properly will be significantly enhanced by an operator team that has been well trained and possesses the skills and confidence to operate and maintain the new facility. To accomplish this goal the owner’s education and training program needs to start long before the facility is mechanically complete and should continue throughout the pre-commissioning and commissioning phases and throughout the life of the operation. Depending upon the size of the operation, a full-time training coordinator should be an integral part of the owner’s operating team. To be effective, a comprehensive training program should consist of the following elements; classroom training, on-site practical field training, and possibly off-site training. As part of the overall training program, the training opportunities that exist during mechanical completion and pre-commissioning should not be overlooked. Examples of the many different training opportunities that exist as part of mechanical completion or pre-commissioning are summarized below: During the functional checkout or bumping and running in of motors the control room operators can participate in the checkout by calling up the appropriate PLC and/or DCS screen to initiate the start or stop signal for the motor. As the plant checkout progresses from functional checkout and the bumping and running of motors to the precommissioning activities with entire systems or sub-systems, the control room operators can obtain significant hands-on experience prior to plant commissioning or start-up.

2282

More and more in the design of new facilities the owner’s people responsible for the plant control system are participating in the initial configuration of the PLC and/or DCS. Their participation may include the development of the operator interface displays, controller configuration or possibly the configuration of the entire control system. Whether their involvement in the initial configuration is minor or significant, they should be expected to participate in the verification of the plant control system during the functional checkout and pre-commissioning phases. Having the owner’s personnel participate in PLC and/or DCS verification will provide them with the confidence that the plant control system was configured as intended and the various input and outputs (YO) are reporting to the correct locations within the control system. During pre-commissioning many equipment vendors are brought to site to assist with the checkout and initial operation of individual pieces of equipment or entire processes. When the vendors are onsite, it is an ideal time for both the operations and maintenance staff to receive additional classroom training or field training for that specific piece of equipment directly from the factory representative. For electricians and instrument mechanics there are a number of different precommissioning activities they can participate in that will enhance their training ahead of commissioning, for example; - Field calibration or calibration verification of instruments - Trouble shooting of field devices that may not be functioning properly - Assist with recording data during the initial equipment run-in - Assist with trouble shooting protection relays for medium or high voltage switchgear or large horsepower motors. Similarly for mechanics and millwrights hands-on training opportunities exist during precommissioning, for example; - Assist vendor representatives with checkout and start-up of specialty equipment - Assist in collecting and recording baseline vibration data that will be used as part of the ongoing condition monitoring - Many pieces of equipment need to be re-greased or have the oil changed after the first few days or few hundred hours of operation. Assisting with this exercise can provide hands-on experience with many different pieces of new equipment - During the initial run-in of conveyors the belts will need to be trained. In summary, the functional checkout period of mechanical completion and the various activities during pre-commissioning can provide excellent hands-on training opportunities for the owner’s operation and maintenance staff. Coordinating these activities with the owner requires a considerable amount of effort as many other activities will be ongoing simultaneously as the owner is preparing his people for start-up. Safety

The conditions that exist during the construction phase of a project are well understood by the construction and safety professionals within the industry and the necessary work processes and procedures are implemented on construction sites to assist with worker safety. The hazards that arise during the pre-commissioning and commissioning phases of a new facility are different from those encountered during construction and special precautions should be taken to safeguard the remaining construction workers and operating personnel. On most sites, a lockout tag-out procedure is used to assist with the identification of responsibility for an individual piece of equipment. The same procedure also identifies the potential hazards associated with a piece of equipment during the testing phase. For example

2283

purposes, a three tag lockout tag-out system utilizing the colors red, orange, and green will be reviewed. As construction progresses and the construction testing phase is approaching, red tags should be attached to all pieces of equipment - mechanical or electrical, all pipeline isolation valves, and all instruments. The red tag indicates the piece of equipment is under construction and under no circumstances can the piece of equipment or pipeline be energized or put into service. As construction progresses and the piece of equipment goes into the testing phase, the red tag is replaced by an orange tag. The orange tag indicates the piece of equipment - mechanical or electrical, valves, or instruments - can be energized or put into service at anytime. The orange tag is a visual indication to the construction workers that they are near a live or potentially live piece of equipment. Should they be required to work on that piece of equipment, a work permit is required. As an additional safeguard during the testing phase, precautions such as barrier tape should be used when running in a motor or performing a load test on a piece of equipment. Upon the completion of the testing phase and the individual piece of equipment being turned over to the owner, the orange tag is replaced with a green tag. The green tag indicates that the piece of equipment - mechanical or electrical, pipeline or pipeline isolation valve or instrument - has been turned over to the owner. Should the contractor be required to work on a piece of green tagged equipment, a work permit should be obtained from the operations group. Whenever a work permit is required to work on either an orange or green tagged piece of equipment, it is important to follow the appropriate lockout procedure. The potential hazards associated with the initial commissioning phase of a new facility may be far greater than a facility that has been in operation for an extended period of time. The increased risk may be associated with lack of familiarity of the new facility by the operations and maintenance staff, or increased congestion on-site due to the additional personnel retained to assist with plant commissioning. During this initial commissioning period, it is essential that the owner’s safety personnel be fully trained in all aspects of fire fighting, rescue and evacuation, utility isolation, and the health hazards associated with various reagents that may be used and stored on site.

2284

Plant Ramp Up and Performance Testing Roger M. Nendick I

ABSTRACT The economic success of a mining project can be significantly impacted by the rapidity with which the plant gets to its full operating performance, both in terms of production and cost. The time taken for the plant to get to full production from the time ore is first introduced into the facility is referred to as the ramp up period. In extremes this time can range from days to years. The financial institutions providing the capital for new mining projects recognize the importance of this ramp up period to the overall project economics and usually insist on a performance test to demonstrate that the facility is operating as per design. There is usually a significant financial incentive for the owner to pass this test and so, in turn, the owner normally has a performance test as part of his contract with the E&C Contractor. The organization of the Ramp Up, and the execution of these performance tests becomes a mini project in itself.

INTRODUCTION A mining company makes a decision to proceed with a new mining project. Funding for the project has been arranged with a consortium of financial institutions who, in an attempt to place some assurance on receiving their money back, have imposed a performance test on the mining company to ensure that the property performs according to its financial parameters. As an incentive for the mining company to pass that performance test, there is usually a financial incentive in the form of a reduction in interest rate, or changing of the loan status from recourse to non-recourse once the test has been completed satisfactorily. The mining company then negotiates a contract for the construction of the facility with an engineering and construction company (contractor). In order to reduce its risk, it passes on as much as the performance test as it can to the engineering and construction contractor; how much it is able to pass onto the contractor is all part of the contract negotiation. With the contract signed, the project proceeds to construction stage. The sequence of events during the course of construction are as depicted in the bar chart shown below.

1

Vice President, Consulting & Studies, Fluor Mining & Minerals

2285

The final months of construction are typically a frenzy of activity on both sides: The contractor is busy trying to finish the physical construction and to do the precommissioningof the various units of the plant in order to hand them over to the owner; The owner in turn is trying to recruit the necessary labor force and to order in the spares and operating supplies for the operation of the plant. Often, but not always, insufficient attention is given to the Commissioning of the plant, the ramping up of the plant and the execution of the performance testing. As a result the commissioning and ramp up take place with much anguish and acrimony, and the performance test is forgotten about until some one turns a page in the contract and realizes it has to be carried out. The above describes the nightmare, but one that is all too frequent in the history of new mining projects. For the rest of this paper I would like to deal with how the agony of this kind of start up can be avoided.

THE CONTRACT The ideal situation would be that the owner has identified his metallurgical superintendent and that the contractor has identified his commissioning leader and that both of these people are involved in the contract negotiation, particularly with regard to the performance test. If this is not done, then frequently the terms of the performance test will render it extremely difficult and costly to perform. The practicality of executing the performance test needs to be considered at the time of writing the contract. Lawyers are ignorant of plant operation. The test needs to be designed to minimize the disruption of normal plant performance. As an example, tests that require samples to be taken from the SAG mill feed belt should consider that stopping, locking out, sampling and restarting the belt can take as long as 30 minutes, during which time there will be no feed to the mill. This alone is equivalent to over 6% downtime at only one sample per eight-hour shift. Test Duration Typically Performance Tests are specified for two time frames: A long duration test typically for 60 days but sometimes 90 days or more, which requires the plant to operate at an average of 90% of its rated capacity during this time; A short duration test where the plant is expected to perform at 100% or even 110% or 120% of its average rated capacity. The purpose of the latter test is to ensure that the plant has some catch-up capacity to make up for occasions when it has performed below the average rate.

Definitions It is absolutely vital that the expectations of plant performance are clearly spelled out in the Contract. Throughput must be specified in terms of average values and at peak design values, and the derivation of the plant availability should be clearly spelled out. The performance test expectations must clearly state which values are being measured for each test. Where a recovery and product quality performance test is also to be provided, the ore grades and the expected impurity levels, as well as the physical and lithological characteristics of the ore must be clearly specified. It is prudent for the owner at this time to make sure that the mine can supply ore with the required physical and chemical characteristics at the time that the performance test will be conducted. If the mine plan does not permit the mining of this typical ore when required, then further discussion and debate will be required at the time of the test in order to agree upon acceptable feed material.

2286

As well as specifying the main throughput parameters for the entire plant, the contract should also spell out the expectation for the various unit processes within the plant. In a typical flotation concentrator, for example, the capacity of the concentrate filtration plant may well be significantly different from the design average, in order to account for varying ore grades during the life of the mine. Responsibilities The contract must clearly spell-out who is in-charge of each stage of the project. Typically the contractor has responsibility for the plant up until the end of pre-commissioning. When the precommissioning packages are signed over to the owner together with care custody and control, they become the owner's responsibility. When all of the pre-commissioning packages have been handed over the owner generally takes control of the complete facility along with responsibility for the commissioning. An exception to this is if the plant has been built on a lump sum turn key basis, in which case the contractor may have responsibility up until the completion of the performance test. Whatever the case, it is important that both parties understand exactly who is responsible so that responsibility for equipment damage, along with insurance responsibilities, is clearly understood.

ENGINEERING Once the contract is signed and the legal details of the performance test agreed, it is important that it is not forgotten during the engineering process. Frequent changes may be made to the plant design, or to the scope of the facilities, that can impact on the performance of the test. The owner and contractor should schedule meetings approximately monthly to discuss progress on how the test will be executed, as well as any engineering matters that have arisen that impact on the test. Planning for how the test will be executed can be started at this early stage. At a point midway through the engineering, the metallurgical superintendent on the owner's side may wish to appoint a performance test manager whose sole responsibility is the execution and passing of the performance test. Similarly on the contractor's side the commissioning manager for the contractor may also wish to appoint an individual to specifically look after the performance test. The planning and execution of the test is a mini project in itself. As an example of this, the performance testing for a major concentrator in South America which was undertaken recently, necessitated the hiring of 18 additional temporary staff on the owners side in order to handle the sampling, sample preparation and analysis of the samples required for the test. The requirement for sample storage in the event of dispute over the test result required the purchase of 1800 steel drums for bulk samples, and a shipping container for pulverized samples. The tests themselves required individual tests at locations at the concentrator, the filtration plant and the port shipping facilities. These facilities were geographically spread over a IOOOkm distance. COMMISSIONING PLANS It is never too early to start the commissioning plans for a project. Although the commissioning plan is typically the responsibility of the owner, it is as well that he also involves the contractor to get his buy in to this important phase of the project. It is often claimed that commissioning plans are futile because "we don't know what is going to happen once we push the button". The inference of this is that it is impossible to plan for the unknown. There are many advantages to the preparation of a commissioningplan: It focuses the attention of all parties on the task that lays ahead; It enables plans to be made for the recording of the many items of data that need to be collected about the performance of individual pieces of equipment; It provides the basis for a platform from which to build in the eventuality that something does go wrong;

2287

It provides a document that will become a written record of all the events that took place during commissioning. Items that need to be considered in a commissioning plan, include: If there are is more than one module in the plant, which module will be started first (obviously this will require liaison with the construction and pre-commissioning teams); Will the plant be started at its full tonnage for each module or will it be started at a reduced tonnage; Are there any special ore feed requirements that need to be addressed, for example special ore requirement for bedding in of stock piles or bedding in of thickeners etc; What will be the water supplies for start up of the plant? Frequently there is no return dam water available during the initial operation of the plant and this requires additional fresh water supply to be available.

RAMP UP The period from the first introduction of ore into the plant through until the plant reaching its ultimate design capacity, both in terms of quantity and quality of product, it is referred to as the ramp up period. Depending on the complexity of the design and the degree of preparation and planning that has gone into the start up, the time necessary for this to be achieved can require anything from a few days to 2 or 3 years to reach its full capacity. There are no hard and fast rules for defining this period, but an excellent paper by McNulty' provides good guidelines. A graph from that paper, reproduced below, shows the ramp up times derived from studying the case histories of various types of plant. Series one through series five shows increasing plant complexity from simple flotation concentrators to integrated hydrometallurgical facilities.

RAMP-UP TIME

2

"Developing Innovative Technology" by Terry McNulty, Mining Engineering, October 1998.

2288

As the tonnage is gradually being increased, bottlenecks will appear in various areas of the plant. Below is a listing of some of the more frequent bottlenecks to production, but each plant ramp up will have its own particular array of problems. General items include: Wrongly sized pumps (This is not always that pumps are too small. Frequently pumps that are too big can be as big a barrier to production); Undersized motors for equipment. Poorly designed transfer chutes that either plug or wear out too quickly; Pipelines that sand out or wear out preventing reliable material transfer; Excessive spillage and inadequate sump pump arrangements (Note: this is difficult to design for, as some pumps that are adequate for handling routine spillage when a plant is in normal operation may be inadequate for the excessive spillage that often occurs during the start up. Although this can be an irritation during a plant start up it is as well not to overreact and install massive pumps that become a problem when the plant reaches routine operation).

RECORD KEEPING It is extremely important that both the owner and contractor keep good and accurate records of events during commissioning. The owner should invite the contractor's representative to attend the daily production meeting for the operation and accurate minutes should be kept at these meetings. A further meeting should be held on a regular basis between the owner and the contractors commissioning representatives to set the priorities for repair, rectification and punch-listed items. Demands on maintenance crews during the project ramp up are often excessive. It is prudent for the owner to arrange for a temporary SWAT Team of maintenance people to attend to project work arising from the ramp up. In this way the regular maintenance requirements of the plant do not get behind such that the long-term reliability of the equipment suffers. THE PERFORMANCE TEST By the time the performance test is ready to be conducted, the plant should be operating in a reasonably stable manner. Notification of the date of the test needs to be agreed a few weeks in advance, so that the necessary arrangements can be made for the test. Also the Independent Engineer for the financial institutions needs to be given time to get to the property to hold some preliminary meetings. The independent engineer becomes a third member of the team conducting the performance test and will need to be given full sets of records for each day's production. Frequently the performance test also requires the plant to perform within certain cost parameters and this will necessitate the Independent Engineer being given data on reagent consumptions, power consumptions and other components of the operating cost. Hopefully sufficient trust is built between the three members of the team, such that all individuals do not consider it necessary to be present on site throughout the entire test, i.e. 60 days or more. At the end of the test period it becomes a major exercise to gather together all of the information and produce a performance test report. Explanation of any unusual events that have occurred during the performance test may be required if the requirements have not been exactly met during the 60-day period. A well documented and written report may obviate the need for reperformance of the test with its attendant cost. CONCLUSION The success of both plant ramp up and performance tests can be reduced to two 'P's: People and Planning The Planning needs to start before the contract is signed and continue right through the execution. The People can cause failure or create success, despite everything that has gone before. It is close to impossible to successfully ramp up a plant and conduct a successful performance test unless there is close cooperation between all parties.

2289

Preparation of Effective Operating Manuals to Support Operator Training for Metallurgical Plant Start-ups Stephen R. Brown

ABSTRACT Effective plant operating manuals used in a formal training program can make the difference between a successful start-up and a failure. Once the plant process design and control strategies have been fixed, equipment has been ordered, and the plant is under construction, the only major variable affecting success is the capability of plant operating personnel. It is essential that the myriad details concerning plant operation are documented in comprehensive operating manuals suitable for training the non-technical personnel who will operate the plant. This paper describes the best approach for producing the operating manuals and conducting operator training.

WEUT OPERATORS NEED TO KNOW Performance Associates has been in the business of assisting mining companies to successfully start up new mineral processing plants for the past 18 years. Based on this experience, it is obvious that certain, key information must be known-and applied-by the operators and front-line supervisors during the start-up. Failure to impart thls information, and to apply this knowledge during the commissioning phase, will likely result in either outright failure or a long, agonizing, and protracted effort to achieve design capacity-if design capacity can be achieved at all. This key information consists of several elements. Process Unit Operations First, it is essential that the operators have a conceptual understanding of the process and the principle of operation of each major unit operation in their area of responsibility. Conceptual knowledge allows for more effective reasoning when process upset conditions occur. Rather than attempting to provide a recipe covering any conceivable upset, the operator’s conceptual knowledge will allow for drawing the appropriate conclusions based on the situation at hand. Specifically, the following elements concerning the functioning of individual unit operations should be documented and thoroughly understood by front-line supervisors and operators. Objective. Describes the purpose of the unit operation. For example, the objective of a ball mill circuit is to reduce the size distribution of feed material to allow for liberating minerals locked in the host rock. Basic Theory. Describes the chemical, mechanical, electrical, etc., methodology (e.g., magnetism, chemical reaction, mechanical action, differential density) used by the unit operation to effect the objective without reference to the physical layout of the equipment. Principle of Operation. Describes the physical layout and how the basic theory is applied by the actual equipment being described. The inclusion of diagrams, photos, etc., as necessary, to illustrate the important principles of operation is required. Critical Variables. In any unit operation, the output quality is a function of certain critical variables. For example, cyclone feed density and pressure drop affect how effective a cyclone is at making a size differentiation of the feed slurry. This element identifies the important variables associated with each unit operation described.

2290

Safe Job Procedures To ensure each employee works safely, information on the correct methods for performing potentially hazardous jobs must be learned. Each new plant operation will contain potential hazards that must be understood by every employee. These hazards include working with various reagents, such as sodium cyanide, caustic, sulfuric acid, etc. They also include working around various types of moving equipment. Process Control Each operator must also fully understand each control loop in his area of responsibility. This understanding includes the variable being controlled, the instruments and control strategy employed, how to recognize when control problems occur, the backup options available, and when it is appropriate to exercise those options. Distinguishing between process problems and process control problems is also important. This understanding has become more difficult over the recent past since control strategies have become increasingly complex with the advent of more and more powerful process control software. Interlocks In addition to the control loops, all interlocks must be thoroughly understood, including how interlock logic is affected by various operating parameters, such as remote operation, local operation, maintenance operation, etc. We have seen a very significant increase in the complexity of plant interlocks being designed into new plants.

Alarms Once the process and its critical variables are understood, along with the controls and interlocks, the operator must then learn the fault, cause, and remedy associated with each alarm. This learning can be a tall order since many of today’s new plants have literally hundreds of programmed alarms in each plant area. Start-up and Shutdown Procedures Each operator must also learn the correct steps to start up and shut down the plant under various conditions. These conditions normally include: start-up from complete shutdown, start-up from standby shutdown, start-up fi-om power failure, and start-up from emergency shutdown. Additionally, each operator must know how to manipulate the distributed control system @CS) to determine what is happening in the process, to take control of a particular controller, to adjust set points, etc. Operators must also know how to effect control using any local control panels in the plant. These panels are typically used for packaged boilers, samplers, and solution heaters. The new complexity of metallurgical plants make all of these procedures much more involved than they used to be. Operator Tasks Finally, each operator must learn other operating procedures associated with optimizing the plant. These procedures include such tasks as checking pulp density, optimizing flotation cell performance, tapping a furnace, taking solution samples, conducting routine inspections, and ensuring that the plant operates within permit requirements. Conclusion Operators having any limitations in the above described knowledge will cost the company during the start-up and subsequent initial operation-the more the knowledge gap, the more it will cost. In plant start-up situations where very little of this knowledge has been transmitted to operators, the start-up can be little short of disastrous. In a typical start-up scenario, personnel react inappropriately to process upset conditions, causing further upsets. These upsets result in a new series of inappropriate reactions, sometimes including physical plant changes. Many times a never-ending cycle of operator reactions causing problems-resulting in different reactions-

229 1

causing more and more serious problems, occurs. Once this cycle has started, it can quickly get out of hand. Just getting back to the base plant condition can be virtually impossible. To add to these problems, once the plant actually starts and problems develop, there is no time left to train the operators. The problems begin to multiply; since everyone is working extra hours to deal with the problems, there is no chance to catch up. Operators are left to absorb the necessary information by trial and error while dealing with the start-up problems. In some cases, more complex plants never do successfully start up. Simpler plants may eventually operate at production capacities approaching design, but only after long, arduous start-up periods. Generally, operators of plants started under these conditions all have their own pet methods for controlling the operation. We have observed many plants where critical variables are In some cases, even the controlled with entirely different home-grown strategies on each sM. target values are different.

AN ACTERNATIVE APPROACH Introduction Few people would dispute the necessity of transmitting literally thousands of critical pieces of information about the new plant to the operator. In fact, there is really only one way to actually do it. We have found that writing a series of custom plant operating manuals, specifically designed for an education level of the target employee pool, is the correct approach. These manuals are then used in a formal classroom training program, complete with graphc support, workbooks, and tests. The training must occur prior to mechanical completion. Ideally, the trained operators complete the class and field training and then assist with the final stages of preoperational testing. Only then are they ready to introduce feed and perform their normal operating functions. For our plant operating manuals, we have found that the following contents work well, both for training, and as a continuing reference. Operating Manuals Contents Introduction. This section describes the purpose of the manuals and identities those volumes in the set. It also illustrates the scope of the particular manual volume. Use of the Manual. Ths section describes how pages are numbered and how to find information. It is important that the manual contents are well organized and it is easy for an operator to find the informationneeded. Safe Job Procedures. This section provides formal written procedures, including any special equipment required, to be followed by the operator when performing potentially hazardous job functions. Process Design. Thls section provides a written description, complete with necessary schematic dagrams and illustrations, describing the process. It also includes principles of operation for all of the major process unit operations, Refer to Figure 1 for a typical graphical illustration. Unit operations such as vacuum pumps, flotation cells, furnaces, dryers, filters, etc., are described. It is essential that the operator is provided with the necessary information so he or she can describe the key operating principles. This section also provides an equipment list and color flowsheets. Each process flow stream is illustrated with a different distinctive color. Process Control. This section provides a table i d e n m n g each critical process variable, such as temperatures, flows, pressures, densities, etc. It also summarizes their target values, methods of control, and their impact on the process. Following the process variable table, each control loop is described using text, a simple block diagram, and a simple loop diagram extracted from the piping and instrument diagram @&ID). Each method of control such as automatic remote set point, automatic local set point, and manual, is discussed, as applicable.

2292

Figure 2 Typical Illustration of a Process Control Loop Interlocks. This section provides tables identifling all interlocks and permissives for each motor and affected instrument, along with cause-and-effect diagrams. The diagrams cover the

2293

same information as the tables and are used to complement the tables. The interlock tables and diagrams are organized by logical process system.

Figure 3 Typical Illustration of an Interlock Alarms. This section illustrates each alarm in the process sorted by tag number. It is in a tabular format and includes the affected equipment, the fault, the potential causes, and the steps to take to remedy the alarm. Operating Procedures. This section is divided into three sections: Start-up, Shutdown, and Operator Tasks. Start-up describes the detailed procedures necessary to start up the plant from a complete shutdown, from a standby shutdown, from a power failure, and from an emergency shutdown. Shutdown describes the procedures necessary for a complete shutdown, a standby shutdown, and an emergency shutdown. It also describes the effect of a power failure and any specific procedures the operator should perform. Operator tasks describe additional procedures required to operate the plant. These procedures always include preoperational inspections necessary to set up the plant for start-up. They also include procedures necessary for the operator to perform his job function. They may also include any steps the operator must take to manually control key process variables. Typical operator tasks for various job functions are: 0 0

0

Manual boiler blowdown. MeasuringpH. Preparing a batch of flocculant. Furnace tapping. Shift inspection.

An operator task procedure is important whenever consistency is critical. DEVELOPING THE MANUALS Writing each manual is a tedious and involved process. The following source material is needed:

2294

Process flow diagrams. Piping and instrument diagrams. Equipment operatingand maintenance instructions. Functional descriptions. Motor control schematics. Control valve specification sheets. Design criteria. Equipment list. Alarm list. This material is used to develop each of the sections previously described. Manual writers must be experienced in plant operation and should be good writers; this is often a difficult combination. Typically, several months must be dedicated to the manual writing process. In many cases, engineering changes are still occurring as the manuals are being prepared; this adds to the complexity. Once the manuals are completed, we suggest preparing an accompanying training module for each manual. The training module optimizes use of each of the manuals in a formal classroom instruction setting.

TRAINING MODULE Learning Objectives and Mo.dule Outline This section provides a list of the objectives that the trainee should be able to accomplish once the training is over. The module outline provides the instructor with an outline of the manual and a suggested time duration for training on each section. Overhead Transparencies or Computer Projector All graphics in the operating manual are made into overhead training aids for use during classroom instruction. Workbook The workbook is a learning device comprising a series of fill-in-the-blank-type questions which the trainee answers while referring to the operating manual. It is used to reinforce learning after the material is covered in a traditional lecture. The instructor is provided with an answer sheet. Knowledge Assessment Test (Theory Assessment) The knowledge assessment test is a validation device designed to determine how much of the material was learned by the trainee. It comprises multiple choice and true-false questions and is given after the module’s classroom instruction is completed. Results can be used to determine if remedial training is required; they can also be used to determine where individual operators are ultimately placed in the operation. Qualification Checklist (Practical Assessment) The qualification checklist is designed to validate that the trainee can apply the theory learned in the classroom on the job. It is completed by the trainees’ immediate supervisor during the initial stages of operation. It can be used in combination with a probationary period during which the operator proves he or she can accomplishthe job functions required. Once the modules have been completed, the next step is to conduct classroom training. CLASSROOM AND IN-PLANT TRAINING It is important to use credible personnel with previous experience in plant operations for training instruction. Ideally, the personnel who have prepared the manuals and modules should carry out the classroom instruction. In many cases, we use our personnel to conduct train-the-trainer

2295

classroom sessions for the client’s trainers, then they, in turn,train their plant operators. We have found this approach very successful. The formal training consists of three components: 0 0 0

Classroom lecture. Trainee completion of workbooks. Site visits to observe the plant equipment and instrumentation.

We have found that the lecture, workbook, and site visits work best when they are distributed throughout the training day. Too much time in the classroom can dull the learning process. During the classroom phase, it is important to get the trainees involved. Trainee participation results in better retention and makes for a more interesting experience. Near the end of each module, simulation dnlls can be held. These drills require that the group is split into teams. Each team then attempts to determine the cause of hypothetical process upsets postulated by other teams or by the instructor. The simulation drills require knowledge of the full breadth of information contained in each manual. Once the formal classroom training sessions are completed, additional time can be spent in the field tracing pipelines, identlfying every control valve and instrumentation element, and generally marking up P&IDs as instruments, equipment, and pipelines are identified. The fmal phase of training is trainee participation in preoperational testing prior to introduction of feed. Operators, having completed training, are extremely knowledgeable about the new plant. They make ideal personnel to walk the plant and prepare punch lists of discrepancies. When functional testing of completed plant systems occurs, operators can also participate in that testing. Ideally, the new operators can use the distributed control system or local PLC controls to operate the equipment necessaq under the direction of appropriately qualified engineering personnel. As problems are idenMied during the testing phase, the new operators and supervisors can participate in problem-solving teams investigating the problems. We have found that there is no substitute for highly trained operating personnel during the testing and start-up phase of any new plant. The training program described above will provide those highly trained operators and supervisors. We know of no other satisfactory method for ensuring that your personnel are ready to operate the new plant. COMPUTER-BASED TRAINING The current state of computer technology allows for taking the plant operating manuals justdescribed to the next step-either a web browser-based interface allowing for hyperlinks to navigate the manuals on a company intranet or a full-blown, interactive multimedia interface. Web Browser Interface Using a web browser version of the manuals, the user can click on the manual desired to obtain the detailed manual table of contents. The user then clicks on the hyperlinked table of contents item to access that section of the manual. Lnhvidual manual subsections are accessed in the Same way by cliclang on hyperlinks. The text and graphcs in the hard-copy manual are the same as those accessed by the computer. Hyperlinks can also be added to any references. In other words, a hyperlmk could be provided whenever another section of the manual is referenced such as a safe job procedure. Therefore, the user can simply click on the hyperlink reference to go directly to the referenced section. The user can then return to the previous section by using the Back button on the web browser. The web browser version is essentially the same as the hard-copy version; the difference is that it is electronic and all volumes, sections, subsections, etc., are accessed by hyperlinks.

2296

Interactive Multimedia Version The multimediaversion contains the same information contained in the hard-copy manuals, but it is in a Microsoft Windows environment. Specifically, the multimedia interface is designed as follows. Safe Job Procedures. Each safe job procedure is selected from a drop-down menu selected from the main bar menu for each area. The safe job procedures are in text format, with hyperlinks to graphics, if appropriate. Process Description. In the multimedia application, the process components of each sitespecific area, such as grinding, flotation, etc., are divided into process systems. A text box containing the process description for each system can be scrolled down the left side of the computer screen. The center of the screen is used to display graphics associated with the process description. The graphics box displays color flow diagrams from the operating manual, along with any other appropriate illustrations including schematic diagrams and principles of operation. In addition, hypertext links are accessible and provided at pertinent points in the process description text to allow the user to link to glossary definitions, relevant principles of operation and their associated graphics, full-motion video, digital photographs, and other useful illustrations. These hyperlinked objects appear in the graphics box. This version can also be provided with a voice-over narration of the process description. The script for the voice-over appears in a text box on screen so the user can follow along. As the voiceover proceeds, the graphics in the graphics box automatically change to illustrate what is being discussed in the narration. Process Control. Process variables can be selected from a menu box that appears on the main screen for each site-specific system. Each variable is shown in a list box, and text boxes provide information regarding the target range, control method, and impact on the process for the variable selected. Control loops in the system are selected from a menu box. The loops each include text and diagrams. As in the process description, the user can scroll down the text on the left side of the screen. The graphic box illustrates the loop dlagram or a simple block diagram depending on the user’s selection. For automatic sequence controls, if applicable, animations of the sequence can be provided. Interlocks. Each site-specificinterlock can be selected from the menu box. When an interlock is selected, the text box on the left side of the screen shows each required logical input. The graphics box shows the corresponding interlock diagram. Alarms. Alarms can also be selected from the menu box. For each site-specific system, a list box containing the relevant groups of alarms (for example, gnnding lubrication alarms) appears. A graphic showing all of the alarms in that group on a flow diagram background is displayed. Once a group has been selected, each individual alarm within that group is listed in a second list box. As the user selects an alarm from the second list box, the selected alarm is highlighted on the graphic diagram. Text boxes then illustrate the fault, cause, and remedy for the selected alarm. Alternatively, the user can click directly on any alarm on the diagram to link to text boxes containing the specific fault, cause, and remedy associated with it. Start-Up/Shutdown. From the main site-specific area bar menu, the user selects operating procedures and then start-up. The Merent types of start-up are listed. The user selects the kind of start-up-for example, start-up eom complete shutdown-and can then move through the procedures using the mouse. Each step is displayed, as are any observations, cautions, warnings, or notes associated with that step. Operator Tasks. Operator tasks are also selected from operating procedures on the mainbar menu for each site-specific area. Once selected from a drop-down mem, they appear in scrolldown text boxes and contain hyperlinks to appropriate graphics. Workbook. Workbook questions from the hard-copy module are provided in a separate window below the main multimedia window. The user can progress through the workbook

2297

questions while finding the answers in the main window above. The workbook questions lead the trainee through the technical information to be learned. Tests. Knowledge assessment tests are provided to validate that each trainee has learned the material. The tests are composed of multiple choice and true-false questions. Once each test is started, the trainee must finish it without reference to the reference material. Training Curriculum and Data Tracking System. A database traclung system can be integrated with the multimedia system. This component allows for establishing a custom curriculum for each defined job position. Personnel are then assigned to jobs resulting in each person’s learning hierarchy. Test results and qualification checklist results are tracked in the database for each individual trainee. Videos and Digital Photographs. Full-motion video and photos can also be included with appropriate hyperlinks. Videos are best used for illustrating operating equipment and for illustrating the correct performance of procedures. Interactive Simulations. Interactive simulations of unit operations are also possible. These simulations provide the operator with the opportunity to change set points or operating conditions and observe the effect on the process.

CONCLUSIONS There are literally thousands of specifics that must be learned by each operator involved in a new plant. In addition to facts concerning the new plant, operators must also learn principles and theory associated with the new equipment, controls, and methods. No matter how carefully the plant has been designed, or how well the new equipment works, the start-up will not be a success until the operator has completed this leaming. There are only two ways for operators to learn the material necessary. They can learn it in a controlled classroom environment as has been discussed in t h ~ spaper. Alternatively, they can learn it as they are attempting to operate the plant by trial and error. The cost of the former approach, while certainly not inexpensive, is very low compared to the lost production and damage usually associated with the latter approach.

2298

Planning and Staffing for a Successful Project Start-up Kenneth A. Brunk, Larry J. Buter and K. Marc Levier

ABSTRACT A successful project start-up can be defined as the culmination of a series of tasks that result in an operation that produces the saleable product at designed tonnage rates, to design specifications and at designed recoveries from design feed at designed costs to meet schedule. Successful project start-ups are the result of vision, cooperation, commitment, planning, and execution involving every area of corporate responsibility. Such successful start-ups have their roots in the executive offices where the philosophies and standards for ethics, health and safety, environmental and cultural responsibilities, training, design, and closure are set. The effective communication of these philosophies to the balance of the corporation and the project team, with the expectation that they are complied with, forms the basis for successful project execution. This chapter will identify the teams of people, the project organizations, and tasks that are needed to accomplish the project start-up. Additionally, the chapter will discuss the integration of the various engineering and corporate disciplines required to design, construct and start-up the project. Schedules for key events such as metallurgical treatment, final design, mine design, prestripping, project engineering and design, site pioneering, site construction and staff hiring, training manual development, training, procurement of project*consumablesand start-up spares, pre-commissioning,commissioning and finally start-up itself will be presented and explained. Definition of a Successful Start-up What is the definition of a successful project start-up? It can be defined as the culmination of a series of tasks that result in an operation that produces the salable product at designed production rates, to design specifications and at designed recoveries on schedule and within budget. The importance of a successful start-up is plain and simple economics. However, the morale of the project team and the new operations team is equally important in creating continued operating success. Successful project start-ups are the result of vision, cooperation, commitment, planning, and execution involving every area of corporate responsibility. Such successful start-ups have their roots in the executive offices where the philosophies and standards for ethics, health and safety, environmental and cultural responsibilities, training, project design, and operations closure are set. The effective communication of these philosophies to the balance of the corporation-and- €he project team forms the basis for successful project execution with the understanding and performance expectation that these philosophies will be followed. The start of success for a project begins with the net-present-value (NPV) and cash flow calculations which must meet or exceed corporate minimum standards before the corporate board of directors will approve the project for construction. These economic numbers are important to the corporation because personnel and monetary resources available are limited and the corporation must assure that these resources are utilized on projects that will provide the best return on the investment. Every project submitted to the headquarters office for consideration is in direct competition for funding approval. The funding for a project may come from the corporate funds, standard loans from outside sources or loans where money is borrowed from the sale of future production. The success or failure of a project start-up has a major and critical impact on meeting these expectations. Many projects have failed to meet the required financial criteria and this has had a direct impact on the ability of the parent corporation to remain in the mining business.

2299

An article published 1984’ displays impact on the NPV and DCF on delayed start-ups, Although it is now dated, it still is a good example for today’s projects. An excerpt from the article is shown in the next several paragraphs. “Based on recent experience, it is prudent to consider the following for estimating cash flow in project feasibility analysis: New mines may be expected to have average annual production equal to 50-70% of the designed capacity during the first year from startyp, 80-100% of designed capacity during the second year, and to be near or at designed capacity after the third year. New beneficiation plants may expect to have average annual production of 40-6096 of designed capacity during the first year from start-up, 80-100% of designed capacity in the second and third year from start-up, and to be near or at designed capacity after the fourth year. New processing plants may expect an annual production of 40-60% of designed capacity in the first and second years from start up, and 80-90% of the designed rate during their third and fourth years. The CRA study for the World Bank clearly indicates that firms do not realistically estimate the time necessary to bring a project to full capacity. Based on the record of recent projects-in the Americas, Europe, Asia, Oceania, and the Mid-East-investors should be prepared for long delays. On the average, most new mining and smelting operations have not produced positive cash flow in the year following start-up. In fact, many projects not only take about two years to achieve design capacity but, in the interim, increase the cash flow exposure of the sponsor. As an illustration of the effect of start-up delays on project profitability and cash flow exposures, consider a copper minehmelter complex scheduled to produce 100,000 st/yr starting seven years from today. Considering all costs in constant dollars, the investment required is roughly $700 million. Assume that the project was justified on the basis of a 20% discounted cash flow rate of return (DCF-ROR) and the net present value (NPV) of $760 million at a 10% real discount rate, for example. As shown in Fig. 4, with no additional costs, a one-year delay decreases the DCF-ROR to about 19%, while an extreme five-year delay reduces it to about 15%. When annual costs equal to 15% of total investment are added to the delays, the DCF-ROR will drop to 17% with a one-year delay and to about 12% with a five-year delay. The impact of start-up delays will also be disastrous, because of the following cash drains: 1) high operating costs, due to curtailed or no production; 2) low revenues from lower-than-forecast production; 3) low revenues from off-specification production during prolonged start-up; 4) special unplanned start-up crew expenses, 5) fix-up costs; and 6) increasing debt due to inability to service the debt.

’

Agarwal, J. C., Brown, S.R., Katrak, S.E., “Taking the Sting Out of Project Start-up Problems”, E&MJ, September 1984, pp. 62-66.

2300

The foregoing cash draw-downs are all cumulative. In the copper project mentioned, the cash flow implications of a delayed start-up may be as follows: Original capital expenditure Interest charges for 3-year delay at 10% ........................... Fix-up expenses ........................... Operating costs (net of any revenues) for 3 years at 40$/lb of copper at an average of 50% capacity ............

Total

....................................

$700 million

220 million 100 million

120 million

$1,140 million

What had started out as a $700 million project with a 20% DCF-ROR will now have roughly an 1 1% DCF-ROR (Fig.5). More importantly, the total cash exposure for the project has increased from $700 million to $1,140 million. Therefore, even a modest improvement in decreasing start-up delays can have impacts of hundred millions of dollars. If corporate executives use similar start-up data for production scenarios as a guide for timing future cash flows, they stand a better chance of meeting ROR targets. A baseline schedule reflecting the reality of start-up delays justifies a comprehensive analysis and training program that: 1) examines the various pitfalls in scale-up design and allows for adequate safeguards; 2) reviews the start-up procedures; and 3) trains the operating and supervisory personnel to meet the challenge of a preplanned start-up schedule."

The co-authors of this paper have been involved in start-ups that have met designed production goals in less then a day and in others that have met the above stated "normal" industry

2301

standards. The difference in the two time lines, is that in the one-day cases, the start-up was given a very high level of priority with a great attention to detail and forward planning.

CORPORATE STRUCTURE AND PHILOSOPHY FOR PROJECT EXECUTION Corporate Standards Large or small companies have some of the same issues in common that must be addressed at the senior level of management. The first of these is the company’s or corporation’s standards. These include the philosophies and standards for ethical conduct, health and safety, environmental and cultural responsibilities, training, design, and closure. The effective communication of these philosophies to the balance of the corporation and the project team with the expectation that they are complied with forms the basis for successful project execution. These items are very important to projects since each can have a major impact on the project economics as well as staying in operation. Major violations of most of these items can lead to the project being delayed or temporarily closed until brought into compliance with the standards. The company executives must set in place a system that audits the compliance of their standards to assure that all of the project objectives will be achieved. A method of correcting noncompliance issues must be in place before the project is developed to assure that start-up will not be delayed. Management must also provide the support, both financially and in commitment, to allow their standards to be met. If they are stating that the standards will be met but do not provide the resources of management and financial support, failure to some degree will result. An example of this would be to state that every employee will be properly task trained before start-up. This is a requirement for all operations in the United States by OSHA and MSHA law but may not be required by law in foreign countries. However management will often proceed to cut funds needed to develop a good training program and have the proper personnel hired early enough for the training to be completed if expenditures are higher than originally projected. Such conduct serves only to undermine the goals of the corporation and the project. Management must also define the reporting structure of the project team to senior management. This can range from the team having total autonomy to total control of the project by senior management, or more normally, somewhere between the two styles. When the company management establishes a clear line of responsibility and authority for the project team, the project is given an increased probability of success. This includes the level of involvement that the senior management will have as the project develops. Without clear lines being established before the project starts, confusion will result around the project direction (Who is in charge?) and will destroy the team morale. This will only cause frustration for both parties. The project will also have an increased probability of being over budget and have schedule delays. Both of these will greatly increase the project costs and reduce the ability of the project to meet the project economics. Projects can be located near the company’s home office, another company operation or in a remote part of the world. The location of a project can contribute to confusion within the company as to how the project will be developed. Projects close to the home office or another company operation can be the easiest to develop since the company’s standards, ethics and culture can be more easily imposed on the new project since these are a part of the local culture. If the project is in a remote location or in another part of the world, the local customs, business ethics, and worker productivity must be addressed when calculating the project economics. A standard of operation for all of the company’s facilities should be the goal, but local customs or decrees may require the standards to be modified to meet them. The problem then becomes that not all of the company’s units are being held to the same standard and it is easy for a unit to argue that other standards should also be modified to meet what they feel is needed for success. Caution must be exercised when corporate standards are modified! An example of this is the cost of providing benefits for employees. In some areas of the world, it is required that the company provide a meal for the employees on every shift. This is a cost not normally expected if the company only has operations in the United States; but it must be taken into account in other locations of the world. In some instances this may amount to thousands of meals per day which increases the food supply requirements for procurement and logistical

2302

systems that may already be challenged to provide the necessities for the project team and construction workers.

Productivity Another often overlooked factor is worker productivity as compared to U.S. Gulf Coast standards. In dealing with a project in a former Soviet Union country, the government officials gave guarantees of their excellent construction capabilities and worker productivity. On advice of the engineering company “experienced” in these matters, all estimates of labor man-hours were increased by a factor of 2.5. In fact after the project concluded, the correct factor was more like 5.5. This in itself led to significant budget overruns and further start-up delays. Project Management Approach Each corporation has a culture and a method for developing projects. This may vary from small companies, with no development resources to large corporations that have dedicated teams that move from project to project. Small companies may need to depend totally on outside resources to develop a project. This requires extensive evaluation of the outside project team to assure the company’s economic objectives are met, since the control of the project is now outside the company’s direct control. Large corporations that have dedicated project teams may have more control of their success but this is not always the case nor is it guaranteed. They too must undertake extensive evaluation of all outside resources to be used on the project to assure project success. Large corporations must also safeguard that the project team is thorough and acts in the best interest of the company. The major reason for concern here is that the project team must make projects viable in order to stay employed. This can sometimes lead to over optimism and bias on the part of the team leaders when only the upside conditions are used to produce acceptable economics for the project. This can lead to failure of the project since it is unrealistic to assume that everything will meet the most optimistic conditions. The project leaders and team members must be incentivized regardless of company size to maintain objectivity in their evaluations and recommendations. For example, if a project that has been optimistically estimated is approved by management and later it is learned the costs escalated then the project may become uneconomic. Often this happens when the project is well under way. At this late stage of the project life, most of the money for engineering, design, equipment procurement and construction has been spent or committed and only the owner’s cost portion of the project cost can be targeted for spending reductions. This can lead to project management taking short term cost cutting measures in the area of hiring, training, and procurement of operating and maintenance supplies. This can also contribute to redesign of the plant to incorporate “cheaper” equipment that appears to save money but in fact is being misapplied in its application. This is false economy, since this will cause a successful start-up to be almost impossible to achieve. The simple advent of equipment improperly designed and misapplied can result in millions of dollars in loss, for the most simple problem, when you add up the time and effort that will go into analyzing the problem, attempting to make the original design work, the evaluation of a correct fix, the re-engineering of the fix, and the installation and start-up of the fix. Additional cost consideration must also be added in for the loss in production and plant availability. Of course at this point the operator is stuck to solve the problem while the engineering team has already departed for the next job and spent their bonus! (The engineering team is available to come fix the problem they created for a price! ! !) It is up to management to assure such optimistically engineered projects do not make it off the drawing board and into the field.

2303

PROJECT TEAM OVERVIEW Team Makeup Projects can be developed using a fixed cost or cost plus basis for engineering, design, procurement, and construction management (EPCM). Depending on the company's philosophy of project development, one of these methods will be chosen. Either requires an individual or a team of individuals to interact with the EPCM firm in order to track the cost and schedule of the project. To develop and start-up a project successfully, a group of individuals must be chosen to form a project team. This team must have a leader commonly referred to as the project manager. This person must have experience in dealing with the size and type of project that the company intends to develop. This individual may already exist in corporations that have dedicated project teams or, as the case with a small company, an individual may need to be selected from outside the company. In either case, for the project to be completed on time and within budget and start up successfully, the manager must have experience with similar projects. It seldom works to allow an individual that has only managed a $50 million project to manage a $500 million project or vice versa, since the level of activity for the projects are very different. In small projects, normally the project manager has a small number of individuals to assist in the design and construction activities. These people must have multidiscipline or varied backgrounds in order to review and assist the engineering company to provide a project that will start-up successfully. In a very large project, the project manager can have an individual with specialized expertise in each area, since the level of activity for each area is much greater and requires a full time person to maintain the project schedule. Management must make it clear that the project manager will oversee the project under the direction of the general manager who will be responsible for the project once it is started. Both managers need to be dedicated to the overall objective of the project, i.e. a successful start-up, and be willing to cooperate with each other to fulfill this common goal. Neither the project manager or the engineering company can be allowed to believe or act as if they have supreme control over the total project because their actions can have a very large detrimental impact on the long term operation of the project. For example, when the project manager and engineering company are allowed to manage all affairs in the community and make promises of long term employment, building hospitals, churches, recreational facilities, or providing free transportation to the project site, the GM and operations people are left holding the bag and must deliver. The general manager will be the person ultimately held accountable for the projects cost and operating performance and will be held accountable by the local people if the promises made during development are not met. Once the project manager has been assigned to the project, the remainder of the team can be selected. These individuals may be from within company operations, or they can be selected from outside of the company. If they are selected from within the company, they should be relieved of their current position obligations in order to concentrate on the new project. An individual who is an excellent employee in an operating situation may not be successful in the role of a project team member since the level of control for each position is quite different. If the project is being developed in a country foreign to the company, the project manager and general manager must at least have been issued a passport and traveled out of the country as part of the job qualifications. Individuals that have zero experience outside of their country, will find it a challenge to succeed. They may have been put in a position to fail by their company. If members are selected from within the company for the project team, they should be assigned to the start-up and operating group after start-up. This allows them to have ownership of their decisions during the development process. This also leads to a better start-up since they are very familiar with the project and the requirements that need to be met in order for it to be successful. They must have an adequate level of expertise to allow them to succeed in their assignment. A means of retaining them for the duration of the development and for at least a year after start-up needs to be addressed by management before the individuals are selected. If incentives are granted to the team, they must be fair, clearly understood by both parties and must be in writing to prevent problems as the project progresses. If contract individuals are hired outside the company, their contract should clearly state the duration of their contract is through start-up and for a fixed time thereafter. An incentive needs to

2304

be in place in the contract to make sure they complete the assignment. These people will be critical to having a successful start-up and an incentive is one way to prevent disruption of the continuity of the project. With a large turnover of either in-house or contract people, the project continuity is lost. The location of the project team is a decision that must be made early in the development process. Facilities should be obtained in the engineering house for the project team. These facilities need to be close to the engineering team's location but yet separate enough for the project team to have some privacy so discussions can be held without disrupting the engineering team. The goal of the total project team is to have a relationship that is open and beneficial to the successful start-up of the project. A clear method of communicating between the owner's team and the engineering team is needed to prevent confusion and added cost. All change orders need to be approved by several members of the project team including the project manager to allow tracking of costs associated with design changes. At some time in the process, all changes must be stopped or the project will never be constructed on time. If the practice of change is allowed to continue, the project can never be developed at the estimated cost. After the completion of the flowsheet, P&ID's, G.A's, and equipment specifications the operating teams involvement is changed from one of providing design to insure the criteria are included into the project. This phase of effort needs to be carefully managed and controlled to provide oversight while not being disruptive to the overall project design effort. On every project design issues will come up that need to be thought through and evaluated on a very rapid basis to contain costs and meet schedules. An adept project manager needs to manage the situation with the goal of providing a process plant that works. At the same time an observer from the operating group needs to understand cost, schedule and be able to evaluate alternative means of accomplishing the task. As project development is nearing completion, the method for transition of members from the project team to the start-up and operating teams located in the field needs to be decided. The timing of this transition is determined on a case by case basis. Close communication between the general manager (who should be part of the project team from the start) and the project manager is a must for this to happen smoothly. As important as having facilities in the engineering office, are having facilities for the team in the field. Construction of the permanent site offices needs to be a high priority. This allows the key operating people to be hired early in the process and allows them to become familiar with the facility they will be held accountable to operate. Temporary facilities will also be required for the owner's project team that will be onsite earlier than the permanent office completion. If the project is remote, housing arrangements need to be in place or contract man camps hired to house both operating and construction overseers. If family housing is part of the project, these must be constructed very early in order to allow families to become part of the project and to reduce turnover and extensive travel costs. This is not normally considered the best use of project cash but in the end can contribute significantly to a committed team and allow a successful project completion. There is no worse situation in the world than contractor's project personnel, and operations personnel fighting over bunks, office space or food!

Frozen Flowsheet Successful projects result from the integration of the geology, mining, and metallurgy to produce a mine design and process flowsheet that work together. Ample test work needs to be done in the mine design and process plant design to develop process criteria and equipment specifications that can be translated into engineering drawings and specificationsneeded to construct the facilities.

2305

Successful projects also begin with a flowsheet that is completely fixed or “frozen”. This metallurgical l and process related test work has been completed prior to commencing means & process plant design engineering. From this test work all process parameters such as flow rates, temperatures, pressures, retention times, pipe line velocities, slurry densities, rheology, bulk densities, etc. have been determined to enable material of construction selection and equipment selection that is compatible with those characteristics. Coupled with the frozen flowsheet is the development of a plant operating philosophy that fully describes how the plant is to be operated and controlled. For example, a detailed description of a process circuit is needed to enable the design engineers to include the correct instrumentation and control package necessary to control the process. Good project practice should include the plant operating personnel in the development of the general arrangement drawings, the flowsheet, the process instrumentations diagrams, the project equipment specifications (P&ID’s) and spare part selection. By including the operating team in the above, the project manager has tapped the most talented people in those disciplines to define the project details. This inclusion will ensure that issues related to personnel safety, traffic flow, operator access, maintenance access, flow sampling, and so on have been addressed.

ENGINEERING COMPANY SELECTION Overview After the flowsheet has been frozen, the first order of business is for the project team to select an engineering company. Engineering companies with demonstrated technical capabilities for the type of project and project size and complexity should be placed on a bid list. Each company should be researched on the projects of the anticipated size that they have completed in the location of the project. This is very important if the project is located in a remote location because logistics of manpower and supplies will be critical to the project success. Companies without foreign or remote project expertise should be removed from bid lists for foreigdremote projects. A list of specific criteria that will be evaluated for each engineering company must be developed and weighted so the evaluation of engineering firms can be done objectively. Higher weighting should be given to critical areas of experience which will make or break the project’s success. When several companies have been selected, the project manager and several of the company’s senior staff must visit the engineering companies and personally evaluate this group of f m s . This decision as well as the selection of the project manager is critical to the project success. Design Phase Criteria Corporate objectives for the project must be clearly understood by all parties before design is started. It must also be decided if the capital cost will be the important driving force or will operating cost be emphasized. Obviously, management wants both the low capital and operating cost but this is normally not possible. Normally capital cost can be increased slightly to slightly lower the operating cost. An example of this is in the general arrangement of equipment. If operating, cleanup and maintenance accessibility is included in the design process, it will require more floor space than if only equipment installation is considered. If not included, it will increase the operating costs will increase due to downtime and excess manpower required to perform normal operating and maintenance functions. These decisions must be justified through trade off studies and properly documented. Accounting procedures must be determined by both the owner and engineering fm in order to track the cost and progress of the project. The procedures need to be adequate to allow management of the project but not so complicated that a great deal of cost is incurred to manage a very small amount of money.

2306

The same concept is true for the security of the site during construction and operation. The amount of money that is allocated for security should be evaluated against the potential cost of loss. In locations where there is civil unrest, the safety of employees must be included in the evaluation. In this case evacuation planning and cost must be included. Losses of expensive key components during construction will delay start-up and increase costs. Loss of simple items such as zerk fittings or stainless steel valves and piping can cause a major delay in start-up in remote areas where these items are not available locally. A good rule of thumb in these locations is that the item will disappear if it is not bolted to the ground, and even then they sometimes disappear too! Corporate, national and local safety standards must be determined and the standard for the design must be established. In some cases, the corporate standard may be different than what the local or national law requires. For North American companies doing project internationally the corporate standards are generally more stringent. The standard that provides the proper protection should be used as long as it meets the law. If the project is located in an area that does not speak the same language as the engineering firm,then the documents must be provided in each of the major languages. The interpretation of the documents in an accurate and timely manner is necessary for the local people to be involved in the project review and design. Failure to recognize the importance of this will occur in the training of the employees who must operate the plant. An example of this is the labeling of equipment. If the local language is different than the label it can be confusing and dangerous. Dual labeling may be required for mixed language locations. Projects located in remote areas may require that mancamps be established as the project is being constructed. If this is the case, then the company must decide how the camp will be operated. Included are the security, health and safety of all camp employees. Most camps should be operated where alcoholic beverages are not allowed. This keeps the in-camp trouble minimized. Other forms of recreation must then be included in the camp setting to occupy the employee's free time. Control of outside visitors must be done to limit the visitation of unauthorized people. Camp facilities during construction and start-up must be adequate to prevent people turnover and to prevent sickness. Both of these conditions will cause labor unrest and may delay the project. Transportation of employees during and after construction must be decided before and during construction. Of importance here is if buses or cars will be used to transport people. If cars are used, then an established parking area must be located to allow construction and still provide reasonable access and protection of employee's vehicles. If buses are used, then established routes and stops must be determined. Parking at pickup points must be developed at the pickup points if it is not available.

ACTIVITIES PARALLEL TO THE DESIGN AND CONSTRUCTION PHASE Overview The project will have many activities that are being done concurrently with the construction. Some of these include in-fill drilling, pre-production mine development, background environmental sampling and monitoring, staffing of operating personnel, and procurement of the project's supplies and equipment. These activities need to be coordinated with the engineering and construction contractor to prevent delays to start-up. The site manager and his early key people must coordinate and communicate with the construction personnel to prevent delays and hazardous conditions. An example is the blasting required for pre-production stripping of an open pit. Since the mine is normally located near the process facility to prevent extra cost of long ore hauls, blasting during construction must be coordinated with the construction firm. If possible, blasting should occur when construction crews are changing shifts or at lunchtime when work has stopped. The production people must be adaptable and accommodate the contractors to cause limited delays of construction. The owner and his people, with the help of the engineering firm, develop a plan for the training of employees and a start-up plan. As the plant is completed in sections, a clear plan must be developed to transfer the sections from the contractor to the operator. Operations will want to commission sections and systems within the plant before the project is completed. These areas,

2307

however, must be signed off by the contractor and owner to prevent disagreement if damage occurs to the section after turnover. Since no plant is totally without items that need correcting at the end of construction, a list of unfinished or unacceptable items must be developed (punch list). A crew from the contractor then must complete corrections to this list, usually while the plant is running. The operators must coordinate with the construction crew to prevent hazardous conditions for the employees. An example of this is the use of cutting equipment by the contractor in modifying conveyor transfer chutes. This can cause fires and do major damage to plant equipment causing start-up delays. The site general manager is responsible for coordinating the activities described above as well as many other business related activities. For a grass roots operation the efforts needed are to identify, organize, design and implement all the items necessary to conduct the operations of the business. These include items such as: 0 0 0

0 0 0 0 0

0 0 0 0 0 0 0 0 0 0

0 0 0

0 0 0

0

Accounting Purchasing Payroll Employee benefits Communications Computer and data handling systems Permits Licenses Human relations Warehousing Maintenance systems Mine development and operations Taxes Safety or loss control Insurance Shipping and traffic Off take agreements Loan compliance documents Inspections Training, salary, operations, maintenance Community relations Management reporting Operations reporting Process plant operations Analytical support

While this list is not complete it serves to point out that a successful start up and operation require the coordination and input of many people. None of these areas can be neglected or the operation will suffer. Also, these operations need to be coordinated with the project manger to minimize expenditures and obtain the most efficient use of manpower as possible. Below is a sample organization chart for a large project. It may not include all the key positions needed for every project as well as it may contain additional positions above what is required for a smaller project.

2308

I BoardofDirectors I I

I Corporate President I Site General Manager

i Project Manager

Controller / Office Manager

Operations Manager

.c

Process Manager

Health, Safety & Resources Manager

Manager

Manager

Manager

Community Relations Manager

The table below is a summary of the anticipated schedule to hire people for the project. This schedule allows for people to be allocated to the engineering effort as well as to allow the site to have employees to become familiar with the project. Employees must be on the site early to allow policies, procedures and training efforts to be completed before start-up.

When Project has Approval Site General Manager Project Manager Selection of Engineering Company for Final Design Controller Human Resources Environmental Manager Process Manager Maintenance Manager Start of Detailed Design Mine operations Manager Community Relations Manager Health, Safety & Loss Control Manager Computer & Information Manager

2309

Completion of Detailed Design Mine Planning Engineers Logistics Manager 6 months before Start of Mine Development Mine General Foremen Planning and Scheduling General Foreman-Mine Accounting Personnel Warehouse General Foreman Environmental Engineers Health, Safety and Loss Control Personnel Analytical Department General Foreman Mine Operation Foremen 3 months before Start of Mine Development Analytical Personnel Warehouse Foremen Maintenance Foremen-Mine Planning and Scheduling Foremen-Mine Mine Operators Mine Mechanics Mine Electricians

9 Months before Process Mechanical Completion Maintenance General Foreman-Process Planning and Scheduling General Foreman-Process Maintenance General Foreman-Facilities Planning and Scheduling Foremen-Process Planning and Scheduling Foremen-Facilities

6 Months before Process Mechanical Completion Process General Foreman Planning and Scheduling General Foreman-Facilities Metallurgical Engineers Maintenance Foremen-Process Maintenance Foremen-Facilities 3 Months before Process Mechanical Completion Process Operating Foremen Process Operators Process Mechanics Process Electricians Facilities Electricians GENERAL DESCRIPTION OF FUNCTIONAL RESPONSIBILITIES Mine Development Team Schedules for mine development, including equipment delivery, crew training, pre-stripping, underground development, de-watering and the like must be meshed with the facilities

2310

construction schedules. Such coordination will ensure a source of proper plant feed for start-up and operation. Generally the owner’s team carries out the mine development. This team is specialized in mining techniques and mine planning. However, the team must be very diligent and keep the project manager informed of progress, problems, and performance against the overall schedule. Also, often times much needed earthworks or building materials are supplied from the early mine development. Therefore, adherence to schedule and quality standards for the materials are a must. Accounting The areas of accounting, payroll, data handling, taxes, insurance, management reporting and operations reporting can often be placed under the direction of the controller or chief accountant. This person must be articulate and knowledgeable in the software systems necessary to meet the daily requirements of a mine site while still complying with the corporate reporting requirements and standards. The individual will need to hire and train a staff composed of as many people local to the area as possible. Purchasing and Warehousing Purchasing and warehousing are two areas of an operation that are also suited to report to one individual. This person must be honest, understand the expectations of performance of the corporation and knowledgeable in the general business of mining. Furthermore, the individual must be able to analyze situations, able to handle conflict, and create a can-do group of employees in the department. It will be important for the individual to understand lead times, shipping routes and cost and contract types and terms. Purchasing and warehousing are service centers on a mine and they need to operate as such. This group needs to have a place at the start-up to store the necessary spare parts for the plant and mine equipment. These items must be arranged in an orderly manner and stored in a place that prevents contamination by the environment they are in. Also, if the first fill reagents are taken as part of the owners responsibility and cost, they must be ordered and delivered in time to prevent start-up delays. Human Relations-Employee Benefits Human relations and employee benefits are two areas that historically are placed together. This area needs to be headed up by a person with a can-do attitude who is knowledgeable in labor law and local labor issues and practices. Additionally this person must be capable of building and leading a team to staff this mine facility and provide for new hires when vacancies occur. Criteria for hiring must be established through interfacing with the various department heads. Also, training programs and training critique should be coordinated by this group. This is especially true in regions where skills training is a determinant in pay scales for individuals. Record keeping is also a must for this aspect of the company as good records can be an asset in workers compensation cases, etc. Employee benefits must be designed to meet the criteria of the corporation while recognizing local customs. Where necessary dual standards of compensation and living may prevail. The hiring of a competent human relations professional is one of the first orders of business for today’s operations manager. Prior to full scale hiring of the employee contingent for a mine site it is necessary to have in place the wage ranges, benefits plans, and employee qualification guidelines to enable successful hiring and team building characteristics. Such planning will help ensure the success of the operation and mining future employee related issues. Maintenance and Maintenance Systems For today’s large and complex operations it is generally necessary to hire an experienced maintenance professional early in the project life. Ideally this individual would lead the effort of establishing design specifications for the mine and process plant equipment. The individual is also

231 1

invaluable for overseeing the design of the maintenance shops, workshops and selection of spare parts and tools. This individual is also responsible to specify, evaluate and select the maintenance planning software to be utilized to ensure the equipment reaches and maintains design availability. The person must be able to critically evaluate problems, and make decisions. The Maintenance Superintendent will have in the crew individuals who are organized, who can evaluate problems and who have a can-do attitude. In addition to knowing maintenance skills they must have a working knowledge of the processes in the plant to enable good communications with the operators. Mine Operations and Development The mine superintendent must be on the project team almost from inception of the project. Mine production for today’s complex ores and environmental constraints requires that mine development work occur often before plant construction. To achieve the mine production rates and ore blends necessary to meet the project requirements, mine development needs to be evaluated and planned very carefully. Indeed, process plant location should be such to minimize haulage of material, ore and waste, from the mine. This decision can only result from high quality mine plans. As alluded to earlier, mine operations are often integrated into the plant construction effort to provide fill material, construct tails dams, fresh water ponds, access roads, and to provide major cut and fill operations as required. This means that mine engineers, mine geologists, equipment operators and maintenance people must be on site early in the project life. These people need accommodations, training, equipment, facilities, and so forth to function. The mine superintendent must be knowledgeable in the mining method to be used, be appreciative of the construction requirements for the project, be cooperative with the project manger and be accountable to the general manager. The individual needs to be innovative to be able to direct crews in often harsh environments with minimal facilities to work with, especially in early stages of the project. Computer Systems and Communications Mine sites require significant computing power for the technical support systems, mine design systems, business systems and accounting systems. Also, associated with this computing power is the necessity of fast, reliable communications within this site and to the outside world. Proper system sizing and design requires the identification of the various software packages and systems that will be used on site as well as those necessary to communicate management information off site. Approximate data and communications requirements need be developed and systems sized accordingly. Following that exercise the requirements for system support and training need to be determined. Also, the philosophy of how the training will be accomplished and by whom needs to be addressed. Therefore it is necessary to place on the payroll a technical systems administrator who understands the needs of the business and the uses of the programs. The various systems would ideally be integrated to eliminate multiple data entry and efficient reporting. This individual needs to be on the team in time to organize the soft and hardware acquisition, and to establish training programs for the users. The systems person should also specify the communications system to be used on and off site and enable their use during construction to eliminate cost duplication. Environmental Control Environmental permitting and compliance are cornerstones of today’s mining operations regardless of what country in the world the mine is located. Permits must be obtained for everything from operating the actual mine and mill to establishing the landfill in which to place garbage. The permit effort must be timed such that when the project construction is complete the operation can commence start-up and construction.

2312

Permitting effort must start as soon as the project appears to be viable, even at the conceptual stage. Studies to obtain background information to provide site characterization and base line information must take place even during the late stage exploration on the property. Also, the road map of effort and timing of the studies that will be needed to enable the permitting of the project must be established and evaluated from a risk point of view before the commitment of major dollars to develop the project. Therefore the placement of a high caliber individual whose expertise is the permitting of projects needs to occur early on in the project life cycle. This person should be available to see the project permitting through to completion and be available for consultation at least for a year after start-up. The individual will need to develop environmental compliance and monitoring plans to assure the process design takes into account the environmental constraints and conditions of the permits. Items such as the allowable treatment rates, discharge water quality, pounds per hour of emissions from stacks, and so forth have direct impact on plant design and cost. Additionally the individual must have hired and trained a team of professionals on-site to monitor the operation and assure compliance with the permits and environmental laws. These people will need to be skilled to work with the operators at site and have first hand knowledge of the operation. They will need to interface with the local officials and community to establish and maintain workable relationships over the life of the project and its ultimate closure.

Health, Safety and Loss Control Industrial health and safety responsibilities for an operation must be addressed early in the project to assure the process plant and mining operation meet the minimum standards established by local governments, the insurance companies, the lenders and the corporation. Identifying problem areas early in the project allows for the design to address the issues. Items such as chemical storage, ventilation, equipment access, fire protection, fire fighting systems, ambulance requirements, medical evacuation requirements, personnel evacuation and security requirements, employee education and training and the like need to be designed into the project and available upon startUP. The crews and management need to have been trained in the operation of the facilities and emergency response systems prior to start-up. Fire response teams, and first aid response teams need to be in place and functional at the start-up stage of the project. Also, the safety officer needs to have established disaster plans and coordinate the plans with the local medical facilities, should they exist. Therefore the person to head this aspect of the operation needs to be on board and functional well prior to start-up. The individual must be involved in the employee training at all levels to assure common implementation of the corporate expectations and laws of the region. The person needs to be articulate, able to think on his feet and able to manage in emergency situations. Analytical Support Today’s ore bodies require a high degree of rather sophisticated analytical characterization on a daily basis to assure the delivery of ore that meets the requirements of the processing facilities. To accomplish this requirement requires that a laboratory capable of the minimum analyses within rapid turn around time be located on or near the mine site. The laboratory needs to functional prior to the commencement of the actual operation of the process plant. This lab may also serve to monitor various flow streams to maintain permit requirements, even if an outside lab is used to obtain the results that are officially reported to the watchdog agencies. The chemist and the staff will need to be on board and functional at start-up to provide the analytical information and sample turn around that is needed at this time of the project. Plant Operations Superintendent The plant operations superintendent will be a very key individual in the life cycle of the project, and needs to be on board, and involved immediately after the project is deemed ready for flowsheet design. This person will provide valuable operating input to the design team in all

2313

aspects of the site layout and process plant design. The superintendent will provide guidance in areas such as operator and maintenance access, ergonomics and traffic flow, operating philosophy, equipment type, selection, equipment configuration, and process control. While the plant is in detail design, the individual will monitor the drawings and equipment selection process to ensure the layout and equipment is in compliance with the concepts agreed with, He is there to ensure that the engineer provides an operable plant. The superintendent is also involved in the interviewing and hiring of his staff of foremen, general foreman, metallurgists, clerks, etc. The individual is responsible to have in place a functional crew of plant operators, and technicians capable of running the operation during the start-up exercise. Additionally the reporting systems need to have been designed and tested prior to start-up to provide management with the production they need.

Transition from Constructionto Operations As construction of the project nears completion, some construction employees will desire to become part of the operating team. This can be good for both the contractor and the operation if done properly. Operating staff can evaluate the employee’s work performance during construction and decide if it matches the expectations of the operating staff. For the contractor, it can be positive since a good employee can be given a permanent position rather than being laid off at the end of the project. It can also be a negative for the contractor if the transition timing is not agreed to in advance, and the contractor is left without key employees to finish construction. The answer is to have very open communication between all parties and a well thought out plan before the construction is completed. CONCLUSION Successful plant start-up does not just happen! It occurs because a tremendous amount of planning and cooperation of many different factions has happened due to excellent resource planning. Communications between all people involved is required to prevent misunderstandings and delays. The project must be a team effort by all members if the project is to start-up successfully.

2314

Maintenance Scheduling, Management and Training at Start-up: A Case Study Peter Vujic’

ABSTRACT We all subscribe to the philosophy of thinking about maintenance before the start-up of any new plant. To those of us in the maintenance fraternity the resultant benefits are obvious. Activities include: having maintenance input at the design stage; ensuring maintainability in design; ensuring quality information is received from vendors; optimizing, selecting and purchasing start-up and one-year spares; developing maintenance plans; and having selected and trained (educated) maintenance personnel prior to start-up. Questions remain of how to go about doing all this. How many people are needed to make this happen? When is the best time to introduce a maintenance development team to the project and at what cost? Is there a typical plan and time frame for all these activities? What problems will be confronted and what are some of the solutions? This paper, by illustration of case studies within the BHPBilliton group, will cover the above issues and provide a better insight into what is meant by ‘Maintenance Scheduling, Management and Training at Start-up’.

INTRODUCTION The last few years, there has been a change in the thinking and commitment to new business ventures. The traditional approach to projects has been: Equipment Design and Delivery, Construction, Commissioning and Handover. The maintaining aspect of the project has often been neglected. Today BHPBilliton no longer considers these projects as “projects” but “business ventures” and include as part of the venture, the maintenance function. The objective is to reduce overall venture development costs whilst maximizing start-up effectiveness and subsequent whole of life venture costs and the key theme in using this approach is do the right things, at the right time, in the right sequence and to the right level of detail2. HATCH, in conjunction with the Global Maintenance Network of BHPBilliton, has developed the ‘Venture Maintainability Guidelines’ in order to realize the overall objectives. The development of the Maintenance Function is a project in itself and with each new venture there is a trend to appoint a Maintenance Development Manager to deliver the maintenance function. Some examples of such appointments have been at Escondida Oxide plant in Chile in 1998, and the Oxide plant at Tintaya in Peru and the Phase 4 Expansion at Escondida in Chile in 2000. The venture maintainability guidelines become the road map to achieve the set objectives. Figure 1 below identifies the tasks ahead and also shows where the EPCM approach interacts with the Maintenance Function.

This paper follows the activities as they occurred during the process of developing the maintenance function at BHPBilliton Tintaya Oxide project.

Maintenance Manager - HATCH, Peru.

* From ‘Venture Maintainability Guideline’,May 1999, HATCH & BHPBilliton GMN.

2315

Figure 1: Venture Maintainability Process Flow This paper is a “worked example” of using the Venture Maintainability Guidelines at the Oxide Leach Plant venture at Tintaya in Peru. The paper looks at how the Venture Maintainability Guidelines were applied, the issues confronting those responsible for setting up the Maintenance Function for a new project, some helpful hints and ideas, and opportunities to improve for the next project.

What is Venture Maintainability? Simply put, it is about minimizing the total venture development and early life operating costs, whilst maximizing start-up effectiveness and subsequent whole-of-life venture performane. We use Venture Maintainability and not Project because Venture implies the development of a concept through financial justification, Board approval, design, construction, set-up and then sustainable and profitable business operation. Whereas Project has traditionally implied a more specific focus on the design, procurement, construction and commissioning of the plant and equipment associated with a venture. In the past, the relationship between “the Project” and “Operations” was at best poorly integrated, and could even be adversarial. A Venture approach aims to pull all these players together. In practical terms, Venture Maintainability aims to ensure plant and equipment “operability”, “maintainability” and a logistics start-up plan are considered and incorporated into capital ventures from the time of conception; not left until after commissioning. It is not necessarily about doing something new, nor is it about increasing project costs - rather it is designed to leverage what we know now and do it better. The underlying premise is to carry out the venture development by doing the right things, at the right time, in the right sequence, and to the right level of detail. In the context of the Venture Maintainability process, the concept of maintainability can be defined by the desired outcomes of the maintainability approach: The provision of capable, motivated and dependable people at the right time to be fully prepared to take ownership and responsibility for carrying out the work required. The provision of an appropriate organizational learning environment within which those people can develop and work effectively. The provision of appropriate information and knowledge in the right form, at the right time. 2316

The provision of appropriate business systems and support facilities to enable those people to work together effectively. The provision of plant and equipment which is accessible and safe to operate and maintain. The provision of appropriate operating and maintenance plans for effective management of equipment condition and performance. The provision of the appropriate tools and facilities to support safe and effective operations and maintenance. The provision of an appropriate spares support strategy.

THE TINTAYA OXIDE PROJECT The Oxide Leach plant was built along side an existing Copper Concentrate plant at BHPBilliton Tintaya in the hills (4,200 meters above sea level) of the Andes in southern Peru. Cathode copper production rate is planned at 34,000-tpa in the first two years and rising to 40,000-tpa for the next five years. The plant economic life is seven years. The Oxide Process Flowsheet is shown below:

Figure 2: Tintaya Oxide Process Flow diagram. The overall workforce, operators and maintenance personnel number eighty. The maintenance team is made up of one Maintenance Superintendent, two planners, one reliability engineer, five electrical and six mechanical trades staff.

DEVELOPING THE MAINTENANCE FUNCTION FOR THE OXIDE PROJECT The paper is titled Maintenance Scheduling and Training at Start-up, however as you will see, there is much more involved than the title implies. For this reason, The term “Maintenance Function” has been chosen to explain all the maintenance type activities that are necessary before start-up. The Maintenance Development Manager and his team The project had a number of stop/starts (due to poor copper prices) but finally was given the green light to proceed in February 2000. The project was resurrected at 35% design completion. The Maintenance Development Manager was appointed to the design team almost three months after the restart of the project and at 40% design completion.

2317

The overall initial objective was to have all aspects of Maintenance Function (MF) in place before start-up. Activities to support this objective were to: 0

0 0

Develop an overall Maintenance Function Development Plan (MFDP) of how to plan to develop the MF and clearly define what was required. Develop a Capital Budget to realize the Maintenance Function that was submitted to management for approval. This budget also included the cost of other engineers required to assist the Maintenance Development Manager achieve the objectives. Identify manning requirements for maintenance. Carry out a Maintenance Criticality Assessment on all equipment in order to obtain an agreed list of critical equipment. This agreed list of equipment will be used initially, to prioritise maintenance development efforts and then prioritise ongoing maintenance and operational activities during production.

An additional objective was to try and optimize/minimize the overall maintenance development costs by taking advantage of work that had already been carried out by the Maintenance Development Manager for the BHPBilliton Oxide plant at Escondida in Chile.

Results: 0

Maintenance Function Development Plan (MFDP) was completed within the first two months and was the blue print for the work ahead. Advantage was taken from the MFDP developed for Escandida Phase 4 project. The MFDP document incorporated “best practice” ideas from BHPBilliton Global Maintenance Network (GMN)’, PHASE IV PROJECT , moreCLASS* ideals and HATCH’S Venture Maintainability. This document is available to the next Maintenance Development Manager and can be easily configured to suit the requirement. Maintenance Capital Budget - once the MFD Plan was completed the preparation of the capital budget was relatively straight forward. Details included: 0 0

0 0 0

0 0

0 0

0 0 0

Cost of two Maintenance Development Engineers for spares, special tools, vendor manual review and specific equipment data collection. Development of equipment codes and preventive/predictive plans - the maintenance strategy for the equipment. Development of an overall Condition Monitoring Strategy. Development of first year and critical corrective procedures. Development of instrument calibration procedures. Development of lifting procedures Development of isolation procedures and risk analysis review. Development of an electronic technical library. Translational service costs. An allowance for special development. Training activities. First year operating and insurance spares.

An equipment criticality assessment was conducted using a maintenance criticality assessment system called MCAS3. The assessment was conducted with the operations manager and selected process and equipment experts. An example of the end result of the criticality assessment review is seen in Figure 3. GMN - A dedicated grouphetwork of maintenance practitioners within BHPBilliton.

* moreCLASS -Minerals Operating Excellence Capability Assurance Strategy - name given to the strategic direction for maintenance within BHPBilliton. MCAS - a HATCH proprietary decision support software.

2318

Figure 3: Graphical output from MCAS showing final criticality priorities of equipment Issuedtips MFDP Overall Timeline Plan - a timeline model (developed at Phase 4) was used, see figure 4 below. No detailed, project type plan was developed. The bulk of the project was managed from this initial timeline, however a detailed plan was developed for the last three months of the project. Some people however may feel more comfortable with a detailed plan and that is entirely up to the individual.

2319

0

Determining development costs of projects within the plan - Previous experience in this area proved very valuable and close to the mark. Using local companies in some areas of the maintenance development work proved successful from a cost saving point of view in particular and the work was also of a high standard. (Some money was given to back to the project manager). Determining the initial spares budget - ideally, the “first year operating spares” (The term “first year operating spares” has been used for convenience only, however it should be noted that the term “first year spares” implies initial purchases of spares. Spares may be insurance type or on consignment or required for the first year of operation and are determined after discussions with vendors on equipment function and failure modes) budget should be created from a detailed study in conjunction with the development of the equipment maintenance plans, however this is not realistic. The detailed maintenance plan cannot be developed until: o all the vendor manuals including data specifications are available, o after discussions with vendors about their equipment, related to reliability, past experiences, failures and failure modes. (this is a time consuming exercise and needs to be carefully managed, most vendors reside in the states). Face to face contact is ideal however telephone conference calls were also used and proved reasonably successful. o other sites with similar equipment have been canvassed, (again a time consuming exercise but very valuable) o final equipment design and selection. This was a problem as there were ongoing design changes on some equipment well into the project.

The spares budget was developed by reviewing the equipment list and making a considered judgement on what spares would be required and by consulting experienced people from within the organizations. There is a rule of thumb and percentages from past projects but we opted not to use them. There was sufficient doubt, due to lack of historical evidence, to substantiate these percentages. One of the objectives was to optimize on spares purchase. With dedicated people and well defined maintenance strategy, spares should be a minimum. Spares requirements and costs could be a function of the equipment expenditure and not the overall project value. The initial spares recommendation was accepted but was also criticized based on previous projects that it was too light. Advice given was found to be closer to the real value of spares. The spares cost was approximately 7% of the cost of the equipment and for future projects a suggested percentage of 7.5% is recommended for the initial spares budget. The spares issue is further discussed later in the paper. Determination of the number on the maintenance team and the operating budget There are a number of schools of thought here. One is to use benchmarking and industry standards. The danger here is the competitive nature of being the best, each business trying to get the lowest ratio of employees per output unit. Are we sure we have enough people to do the job efficiently and effectively or are we just trying to compete? Do we really understand the demands on the maintenance team? The approach used was to work from first principles, a simpleapproach. It is the number of equipment and components that will determine the workload for the maintenance team. The maintenance objective is 100% reliability of equipment - zero failures (during the required operating period). The only way to achieve this is to be able to understand or know equipment condition at any given time. To understand equipment condition we have to understand the failure behaviour of each component because the ’equipment’ doesn’t fail as such, it is the individual components that make up the equipment that fail. If we appreciate the number of components and each component has an average of three failure modes then there is an inspection activity associated with each failure mode - we can soon calculate the required time and people required for inspection tasks alone.

2320

In addition to the inspection tasks are shutdowns and repairs, which again can be calculated by understanding the failure modes. Inspections make up 30% of the workload with another 60% of the workload as planned corrective maintenance. (These percentages are consistent with studies conducted by BHPBilliton Global Maintenance team). Consideration should also be given to a commitment of approximately four weeks per year for ongoing training of personnel (implies a working year of approx 1800 hrs).

Timing of the start of the Maintenance Development Team. o Maintenance Development Manager: Initial thoughts on the appointment of a Maintenance Development Manager were for himher to be appointed at the beginning of the project full time. That appointment should be made a the beginning of the project and involvement on a part time basis early in the project, to be involved .in feasibility studies, involvement in the bidders requirement documents (to ensure that maintenance requirements are met). A fulltime involvement at no later than 30% engineering completions is recommended. - Alternatively, for the BHPBilliton group (possibly via GMN) to create a specialist “Maintenance Feasibility Development” role that who is involved fulltime in the feasibility phase and initial bidders requirements to ensure the Maintenance side of the business venture is understood. Maintenance Development Engineer(s1: One engineer (as a minimum) commenced approximately four months after bids were received. His main role was to review spares (start-up and first year spares) recommendations and centralize all equipment maintenance related data, i.e. folders were created against Purchase Order numbers and relevant information was copied and included in each folder. This exercise proved worthwhile and saved valuable time later in the project. This engineer continued to work full time on spares review and ordering (via the existing Tintaya stores group). Spares review will be considered separately later in the paper. Maintenance Development Eneineer(s1: A second engineer commenced once the project was formally approved. His role was document control and management and included review of vendor manuals; collection of lubrication data; management of the contracts for corrective procedures and the development of the technical library. Spares Requirements for the first year of operation This by far was the most interesting part of the whole project and also presents an area of opportunity to save time and money by getting the vendors to change their ways. The overall objective was to optimize on the number (and cost) of the first year spares requirements. It was decided that the purchase of the first year spares would be organized by the owners as apposed to the project contractors. Note that the start-up spares were the responsibility of the project contractors. The first task of the Maintenance Development Manager was to issue to the project contractors a Maintenance Bid Requirement Document 4as an addendum to the purchase orders. This document spelled out in detail what maintenance type information was required with the bid and included, format and presentation of spares details, format and presentation of lubrication data, standard of vendors manuals and requirements of maintenance recommendations. Unfortunately only a handful of vendors complied which made the task of spares review a very lengthy exercise.

4

Maintenance Bid Requirement Document - developed by the Escondida Phase 4 Maintenance Development Manager.

232 1

Reviewing the data submitted by vendors An introduction to world of spares. Spares details were submitted in various formats which included, faxed copies of outdated recommended spares, word documents, different Excel type spread sheets, ink jet copies with hand written changes, other copies with the description truncated (printer set incorrectly) and all fell well short of the requirements. This format was supposed to save time and money during review and it is still believe it will provided that vendors comply. As can be expected the next step was to contact each vendor in turn to seek clarification of details and cost of all spares. This continued for the length of the project (into 2002). Reviewing the spares recommendations with the vendors. The objectives here were threefold. First, to take advantage of the equipment vendor’s knowledge and experience with his equipment to gain valuable information of the possible failure modes and past experiences that they may have in order to develop the best maintenance strategy possible for the equipment. A modified Reliability Centred Maintenance (RCM) approach was used and well appreciated by the vendors. Second, to review the recommended spares submitted by the vendors in light of the discussions and come to a mutual agreement on spares requirements. It was interesting to note that many vendors willingly agreed to change the number of spares required after the review. Third, to clarify all spares details. These tasks took time, organizing discussions and countless email communications. To date the benefits of the review have been: o o o

o o

A “saving” of almost US$200,000 on vendor recommended first year spares. Identification of critical spares overlooked by vendors. Correction of details that may have resulted in the wrong spares being ordered. Having all information available so that spares can be catalogued and ordered with a minimum delay by Tintaya’s purchasing department. Agreements on ’on consignment’ and vendor held stock - mainly related to known wear items.

Purchase of Spares by Tintaya. Due to the detailed analysis conducted, the ordering of spares was relatively straight forward. The detailed review gave us confidence that the correct spares will be available when required, eliminating one of the possible failure modes of equipment and often cause for prolonged equipment delays - the wrong spare! In addition, in order to save time and money and being conscious of equipment warranty, Oxide took the decision to buy ’first year operating spares’ direct from the original vendors. This policy saved many hassles and enquiries from local agents who were trying to sell us equivalent spares, references and requests from ’friends of friends’ and enquiries from our own stores personnel. A review of the equipment under operating conditions is the responsibility of our maintenance team and is part of the continuous improvement cycle. They will decide (cost and reliability point of view) later if other vendors will be invited to submit alternative spares. The information collected will also be valuable when the contractor hands over their start-up spares. Purchases by the contractor were based on the initial information included with the purchase orders this was shown to be lacking and on handover to Oxide the maintenance development team had to go through the same process in getting detailed information required to catalogue the spares in the Stores. Condition Monitoring Strategy The focus today is on predictive maintenance, also known as Condition Monitoring. HATCH was commissioned to develop an initial strategy and then to assist in its implementation (assisting the nominated reliability engineer) just before, during and after start-up. The development of the strategy was made easier as the maintenance development team had all relevant equipment data available in separate folders.

2322

There was a strong fopus on “clean oil” and monitoring of same. During the development phase of the condition moriitoring strategy it was found that many gearboxes lacked oil sampling points and/or sight glasses. In hind sight, it would have been more appropriate if this level of detail was carried out during the equipment selection phase to ensure when equipment arrived on sight they had the appropriate facilities to monitor oils easily.

Development of Corrective Procedures. This activity is a ,time consuming one and a local Peruvian engineering company was commissioned to develop these. Corrective procedures take an average of two to three days to complete and considering that some three hundred initial procedures were required, the estimated time for a team of three people was one hundred days. The timing of this activity was approximately four months after the project approval when some 50% of the vendor manuals were available. Tintaya Oxide has a preferred format for standard procedures and the contractor’s task was to extract the information from the vendor manuals and develop the Tintaya standard. There were two stages to the development of these procedures. Approximately 80% of the corrective procedure details were obtained from the vendor manuals, the other 20% was a sanity check of the procedure once the equipment was physically on site. Development of the equipment maintenance strategy. (preventive and predictive plans). Previously developed guidelines for the development of equipment maintenance strategies and these guidelines were used on the Oxide project. The approach is one based on the RCM methodology and was supported by the software system called MPDS’ (Maintenance Plan Development System). A team of engineers was contracted to develop the maintenance strategies. An overview of the process is seen in figure 5.

Concepts in the Maintenance Strategy Development Process

Figure 5: Concepts used and applied when developing equipment maintenance strategies. Issues - Ownership of the Maintenance Strategy Ideally, this process is carried out with the maintenance team so that they have ownership of the strategies developed. A compromise was required at Oxide as there was insufficient time for the maintenance team to be involved in all the analysis due to process and specific equipment training programs. The compromise reached included that the initial equipment strategy reviews were conducted with the maintenance team with the objective for them to appreciate the review, input into the strategies and to gain confidence not only in the approach but the strategy development

’MPDS - maintenance decision software support tool provided by HATCH was developed by BHP/BHPE. 2323

team who would continue with the project. The data collected from the vendor spares reviews proved very valuable during the development of the equipment strategies. Development of Calibration Procedures. Instrumentation has been, for a long time, neglected in terms of having appropriate maintenance strategies. Today, with automation of our plants, instruments must be given the attention they deserve. They are controlling devices, protection devices, safety devices, environmental protection devices and quality devices. The reliability and accuracy of these devices is vital to reliable production and its quality. Regular calibrations of instruments is an important part of the maintenance strategy for the plant. HATCH were commissioned to undertake this exercise as they have had past experience in this type of work (in particular at ESCONDIDA Oxide in Chile) and have built up a library of calibration procedures for a large number of instrument types. Development of Lifting Procedures. Any lifting activity should be clearly defined by a procedure because of the inherent safety risks with the activity of lifting equipment. Vendor manuals were surprisingly lacking in detail and clarity of lifts in this area. Without lifting procedures, the risk of accidents increases, as people may try anything to carry out a lift. Each corrective procedure was reviewed and any activity that required lifting was developed into a procedure. The procedure included:

0

calculations to determine the correct wire rope and lifting angles. weight calculations and nominated hoist or crane to carry out the lift safely. clear photos or sketches to identify the lifting lugs or connection points. analysis of risks during the lift.

Development of an Electronic Library. It is important to get standardization of manuals from the vendors because at the end of the day you want anyone to be able to search and FIND information quickly, in particular the planners. Simply storing electronic copies of vendor manuals is not the answer to an electronic technical library. Each manual differs in format, each manual tends to use different maintenance terminology, each manual has maintenance information in different locations throughout the manuals and are not conducive to finding specific information quickly. Where electronic manuals were not available from the vendors, the manuals can be digitized (Oxide project utilized a company called metanoia6 in Chile). The digitized manuals were then edited to a standardized format. A standard contents page was then developed to facilitate the finding of information as required. This exercise may appear to be costly but the opportunity exists for substantial savings for future projects especially if we can get vendors to work towards standardizing their manuals. Special Development Projects. It is a good idea to reserve some funds for special development work. ’Special Development’ means; bright ideas, enhancement of procedures, special guests, training and leaving the door open for the ability to fund improvements and pursue those bright ideas. For the Oxide project there were three special development projects and they consisted of Development of Posters The posters were designed to reinforce the maintenance philosophy and maintenance message. Motivational posters such as: TLC’ - Tighten Lubricate and Clean, a very basic but extremely important maintenance philosophy. mefanoia - Metanoia Ltda of Santiago in Chile.

T L C - The origin of TLC comes from Tender Loving Care and an analogy is drawn with the need to look after equipment in a similar manner.

2324

Taking photos and mounting them next to the new equipment. The challenge to the team is to keep the equipment in the as new condition. The photos are a motivator and challenge for the total operating crew and reinforces the TLC philosophy. I NEED ATTENTION (INA) tags. Tags that had two parts. The top part is detached and using a tie, is secured on or close to the equipment that needs attention. The detached part has a unique number, the equipment tag number is identified, date and signature is included. The bottom portion also has the same number and tag number but also room for text to further describe the attention required. Main benefits include: o All equipment that needs attention has an identifying tag (no equipment is forgotten) o Reduced safety risk of working on the wrong job - the work order for repair references the INA tag number. o Provides a visual view on the condition of the plant and valuable for the Maintenance Supt on his walk around inspections.

Equipment and Component Tagging. The major equipment had contractor tag numbers installed (part of the contract) however there was much equipment (valves, instruments, miscellaneous pumps etc) that did not have specific equipment tag numbers. Unique equipmentkomponent identification is important, first and foremost from a safety point of view and then from a point of view of equipment history and costing. Equipment tagging would be carried out once the ‘Functional Location’ (term used in the maintenance module of the SAP business management system used at Tintaya) for equipment had been completed. Each component would have a tag prepared with a dual Functional Location and Equipment Tag No. These would be placed adjacent to the components. People Selection and Training Activities. Capable people are an important ingredient in ensuring equipment and process reliability. It has been said that the root cause of any failure can be ultimately traced to ’people’. People Selection Selecting the “right” people is half the battle in ensuring there are capable people. At Tintaya, a DDI’ selection process was used. The process proved successful. It was the first time Tintaya had used the process and the opportunity was taken to review the selection process itself. DDI were given feedback and agreed with the recommendations. Some recommendations were to streamline the process as it is a time consuming task. Oxide has one dedicated person responsible for -coordinating selection and training activities. The job was very demanding but well supported not only by all of us on the management team at Oxide but by training personnel from Escondida Phase 4. ~~

8

DDI - Development Dimensions lntemational - Addison, Texas, USA.

2325

Training Activities The Escondida Phase 4 project currently underway has a dedicated team of “training” specialists. Their focus has been to ensure that any learning activities undertaken will be prepared, delivered and assessed in a way to maximize benefits to the participants. They developed learning guidelines which were sent to selected vendors of equipment who were invited to offer training. Oxide also made use of these guidelines when inviting vendors. The response to the guidelines was very encouraging and some vendors commented that the guidelines really made them think again on how they were traditionally delivering this particular service. We also made extensive use of the local technical college, TECSUP with refresher courses on basic mechanics, electrics, hydraulics, pneumatics and instrumentation. Vendor training was also carried on targeted critical equipment. Further equipment specific training is planned after start-up of operations. Managing Maintenance. The responsibility of ’managing’ maintenance falls on the shoulders of the planners. Everyone has a part to play of course but the planners are the central figure. The overall process is seen below in figure 6.

Figure 6: The process of managing maintenance and the interrelationships with other members of the team. The computerized maintenance management system used at Oxide will be SAP. A “pipeline process” was developed to re-educate planners in the process of managing maintenance with a strong emphasis on correct planning and scheduling. Another strong theme and objective for Oxide is that b l l work will be planned’ and it was imperative that the planners understood the function of planning and scheduling. During the process KPI’s (Key Performance Indicators) were identified and set. The exercise proved valuable and is recommended for all planners. Figure 7 below demonstrates the ’pipeline process’ for planning.

2326

Unplanned

n pAuthorise ..

I'

ldx A

planned Jobs (Reventive Maintenance)

Total

A

Done

4

4

ACCC

Bredkdowns

Continuous

I Improvement I

Figure 7: The 'Pipeline Process' for planning and scheduling. Project Audits and Reviews. Three audits or reviews were planned and conducted on the Oxide Maintenance Function Development project. They were:

0

Venture Maintainability Review early in the Project Equipment Maintainability Review at about 80% design phase Mid term Venture Maintainability Review

The reason for the reviews are: to demonstrate support to the project manager by the maintenance fraternity in BHPBilliton, to identify issues within the project so that corrective actiods can take place, to ensure a consistent approach to developing the maintenance function and to identify opportunities to improve the overall process of developing the maintenance function. GMNMATCH developed what is called the VME, Venture Maintainability Evaluation and looks at all aspects and interrelationships within the project. This evaluation can be conducted as a series of interviews with project personnel or as a self evaluation. The Equipment Maintainability Review process looks specifically at equipment maintainability and considers such things as, accessing equipment for both predictive and corrective maintenance, standardisation of equipment, facilities to repair equipment . The timing of this review is recommended at approximately 80% design. If the review is conducted too early there is usually insufficient information to give justice to the review. If it is done too late, example design completion, all information is available however it is virtually impossible to influence any changes. CONCLUSION For the uninitiated, developing the maintenance function or preparing 'maintenance' before startup can be a daunting exercise. This paper has shown that there are a multitude of issues to consider but with proper support in particular the standards and guidelines, the task is no longer daunting. It is still complex and involved as all projects are but very manageable. The paper also shows that there is opportunity for improvement and potential to reduce the overall capital investment of developing the maintenance function and for the venture itself. ACKNOWLEDGEMENTS 1. BHPBilliton for granting permission for me use the Oxide project as my worked example. Alan Pangbourne (Oxide Project Manager) for his support and encouragement. GMN (Global Maintenance Network) for their support and input. Mike Duggan (Maintenance Development Manager for the Phase 4 project) of whom I hold high regard in this area of developing the maintenance function. Many of the ideas and methodologies developed by Mike and his team I have used to best advantage in the Oxide project. 5. The whole Tintaya Oxide team and their commitment and support to what we a trying to achieve with developing the 'maintenance function'.

2. 3. 4.

2327

Operator Training And& Vim'

ABSTRACT Training, taken in the broader context, actually involves two separate functions: education, whch teaches knowledge based on principles; and training, which teaches skills based on procedures. Both are essential to shorten the learning curve during the start-up of a new piece of equipment, and indeed of an entirely new plant. Shortening this learning curve has a significant impact on attaining, and possibly exceeding, design production targets sooner. Combine this with the requirements to have a significant portion of indigenous workforce, which may not have had any exposure to mining, the necessity to undertake an educational program is heightened.

INTRODUCTION When we buy a piece of process equipment it is critical that the mechanical installation be done properly to ensure that the device will operate at peak performance on start-up. Once the equipment is running, some form of preventive maintenance is required to ensure continued efficient operation. The same philosophy should be applied to operator training, as they are also elements of the entire processing system. That is, operator training should be considered an integral part to circuit operation. Using the analogy of a piece of machinery, operators need proper installation (initial training) and preventive maintenance (continuous training - or refreshers - after start-up). In the mineral processing industry, operator training has been quite variable. Their knowledge and slulls are very much a function of training and experience, which are clearly corporate and site specific. In fact, it can be crew specific. A quantitative way to assess this is to conduct a statistical analysis of historical production data (Vien et al. 1994a). If one operator or crew is statistically superior to the others, then the performance difference is a measure of the benefits of better training, i.e. bringing everyone to at least the same level as the best operator/crew. Variation of performance between operators is an indication of training deficiencies. The cost of these training deficiencies can be very significant, and in some cases in of the order of $US500,000 per year (Vien et al. 1994a). Potential benefits of this magnitude seem to be the rule, and not the exception. Another telltale sign of training deficiency is that in many plants the operating crews employ quite different strategies, all in pursuit of a common operating goal. This is typified by the changes made to set points just after a shift change. Although the different strategies may produce similar metallurgical results, typically, one will be more economical than the others. Additionally, changing the set points will create a disturbance and hence, for a period of time, circuit operation will not be optimal. The existence of different operating strategies is therefore a clear indication of the need for better training. As the trend for newer mines is to be located in regions where mining is foreign to a significant portion of the indigenous workforce, an educational program to teach the fundamentals of mining and mineral processing is now essential to most new mine sites. This educational program must start at the very beginning by describing the series of transformations required to go from a rock to a metal product. Ths provides a foundation to provide context and relevance to the training on operating procedures and operator duties.

I

Metso Minerals, Business Line Mineral Processing - Process Technology, Kelowna, B.C., Canada

2328

Under the current economic pressures, a good operator must possess good knowledge and good slulls in order for the company not only to perform, but indeed to survive. Knowledge and skills are taken to be equivalent to theoretical and practical abilities, respectively. To fully develop in these areas an operator requires a good grounding in both procedures and principles, as well as the opportunity to apply and hone h s h e r knowledge and slull sets. As Figure 1 indicates (the text appearing to the far right is simply the dictionary definition of the corresponding term on the left.), knowledge is imparted through education whereas skills are imparted through training. This difference is explored in more detail in the next section. Since the reader may not be familiar with educational material aimed at the mineral processing plant operator, selected material of such a course is presented.

Figure 1: A perspective on education and training in the mineral processing industry (Vien et aL 199413)

As a separate topic hom the education vs training issue is the question of mode of delivery. Should the material be delivered in a classroom setting or through computer based training or a combination of both? The section on computer based training addresses this issue and the potential benefits and pitfalls of such a system. Finally, an overview of the contents of a training manual (procedures) is presented. The purpose of h s section is to show that there is little overlap in the scope of material in a training manual compared to that of an educational program. They really serve two different purposes and both are essential. EDUCATION VS TRAINING As mentioned in the introduction, a distinction is made between education, which teaches knowledge based on principles; and training, which teaches skills based on procedures, as depicted in Figure 1. This is an important distinction. A trained operator will try to minimize upsets after they occur (applying cause and effect reasoning) whilst an educated operator can anticipate and mitigate upsets before they occur (applying conceptual models faithful to the physics and chemistry of the process). Traditionally, the mining industry has been involved in operator training, and it is only recently that educational programs have emerged as an important aspect of operator development. However, and probably for hstorical reasons, the educational aspects are still combined under the general umbrella of training. Whle this conventional terminology is acknowledged, it is important to understand the roles of these two distinctly different activities. A truly effective training program will include the proper balance for the needs of the operation, and these two components should be properly integrated to support learning. Training is aimed at issues such as safety, regulations and operating procedures (startup, shutdown, etc.). Education provides an understanding of the principles of operation behind the process, and of the principles of process control. Management's expectations of plant performance

2329

have been continually increasing in response to an increasingly competitive world. In order to meet these expectations, operators must actively endeavor to optimize the process, rather than to just get through the shft without a disruption. The level of knowledge required of the operator to meet today's performance expectations can only be achieved with a more thorough understanding of the process. This understanding also gives the educated operator the knowledge necessary to troubleshoot unforeseen situations. One of the major differences between training and education is that training is explicit, e.g. a startup procedure, whereas education is implicit, e.g. reasoning about the current situation. This reasoning is only possible by providing the trainee with the fundamental blocks upon which our current understanding of the principles of operation is based (Vien et al. 1994a). To understand grinding circuit operation, one needs to first understand unit operations, such as a grinding mill. To understand unit operation, one needs to first understand fundamental concepts, such as particle size measurement, rate of breakage, fragmentation, liberation, etc. An educational program must therefore start by covering fundamentals, which include basic physico-chemical properties and engineering concepts (e.g. work index, liberation spectrum, rate of breakage, etc.). Only then can the design, operating and control characteristics of unit operations (e.g. grinding mills, cyclones, pumps, etc.) be added to the curriculum Finally, the knowledge gained flom the previous sections is used to perform reasoning on the dynamic impact of interactions between unit operations in the circuit, both for changes in process variables and for diagnostic reasoning. As can be seen, each level builds on the previously acquired knowledge to form a complete understanding of the principles of operation of a circuit. Since the principles of operation are generic in nature, the knowledge imparted to the trainee is transferable to another plant or flowsheet. It can also be more easily transferred to entirely different situations. For example, a knowledge of the property of viscosity, as introduced in the context of milling, is useful in the understanding and diagnostics of other processes such as flotation, pumping, dewatering, etc. In contrast, the training on the site-specific procedures is not directly transferable to another flowsheet. Indeed, being skilled in the startup and shutdown of a grinding circuit does not help for the startup and shutdown of a flotation bank. A beneficial side effect of implementing an educational system is that it provides a platform for unification of knowledge standards, development of a common jargon and more harmony in the conceptual understanding of milling processes between operators, management and process engineers. This, in turn,fosters more and better communication and interaction between all levels of personnel.

Illustration of Educational (Knowledge-Based) Content (after Rybinski et al. 2001) For the purposes of illustrating knowledge-based content, the author has elected to describe the Metso CBT software package. The origins (Vien et al. 1994a), prototype evaluation test work (Vien et af. 1994b) and a more detailed illustration of content (Vien and Grondin 1996) have been described elsewhere. In this section, only a brief description and example' is presented. As a background, it is interesting to note that the educational system was originally conceived as something that would enhance existing training programs in North American operations, i.e. where basic training programs already existed. The objective was to help operators extend their understanding of the principles behind their circuit operation to enable them to operate the circuits closer to optimum. However, where an educational system has perhaps been even more helpful is in new mining operations where the operators are recruited from local communities, and they have essentially no training or experience in the industry. In these cases, feedback from users indicates that the educational system is probably best used first, to provide a basis of understanding (context) for the normal skills training efforts. The development approach for a given module consists of utilizing focus groups comprising: (a) academics and technical practitioners to set the curriculum and develop explanations, and (b) The example is from a course that is designed for an interactive multimedia system and thus it can only be roughly approximated here.

2330

operators and subject matter experts to test relevance and refine explanations. The magnitude of the effort for development of educational material should not be underestimated. The grinding and flotation modules offered by Metso contain over 400 and 800 pages, respectively, and took several man-years to develop. The development philosophy was designed to: educate operators - i.e. faithfully teach engineering principles without any mathematics, ensuring the explanations were based on sound science 0 promote communication among operators and between operators and technical staff, by introducing a common conceptual understanding of the science and technology, and establishing a common vocabulary. To try to illustrate the content accuracy and depth, as well as the level of the presentation, an example is drawn from the Metso CBT Flotation Module. This particular example, starting in the Fundamentals section, combines the processes of grinding and flotation using the liberation spectrum - as would be the case in most university courses on this subject. (The reader will have to accept that terms such as grade, recovery, etc. have already been explained elsewhere in the software.) Figure 2 includes two of the numerous images from a narrated animation on a page in the Mineralogy topic area, introducing the idea of a Liberation Spectrum. The illustration shows that one can take broken(partic1esand sort them on the basis of their grade (in this case expressed in terms of the copper mineral, chalcopyrite). It also shows that with finer grinds one tends to get better liberation, i.e. more material ends up in the two extreme regions, and less in the middle area. The resulting distribution of particles by grade is the liberation spectrum, and it is a handy practical (and theoretical) way to think about the role and results of comminution in mineral processing. 0

Figure 2: Introduction of the notion of liberation spectrum for grinding a copper ore Having introduced the liberation spectrum, it now provides a basis for other explanations, including an illustration of how a perfect flotation process would operate; an explanation which is conceptually consistent with those provided for other separation devices, such as the hydrocyclone. Figure 3 is again a series of images taken from a narrated animation on a page dealing with the ultimate grade-recovery curve. The graphic in Figure 3a simply illustrates the notion of perfect separation by drawing a vertical line, which represents the division of the feed material into a tailings and a concentrate product. Figure 3b expands on this idea by showing that different separations can be made by drawing the vertical line in different spots. The accompanying narration would point out that, for example, in the case of point B, the separation would produce a fairly high grade concentrate lots of pure and near-pure chalcopyrite, but the recovery would be low, as quite a lot of the locked

2331

chalcopyrite would be lost to tailings. The third graphic (Figure 3c) shows the various combinations of grade-recovery associated with the four points in Figure 3b, thereby linking grade-recovery to both the grinding (liberation) and separation performance. The fmal figure is simply an observation that because of inefficiencies in separation, the real grade-recovery curve lies below the ultimate grade-recovery.

Figure 3: Using the liberation spectrum and a simple separator to explain the graderecovery curve

In the section on unit operations, one of the effects explained relates to changing air flow rate. (Elsewhere the ideas of collision, attachment, detachment and entrainment are explained.) Increasing air results in increased recovery (see Figure 4), and because of the grade - recovery relationship, this means lower concentrate grade. The link between air flow rate and grade (recovery) control is thus established. Finally, in the section on circuits, the impact of changing air flow rate on the entire cleaner circuit is illustrated (see Figure 5). Increasing air flow rate will increase cleaner concentrate recovery (point 2) at the expense of grade (the explanation would summarize and refer to the detailed description in the unit operation section described above). The explanation would continue by indicating that the cleaner tails grade and mass flow would drop (point 3). The consequence is that the cleaner scavenger concentrate grade and mass flow would drop (point 4), as would those of the cleaner scavenger tails (point 5). The end results is that the overall recovery is increased, but at a lower grade. The link between air flow rate and the overall cleaner circuit recovery is thus established, based on the descriptions in the unit operation section as well as the fundamental engineering concepts such as the grade-recovery curve and liberation spectrum.

2332

Figure 5: Illustrating the effect of air flow rate on recovery and grade in the cleaner circuit Hopefully thls example achieved its objectives. While it was short, the description conveys a reasonable picture of what operator education should entail. Perhaps this will also help to explain why it is that such systems are useful to bring operators and engineers together as a technical team, to some extent eliminating the traditional hierarchical relationship. It is important to reiterate here that such educational systems are entirely focussed on these kinds of knowledge concepts, without regard for the basic safety and operations issues. These should be handled by a complementary (skills) training system, as described in the training manual section. There can be little doubt that the reader would have had a more enjoyable time working though this example on a computer based system, rather than reading it here. Computer based training is the subject of the next section.

2333

COMPUTER BASED TRAINING To achieve this goal of increased process knowledge, principles of operation must be added to the list of operator training educational requirements. Since the principles of operation are generic in nature, the educational material can be standardized throughout the industry. This is desirable because it saves time and money by avoiding duplication of effort, and it provides the best material possible to the trainee. This can also apply to training on industry-wide standardized procedures such as first aid, lockout and rigging. The best way to disseminate this information widely is through computer based training. Two major benefits of computer based training are the use of multi-media presentations and the ability to use simulation. Rather than have the trainee learn on-the-job with all the risks that entails, simulation can teach the trainee how to address a specific situation. It also allows the trainee to experience situations that are uncommon, without having to upset the plant or wait perhaps years before a certain problem occurs. Computer Aided Instruction (CAI) allows the trainees to see, hear and do! Computer based instruction is the only method that provides a high level of retention of the material. This is principally due to the engaging and interactive nature of such a package. It keeps the interest level of the trainee high throughout the learning experience as compared to a classroom setting where the mind of the trainee may wander at times. As a result, the amount of time required for training is shorter. Different people learn at different rates. Computer aided instruction provides an individualized pace which is better than the common pace of a classroom setting, keeping the trainee alert throughout the learning process. However, a computer based training system cannot anticipate all the questions and cannot provide interaction outside the preset system configuration. Discussions with a trainer are invaluable in &IS regard. In fact, a computer based training system, meant for individual use, has been successfully used in a facilitated group discussion mode (Rybinski et al. 2001). This ingenious use of computer based training assisted the trainer by providing the best of both worlds: a software package providing top-notch educational material and interaction with the trainee to foster better communication, deeper understanding and a greater sense of team work. The end result was a faster assimilation of the knowledge required for the operation of new equipment involving technology not previously used at the site. Of course, slulls practice, which can only be performed by manipulating the actual equipment, can only be done in the field and not with a computer based (or paper based) training system. Undoubtedly, preparation for this skills practice through computer based training makes the practice more effective and less time consuming. In the introduction it was stated that training should be provided on a continual basis to maintain a h g h level of competency. A computer based training system can be made available twenty-four hours a day. Why provide training only during eight hours a day on week days when the operators are there twenty-four hours a day, seven days a week? Additionally, night shifts tend to be less busy than day shift. Operators could take advantage of the slack time for training. This also means that training can be provided in manageable chunks that are smaller than the saturation level of the learner (when the learner can no longer absorb the material being presented). Additionally, timely accessibility to training means that the material can be reviewed as required. It provides an on-going source of technical support to operators for diagnostic and decision support purposes. Automated instructional software is readily available to create the course content. However, trainers should heed to the following warning:

We are here to tell you that an authoring package does not a multimedia developer make. Please do not subject your trainees to a program developed with no more skills than word processing requires. This is cruel and unusual punishment. (Salopek 1998)

2334

In other words, it is more important to focus on the contents and the judicious choice of media to enhance the learning experience as opposed to a glitzy presentation that is no more than a flash in the pan. “There’s a social behavior of staying in the classroom. There’s no such compunction against leaving a bad screen ... The tolerance for mediocrity is extra low.” (Salopek 1998) To avoid this possibility the creator of the material should have a good grounding in instructional system design and understand the principles of adult learning e.g. when to use animations, when to use videos, what length should they be, how much information to put on a page, where on the page should it be, etc. There are several books on the topic, including that of Lee and Mamone 1995. One of the keys in the design of a successful computer based training system is to recognize the various learning styles: A visual learner learns information best by seeing it, an auditory learner by hearing it, a kmesthetic learner by feeling or experiencing it. Effectively designed multimedia environments accommodate all three styles. The use of graphics, animation, video and audio appeal to people who learn by seeing or hearing. Getting kinesthetic learners to “feel it” is trickier, although good results can be obtained with simulation techniques. (Cohen and Rustad 1998)

In the end, it may be more effective to buy commercially available packages andor contract out the work. In this role, the trainer becomes a manager of the training system rather than a developer. This benefits both the trainer and the trainee as it provides the trainer with (arguably) higher quality material and more time to devote to assessment and management of training needs.

TRAINING MANUALS The purpose of training manuals, whether in printed or electronic form, is two-fold to provide the initial skills training i.e. learning the safety, operational and emergency procedures; and to be used as a reference for refreshing and maintaining those skills. Unfortunately, the latter seldom occurs. This is partly due to difficulty in accessing the manuals at the time of need (locked up in an office as opposed to an open shelf, or available on a computer, in the control room) and partly due to the manuals being outdated (e.g. flowsheet or procedural changes). Training manuals must be kept current at all times so that they remain relevant to plant operators. The training manuals should contain the following information (Wilmot, Sass and VanDeBeuken 1986): Cover letter (statement of objective and management’s commitment) Table of content (to facilitate access to specific topics) 0 Overview (process area covered, summary description, flowsheet) Safe job procedures (operational, maintenance, emergency, hazards, first aid) 0 Process description (design basis, flowsheet, general arrangements, equipment description, brief explanation of how equipment works) Process control (loop narratives, interlocks, alarms, P&IDs, control system interface, instrument location) Operating procedures (start-up from long shutdown, start-up from short shutdown, start-up from emergency shutdown, short term shutdown, long term shutdown, emergency shutdown, changing circuit configuration) Operator duties and responsibilities Trouble-shooting guide. Although the initial manuals cannot be completed until detailed engineering is performed, the basic operating guide can, and should, be created at the time of the P&ID review (Wilmot, Sass and VanDeBeuken 1986). Not only will it make the P&ID review easier but, perhaps more importantly, it can be used as a tool to make the final design more operation friendly. This is where extensive plant operating experience on the part of the design team is essential. No amount of operator training can compensate for a flawed design - they can only learn to cope with it.

2335

CONCLUSIONS Companies spend very large sums of money building mineral processing plant. To make the most of th~sinvestment it is necessary that it be operated well and safely. Both operator education (whch teaches knowledge based on principles) and operator training (which teaches skills based on procedures) are essential components required to meet this objective. Both knowledge and skills are vital to shorten the learning curve during the start-up of a new piece of equipment, and indeed of an entirely new plant. Shortening this learning curve has a significant impact on attaining, and possibly exceeding, design production targets sooner. Training should be undertaken before, during and after plant construction. Even before detailed engineering is completed, but after the flowsheet has been selected, the educational portion of training can commence since the knowledge imparted is generic in nature. This provides the necessary foundation to minimize the time required to learn the site-specific skills, which can only be undertaken once detailed engineering has been completed (and based on current practices this means that the construction phase is well underway). After start-up, the knowledge and skills can be honed by providing continual access to education and training. With fewer operators on the plant floor, each decision made has a greater economical impact since there is a lower chance for correction or review by other frontline personnel. Combined with the requirements to have a significant portion of indigenous workforce, which may not have had any exposure to mining, the necessity to undertake an educational program, possibly involving computer based training, is heightened. The material taught should include the fundamental principles of operation to provide a deeper understanding of the process. This will allow the operator to not only stabilize the process, but to optimize it as well. Studies have shown that a higher retention of the material is obtained when a computer based instruction system is used. Computer based training can therefore be more effective. However, this is an aid and not a replacement for trainers as interaction with the trainee is essential to complete the learning process. A computer based instruction system can be made available in one of several plant locations twenty-four hours a day. This access would allow operators to train at times other than during the normal staff working hours. It is imperative that training manuals be kept up to date to remain relevant to the plant operator.

REFERENCES Cohen, S.L. and J.M Rustad. 1998. High-Tech High Time? Training & Development. December 1998:31. Lee, W.W., and R.A. Mamone. 1995. The Computer Based Training Handbook: Assessment, Design, Development, Evaluation. Englewood Cliffs: Educational Technology Publications. Rybinsh, E., R. Zunich, M. Grondin and B. Filntoff 2001. Operator Education: An Important Element of the Corporate Knowledge Management Effort. SME Annual Meeting, Preprint 01156, Denver, February Salopek, J.J. 1998. Coolness is a State of Mind. Training & Development. November 1998:22 Vien, A., D. Willett, B.C. Flintoff and D.H. Hendriks. 1994a. Mill Operator Training - Where Do We Go?. Proc. 26'h Canadian Mineral Processors Mtg., Paper No. 16, Ottawa, January. Vien, A., D. Willett, B.C. Flintoff and D.H. Hendriks. 1994b. Mill Operator Training: Prototype Evaluation. Paperpres. 1 6'hCIM District 6 Mtg.. Vancouver, October. Vien, A. and M. Grondin. 1996. Conceptual Model of AGISAG Operation for Operator Training. Proc. International Symposium on Autogenous and Semi-Autogenous Grinding Technology. Eds. A.L. Mular, D.J. Barratt and D.A. Knight, 2:729. Wimot, C.I., A. Sass and J. VanDeBeuken. 1986. Preparation of Operating Manuals. In Design and Installation of Concentration and Dewatering Circuits. eds. A.L. Mular and M.A. Anderson, Chapter 50. SME

2336

Safety and Health Considerations and Procedures During Plant Start-up Louis A. Schack

ABSTRACT The safety and health of employees and contractors is of primary importance before and during any plant start-up. This is especially true with regard to the start-up of complex processing plants staffed by large numbers of employees with differing skills and levels of experience. The early assignment of employees to plant operations crews provides for detailed review and adaptation of start-up manuals and the development of safe practices well before actual start-up. Teams of employees must thoroughly examine and trace the production circuit as construction proceeds. Only well-trained, knowledgeable and safe operators can achieve or exceed the production targets envisioned by plant designers. INTRODUCTION Newmont Mining Corporation operates both surface and underground gold mines on Nevada's Carlin Trend. This geologic feature is at the center of most productive gold mining district in North America. The modern practice of mining low-grade deposits of disseminated gold began at Newmont's Carlin pit in 1965. Today, several surface and underground mines remain in production on the Carlin Trend, yielding more than three million ounces of gold annually. Historically, on-site ore processing methods include cyanide leaching of low grade oxide ores, traditional milling (semi-autogenous grinding and ball mills) followed by gold recovery through carbon-in-leach (CIL) and carbon-in-column (CIC) circuits. As mining in several surface pits expanded throughout the 1980s, exploration drilling revealed substantial deposits of refractory ores in which gold is associated with sulfide and carbonate mineralization. Such ores are not generally amenable to the traditional processing methods applied to oxide ores. These refractory deposits included ores accessible from existing surface pits such as Gold Quarry ( I 984) and ores of higher grade that presented opportunities for the development of major underground mines such as Carlin East (1993) and the multi-deposit Leeville Mine, now under construction. Gold recovery from refractory deposits requires the application of complex milling and chemical processes including high-pressure oxidation (autoclave process) and fine grindinglhightemperature oxidation (roasting process). Newmont's 1997 acquisition of Santa Fe Pacific Gold Corporation included the purchase of two additional refractory processing facilities in Nevada: the Lone Tree Autoclave (near Valmy, Nev.) and the Sage Autoclaves (Twin Creeks Mine). Both facilities are located west of the Carlin Operations in Humboldt County. Prior to the Santa Fe merger and in coordination with sequential mining plans for a minimum of 20 years of refractory ore production, Newmont Mining Corporation designed and built a Refractory Ore Treatment Plant (ROTP) adjacent to existing oxide milling facilities at the company's Carlin South Area Operations. Construction of the ROTP began in 1993. The ROTP consists of a secondary crushing circuit, twin roasting circuits (north and south) fed by a dry grinding. double rotating mill with a nominal capacity of 352 stlhr. (2.8 Mtlyr.). The plant also includes air and gas preheating circuits, gas recovery, cleaning and cooling circuits and an acid recovery and storage circuit. The chemical

2337

constituents of many refractory ores create large volumes of recoverable sulfuric acid during the roasting process. The roasters are of a Circulating Fluid Bed (CFB) design. Commissioning was completed and refractory ore processing began in 1994. At the time of its commissioning, the ROTP was the largest plant of its kind in the gold mining industry. The designs of many of its circuits and related components are unique to this particular mill. Initial design and construction capital approached $400 million. The immense size, complexity, and intrinsic operational hazards of the ROTP and its ancillary facilities required a long term, comprehensive approach to employee training, operation and maintenance procedure development and coordination of activities between departments, vendors and construction contractors. Of particular importance was the integration of safety considerations and situational response protocols into all operations and maintenance procedures. Beginning in mid-1993, a team of four experienced operations, electrical and maintenance supervisors were charged with establishing a training program for all employees involved in the start-up of this complex mill. To most involved, the ROTP and its accompanying processes presented many new and unknown challenges. This team, dubbed the "Vector Group," eventually created more than 20 training modules covering all areas of operation and maintenance. Many others contributed to this core group's efforts during the months leading to start-up in late 1994.

STRATEGIC APPROACH The acronym ROTP took on a secondary meaning as start-up preparations progressed - Reliable On-target Training Program. The process began with a comprehensive study of construction drawings, circuit schematics, flow charts and available manuals. In many cases, specific systems were broken down for detailed study of individual components. As each component was analyzed relative to its function in the larger combination of systems, the team developed a broad understanding of the plant's complex operation. Throughout this early phase, team members attempted to identify and assess potential hazards, establishing safe procedures and emergency response protocols. The Vector Group also traveled to similar operations at various U.S. locations to study safety practices and to identify application opportunities for the ROTP. These visits included tours of acid plants in Idaho and Tennessee and a cement plant in Iowa. Special attention was paid to personal protective equipment and engineering controls. Rather than developing procedures that support only availability and production goals, the teams were charged with building safety considerations into all operations and maintenance practices. Any procedure in which safety was not the primary concern was rejected. Throughout this process - up to and beyond start-up - vendors provided technical staff to assist Newmont teams in deciphering drawings and documents, identifying hazards and developing preventive measures and response plans. The inclusion of those who would eventually manage the plant in the early development of all operations and maintenance procedures created a sense of ownership toward the final procedures manuals and a healthy respect for the process and its potential hazards. These advantages proved valuable as former start-up team leaders assembled crews and trained employees using materials they had personally developed and tested. Several of the larger vendors, including Lurgi - supplier of the roasters, gas cleaning components and acid plant - and Krupp - supplier of the grinding circuit components - provided commissioning groups to participate directly in Newmont's start-up preparations. These vendors contributed to detailed investigation of all processes and shared anecdotal information from previous installations of similar processing systems. The role of these vendors' representatives proved vital to safe start-up of the plant. Two of the initial four members of the Vector Group were designated liaisons to Lurgi and Krupp, sharing information on a daily basis. Newmont's internal safety department assigned a safety representative to review and contribute to the development of all safety practices. This representative also assisted in the testing, selection and purchase of all safety supplies and personal protective equipment.

2338

OBSERVATION AND ADAPTATION While initial efforts focused on studying technical materials and building a common understanding of the process, components and controls, it was recognized that technical knowledge is no substitute for practical, visual study of the complex process under construction. Throughout the construction phase, team representatives toured the project twice weekly. As equipment was installed, photographic records created a permanent reference for many components that are enclosed or inaccessible during plant operation and routine maintenance. Where needed, drawings and flow charts were modified to reflect slight variations in component construction, circuit flows and integrated systems. Controls, sensors and other support equipment installations were verified and evaluated as well. In most cases, operating and troubleshooting controls and software were provided by contracted suppliers. All control software was adapted and tested for applicability and reliability by Newmont Information Systems programmers. Software adaptation is a continuous process that allows for improvements to safe practices, systems integration and operational stability while minimizing downtime and lost production. Such improvements will continue throughout the operating life of the ROTP.

TRAINING MANUALS A training manual covering operation, safety considerations, start-up and shutdown procedures and other activities was produced for each of the many processes and circuits that make up the ROTP. A typical manual includes 14 sections, the first of which includes this disclaimer: "The instructions and procedures represent decisions based on limited knowledge of equipment and design criteria. As the construction evolves and more precise information becomes available, the procedures should be reviewed for correctness and applicability. This is especially critical for personnel safety, as changes may impact previous safety considerations." The training manual for employees in the Fine Ore Handling circuit begins with an Organization and Application section that details the manual's purpose - "to provide the information that operators need to know to operate the equipment properly, efficiently and safely." Also included in this section is a discussion of hazard recognition and the safe work practices each employee is expected to follow in each of seven operational conditions. Section 2 includes a general description of the circuit and the function of each piece of equipment, including dust collection and air filtration. Design capacity and operation limits are detailed as well. Section 3 provides general safety procedures, with descriptions of personal protective equipment and basic procedural guidelines. Section 4 describes hazardous tasks and begins with lockout procedures for all equipment within the circuit. The section covers safety considerations for working with compressed air, working inside an ore storage bin, using hand tools in confined spaces, cleaning plugged transfer points and other situations. Sections 5 through 11 explain in detail the steps for safe operation of the circuit in each of seven operational conditions: 1. start-up from complete shutdown 2. start-up from standby shutdown 3. start-up from emergency shutdown 4. complete shutdown 5 . standby shutdown 6 . emergency shutdown 7. normal operation Each section includes cross-referencesto applicable drawings and flowcharts. Section 12 details the effect failures of or upset conditions in one piece of equipment or area can have on the circuit as a whole. Several tables guide the trainee through a troubleshooting process that includes the alarms and control features that are activated under each condition.

2339

Section 13 provides appropriate procedures for inspection and preventive maintenance of major equipment before start-up and during normal operation. Section 14 is a series of drawings of each piece of equipment and flow charts of the Fine Ore handling circuit. As stated above in the disclaimer, several years of experience in operating the plant have required varying levels of modification to the training materials for each ROTP work area and process circuit.

EMPLOYEE HEALTH AND SAFETY The ROTP presents a great number of health and safety concerns for employees and contracted workers. These include various corrosive acids, chemical by-products, gases and reagents. Many of these materials are contained at high pressures and temperatures well over l,OOOo F. Other important hazards include burning fuels, high voltage electrical sources, rotating equipment, fall hazards and confined spaces. Maintenance employees undergo extensive training in confined space and "hot work" safety. Members of the start-up teams developed specific safety guidelines and personal protective equipment (PPE) requirements that reflect the hazards present in each area of the plant. PPE requirements range from respirators for all employees and plant visitors to acid-resistant suits, boots and face shields for employees in acid handling circuits. Consultation with operators of similar plants and various equipment vendors were invaluable in providing the appropriate safety equipment to ROTP employees. Examples include: A fiber metal hard hat in combination with a specific face shield and visor for employees exposed to acids GoretexTM suits for daily work in the acid plant Heavyweight acid-resistant suits for close maintenance work in the acid plant 0 Safety harnesses rather than belts and lanyards for fall protection

START-UP As construction proceeded, ROTP operations personnel were divided into four crews in early 1994. At this time, a bidding process began which filled the remaining operations and maintenance crews, drawing many experienced employees from other mills on the mine site. Their training began with classroom instruction and plant tours during the final phases of plant construction. These employees observed plant components and modified training materials as they became familiar with their new work areas. The operations crews performed small-scale start-up simulations in preparation for coordinated simulations involving other ROTP functions. Many of these simulations included tracing and identifying equipment and control systems. Initial mill start-up began with oxide ore feed through the grinding circuit as construction of the roasters was completed. Refractory ore processing began soon thereafter.

CHALLENGES AFTER START-UP Start-up and operation of the complex plant led quickly to the formation of a team assembled to identify, prioritize and develop solutions to a variety of operations and maintenance challenges. The ROTP Phase I Maintenance HAZOP Review (L. Davies, et al. 1995) was completed over a period of three weeks in June 1995. Increased plant availability, higher gold production and improved safety were the stated goals of the HAZOP review. Using a methodical approach, 13 areas ranging from the ROTPs acid pumps to its propane system were analyzed. Of the problems identified, three levels of priority were established. Priority 1 items included all safety or environmental concerns and those that had a significant impact on availability and production. Priorities 2 and 3 included items of less acute concern. Phase 1 addressed only Priority 1 problems. These were further identified as High probability and High severity (HH), High probability and Medium severity (HM) or Medium probability and High severity (MH). The Phase 1 review led to 140 solutions that marked the beginning of a continuous improvement process that remains a driving principle in maintaining and operating the ROTP.

2340

RESULTSBENEFITS Despite numerous challenges and concerns encountered early in the R O T S operational life, a history of effective troubleshooting and solution implementation has proven its worth as plant availability, ore throughput, gold recovery and safety performance continue to improve. Scheduled major maintenance outages that originally consumed - at minimum - one month of production each year are now shortened to a single 3-week outage. Unscheduled maintenance has also been dramatically reduced. Ore throughput averages well above 3 million tons annually and gold recovery consistently exceeds 90%. The ROTP produces about 700,000 ounces of gold each year. Beyond the production improvements, the safety records accomplished by the R O W S crews are the most significant result of a comprehensive and effective training program. As of May 2002, the mill's maintenance crews marked 6 years without a lost time accident while the operations crews achieved 5 years versus the same measure. Efficient and safe performance also support environmental performance. The ROTP boasts a commendable performance record with regard to its air quality and operations permits. From a broader perspective, the successful start-up and operation of the ROTP minimizes potential obstacles to the financing and permitting of projects of similar scale at other Newmont operations. The ROTP provides hundreds of Newmont employees with invaluable experience in the safe and efficient processing of refractory ores. This produces a commodity more valuable than gold itself - a steady supply of expertise that supports the growth of a minerals company that now leads the world in gold production. ACKNOWLEDGEMENTS I am grateful to Tony Gunter, former Process General Foreman at Newmont's Refractory Ore Treatment Plant, and to Rick Folkmire, Process Operator, for providing much of the information contained in this manuscript. REFERENCES L. Davies, et al., H.A. Simons, Ltd. 1995. Newmont Gold Company Refractory Ore Treatment Plant Phase I Maintenance HAZOP Review Report.

2341

SUNRISE DAM GOLD MINE - CONCEPT TO PRODUCTION WR Lethlean' and PJ Banovich'

ABSTRACT The Cleo deposit in the Eastern Goldfields region of Western Australia was discovered by the Shell Company of Australia in 1993. In 1994, Acacia Resources was floated fiom Shell, taking with it the 500,000-ounce Cleo resource under the name of Sunrise Dam Gold Mine (SDGM). This paper describes the metallurgical testwork, plant design and initial development that led to pouring of first gold in March 1997 and the subsequent Stage 1 plant upgrade that occurred in 1999. INTRODUCTION/PROJECT HISTORY The Sunrise Dam Gold Mine is located approximately 55 km south of Laverton and 730 km north west of Perth in the Eastern Goldfields region of Western Australia. The Placer-Granny Smith Joint Venture has exploited the Sunrise resource located on the adjacent mining lease to the Cleo resource via their Granny Smith plant located approximately 30 km to the north of SDGM (Figure 1). Initial tenements in the Sunrise Dam area were acquired by Shell in 1987. Assessment of preliminary geophysical and drilling data throughout the tenement area during the period 1988 to 1991 resulted in the identification of the Golden Delicious, Pink Lady, Red Delicious and Cleo Prospects. Considerable difficulties were experienced in penetrating the lake sediments which blanket the western half of the tenement area to a depth of 60 to 90 m. Exploration activities focused on the Golden Delicious Prospect, where, by late 1990, a large, predominantly low-grade mineralised system was established. Minimal exploration activity was conducted outside Golden Delicious until early 1993 when additional aircore drilling was conducted in the Cleo area. This program identified the need to extend drillholes into fiesh rock, and a number of anomalous intercepts were obtained. Follow-up drilling was conducted during 1993 at Cleo, with a number of encouraging intercepts, including 55 m at 10 g/t gold. A program of RC drillholes, with seven diamond tails, was completed by the end of 1993, resulting in the discovery of the Cleo Resource. Extensive RC and diamond drilling programs were conducted at Cleo during the following three years, resulting in a reserve of 600,000 recoverable ounces being considered in the May 1996 Feasibility Study.

1

2

Chief Metallurgical Engineer, AngloGold Australia Ltd Project Manager, Sunrise Dam Gold Mine

2345

/ /

/’ QLEINSTER

/ 0 LAVERTON

-‘$&,

l@GRANNYSMITH

LEONORA 0 ’

@SUNRISEDAM

/‘

OhlENZIEs

.

WESTERN AUSTRALIA LOCATION PLAN Figure 1 Sunrise Dam Gold Mine location plan GEOLOGY The Sunrise tenements are located withm the southern portion of the Archaean age Laverton Tectonic Zone, within the Yilgarn Block of Western Australia. The geology of the region is poorly exposed, deeply weathered, and extensively covered by surficial sediments and deep soils. The region consists of a north-trending greenstone package bounded by undifferentiated granitoids to the east and west. The central portion of the greenstone package, in which the Sunrise tenements are located, consists of acid to intermediate volcanics and sedimentary rocks, sandwiched between two corridors of predominantly mafic and ultramafic extrusive and intrusive rocks. Late stage intrusives of loosely granitoid composition occur throughout the area. Cleo Geology There is no outcropping geology or surface expression of the mineralisation and the geological understanding of Cleo is based entirely on drilling information. The area is covered by 20 to 90 m of lake sediments and aeolian sand dunes. Beneath the lake sediments is a sequence of interbedded Banded Iron Formation (BIF), Volcanoclastic (VC) and intermediate Volcanics, which have been intruded by narrow quartz felspar porphyries. Primary gold mineralisation is associated with shallowly and steeply north-westerly dipping structures and is hosted by all major rock-types. Shallow dipping structures are characterised by intense pervasive carbonate-sericite-chlorite alteration, pyrite and quartz carbonate veining. These zones exhibit intense shearing over narrow zones. Higher gold grades often occur where these zones intersect BIF structures.

2346

Steeper dipping zones are characterised by zones of quartz carbonate veining and breccias with pyrite and arsenopyrite.Wallrock alteration is generally less intense and pervasive relative to the shallow dipping zones. Gold occurs as coarse visible gold within veins and breccias and in association with pyrite, arsenopyrite and arsenious pyrite. The gold associated arsenic minerals lead to some refractory behaviour. The Cleo deposit is divided into three general ore types depending on degree of weathering, namely oxide, transition and fresh. Mining would commence with oxide ore for a period of two to three years, followed by treatment of the transition and fresh material.

Annual Throughput Mine Life Head Grade Gold Recovered Total Operating Cost Cash Operating Cost Initial Capital Cost

Mtpa Years Au g/t oz $Alt $MOz $AM

1.o 5.8 3.8 592,000 3 1.20 307.03 65.2

METALLURGICAL TESTWORK The SDGM metallurgical testwork program was carried out in three basic phases as outlined below. Resource Definition Phase During this phase the geological crew were in the process of confirming a resource that was considered to have the potential for further development. The metallurgical support consisted of preliminary leach tests to determine approximate gold recovery, cyanide and lime consumption would be formed. and a determination if weak acid dissociable cyanide (CNWad) Specifically, the aim of the testwork was to determine if any of the three generalised ore types, oxide, transition and fresh, were refractory or not. The information assisted in deciding the size of the resource required before proceeding to the pre-feasibility phase of the project.

Prefeasibility Phase The prefeasibility phase is the critical stage that determines the fate of the project. The key metallurgical work centres on determining gold recoveries for all identified ore types, developing data to feed into the design criteria, generating operating costs for all ore types and identifying environmental issues to be considered. Sample selection was achieved by reviewing the drilling logs with the geologists, and agreeing on the intercept composites to be selected to represent the ore types throughout the known ore resource. In addition the locations of the PQ diamond core samples required for comminution testwork were determined.

2347

. The principal areas of the testwork program carried out for the SDGM project were: 0

Gravitylleach testwork program carried out on numerous intercept composites to determine the gold recovery percentages by gravity and cyanide leach, cyanide and lime consumption, oxygen requirements, leach time required and any C N w a d generated. The testwork program was designed in-house with the testwork being camed out at AMMTEC, a commercial metallurgical laboratory located in Perth Western Australia. The data generated in this program, as summarised in Table 2 below, was used in financial analysis and design criteria.

Bench G L Test Fresh Ore Pso= 53pm Tests in te water YO

HEAD GRADE assay calculated RESIDUE EXTRACTION

6.00 5.21 1.23 3.97

GOLD DISTRIBUTION Gravity Leach Residue Calculated head

1.49 2.48 1.23 5.21

REAGENT CONSUMPTION NaCN (kg/t) CaO (kg/t)

0.80 2.20

0

Bench G L Test Transition Ps0= 75pm Tests in site water

Bench G/LTest Oxide Ore Pso= 75pm Tests in te water

LI

Y O

76.3

3.94 4.39 0.77 3.62

28.6 47.7 23.7 100

0.91 2.71 0.77 4.39

0.70 4.80

82.4

3.57 3.73 0.3 3.43

92

20.7 61.6 17.6 100

1.05 2.38 0.3 3.73

28.2 63.9 8 100

0.70 4.00

Comminution testwork carried out on numerous intercept composites to determine the traditional Bond Work Index (Wi) values for rod mill, ball mill and abrasion (Ai), JK Tech data for SAG mill simulations, Unconfined Compressive Strength (UCS) and crusher work index by size (not shown). A summary of results is shown in the Table 3 below.

2348

Table 3 Comminution summary

Oxide ORE SAMPLES Average BIF Average VC Design Oxide TRANSITION SAMPLES Average BIF Average VC

ucs

Ball Wi kWNt

Rod Wi kWNt

Ai kWWt

MPa

14.54 7.30

19.67 7.30

0.284 0.01 1

nd nd

10.92

13.48

0.148

nd

16.97 11.63

nd3 nd

nd nd

nd nd I

Design Transition FRESH ORE SAMPLES 1 Average BIF Average VC

14.30

nd

17.93 18.63

24.42 26.08

0.49 1 0.168

Design Fresh

18.28

25.25

0.330

142

Tailings characterisation testwork to generate data for designing tailings disposal, both physical containment and any potential acid mine drainage (AMD) issues. Geochemical characterisation testwork on both ore and waste material to identify potential AMD, either within the waste dump or withm the tailing storage facility. Feasibility Phase Limited leach testwork was camed out during t h s stage and was restricted to drill core intercepts not included in the previous program if it was considered to add to the database. Other minor testwork programs camed out included characterisation of the impact of hypersaline water on reagent consumption and the fate of cyanide in the tailings system, both important issues to be considered in the feasibility study. A program to determine a suitable and cost effective refractory process route to lift primary ore recovery from 76% to 95% was also commissioned. Because of the degree of difficulty of this task, the program remained active over the next four years. The key obstacle to progress was the poor quality of process water. The water available for this project contained more than 250,000 ppm total dissolved solids. PLANT DESIGN AND LAYOUT The plant design and layout was carried out in two distinct stages. The fxst stage was the prefeasibility study where design concepts and criteria were developed and costed to a f 25% accuracy. The prefeasibility information was used in the initial project financial analysis to determine project viability. The second stage was the feasibility study stage, where advanced engineering drawings were developed and costed to f 10% to confim the accuracy of the previous capital cost. Generally very few new concepts were added to the project during this stage. Capacity The key design criteria of the SDGM process plant capacity was decided by the then owners, Acacia Resources. Their corporate strategy had gold production targets, for which SDGM was the major building block. Prior to project approval, the two objectives were to produce 100,000 ounces per year, and establish a minimum life of seven years. The project development approval not determined

2349

was given in May 1996, at completion of the feasibility study and c o n f i i t i o n of the corporate objectives.

DESIGN CONSIDERATIONS Mine Development The SDGM project was based on being developed as an open pit mine. The mine design dictated that ore processing would commence on 100% oxide material for approximately two years followed by a period of blends of oxide, transition and fresh until the mill feed would revert to 100% fresh ore in year four. Comminution

The decision was made early in the prefeasibility phase to copy the successful comminution circuit concept installed at Acacia Resources’ Union Reefs Gold Mine (URGM) and install a two-stage crushing plant followed by single stage ball mill for the oxide ore processing. Equipment was sized to achieve an annual throughput of 0.75 Mtpa of oxide ore. The throughput was subsequently upgraded to 1.OO Mtpa during the feasibility study.

In making this decision, due cognisance was taken of the potential problems that can occur with wet clay based ores and secondary cone crushers and screens. The design considerations supporting the decision to proceed with the two stage crushing circuit were: The experience gained at URGM The lithology of the deposit that showed the oxide ore zones were generally composed of thick widths of competent BIF material associated with the soft “clayey” volcanoclastic ore, hence competent material was available to assist in keeping the clayey ore moving through the secondary crusher The secondary crusher would be operated with a wider gap than normal The product screen aperture would be set at a relatively coarse 16 mrn Acceptance that the ball mill would scat the coarser material and that the scats would be recycled when processing primary ore. At the time that the decision was made to proceed with the oxide comminution circuit, it was recognised that major changeslupgrade of the circuit would be required to enable the competent primary ore to be processed. The two options considered were three-stage crushing followed by two-stage ball milling or SAG and ball milling with scats crushing (SABC). The plant layout allowed for either option to be installed. The reason to not proceed fkrther with the design was the strong possibility that the ore reserve would increase sufficiently to allow throughput to increase substantially above the start-up 1.OO Mtpa rate. The decision not to lock in on throughput capacity, other than for the oxide ore, which had a defmed tonnage, was proven correct as outlined later.

2350

Figurt! 2 Process flow sheet

2351

Gravity Recovery The testwork indicated that approximately 20% of the gold in all ore types could be recovered in a gravity circuit. Diamond drill core consistently contained free visible gold in ore intercepts. The decision was made to copy the URGM flowsheet with the use of Knelson Concentrators as primary concentrating units and a Gemini shaking table as the cleaning unit. The circuit design was based on takmg approximately 20% of the ball mill discharge from the front of the ball mill discharge trommel screen protected by a section of the trommel screen at 5 mm aperture. A Warman pump would deliver the required pulp flow rate to the 2 mm aperture vibrating screen ahead of two 30 inch automatic discharge Knelson Concentrators. The concentrate would discharge at set intervals to a storage hopper above the Gemini table located in the Gold Room. Knelson Concentrator tail, at low pulp density, would be discharged into the ball mill discharge hopper. Gemini table tail would report to the ball mill discharge hopper via the Gold Room sump pump. Historically, the gravity circuit has consistently recovered 40% of the gold. Leach The testwork indicated that a 24-hour leach time was required at cyanide concentrations of 200 ppm in the feed tank. A hybrid CIL circuit consisting of one leach tank followed by six CIL adsorption tanks was chosen. Cyclone overflow at an average 44% solids pulp density would be screened at 0.8 mm for trash removal prior to being fed to the leach tanks. The carbon retention screens were the standard rotary swept vertical screen. Gold Recovery Gravity-recovered gold would be produced by direct smelting the Gemini table concentrate after calcining the concentrate at 700 "C for 16 hours. The calcining stage was required to oxidise iron from the balls and sulphides (pyrite and arsenopyrite) recovered into the Gemini table concentrate.

A refinement to be installed later was to process the Knelson Concentrator concentrate through the inhouse-developed ACACIA Reactor and recover the gravity gold via an electrowinning cell. The major benefits acheved by the ACACIA Reactor were: 0 0

0

Improved security by designing an automated hands-free operation Improved safety by eliminating arsenic fumes generated during calcining and smelting Improved gold recovery by recovering the fine free gold circulated back to the grinding circuit in the Gemini table tail Improved gravity gold accounting.

The elution circuit design was based on the split Anglo American system principally for conserving precious potable water. The column was sized to enable the expected batches of high grade feed material to be treated without the need to reduce tonnage or sacrifice recovery. The eluted gold would be recovered in electrowinning cells with the cathodes calcined at 700 "C for 16 hours prior to smelting in a gas-fired pot furnace. Tailing Storage Facility A paddock style tailing storage facility was chosen for the oxide ore treatment phase. The structure would be built to water retaining specifications with underdrainage to ensure consolidation of the settled solids. As described later, due to process plant capacity increases, a Central Thckened Discharge tailing storage facility was installed after the Stage 1 upgrade.

2352

Metallurgical Balance The requirements of the metallurgical balance were considered during the design. The mill feed conveyor weightometer would record tonnage and the tail sample would be collected via an automatic sampler. The issue of sampling the mill feed was debated at length with the main considerationbeing the influence of coarse gold on sample size required. In the end, it was decided to place an automatic sampler on the leach feed stream and back-calculate a mill feed grade by including the gravity-recovered gold into the calculation. Reagents The key considerations in the reagent design were safety and environment protection. The four major reagents to be used in the plant were cyanide, quicklime, sodium hydroxide and hydrochloric acid. The characteristics of each reagent dictated individual treatment from delivery to storage through to distribution within the process plant. Cyanide, as solid briquettes, was initially designed to be delivered to site in one tonne boxes and the two-month consumption equivalent would be stored within a fenced and locked compound. The cyanide mixing and liquid cyanide storage tank was located within the plant structure, but isolated from the daily human traffic routes. The standard mixing procedure was to add water, sodium hydroxide and the solid sodium cyanide briquettes to the mixing tank. The procedure was designed to prevent hydrogen cyanide gas generation. An additional safety feature is to add a dye colour to the mixed cyanide solution so that any failure that results in cyanide solution spillage is easily identified and appropriate safe clean-up procedures can be applied. A change to liquid cyanide delivery occurred after three years operation. The liquid is delivered at 30% concentration and offloaded into a storage tank with a one-week capacity.

Sodium hydroxide would be delivered to site as a 50% sodium hydroxide liquid and stored adjacent to the cyanide-mixing unit. Quicklime would be unloaded and stored in a 300-tonne silo located adjacent to the ball mill feed conveyor such that the quicklime could be added to the circuit via this conveyor, utilising the ball mill as a slaker. Quicklime storage was equivalent to one month’s consumption. Hydrochloric acid would be delivered as a 35% HC1 solution. The storage tank was located remote from both the cyanide and caustic storage tanks. In designing all liquid reagent storage tanks, the principal consideration was that they were sized to hold the equivalent of one and a half truck container loads. The reasons behind this concept were that The re-order level was such that the operation did not run precariously low while waiting for deliveries On delivery, it was guaranteed to hold the full truck container load without overflowing the storage tank. In addition, each storage tank was located within a concrete bunded area designed to contain 110% of the largest tank within the bund and capture any spillage from a rupture in any tank based on the requirements of the Dangerous Goods Act. In designing the reagent solution distribution systems, it was decided to use high-density polyethylene pipes colour coded on the outside to identify the reagent being distributed within the pipe. The use of tags was considered to be ineffective within the pipe trace and only effective at feed and discharge ends. The concept has proven very effective at the mine site.

2353

Infrastructure - Process Plant The remoteness of the SDGM site meant that the facilities were needed to provide power, water, communications and an access road surface suitable for equipment and supply transport. A power requirement of 4 MW meant that six 1-MW generators were required. Due to the climatic conditions at site, the capacity of each generation set was down-rated to approximately 0.8 MW continuous supply. SDGM purchased the generation sets fiom StateWest Power Supply and then contracted that company to operate the station. Telephone communications were non-existent at the site until the development of the project. An arrangement was entered into with the national telephone company to provide voice and data communications to the site via a microwave system. The development of the water supply was not an easy task due to the lack of good quality ground water or suitable terrain to build water storage catchments within a practical distance of the mine site location. Consequently, process water was sourced fiom the pit dewatering bores which produced water at 250,000 ppm TDS. Water for sprays, hose-down and reagent mixing was sourced from a small limited bore field producing water at 30,000 ppm TDS. Potable water was generated on-site via a reverse osmosis (RO) plant fed with water at less than 4,000 ppm TDS from a small borefield approximately 12 km fiom site. Access to site from the nearest bitumen road during geological exploration was via a deteriorated haul road. The priority job after project approval was the upgrading of this road to allow truck access to bring in the mining and construction equipment. The other major consideration was to protect the plant during the heavy torrential rain events that occur in the area. Because the landscape was generally flat with a shallow fall from the east to the west, the mine site and associated facilities would be subject to sheet flooding. Shallow but wide protection drains and bunds were constructed around each of the facilities. Hindsight has proven this to be a wise decision.

Hazan and Hazop Once the design was confirmed and the piping and instrumentation and final plant layout drawings were approximately 90% completed, a detailed HAZAN/HAZOP review was camed out. The review identified a number of issues that had been either overlooked or had not been given sufficient consideration during design. The timing of the analysis was such that the corrections required could be carried out without affecting the project schedules. !,

iThe concept of Hazan/Hazop analysis has been used consistently in the subsequent mill upgrades /to assist in the management of change. I

Iqfrastructure - Administration A’majorconsideration during design was that the mine site was to operate under a fly in/fly out (FIFO) arrangement for the staff. Thls meant that an airstrip capable of accommodating commuter aircraft was to be constructed along with a 156-personaccommodation village. The location of these two facilities presented interesting challenges as they had to be located on the mine lease, in an area that was not prospective for gold mineralisation and that catered for the special requirements of the facilities. Specifically, the airstrip required a “protection” zone around it, the size being dependent on the largest aircraft likely to use the strip. The accommodation village was to be located at a distance &om the mine operation such that industrial noise would not be an issue. The village also increased load on the power and potable water facilities. The most difficult design task in the whole operation was the main administration office. Everybody w i t h the organisation had a view on what was required, but no concept of the costs required to fulfil their dreams. Common sense, logic and cost control prevailed!

2354

able 4 Design criteria Material Characteristics Ore type Grade Mineralisation - BIF - Volcanoclastic Rod mill work index Ball mill work index Abrasion index Specific gravity Bulk density

Oxide 4.0 43 57 13.5 10.9 0.148 2.81 1.65

Production Throughput Gold recovery - Gravity - Leach - Total Gold production Crushing rate Crushing plant utilisation Milling rate Milling plant utilisation CIL circuit configuration Total CIL residence time CIL feed pulp density

1,000,000 20.0 72.5 92.5 100,000 200 55 109 94.3 1 leach + 6 CIL 24 44

Consumables Cyanide consumption Lime consumption Grinding media consumption Total power consumption Process water salinity

0.94 4.00 1.15 25.9 250,000

CAPITAL, COSTS Capital costs for both the prefeasibility and feasibility stages were generated using the following methods:

Obtaining three quotations for major equipment estimated to cost in excess of A$5,OOO with all other equipment costs from the engineering company’s records Material quantities taken from engineering drawings and quotations from suppliers Construction costs from budget quotations from construction companies based on the quality of drawings available at time of quotation Project engineering and construction management costs from the engineering company Owner’s cost estimates supplied by the owner. The cost comparisonbetween the prefeasibility study and the feasibility study is set out in Table 5 below.

2355

Feasibility Study A%

Prefeasibility Study A%

Difference

Plant construction Water Tailings Services Support facilities

15,842,760 1,559,887 4,066,706 5,921,031 7,584,730

12,748,949 2,712,496 2,558,424 7,523,685 4,540,037

3,093,811 -1,152,609 1,508,282 -1,602,654 3,044,693

Construction indirect

5,27 1,55 1 40,246,665

4,882,000

389,551

34,965,591

5,281,074

Description

Total

A%

The reason for the increase between the pre-feasibility study and the feasibility study costs were 0 0

0

0

Increase in tonnage from 0.75 Mtpa to 1.OO Mtpa Increase in tailing storage facility area to account for the reduced consolidation affect due to the use of hypersaline water Village accommodation increased from 80 rooms to 156 rooms due to the operation agreeing to provide mining contractor’s accommodation Increased accuracy of the estimate.

The above prefeasibility costs were developed on the EPCM concept. The feasibility study tender documents requested that the engineering companies submit both an EPCM bid and a Lump Sum bid. The Lump Sum bid by JR Engineering Services (JRES) was too attractive to ignore, and following clarification meetings, the JRES Lump Sum tender was accepted. The final capital cost for the process plant and associated facilities was A$38,967,431 which included a sum of A$1,684,538 for additional water bores required because of the poor production performance of the original bores. In addition to the above costs, A$3,585,000 was spent to purchase the power generation equipment rather than purchase power under contract at an elevated rate that included the capital component as well as the operating cost. OPERATING COSTS The prefeasibility and feasibility operating costs were developed generally from first principles using data generated in the testwork and information from the engineering design. Quotations from suppliers were obtained for all materials and reagents. The feasibility operating costs are set out below in Table 6.

2356

Table 6 Feasibility study oxide 01 Description

Manpower Consumables Power Liners and Media Maintenance Contract Maintenance materials Outside Services ROM Pad Ore Rehandle !Total

$

operating cost Annual costs A%

Unit costs A%It

2,115,456 3,201,000 3,4 12,167 1,685,889 1,002,222 991,556 289,444 600,000 13,297,733

2.12 3.20 3.41 1.69 1.oo 0.99 0.29 0.60 13.30

OPERATING PHILOSOPHY AND STAFFING The decision was made early in the plant design process to install a high level of process control and automation and utilise a relatively small number of highly trained staff to operate the plant. The control platform chosen was based on Allen-Bradley PLC’s and the CiTect SCADA system. It was not deemed appropriate to utilise expert control for the relatively simple flowsheet, however, wherever possible, routine plant control functions were provided by the process control facilities. This was aimed at providing the process technicians with sufficient time during each shift to focus on process optimisation and variance analysis, with the ultimate aim of improved production and safety. The staffing philosophy was centred around using ‘green’ process technicians, based on the success of this approach at URGM in 1995. A longer-term vision of self managed work teams also led to an extremely flat organisational structure as outlined in Figure 3. Plant maintenance was to be carried out by a contract workforce supervised by company coordinators.

I Mill Superintendent

Production

1L-

Production

I

I

Plant Metallurgist

Maint. Supervisor

I

Maint.

- Coordinator

I Maint. Project Engineer

!

Technicians (contract)

The dotted lines represent functional reporting relationships.

Figure 3 Organisation chart - SDGM processing department SDGM works on a FIFO roster system where employees commute to the mine from their place of residence (generally Perth) for a continuous work cycle then commute back to their place of residence for an extended leave cycle. In the case of technical staff, such as metallurgists and

2357

coordinators, employees work nine days followed by five days off. Production and maintenance technicians work 14 consecutive 12-hour shifts followed by seven days off. The 14-shft block was divided into seven dayshifts and seven nightshifts for the process technicians. The FIFO roster system results in three shift crews of process technicians. At any time, one crew is on dayshft, one on nightshift and one is out on leave. A total of 18 process technicians were recruited to operate the plant, resulting in three shifts of 6 people each, allocated as follows: Crushing Technician Control room technician Wet plant technician Gold room technician Laboratory Technician Services Technician

1 1 1 1 1 1

(working seven dayshifts and seven nightshfts roster) (working seven dayshifts and seven nightshifts roster) (working seven dayshifts and seven nightshifts roster) (working fourteen dayshifts roster) (working fourteen dayshifts roster) (working fourteen dayshifts roster)

No foreman or shift leader was appointed, with the control room technician assuming responsibility for the performance of the shift. Individuals are rotated through the positions on a cycle determined by shll levels and training requirements. Technical staff are on 24-hour call to render assistance on nightshift if needed. Of the 18 process technicians recruited, 3 were transferred to SDGM from URGM as experienced operators. These individuals were responsible for leading the shifts through the commissioning period and assisting in hands-on training of the fresh recruits.

Process Technician Training Program The decision to recruit ‘green’ process technicians was based on the opportunity such a strategy provides to train the individuals in company procedures without having to ‘undo’ previous training that may not be compatible with the company’s objectives. In addition to this benefit, the employees exhibited a large degree of enthusiasm for the job given their exposure to a new industry and technologies. A significantproportion of the technicians recruited were transferred from the company’s exploration division, bringing with them a considerable amount of project development history. In order to prepare the technicians for the commissioning of the plant an intensive training program was developed. An analysis of training needs and skills was conducted and the requisite shlls of a competent process technician were divided into five levels of increasing complexity. These levels were then linked to remuneration in accordance with the local employee relations regulations. The skills and training needs were combined into a training matrix outlining theory training requirements and competency requirements as shown in Figure 4. The matrix was designed around normal operation where technicians would be recruited in small numbers to cope with staff turnover. In general, a technician would be recruited as a Level 1 Process Technician and work their way through the matrix as they received training and developed skills and competencies. For the initial influx, however, all technicians were required to receive all of the training and were then classified into levels and assigned roles according to demonstrated skdls and competencies. This required an intensive six-week off-site training program, coordinated and delivered by the Production Coordinator and Production Metallurgist in Perth. Much of the training material was available in a general format from URGM. The two plants shared many common features. However, in order to customise the training material to SDGM, the services of Normet Pty Ltd, a metallurgical and training consultant, were employed. Normet assisted in producing training modules based on the SDGM plant design, largely interpreted from design criteria and engineering drawings. Ths information was compiled into modules incorporating metallurgical and plant operating theory, equipment descriptions and operating procedures. Extensive use of schematic representation was made to assist in the learning process.

2358

Because the training program was carried out in Perth where there is little opportunity to view real equipment, a video was made of the LRGM operations outlining each unit process and detailing equipment operation and care and operating practices. Original equipment manufacturer technical or promotional videos were used when available. The training program was carried out between January 6 and February 20 1997, after which the teams were mobilised to site in preparation for commissioning of the crushing plant. One of the unforseen benefits of the program was the strong team building that resulted from the team spending 6 weeks together in an intensive learning environment. As is natural in any group exposed to such an environment, leaders emerged and individuals took on responsibility for their team mates to ensure that no one lagged behind in skills acquisition. As this phenomenon was identified and other members of the processing department were recruited (i.e. maintenance technicians, metallurgists), these employees were required to attend certain of the training classes to gain exposure to their new colleagues and enhance the team building aspect of the program. Once the workgroup was divided into shifts, this early contact with the other members of the group paid dividends in terms of inter-shift and inter-discipline cooperation. The approach detailed above was not without its problems. The major downside associated with commissioning a new plant with new operators was the workload imposed on the metallurgists and experienced technicians. These employees were called on to ensure the plant ran smoothly and provide hands-on training to their work groups, which was a demanding role. The overall result, however, was extremely positive and would likely be repeated at any future company projects.

COMMISSIONING In order to capitalise on the learning opportunities presented by plant commissioning, the arrangements for plant start-up were focussed on allowing the process technicians to assume as much control of the plant as possible. To this end, SDGM operations and metallurgical staff were responsible for ore commissioning of the plant, whilst the engineering contractor was responsible for pre-commissioning activities and techmcal support during ore commissioning. The crushing plant was commissioned in February 1997. It was decided that a quantity of stemming material would be crushed for use in drill and blast activities in the mine before crushing any ore, thus providing an opportunity to run the crusher on dayshift for a period of two weeks. During this period, as many of the technicians as possible were rotated through the circuit. Had this exercise not been undertaken, the crusher would only have run for a few days before the crushed ore stockpiles were full, resulting in the requirement to crush on specification once the milling circuit was commissioned, with very little training time.

No major operational problems were encountered during crusher commissioning. Design throughput and availability were achieved immediately, paying testimony to the level of design and the quality of construction. The remainder of the plant was commissioned in early March (grinding, leach, gold recovery). A parcel of low-grade ore was treated initially to fill the CIL circuit and allow reagent levels to be established prior to introducing ROM grade ore. A design fault with the mill feed arrangement resulted in overdesign tonnages being treated from the start. However, the ore from the early stages of the pit was considerably softer than design, so the higher throughput did not lead to any operational problems.

2359

P

> P 2. P P

B

Level 1 Competencies Communicate in the workplace Work safely Apply local risk control process Operate gantry crane Operate load shifting equipment Handle reagents Manage water services

P

Level 2 Competencies Conduct conveying operations Manage crushing process Operate Rockbreaker Manage isolation process crusher only Respond to unplanned shutdown -crusher only Perform control room operations - crusher only plus 2 electives

P

Or Manage laboratoty

P P P

> P P

Level 3 Competencies Conduct leaching process Conduct elution process Conduct electrowinning process Manage isolation process leaching/elution only Respond to unplanned shutdown - leachingklution only Perform control room operations - leachinglelution only Conduct thickening process P

> > > > P

B

>

Level 4 Competencies Conduct milling process Perform control room operations Conduct wet gravity separation Conduct gold room operations Manage unplanned shutdowns all areas Manage isolations -all areas Leadership (Frontline Management Standards)

P

Level 5 Competencies Defined on an individual basis

plus 2 electives

Or

P

Manage Goldroom and Conduct ACACIA Reactor process

P

plus 3 electives

P P b P

Introduction to conveying module Crushing module Isolations module Rockbreaker module Or Laboratoty module Or Goldroom Module

Theory

General terminology lntro to Mineral Processing Safe manual handling Skid steer loader - skills training Forklift - skills training Reagents module Water systems module 4WD lntro to Citect Slinging & lifting and Introductory to O/Hcranes

Theory

B P P P

P

Figure 4 Training matrix

Electives Maintain Cyclones + Cyclone module Conduct pump operations + Introduction to pumping Conduct valve operations + Valves module

Theory

Leaching / adsorption module Elution & Carbon regeneration module Thickening module Oxygen module Isolations module Instrumentation Introduction to process control P

P

>

Electives Maintain Cyclones +Cyclone module Conduct pump operations + Introduction to pumping Conduct valve operations + Valves module

> > > > > >

Theory

Grinding module Isolations module Citect Introduction to continuous improvement Gravity separation module Front line supervisors course Electives

> > B P P P P

Theory

Dataanalysis Project optimisation (mentor) Met. Technician Statistical Process Control Workplace trainer & Assessor Report Writing Presentation Electives

The most disappointing aspect of the startup was the significant extent of problems that were encountered with the process control system. The engineering contractor had not allowed sufficient time in the development schedule to complete the programming prior to startup and many changes were required on the run. This led to much frustration and a less than ideal final product. The decision was taken some 6 months after startup to completely reconfigure the software to provide greater operability and maintainability of the system. First gold was poured on March 27, 1997, one month ahead of schedule. Practical completion and final hand-over of the plant took place on April 19, 1997, following achievement of the hand-over criteria listed below, and applying to a continuous 28-day period 90% utilisation of the mill Average throughput of 90% of design 100% utilisation for 72 hours 90% of design gold recovery. The effectiveness of the training program was evident immediately as the inexperienced technicians rapidly gained the knowledge to operate the plant proficiently.

:able 7 Actual performance versus design Design ~~

Production Throughput Gold recovery - Gravity - Leach - Total Gold production Crushing rate Crushing plant utilisation Milling rate Milling plant utilisation Consumables Cyanide consumption Lime consumption Grinding media consumption Total power consumption

1,000,000

1,O 12,572

1,178,046

20.0 72.5 92.5 100,000 200 55.0 109 94.3

42.0 53.5 96.5 149,844 220 53.2 124 94.4

45.1 50.5 95.6 205,956 3 13 47.7 139 97.4

0.940 4.000 1.150 25.900

0.743 4.430 0.880 23.430

0.478 3.563 0.740 22.001

Annual equivalents

STAGE 1 UPGRADE By mid-1998 the resource had grown from eight million tonnes to 22 million tonnes. In addition, the oxide reserves were scheduled to be largely depleted by mid-1999 and increased milling capacity was required in order to maintain production rates with an increasing proportion of transition and fresh ore content in the mill feed. A study into upgrade options for the comminution circuit commenced in May 1998. The primary objective of the study was to design a flowsheet that provided a throughput of up to 2 Mtpa on blended ores and 1.25 Mtpa on fresh ore, without prejudicing future expansion options in the event that a further increase in throughput was warranted. At the time, it was

2361

considered that the plant would ultimately end up as a SAGh3alVCrush (SABC) circuit, treating in excess of 2 Mtpa of fresh ore. However tertiary crushing and ball milling could not be ruled out, so the Stage 1 upgrade had to provide a circuit that would fit both scenarios in future. Studies into comminution circuit options were carried out by JRES. In-house studies using various specialty consultants (Orway Mineral Consultants, JK Tech) were also carried out to cross check the engineer’s conclusions and further explore available options. The final flowsheet was established in September 1998 in a joint review process between JRES, JK Tech and SDGM. Installation of a second ball mill, identical to the existing ball mill, in series configuration was chosen. The benefits of this choice included ease of adaptation into the plant layout, commonality of spares and vendors and perfect fit into a hture SAJ3C circuit if installed. Modelling of the new circuit indicated a capacity of greater than 2 Mtpa on oxide ore and 1.45 Mtpa on fresh ore, with blends falling between the two extremes. Considerable difficulty in finalising the crushmg flowsheet was experienced due to the fact that the simulation and modelling work all suggested an upgrade was required, but historical plant performance indicated that the existing circuit would handle the new duty. A decision was finally taken to delay any upgrade of the crushmg circuit until some operating experience on fresh ore had been gained. Given the excellent leaching kinetics displayed by the oxide ore to date, it was not considered necessary to upgrade the CIL circuit at the time. Leaching was typically completed halfway through the circuit so even though the throughput was effectively doubling, no loss of recovery was expected. Stage 1 upgrade design criteria are presented in table 8. A lump sum contract to construct the upgraded facilities was entered into with JRES in September 1998, for a total value of $A10.1 M. The new facilities were due to be commissioned in June 1999 with the overall schedule being driven by the ball mill delivery. The scope of works covered by the lump sum contract included:

0

0

Installation of a second 1.9 MW (4.27 m x 6.4 m) ball mill in series with the original mill Upgrade of the classification circuit Upgrade of the gravity recovery circuit by installation of a single 48 in. Knelson Concentrator Upgrade of pumps, pipelines and services to match the new duty Installation of three 1-MW generators.

The expanded flowsheet is presented in Figure 5 In choosing a Lump Sum contractor to carry out the design engineering and construction, it was considered that the original contractor, JRES, would provide a clear advantage due to their familiarity with the plant and association with the organisation and its staff. The project was not bid competitively but rather conducted under an open book arrangement with the contractor providing access to all cost and profit figures. SDGM were responsible for final equipment selection and specification as in the original project.

2362

riteria Material Characteristics Ore type Grade Blend Characteristics - Oxide - Fresh Rod mill work index Ball mill work index Abrasion index Specific gravity Bulk density Production Throughput Gold recovery - Gravity - Leach - Total Gold production Crushing rate Crushing plant utilisation Milling rate Milling plant utilisation CIL circuit configuration Total CIL residence time CIL feed density Consumables Cyanide consumption Lime consumption Grinding media consumption Total power consumption Process water salinity

OxideRresh Blend 4.0

!#

t/m3

10 90 25 19.5 0.300 2.80 1.75

tpa

2.0

%

% solids

30.0 65.0 95.0 >200,000 335 75 250 94.3 1 leach + 6 CIL 11.7 40.8

kg/t kg/t kglt kWh/t ppm

0.98 4.00 1.216 29.401 250,000

% %

kWWt kWt kWWt

% %

ozla tph %

tph %

h

Because the upgrade had to be built in and around an operating plant, cooperation between construction and operating teams was of paramount importance, as was minimising any downtime associated with tying in new equipment. This interface was very well managed by both parties and the project was completed with an excellent safety record and a minimum of interruption to production. Good cooperation between the plant maintenance group and the contractor allowed the majority of the tie-ins to be carried out during scheduled plant outages, with a final 24-hour shutdown being the only additional downtime required. The expanded processing plant was commissioned in June 1999. Official hand-over took place on June 30, 1999. Following the success of the initial commissioning, the same approach was employed for the upgrade, with SDGM personnel being responsible for ore commissioning activities. Once again the plant came on-line extremely quickly and design throughput was achieved without any fuss. Process control configuration was carried out by the contractor that had reconfigured the original system, providing continuity and standardisation. This ensured that the system was fully operational prior to startup and the operations personnel had had ample opportunity to gain familiarity with the new features and equipment. The increase in throughput also required an expansion of the on-site diesel power generating facilities and the tailings storage facilities.

2363

'able 9 Actual performance ve us design

Production Throughput Gold recovery - Gravity - Leach - Total Gold production Crushing rate Crushing plant utilisation Milling rate Milling plant utilisation Consumables Cyanide consumption Lime consumption Grinding media consumption Total power consumption

Design

Month l5 July 1999

Year 1 JuI-99 - Jun-00

tpa

2,000,000

2,023,5 12

1,792,154

% %

37.9 50.1 88.0 115,380 335

46.9 41.8 88.7 223,465 313

62.2

64.1

tph %

30.0 65.0 95.0 >200,000 335 75.0 250 94.3

236 96.0

212 96.9

kglt kg/t kglt kWWt

0.980 4.000 1.216 29.401

0.587 1.936 0.895 21.948

0.5 19 3.008 0.978 25.129

YO ozla tph Yo

Annual equivalents Tailings Storage The original paddock style tailing storage facility (TSF) was designed to accept 1 Mtpa of residue for a period of 5 years, with staged raising of the TSF embankments. When it was recognised that the plant throughput was to exceed 2 Mtpa and that the annual rate of rise of the stored tailings surface would exceed what was considered to be the limit of good practice, options for long term storage were investigated. Australian Tailings Consultants (formerly MPA Williams and Associates) were commissioned to evaluate a number of storage options including multiple paddocks, in-pit disposal, in-dump disposal and central thickened discharge (CTD) disposal. Following a desktop review of the available options, central thickened discharge disposal was chosen for its low cost, low energy and excellent environmental characteristics. The CTD storage facility and tailing thickener were commissioned in December 1999. The testwork carried out during the prefeasibility and feasibility studies showed that AMD and CNwadwere not issues that would occur during tailing disposal. The operation of the original paddock style tailing storage facility confmed the testwork results, with m w a d in tailings return water at less than 10 ppm. The operation of the CTD storage facility has continued to c o n f m the low m w a d content of the return water. Crushing Circuit During the second half of 1999 it became apparent that the crushng circuit was not capable of handling the increased throughput and ore hardness. In particular, maintenance requirements of the secondary crusher increased dramatically and a real risk of extended downtime was present. Because of the capacity constraints, operating hours were increased which in turn reduced the time available for maintenance activities and the overall quality of the operation was jeopardised. The safety and health risks associated with the condition of the plant and the pressures on the operating and maintenance staff were also recognised.

2364

Figure 5 Stage 1 upgrade flowsheet

2365

Surveys were carried out on fresh ore to allow sufficient data to be gathered to accurately model the circuit and the ore. By the end of 1999 sufficient information on future expansions was available SO that the requirements of any such expansions could be factored into any short term upgrades of the crushmg circuit. Approval to upgrade the secondary crusher from the Metso Omnicone 1560 to a Metso HP500 was granted in January 2000 and the crusher was commissioned in April 2000. JRES were again used to carry out all EPC activities. CONCLUSIONS The Sunrise Dam Gold Mine commenced its life as a lMtpa oxide operation with a mine life of 6 years. Between its inception in 1996 and mid 1999 the operation grew to more than 2 Mtpa via a major plant upgrade and ongoing optimisation. Further development of the operation to over 3 Mtpa occurred in 200 1. However the second upgrade is beyond the scope of this paper. From the very start, a quality approach was taken to plant design and every effort was made to ensure that the right equipment was installed and the right level of operability and maintainability were achieved. This did not necessarily mean taking a ‘Rolls Royce’ approach, but rather involving operating and maintenance staff at every stage to ensure their concerns were addressed and an appropriate level of ‘buy-in’ and ownership was achieved. The benefits of this approach are apparent in the performance statistics and the condition of the plant, as well as in the calibre and attitudes of the people that operate and maintain it. The model used to manage engineering activities has also proved to be extremely successful, and the relationship developed between SDGM and the principal engineering contractor, JR Engineering Services, has contributed significantly to the success of the operation. ACKNOWLEDGEMENTS The success of the SDGM project is a credit to the dedication, hard work and vision of many people. The authors would therefore like to thank all staff and contractors, past and present, for their contribution. The management of AngloGold Australia are thanked for their support and permission to publish this paper.

2366

A Case Study in SAG Concentrator Design and Operations AT P.T. Freeport Indonesia Rick Coleman', Andrew Nealez and Paul Staples'

ABSTRACT P.T. Freeport Indonesia (PTFI)currently operate a 760,000 t/d mining and 245,000 t/d milling operation in the remote highlands of Papua, the easternmost province of the Republic of Indonesia. The operations stretch from the dewatering and concentrate shipping facility at the Port of Amamapare on the Arafura Sea to the Grasberg open pit mine located more than 114 km away at 4,000 m above sea level. The four concentrators (Cl, C2, C3 & C4) are located an elevation of 2,800 m above sea level. The isolated location and generally difficult working conditions make for challenges not encountered in most North American operations. In spite of this, the mill design, technical and operating groups have been able to develop and maintain an effective level of plant wide productivity through the appropriate selection of available technologies and the implementation of innovative management practices. The history of the mining operation in general and the expansion of the milling operations in particular have been well documented by McCulloch (1991), Russel and Kieffer (1994), and Coleman and Veloo (1996). Van Nort et a1 (1991) have described the geology of the Grasberg deposit in detail. In 1995 C3, consisting of a single (10.4 m) 34ft SAG mill, two ball mills, and rougher and cleaner flotation circuits, was commissioned. Three years later C4, incorporating a single (1 1.6 m) 38ft SAG mill, four ball mills, and rougher flotation, was started-up. The C3 cleaner circuit was expanded and is common to both concentrators. Today the two SAG mills process approximately 175,000 t/day, with the remainder (70,000 Vday) provided by the North/South concentrators (C1 and C2), which are conventional two stages crushing, with single stage ball milling. This case study details SAG mill design considerations and operating experience since start-up in 1995.

DESIGN OVERVIEW Before starting the C3 design, Freeport engineers, operators and maintenance personnel visited several large-scale SAG milling operations looking for best practice techniques to incorporate into the Freeport facilities. The importance of this exercise cannot be over estimated as Freeport learned from the successes and failures of other operations. The open sharing of ideas and concepts between mining operations is a testament to the spirit of cooperation within the industry. The key design features incorporated into both the C3 and C4 concentrators were as follows: 1. 2. 3. 4.

Variable speed SAG mills. Pebble recycle systems with conventional as opposed to high lift conveyors. No recycling of slurry to the SAG mills. Vibrating SAG discharge screens with built in redundancy (i.e. no SAG trommel screens). 5. A high level of process automation and process control. 6. Rougher flotation utilizing the largest cells available at the time of selection. 7. Flotation columns for the cleaning circuits.

I Senior Vice President Mine Operation. PT Freeport Indonesia 2 Vice President Technical Services. PT Freeport Indonesia 3 General Superintendent, Concentrator Planning and Operations, PT Freeport Indonesia

2367

Key factors impacting design considerations for both C3 and C4 were: Variable ore hardness and feed size distribution indicated highly variable mill throughput rates. Unplanned downtime would have a significant impact on overall operating economics therefore equipment redundancy was essential. 3. Importance of designing for ease of operation and maintenance. 4. The layout of C3 incorporated anticipated the future C4 expansion. 1.

2.

In mid-1995, less than 6 months after commissioning C3, a site-based design team was formed to evaluate opportunities for the C4 expansion. By this time a better understanding of the Grasberg deposit concluded that the C4 expansion would not be a simple duplication of the C3 expansion. During the early days of the C4 expansion project, several additional key design criteria were established: 1. Maximize component consistency with C3 to minimize inventory costs. 2. Project cost considerations would be decided on a Net Present Value (NPV) basis, rather than on a capita1 allocation basis. 3. Competitive bidding would not be required on all processing equipment.

The last point was critical. Sole sourcing can expedite a project and allows the vendor to work as a project partner to ensure that the process equipment meets the overall project criteria. Because Freeport had recently completed a similar expansion, the project team was familiar with current financial and commercial terms, so there was little concern that any particular vendor would try to take advantage of the lack of competitive bidding. Table 1 summarizes key installed equipment, while the following sections summarize the specific C3K4 flow sheets, with further details being provided by Coleman and Veloo (1996) and Coleman and Napitupulu (1997).

-

Table 1 Summary of Key Process Equipment

AKEA EQUIPMEKT

I c3 I 1 Number I

c 3 Size

I 1

c 3 Installed

I c3 I c4 I I Capacity I Number I

2368

c4 Size

I

I

c4 Installed

I

I

c 4 Capacity

Ore Delivery Ore from the Grasberg orebody is crushed to minus 20cm (8”) in three primary crushers located at the 3.400m (11,lSOft) level. Crushed ore is conveyed by two parallel conveyor systems, a 213cm (72”) system that transfers 7,500 t/hr and an 183cm (84”) system that transports 10,OOO thr, to a series of four vertical ore passes. The 3m diameter (loft) ore passes dcliver Grasberg ore to the mill at the 2,800m (9,200ft) elevation with two in operation, and two on standby. Ore is reclaimed at the bottom of the ore passes and directed to a common C3/C4 stockpile and a separate NortWSouth (CllC2) stockpile via redundant conveying systems.

Ore Reclaim Four variable speed hydraulic 213cm (84”) reclaim apron feeders are installed in the C3 reclaim tunnel, while three similar units are installed in the C4 reclaim tunnel. Each of the C3 units has a maximum output of 2,600 thr. and, in general two of the four are operated to provide a blend of fine and coarse ore while the other two are on standby. Both C3 and C4 share a single coarse ore stockpile. The C4 ore reclaim tunnel was designed and constructed as part of the C3 project to ensure that the C4 construction would not interrupt C3 production. However, due to the overall plant orientation, and restrictions on the reclaim tunnel length, only three feeders could be installed. To meet production requirements, the bed depth was increased to allow a unit capacity of 3,500 Vhr, with two of the three feeders running during normal operation. Grinding Both grinding circuits were designed and commissioned as conventional SABC circuits. SAG mill product is screened at 10-1I mm, the oversize routed to a pebble crusher, with crusher product returned to the SAG mill feed conveyor. The SAG screen undersize flows by gravity to a reverse closed circuit ball mill circuit. The C3 SAG mill is a 10.4m0 x 5.2m (34ft0 x 17ft) mill with a 10,600 kW gearless variable speed drive. also known as a wrap-around motor. The C4 SAG mill is an 11.61110 x 5.8m (38ft0 x 19ft) mill with a 19,500 kW gearless drive. The SAG variable speed drives have provided operational flexibility for processing variable hardness ores, ease of starting, grinding-out and stopping the SAG mills, and excellent availability. The mills and bearings were structurally designed to facilitate a 20% ball charge in C3, and a 21% ball charge in C4. Separate vibrating screens are installed behind each SAG mill. The higher capital costs relative to a SAG mill trommel screen is more than offset by the increase in SAG mill running time from decoupling the grinding and screening processes. The original two parallel 3.0m x 7.3m (loft x 24ft) double deck vibrating screens in C3 (one operating, one on standby except during periods of high circulating loads) have been replaced with larger 3.7m x 7.3m (12ft x 24ft) and more robust screens. Three similar screens are installed in C4, with two in operation, and one on standby. The C3 grinding circuit incorporates two 6.lm x 9.3m (20ft x 30ft) ball mills with single pinion 6,400 kW (8,500 hp) drives behind the SAG mill operating at 78% of critical speed. In C4 there are four 7.3 m x 9.3 m (24ft x 30ft) ball mills with 5,250 kW (7,500 hp) dual pinion drives installed. The C3 SAG screen undersize is gravity fed to two of three cyclone feed pump boxes each equipped with a 1,120 kW (1,500 hp) 51cm x 61cm (20in x 24in) variable speed centrifugal pump feeding a cluster of fourteen Krebs D26B cyclones. The third system is a standby unit and the cyclone underflow can be directed to either of the two ball mills. The C4 SAG screen undersize is gravity fed to a four-way splitter feeding four cyclone feed pump boxes with the same pumps and cyclones as C3. There is a standby pump installed on each pump box. While a standby cyclone feed pump has immediate capital and design implications, the benefit in terms of overall plant availability more than compensates. Freeport’s position is that a high capital item such as a ball mill should not be taken out of service due to the lack of redundancy of a relatively low cost cyclone feed pump. Operating experience to date has validated this decision. Flotation The C3 rougher flotation circuit consists of three parallel banks of 12 Wemco 85m3 (3.000ft3) cells providing 26 minutes residence time. The C4 circuit has four parallel banks of 9 Wemco 127m3 (4,500ft3’ flotation cells providing approximately 21 minutes of residence time. The rougher tails is final tails, the rougher concentrates are pumped to dedicated regrind circuits comprised of a 3.96m x 7.62m (13ft x 25ft) regrind mill in closed circuit with a cluster of Krebs D20B cyclones. Regrind cyclone overflow is pumped to a common cleaner circuit comprised of 3.66m0 x 15.24111(I2ft0 x 50ft) cleaner and cleaner-scavenger columns. Both sets of columns produce final concentrate. The cleaner-scavenger column tails are directed to a bank of 12 Wemco 85 m3 (3,000ft3) mechanical cells. The concentrate from this bank is pumped back to the cleaner-scavenger columns, the tails to final tails. Cleaner circuit copper recovery averages 9597% while concentrate grades average 30% Cu.

2369

Dewatering Final concentrate flows to one of two 2.2m0 x 13.7m (7.25ft0 x 45ft) vertical tower mills to reduce the top size to 80% passing 45 microns in order to minimize abrasive wear in the concentrate pipelines which deliver the concentrate to the port site dewatering plant. Concentrate is thickened to approximately 65% solids by weight, prior to pumping to port site. Dewatering at port site is done in conventional vacuum disk filters and rotary dryers to reduce moisture content to about 9% by weight prior to loading onto ocean-going ships. Tailings are thickened in a 75m (245ft) center-drive and a 122m (400ft) tractor-drive tailings thickener. The Freeport milling complex is located at about 2,590 meters (8,500ft) above sea level, and water recovery is key to the operation. Reclaim water provides about half of the mill water requirements. Thickened tailings are discharged into the tailings river transport system, which carries the tailings to the tailings deposition system located in the lowlands.

Process Control Both C3 and C4 are controlled from a single manned control room, supported by satellite control booths located throughout the milling complex. The Foxboro DCS/Allen-Bradley PLC control system currently has in excess of 1,700 analog inputs, 500 analog outputs, 10,OOO digital inputs and 5,000 digital outputs. The PLC network is comprised of ten ( 10) fully redundant Allen-Bradley (AB)PLCS’s communicating between processors over a redundant AB Data Highway Plus (DH+) fiber optic data highway, and with the DCS via appropriate foreign device gateways. There are two Siemens PLC’s controlling the two SAG mill drives and four GE Fanuc PLC’s controlling the dual pinion drives of the four C4 ball mils. The major components of the multi-node DCS include the operator stations, engineering work stations, control processors and foreign device gateways are all connected by a fault tolerant Carrier Band Local Area Network (LAN). By mineral processing standards, this is probably one of the larger control systems in the world. Process control was an integral part of the concentrator design, and significant efforts have been made to incorporate the appropriate level of advanced control throughout the operation as described by Neale and Veloo (1996), Perry et al(1996), and Anderson and Perry (1996). CONCENTRATOR 3 START-UP C3 Ball Mill Start-Up The C3 ball mills and flotation circuit were ready for commissioning in early February 1995, before the SAG mill was ready. In response to this, crusher slurry from the wet screening plant in the NorthiSouth crushing plant was temporarily directed to the C3 cyclone feed pump boxes. A graded ball charge of 250 tomes of 25(I”) to 65mm (2.5”) was used to start operation of the ball mills. The 25 mm balls were rejected within the first week of operation and replaced with the standard 65mm balls. No attempt was made to start the C4 ball mills with a graded charge. Many of the grinding and flotation control loops were commissioned in manual as the operators became familiar with the control system. However, by the end of the first week of operation, most loops were running in automatic, a testament to the efforts of the pre-start-up training team. Cyclone feed pump box level, cyclone feed density and cyclone pressure control required significant tuning efforts, and some sanding of the cyclone underflow tubs, flotation cells and cyclone feed lines were experienced during the start-up period. A significant problem during the ball mill start-up was sanding of the cyclone feed pump boxes and cyclone feed lines. This was due to highball mill operating densities “floating” balls out of the mills into the pump boxes and tripping the pumps. The ball mills had a large-diameter trunnion liner, which made it impossible to increase ball charge as part of the effort to maximize mill power draw. This was resolved by installing a ball-retaining ring made from steel screen meshing at the interface between the trunnion liner and the trommel, and also at the end of the trommel. Higher flights were also installed on the trommel to improve the separation of slurry and coarse rejects and ball mill scats, and preventing short-circuiting of slurry directly to the reject scoop. C3 SAG Mill Start-up The SAG mill was commissioned autogenously in late February. The SAG mill start-up was quite smooth with only minor adjustments required to electrical and mechanical components. Part of the reason was that the operators had almost three weeks experience with the ball mill and flotation circuit, so could focus on the SAG mill start-up, once it was ready for commissioning. Figure 1 below illustrates the changes made to the grinding circuit that resulted in throughput increases during the first eight months of operation.

2370

Figure 1: SAG Mill Average Tonnes per Hour

3.000

2,500 2.000

TPH

38 mm Grates installed

1.500 1,000

7

Autogenous Grinding

20- 6- 20- 3- 17- 1- 15- 29- 12- 2 6 10- 24- 7- 21- 4- 18- 2- 16- 30Feb Mar Mar Apr Apr May May May Jun Jun Jul Jul Aug Aug Sep Sep Oct Oct Oct

Initially it was intended to operate the SAG mill autogenously for 2-3 weeks to allow the operators to gain some experience with the new equipment. However, a combination of lower than expected throughput rates. the demonstratedabilities of the operators, and the fact that the ore was otherep and had to be milled, led to the decision to start charging grinding balls after two days of SAG mill operation. Initially 75mm (3p) balls were charged, but within a few days 105mm (4p) balls were introduced, and charge levels gradually raised to approximately 14% within four weeks.

Grate Design It was recognized during the C3 design stage that the SAG mill feed contained a high levels of fines, as much as 50% -25mm (Ip). As a result, the original grates were designed with 25mm (Ip) slots. However. it was S o o n realized that the grates were actually restricting throughput to 1,500 t/hr at a ball charge of 11% by volume, compounded by the fact that the grates started to peen over on the outer section, reducing the slot width to 20mm (0.75~). Because alternative discharge grates were not available on site it was decided to cut 50mm (2p) pebble ports in half of the installed grates to increase the open area from 9% to approximately 11%. This resulted in an increase in throughput to 1,700 tlh. A further increase in ball charge to 14% ball charge increased throughput to 2,000 tlh. A set of 38mm (1.5”) grates was designed, ordered and expedited to site in late May and installed in June. The open area of the 25mm (1”) grates had been increased to approximately 12% by cutting more of the inner and outer grates. While this allowed for some very crude experimenting to find the ideal open area, the cutting activities eventually led to the failure of the grates. The new 38mm grates provided an equivalent open area of 12.4%. Throughput rates at this time were approximately 2,200 t/h. Larger slot width grates were considered, and a set of 50mm (2”) grates were designed but never ordered. The original 38mm grates were designed with minimal structural support between each slot in an effort to maximize open area. Although these castings did provide the desired mill throughput the early breakage due to excessive steel-on-steel contact, reduced SAG mill availability, and overall mill throughput. A revision was made to the grate pattern to add the structural strength back into the casting, which had the undesirable effect of reducing the open area to 10.3%. The center discharge plates were then redesigned as a half-platehalf-grate to restore the open area to 12.2%. The original grate design incorporated a 35cm ( 1 4 ~ high ) lifter bar. It soon became apparent that this high lifter bar was creating a low-pressure zone on the trailing side of the grate, particularly at mill operating speeds of 80% of critical. This resulted in less material passing through the trailing half of the grate than through the leading side. To counteract this effect the new 38mm grates were designed with a lift section reduced to 20cm (8p).

237 1

V i e r Design The original SAG mill liner concept was to maximize mill availability by designing very large, long-life lifters incorporating a “High-Low” configuration. and near vertical lifter angles. This liner profile provides for wear protection of the low lifter by the high lifter, and allows for the replacement of only half the shell liners at any one time thus reducing downtime hours for a particular liner change. Unfortunately, the 35cm (14”) high lifters with no relief angle, overthrew the grinding media, missing the impact zone and impacting the shell above charge. This resulted in minimal breakage of the critical size material, and accelerated liner breakage. The solution was to remove all the high shell liners and install the spare on-site set of low liners. A new set of shell liners was designed with a greater angle of relief on the leading faces to ensure the grinding media released and impacted where required.

SAG Charge Level The SAG mill was structurally designed to accommodate a 20% ball charge since the mill was expected to act more like a very large ball mill rather than a conventional SAG mill to accommodate the tine feed size distribution. The results of increasing the charge level from 14% to 20% were somewhat inconclusive given all the other adjustments being made. However, the end result was that the ball charge was maintained at approximately 18%with SAG mill throughput averaging approximately 2,600 t/hr. At this throughput the ball mills started to become the constraint therefore the focus was placed on ball mill improvements and optimizing SAG mill product size distribution by adjusting SAG screen apertures and crusher cavity profiles.

SAG 1 Discharge Screens and Pebble Crusher The discharge screens were originally fitted with parallel slots of 9 mm aperture bottom deck panels. The panels very quickly blinded resulting in significant quantities of minus 9 mm material circulating to the pebble crusher. This in turn caused the crusher bowl and mantle to build up with mud and resulted in “ring bounce”. The solution to this problem turned out to be replacement of the screen panels with a zigzag design, which was significantly more efficient. The pebble crusher was capable of processing only approximately 300-400 t/h even though at times feed rates were in excess of 500 tph resulting in a fraction being bypassed back to the SAG mill. C3 Ball Mill Throughput Limitations Ball mill circuit limitations were initially due to high circulating loads resulting from low cyclone feed density, undercharged ball mills and inefficient use of installed ball mill power. Increasing cyclone feed density significantly improved ball mill capacity and increasing the ball charge level to 39% by volume with the aid of the ball retaining screens reduced the circulating load. The original ball mill liners were a very thick single wave shell liners designed for longevity. The single wave liners were redesigned as double wave liners with reduced thickness. This resulted in an increase in mill volume when installed and increased mill power draw at the same ball charge, due to the increased lever arm, i.e. the distance from the mill axis to the inside of the liner.

2372

The success of the expansion project is summarized in Table 2 below.

Cyclones CycloneFeed CopperRec.

1 1 I

% %

I

56.5

56.5

90.5

90.5

I

I

65

67

92

88.8

The preceding table suggests the C3 expansion was a success. SAG mill throughput exceeded expectations while above forecast operating time is indicative of a concentrator which was designed and built with top priority given to operational and maintenance considerations. CONCENTRATOR4 EXPANSION PROJECT By mid-1995. less than 6 months after the C3 start-up, a site-based design team was formed to evaluate the opportunities for a further concentrator expansion. In January 1996, a decision was made to proceed with the project, and by early 1998 site commissioning of the C4 concentrator commenced. Some of the key issues of the expansion project are reviewed below, with further details of the start-up provided by Coleman et a1 (2001). The first phase of the project involved expanding the existing C3 cleaner flotation circuit into a combined C3-C4 cleaner circuit, including: 1. Relocating two Vertimills (tower mills). 2. Installing six column flotation cells. 3. Modifying eight existing column flotation cells. 4. Replacing twelve 42.5m3 (1,500ft” flotation cells with twelve 85m’ (3000ft3’cells.

2373

During the 40 days required to accomplish this work the cleaner circuit tailings were directed to the rougher feed with little impact on existing operations. The first of four 7.3 m (24ft) diameter x 9.3 m (30ft) long ball mills and a section of rougher flotation circuit were commissioned in December 1997 utilizing crusher slurry from the North-South crusher wet screening plant. Initial start-up problems were experienced with the clutch control on the dual pinion “Quadramatic” drives for the ball mills. Alignment issues, which are often a problem with dual pinion drives were experienced, but limited. The 11.6 m (38 ft) diameter x 5.8 m EGL SAG was commissioned autogenously in January 1998 with ball charging following within days. This was a mere three years after the start-up of C3, and only 17 months after breaking ground for C4. The SAG mill was commissioned with 38mm (1.5”) grates and a high-high shell liner configuration. Several days were required to balance the ball charge in the SAG and single ball mill. By early February, after commissioning the second of the four ball mills, the SAG mill was capable of processing between 1,700 and 3,300 t/hr depending on ore characteristics. The third ball mill was commissioned in May followed by the fourth ball mill in June. Shortly after the start-up of the third ball mill C4 achieved a throughput of 100,OOO tonnes in one day. The operating objective at this point was to continuously maximize SAG mill throughput with the following being evaluated during the months subsequent to start-up.

Grate Apertures: The 38 mm grate openings initially peened over to 32 mm apertures and restricted throughput. However, as the grates wore the aperture size increased resulting in additional throughput. Grates with 50 mm openings were ordered and installed.

Pulp Dischargers: During the C4 construction phase it was recognized that the original set of pulp dischargers were undersized and would restrict slurry flow through the mill. A larger set of pulp dischargers was ordered and installed in May 1998.

Pebble Crusher: The MPlOOO pebble crusher was started up once tonnage exceeded 400 t/hr. When choke fed, the crusher ran well but required regular cleaning to prevent build-up below the crusher due to the excessive water carryover due to uneven loading on the SAG discharge screens. A number of modifications to the SAG discharge box were required to address this problem. Mill Power: During the design phase the motor power rating was increased above the original specifications of 18-19 MW to ensure adequate power was available if required. Mill power typically averages 16-18 MW and it is unlikely full power will be utilized. Mill Speed and Sound: Experience with C3 suggested mill throughput was optimized while operating at a mill speed of approx 76% critical. The C4 SAG mill was designed to operate at 76% critical under normal conditions up to a maximum of 80% critical. Optimum mill sound levels were determined by field measurements with a hand held sound meter and comparison with the local sound meters installed near the mill shell. Set points were then adjusted accordingly to optimize throughput. Shell Liners: The original shell liners were designed with 3 inches of plate section and 10 inches of lift with a face angle of 12 degrees. The mill contained 69 rows of liners therefore spacing was generous for a mill this size. Experimentation with shell liners designs began in 1999.

SAG MILL OPTIMIZATION By late 1998, the C4 concentrator had achieved near design throughput on a consistent basis. PTFI concentrator personnel identified areas in each of the major unit processes (SAG mill, ball mills and flotation) that were production bottlenecks. The following describes the most significant opportunities for process improvements identified in each area; see Staples et al (2001).

Shell Liners The steep face (12”) lifter angle resulted in a ball trajectory that impacted the shell above the charge toe at normal operating speeds. This risked liner damage. Consistent with practice at several other operations, and with output from a study using the MillSoft modeling software, F’TFI decided to test shallower face angles. In July 1999 a set of 18-degree lifters were installed and in May 2000, a set of 25-degree lifters were installed. Data analysis suggests that only a small gain in throughput can be directly attributed to face angle changes. However. due to the decrease in shell impacts, improved resistance to packing and steadier operation. PTFI continues to use face angles between 18 and 25 degrees.

2374

Due to the inherent difficulty of interpreting noisy production data over long periods of time (shell liners typically have a life of 9 months in the C4 SAG mill), FTFI has been working closely with suppliers and technology groups to investigate new off-line techniques for optimizing shell liner design. Discrete Element Modeling (DEM) simulations indicate that PTFI is running the correct range of face angle. The final consideration is lifter spacing and simulations indicate that increased spacing will provide some benefit. A 46-row set has been designed and is under consideration.

Discharge End Under fine feed conditions the C4 SAG mill can treat sustained tonnages up to 5,800 f i r plus 1,500 f i r recycle. The ability of the mill to discharge this very large volume of material is unique to F'TFI. Thus, a great deal of focus and energy has been placed on optimization of discharge end design. At start-up, 38mm grates were installed with an open area of approximately 11%. Shortly thereafter, 50mm grates were trialed and by mid 1999 had become the standard. Further grate design changes were made over the following 18 months to first optimize the slot layout and then to increase open area. In addition to these considerations were issues of structural integrity and liner life. With the move to very large double sized grates, designed to reduce re-line time, came casting and metal flow challenges for liner suppliers. It is the experience at FTFIthat good grate design must consider several factors: Grate slot aperture size - Apertures must be sized correctly to discharge critical size material but also must match the recycle systems capacity. Increases in slot aperture size have resulted in significantly increased throughput and further increases are expected as testing continues. Open area - Open area must be balanced with aperture size to match recycle system capacity and obtain the desired wear life. Higher open areas have generally been beneficial at FTFI. However, though open area is important, placement of that area is even more critical. PTFI has tested open areas as high as 15%that wore to nearly 20% after 4 months. Recent modifications discussed below have reduced this slightly. Slot design - A crucial aspect of grate design lies in the positioning of the grate slots. Again, there is a balance with wear life that must be made. Other critical factors that should be addressed are ensuring adequate channel depth for flow in the pulp lifter chamber and positioning of the slots to minimize flowback into the mill. Relief angle - The design must also ensure adequate relief angle to prevent blinding of the slots with grinding media. PTFI experience indicates that 3-degree relief is inadequate and current designs utilize a 5-degree angle. By examining wear patterns in the pulp lifter chambers and on the front and backside of the grates, the decision was made to remove three slots at the inner most radius of the inner grate. It appeared these slots were contributing more to flow-back than to transport of material into the pulp chamber. Figure 2 shows the three inner slots that were removed from the inner grate. Figure 3 shows the most recent design in C4 without these slots. The reader may also note that, at the expense of some open area, there is more steel between the slots to reduce incidence of premature breakage. Tests of the new design are currently underway and initial results are very encouraging. Pulp Lifters The other significant change tested was to increase the pulp lifter depth by 4 inches at the cone and taper the increase down as you move to the outer radius of the mill. It was felt that pumping capacity of the pulp chamber was marginal and that there was significant restriction at the pulp discharger. This made it possible to increase the volume of the pulp discharger without the loss of mill volume at the shell, where most of the power is drawn. This change was made in March 2001 and along with changes to the grate design, indications are that pumping capacity has improved with C4 achieving record throughput levels. Again, PTFI is working closely with suppliers and technology groups to evaluate options to further increase throughput. It is hoped that DEM will provide solid information for options such as converting to a vortedcurved discharge end. Because discharge end pumping capacity is such a major concern with both the PTFI SAG mills, moving to a curved end would likely have significant benefits. However, the risks and downtime associated with such a drastic change are daunting and more confidence is required before proceeding. Development of an entire discharge end model is underway and it is hoped that the necessary information can be gained using this technique.

2375

-

Figure 3 Current Test C4 Grate Design SAG Bypass Project A recent development is the C4 SAG Bypass. First simulated in 1998, this flow sheet change introduced two additional recycle conveyors to transport pebble crusher discharge directly back to the SAG discharge screens, thus bypassing the SAG mill. Simulation studies estimated that for every 4 tonnes of crusher discharge that bypassed the SAG mill, an additional tonne of new feed could be processed, The new system was commissioned in June 2001 and is credited with increasing C4 SAG mill throughput by 5,000 t/day, which is remarkably consistent with

2376

the simulator predictions. This is a particularly exciting development in that it may well define the new flow sheet standard for SABC circuits with separate SAG product screening.

-

Figure 4 SAG Pebble Crusher Discharge Bypass Flow sheet

C3 To C4 Slurry Transfer C3 has 13 MW of installed ball mill power, which at a typical feed rate of 2,800 f i r , translates to 4.6 kV installed per tonne of fresh feed. Recognizing that C3 SAG mill throughput was often restricted by ball mill capacity. C4 was designed with 42 MW of installed ball mill power, a significantly higher 8.4 kW installed per tonne at 5.000 fir. Typically, C3 grinds averaged 23% +212 micron while C4 averaged 15% +212 micron (28 mesh). A 35cm slurry pipeline was installed and connected to the spare cyclone feed pump in order to divert a portion of the C3 SAG screen undersize (SAG circuit product) to the C4 ball mill circuit. Pumping this material to C4 utilized the additional ball mill capacity in the C4 circuit and removed a bottleneck that often restricted C3 SAG mill throughput. This improved overall grinding power utilization, resulting in finer grinds and increased metal recovery. The current system is capable of pumping 600 to 1500 t/hr of slurry to C4 and has resulted in a throughput increase of 3,000 t/day in C3. In addition, the overall C3/C4 combined grind has improved by 1% +212 micron. and copper recovery has increased by 0.3%.

SUMMARY OF LESSONS LEARNED As usual with projects the size of C3 & C4, a number of valuable lessons were learned. Following is a brief summary of some of the more memorable lessons from the grinding circuits. The C4 SAG mill discharge box and the distribution of feed to the vibrating screens was one of the more significant challenges. After various modifications and the replacement of the original vibrating screens most of the problems were solved. The importance of getting this right in an SABC circuit cannot be overemphasized as the penalties include excessive downtime. and poor dewatering of the SAG discharge slurry resulting in poor pebble crusher performance.

2377

0

a

Conveyor speed and CEMA loadings are important features to consider in the design of a recycle system for an SABC circuit. Magnet efficiency is significantly reduced when attempting to remove steel from a fast moving highly loaded belt. Tramp metal recovery was improved by adjusting the idler configuration to flatten the belt and spread the load under the magnet. The pebble crusher circuits were designed without a surge bin. Although it is possible to keep the crushers fed through a combination of improved magnet operation and higher circulating loads, the SAG circuit stability suffers when the crusher feed is inconsistent. The design of the SAG mill discharge end liner configuration is important. Not only the size of the grate apertures and amount of open area but also the volume behind the grates and the cross sectional area at the narrowest section of the pulp dischargers. Due to the very high SAG mill volumetric throughputs the pulp chamber depths were increased during the design phase. Recent changes have further increased the volume of both the pulp chamber and pulp dischargers resulting in additional throughput increases. During start-up it is prudent to have two or more sets of varying aperture grates on the ground in anticipation of throughput issues. Ball mill lining requires the removal of the feed end trunnion liner prior to installation of the lining machine, adding hours onto each lining job. Either lining from the discharge end through the large opening or a redesigned feed trunnion would significantly reduce lining time. Wrap around variable speed motors were originally considered for all four ball mills. However due to the relatively small power grid it was felt that the harmonics could potentially disrupt power supply to all process equipment. As such the twin pinion drives were selected. The soft start capabilities of the wrap around motors would have significantly reduced mill downtime and likely improved grinding circuit performance for all conditions experienced. Flotation cleaner circuits bixes were all designed within large sumps with pumps placed at a low elevation within the sumps. The pumps were susceptible to flood conditions causing the V belts to operate in slurry. It was necessary to raise all the pump one meter to keep them high and dry. The slope of the underflow launder for the 400 ft thickener was set at 1/16 inch per foot. Upon start-up of the thickener, the underflow launder overflowed when transporting higher density slurry (+ 60% solids) containing coarser material (+30% 65 Mesh). As C4 ramped up in tonnage the launder problem diminished through a combination of higher launder velocities and better slurry rheology. On reflection, a slope of 1/8 or 3/16 of an inch per foot may have been more appropriate. The 400 foot thickener center well had a number of design flaws including feed pipe arrangement, lack of rubber lining in the center well and insufficient size, all of which contributed to high wear due to extreme slurry velocity. This in turn resulted in excessive downtime to repair or replace the feed well. The solution is a larger rubber lined center well combined with lowering the feed pipes thus reducing the incoming slurry velocity. The dewatering plant expansion was accomplished using disk filters and a thermal dryer rather than more recent technology (filter presses, ceramic filters etc) without fully investigating the advantages or disadvantages of the latter. Secondly, auxiliary equipment design criteria were duplicated without addressing current operational and maintenance issues. Vacuum pump capacity was one such example while CEMA loading on conveyors was another. Although the expansion equipment continues to operate, Freeport has since decided to replace older, high cost dewatering equipment with a pressure filter and upgrade most of the conveyors.

CONCLUSIONS The start-up of the expansion SAG Plant at P.T. Freeport Indonesia’s Papuan operations has, by all standards, been a tremendous success. This can be attributed to a combination of a good overall plant design and layout, unique mill feed characteristics, SAG mill liner and circuit design, a practical approach to plant wide process control, and the emphasis placed on training prior to the SAG Plant start-up. The success of the expansion can also be attributed to a teamwork philosophy from all those involved in these projects. ACKNOWLEDGEMENTS The authors wish to thank the management of P.T. Freeport Indonesia for their support and permission to write and present this paper.

2378

REFERENCES Anderson, L., Perry, R,& Neale, A.J., (1996), Application of dead-time and gain compensation to SAG feeder control at P.T. Freeport Indonesia. Proceedings, 28th Annual Meeting of the Canadian Mineral Processors, Ottawa, Ontario. Coleman, R.E., Nugroho, S.,Neale A.J., (2001). Design and Start-up of the PT Freeport Indonesia NO. 4 Concentrator, SAG 2001. Vancouver, B.C. October 2001 Coleman, R.E and Napitupulu, P., (1997). Freeport's Fourth Concentrator - A large Step Towards the 21'' Century, AUSIMM Sixth Mill Operators Conference, Madang PNG, Oct 6 - 8, 1997. Coleman, R.E. and Veloo. C., (1996). P.T. Freeport Indonesia Concentrator Expansion, SME Annual Meeting, Phoenix, Arizona, March 11 - 14, 1996. McCulloch, Jr., W.E. (1991). Mill Expansions at Freeport Indonesia four-fold production increase in a decade, Copper 91-Cobre 91, Volume 11, pp. 3-17. Neale, A.J. and Veloo, C., (1996). Process Control at P.T. Freeport Indonesia's Milling Operations, CIM Annual Meeting, Ottawa, Ontario, January 1996. Perry, R., and Anderson, L., (1996). Development of Grinding Circuit Control at PT Freeport Indonesia's New SAG Concentrator, Proceedings of SAG 96 Conference, Vancouver, British Columbia. Russell, R.L. & Kieffer, L.D. (1994). Mill Expansions at P.T. Freeport Indonesia, paper presented at the 1994 S M E Annual Meeting, Albuquerque, New Mexico. Staples, P., Siewert, H., Stuffco. T., and Mular, M., (2001). SAG Concentrator Improvements at PT Freeport Indonesia, SAG 2001, Vancouver, B.C. October 2001 Van Nort, S.D., Atwood, G.W., Collinson, T.B., Flint, D.C. & Potter. D.R., (1991). Geology and Mineralization of the Grasberg Porphyry Copper-Gold Deposit, Irian Jaya, Indonesia, Mining Engineering, March 1991.

-

2379

HIGH PRESSURE GRINDING ROLL UTILIZATION AT THE EMPIRE MINE

David J. Rose, Metallurgical Engineer, Empire Iron Mining Partnership Paul A. Korpi, General Manager, Empire Iron Mining Partnership Edward C. Dowling, Senior Vice President-Operations,Cleveland Cliffs Inc. Robert E.Mclvor, General Manager, Cliffs Technology Center Abstract

The Empire Mine is currently using High Pressure Grinding Rolls (HPGR) to process magnetite ore at its facility in Palmer, MI. The evolution of the flowsheet with emphasis on the needs for and advantages of the High Pressure Grinding Rolls will be outlined and discussed. Plant performance and circuit improvementswill be reviewed. Empire Mine History

Empire Mine is an integrated open pit mine, concentrating plant and pelletizing facility that is currently capable of producing 8.0 million LT of pellets annually. Opened in 1963, Empire was capable of producing 1.6 million LTPY of pellets. Expansions in 1966, 1975 and 1980 have increased the capacity to its current level. Empire Mine is owned by a joint venture of lspat Inland Steel Co. and Cleveland Cliffs Incorporated. Flowsheet Design

Ore at Empire Mine is crushed to minus 9 inches in a Primary Crusher (Traylor 60 x 89). Crushed ore is transported to one of two crude ore buildings where it provides feed to 23 Primary Grinding Lines. Fully Autogenous Primary Milling is used at Empire with double deck screens separating cobber feed size material (-lmm) from the bottom deck oversize (+lmm to -% in.). Bottom deck oversize material is returned to the Primary mill along with excess top deck oversize material (+% in. to 2% in.), which helps provide grinding media to the secondary (pebble) milling circuit. Figure 1 shows the original circuit, before excess pebble

2380

crushers were added to some lines. This flowsheet is still in place for ten grinding lines. Magnetic separators (cobbers) treat the Primary Mill product, rejecting approximately half of the crude ore from the plant. Cobber concentrate is then pumped to hydrocyclones to separate the minus 25 micron fraction from the coarse fraction, which is further ground in pebble mills. The pebble mill discharge is in closed circuit with the cyclones. Cyclone overflow material flows through a siphonsizer / thickener sizer to deslime the slurry and then through a second magnetic separation stage (finishers) to produce flotation feed grade material.

ohherr

II Figure 1

Empire Original Flowsheet (No Pebble Crushing)

The flotation plant at Empire is a reverse float with the gangue (silica) being floated off the top using an amine reagent. Flotation concentrate is thickened in conventional concentrate thickeners and filtered on disc filters to produce balling plant feed for the pellet plant. Figure 2 shows the modified flowsheet for the Empire in which excess pebble crushers have been installed on thirteen grinding lines. Excess pebbles are defined as those produced by the primary mill not required

238 1

to maintain horsepower setpoint in the pebble (secondary) mill. A variable speed feeder provides all the pebbles required in the mill and the excess enters the pebble crusher circuit by means of an overflow from the pebble hopper. The crushed pebbles are returned to the primary mill just as the uncrushed excess pebbles would be in the lines without pebble crushing. The line 1 Primary Mill was converted to a ball mill grinding fluxstone for the pre-fluxed pellets produced in the pellet plant.

Figure 2

Empire Flowsheet With Pebble Crushing

Pelletizing at Empire is done using eighteen balling lines feeding four rotary kilns. Three pellet product grades are produced annually at the plant, which ships these pellets to several different blast furnace operators around the Great Lakes area.

Plant Equipment Three expansions in the years since initial startup have left Empire with four unique plant sections, each utilizing different equipment, in terms of manufacturer, design and size.

2382

Empire I was originally comprised of six Hardinge-Cascade 24-ft by 8-ft diameter primary mills and six Nordberg pebble mills 25 % feet long by 12% ft in diameter. Eighteen banks of three drum cobber magnetic separators (6 ft long by 3 ft diameter) provided the primary concentrating step. These were followed by 18 banks of six 10-in. Krebs cyclones, six 38-ft diameter siphonsizers and 18 banks of two drum magnetic finishers (6 ft long by 3 ft diameter). Final concentrate upgrade was achieved in one bank of eight 500 ft3 Wemco flotation cells. The magnetic separators in Empire I are a mix of Dings and Jeffrey’s units. Empire I1 started with 10 Allis-Chalmers Rockcyl primary mills, 24 ft by 12% ft diameter that are twinducer driven. The pebble mills are the same Nordberg 25% ft by 12% ft units as installed in Empire I followed by two 3-unit 8 ft by 3 ft diameter cobber magnetic separator drums from Dings. Thirty banks of six Krebs 10-in. diameter cyclones are followed by ten 3 8 4 diameter siphonsizers and 30 Dings finisher magnetic separators (6 ft by 3 ft diameter). The flotation section in Empire II is a 5-cell unit comprised of Wemco 1000 ft3cells. Empire Ill is a section of five lines using Allis-Chalmers Rockcyl mills 24 ft by 12% ft in diameter. These are standard pinion drive units. The pebble mills are Allis-Chalmers Rockcyl units 25% ft by 15% ft in diameter, followed by three Dings three-drum magnetic separators (8 ft by 3 ft diameter). Five banks of twelve 15 in. Krebs cyclones are used with five 46 ft diameter siphonsizers and Dings magnetic finishers to provide flotation circuit feed. The magnetic finishers are comprised of eight 6 ft by 3 ft units and nine 8 ft by 3 ft units. Empire Ill flotation is achieved in a bank of five Wemco 500 ft3 cells. Empire IV is the newest and largest section of the Empire plant. It utilizes 3 Koppers 32-ft diameter by 16% ft primary mills and six Nordberg 32 ft by 15% ft diameter pebble mills. There are 18 banks of three drum (10 ft by 3 ft diameter) cobber separators and 21 banks of two drum (10 ft by 3 ft diameter) finisher magnetic separators. Siphonsizers are replaced in the flowsheet by three 85 ft diameter thickener sizers. The original flowsheet included one bank of ten 500 ft3 flotation cells. Each line at Empire is independent of the others. There are grinding units dedicated to a certain section of concentrating equipment. The changes to this came about with the addition of distribution systems, which effectively “disconnected” the primary and secondary halves of the grinding circuit.

2383

Process improvements have been realized over the years by adding excess pebble crushing (Lines 11 - 24), cobber concentrate distribution (Lines 2 - 6, 12 - 16 and 22 - 24), additional flotation capacity (Lines 22 - 24) and the HPGR (Lines 22 - 24). Flowsheet Evolution

From the time Empire first produced pellets, the plant flowsheet has maintained the same basic design. Primary Autogenous milling followed by magnetic separation followed by secondary grinding in pebble mills and then finishing the concentrating process in magnetic separators and flotation cells. Ore hardness and lower iron content in the crude started to make it necessary to find ways to increase the throughput of the grinding section in the Plant in order to produce enough concentrate for the owners. The first addition to the flowsheet was the Excess Pebble Crusher in Empire IV (the last expansion section), where a 7 ft Symons cone crusher was installed to crush the excess pebbles (+% in.) material and return the crushed pebbles to the Primary Mills. The smaller material that is returned to the Primary mill is easier to grind to target size range and therefore horsepower consumption is reduced and the throughputs can be increased. More crushers were added to the older section of the plant in the mid 1990’s as the orebody was expected to get harder. Four Nordberg HP200 crushers were added to the grinding lines from 11 through 21 in differing combinations. Pebble Crushing proved to improve circuit throughputs by approximately 20% for the times when the crushers were running. These improvements allowed Empire to produce over 8.5 million LT of concentrate from 1995 through 1998. In the mid 199O’s, mine planning projections were still indicating a higher work index crude ore material coming into the plant. Another option to be investigated was the addition of another way to release the load in the Primary Mills and increase throughputs. Using the theory of the crushers, high pressure grinding rolls promised to provide better comminution at lower costs and higher productivity.

2384

High Pressure Grinding Roll Design

High Pressure Grinding Rolls have been developed over the last 15 to 20 years and have primarily been used in the aggregate industry. The units that are in use in cement have proven to be powerful workhorses and are capable of running at high availabilities and tonnage rates. In the mining industry, they are widely used in diamond operations throughout the world, such as a recent installation in a diamond mine in the Northwest Territories. More recently High Pressure Grinding Rolls are being installed in applications for gold and phosphate minerals. Also, a wide range of iron ore applications successfully utilize HPGRs for both coarse ore and pellet feed grinding, although Empire's HPGR installation continues to be the only iron ore application in North America. The HPGR design is comprised of a fixed and a moveable roll that run counter rotation to one another at the same speed. The moveable roll is positioned at such a distance from the fixed roll to exert a specific amount of pressure on the material passing between the two surfaces. The feed material is introduced at the top of the rolls in the area where the two surfaces begin to meet. The movement of the rolls and the static pressure from the feed above provides for a positive downward force and ensures an effective nipping of the feed into the gap between the rolls. The material between the rolls forms a particle bed and the force from the rolls causes the grinding action that reduces the material size. The grinding force acts throughout the particle bed, and the main size reduction mechanism is defined as interparticle crushing. This is considered effectively a grinding action between the ore particles as opposed to the roll-particle-roll contact crushing as in conventional rolls crushers. The manufacturer of the machine has stressed that the units are grinding machines, not crushers so the top size feed material must be limited based on the roll size. The roll surfaces are lined with hard metal alloy studs. These studs are designed to create high and low spots in the surface where the feed material collects. This makes the effective grinding surface smooth and provides a so-called autogenous lining. The second effect is that the wear on both the studs and base metal surface is minimized as the amount of metal area exposed to contact with the ore is reduced. The mechanical characteristics of HPGR include electro-hydraulic control systems for the roll positioning and pressure control. A separate lubrication system is included for the rolls and bearings. The entire unit is

2385

housed in an enclosed frame with dust shields to limit the amount of fugitive dust that escapes. On the ends of the rolls there are cheek plates that keep the material between the roll faces from getting squeezed out and bypassing the grinding action of the HPGR. These cheek plates are a wear area and therefore, are comprised of a high wear alloy. These need to be adjusted for position and pressure occasionally. Getting feed to a HPGR is relatively simple, once you have prepped the feed material to reduce the top size. A gravity feed chute can be employed to maintain a choked feed chute and allow for proper distribution into the HPGR. The drives for the rolls are variable feed, which can be used to maintain a level in the feed chute and keep the rolls choke fed. Access doors at the top and sides of the HPGR frame are used for minor maintenance and inspections of the rolls, cheek plates and feed chute. Empire Options For many years, Empire had been expecting the harder ores and looking for ways to increase tonnages (or at least maintain current levels) to produce high enough concentrate tonnages. In the late 1980’s and early 199O’s, Empire IV, with its excess pebble crusher was significantly more efficient than the other lines at treating harder ores. For this reason, a split ore blend program was used to maximize concentrator productivity. In this plan, the harder ores were directed towards the Empire IV section. The logistics associated with trying to maintain this schedule proved to be difficult. Other options for plant productivity included removal of the original DSM screens, which were a secondary screening step between the vibrating screens and the magnetic separators, the use of smaller pebbles in the pebble mills and changing the cobber feed top size to 1 mm from 2 mm. These changes have been kept in place with all the other changes that have been made.

In the mid 199O’s, the orebody was still getting harder and leaner, requiring the plant to treat more ore to maintain concentrate levels. With the success of the pebble crushing in Empire IV, excess pebble crushers were planned and added to lines 11 though 21 in Empire II and Empire Ill. These installations were complete in late 1995 and helped Empire to

2386

maintain production at 8.5 million LT of concentrate for the next three years. Mine planning in 1995 indicated that the silicate content of crude ores would be steadily increasing over the next few years from a historical 30 to 40% of the blend to 40 to 60% in the next few years to greater than 85% in later years. Harder ores reduce grinding rates and excess pebble crushing had helped to offset the shortfall up until then, but the future was going to consist of harder ores, with lower iron content. The only way to maintain high enough production levels was going to be increasing throughput to offset the reduction in recovery that was inevitable. The Empire Solution

HPGR technology had been pilot plant tested on the Empire ores and had proven in the controlled environment to be successful in treating high tonnages of ore and relieving the load in the primary mill, which increase throughput rates. The HPGR installed at Empire was originally intended to be a test for the technology and its application at Empire. Options investigated include a second stage of pebble crushing, but the technology available at the time would not produce the product size distribution that the HPGR was capable of and the energy costs would be higher. A requisition for funding was approved to purchase a HPGR for Empire IV, with performance guarantees in place to ensure the improvements Empire was looking for would be attained. KHD from Germany was the supplier of the HPGR based on the testing program they had completed in conjunction with Empire and Cliffs technical personnel at the Coleraine Minerals Research Laboratory (Duluth University of Minnesota).

The HPGR installed at Empire is a KHD model RPSR 7.0 - 140/80, the rolls being 1400 mm (55 in.) diameter by 800 mm (31.5 in.) wide, and providing a maximum specific pressing force of 6.25 N/mm2(906.5 psi). It is powered by two 670 kW (900 HP) variable speed controlled motors to provide a roller surface speed'of 0.9 to 1.8 m/s (2.95 to 5.91 Ws) through planetary reducers. The ROLVIS control package oversees pressure, gap width and speed, as well as monitoring power, oil pressure, the lubrication system, and other operational features.

2387

Rollers are carried in a machine frame in four-row cylindrical roller bearings as well as two self-aligning roller thrust bearings. One roll is moveable and the other fixed. The moveable roller moves closer or farther away from the fixed roller to meet the pre-set force as required at the surface where the rock is pressed. The system consists of 2 hydraulic cylinders with spherical pistons and accumulators and associated other equipment. Cheek plates are used on the edges of the rolls to keep material from bypassing the roll surfaces. These are adjusted to keep the spacing between the edge of the roll at a minimum. The HPGR was installed in the Empire flowsheet to follow the Symons pebble crusher in Empire IV treating crushed product and returning the fine product to the primary mills, as shown in Figure 3.

Figure 3

Empire Flowsheet With HPGR Installation

HPGR Performance

2388

The HPGR supplied to Empire was guaranteed to meet certain criteria regarding unit availability, tonnage throughputs, product size distribution, liner life and specific power consumption. The HPGR supplied to Empire was initially installed with segmented studded liners that could be replaced without having to remove the rolls from the frame. These segments (six per roll) were held in place by two rows of six bolts that were covered by alloy pucks over the bolt head. The one thing that became an issue with the segmented liners was that the bolt covers would come loose and fall out or cracks would develop and they would break into pieces and fall out requiring maintenance for the replacement of and repair to the bolt covers and holes. The unit availability guarantees from KHD specified that the unit would be ready to run at least 95% of the time. The frequency of the cover changes was getting to be such that the unit availability was suffering. The tire style liners were provided to Empire at a prorated cost based on the hours that the segmented liners lasted. Along with some other scheduled maintenance, the liners were replaced, with help from KHD and the HPGR was put back into operation. The unit installed at Empire was guaranteed to operate at 400 LTPH by KHD. The feed rates in 2001 averaged 323 LTPH. This is a result of the process not having enough material to consistently feed 400 LTPH to the HPGR. The tonnage of excess pebbles generated by the primary mills is not enough to maintain this feed rate to the rolls. This problem is being addressed by some material handling changes in the circuit allowing for recirculating loads around the HPGR. Product size distribution from the HPGR was guaranteed to be 50% passing 2.5mm. Performance testing in the circuit showed that the product was running approximately 47% passing 2.5mm. The HPGR product, as with any grinding operation, is dependent upon the feed size of material in the process. At the time of the performance testing, the feed size was coarser than the specified range from the pebble crusher and the reduction ratio was actually a little better than projected by KHD. Based on these results, the Empire testing was considered to be a success. The liner life for the segmented liners was guaranteed for 12,000 operating hours from KHD. Liner life on HPGR rolls is determined by the height of the studs on the studded surface. Measurements of the stud height are made on a regular basis to determine the wear rates of the

2389

studs and project liner life. Despite the difficulties with the bolt covers, the liner life was projected to meet the guarantee from KHD, so this was not the issue with the segments.

Figure 4

Original Segmented Liners Showing Bolt Covers

Specific power consumption in the HPGR was guaranteed to be 2.5 kWh/LT (3.35 HP/LT). Actual power consumption has declined steadily since the HPGR was commissioned and in 2001 averaged less than half of the target (1.2 kWh/LT).

2390

Figure 5

Damaged Bolt Cover on Original Liners

A significant change in technology allowed the manufacturer to construct a seamless tire style liner that did not require the use of bolts or bolt covers. This eliminated the need for replacing and repairing the bolt covers and bolts as we had seen with the original liners.

The original liners have been replaced with the new seamless liners as part of the optimization of the HPGR circuit at Empire. HPGR performance has not been affected by the new liners, but availability has been much improved. The only downtime taken now is essentially scheduled downtime to perform routine maintenance tasks.

239 1

Figure 6

New “Tire Style” One Piece Liner

Empire HPGR Plant Results

Table 1 HPGR Results 1998 - 2001

I

!. j

-1KJ98 1999 * 2000 2001 *

I

RP-3 Hrs

-\-.

2.375.3 1.971.5 4.198.8 3,943.9

’

’

I

RP-3 LT 570.051 1,305,918 1,273,460

i I

1 1

HP/LT

% Op Time (with feed)

2.02 1.96 1.61

22.5 47.8 45.0

1 1 1

Feed Rate (LTPH)

289.1 311.0 322.9

I

% PC Product Roll Pressed

81.8.L I-- -66.3

1 1

.

90.9 92.3

Table 1 shows the HPGR run data for the past four years. Operating time (defined as with feed) is still low, but the key is the percent of potential feed that is actually fed to the HPGR. This has increased to over 92% indicating that the unit will handle the tonnage Empire can currently

2392

provide. The feedrates have been higher than the 400 LTPH target at times, but this cannot be sustained because of feed material limitations.

Productivity Without Roll Press Operating

2000 2001

2000 2001

1

1

LTPH 356.9 366.6

HPlLT 23.9 22.7

PRIMARY MILL LTPH 417.3 429.3

PRIMARY MILL HPlLT 20.1 19.1

PRIMARY MI LC-LTPH 17.3%

PRIMARY MILL PR HPlLT -16.2%

HPlLT 37.7 35.0

-

~~

TOTAL HPlLT 32.3 30.7

-

~~~~~

TOTAL HPlLT -13.9%

Table 2 shows the effect on primary and total milling that the HPGR product has. Without HPGR product going back to the primary mills, the feed rates are lower and the power consumption rates are higher. The feed rates are a little more than 17% better with the HPGR than without and the specific power consumption rate drops by almost the same amount. Summary Empire Mine has successfully integrated high pressure grinding roll technology into an autogenous grinding circuit treating high tonnages of magnetite iron ores. Being the first application of its kind, the cooperation and support of the manufacturer has enabled Empire and KHD to learn more about the capabilities and characteristics of these grinding units.

2393

The Raglan Concentrator - Technology Development in the Arctic J. Holmes', D.Hyma', and P. Langlois'

Abstract The pre-operational period of the Raglan mine/mill complex was characterised by two significant exploration and engineering study programs, the first during the mid 1970's and the second in the early 1990's. During both periods, the challenges associated with the remote arctic site led to the requirement for novel construction methods together with the adaptation of technology. This paper focuses on three specific metallurgical studies conducted in 1991-1 992, each aimed at defining the most feasible option for the project. The fully autogenous grinding circuit, the concentrate dewatering, storage and transportation system and the tailing dewatering and deposition system are presented in detail. Technology development is then examined by comparing each initial design to the current operation, including a summary of the lessons learned.

PROJECT INTRODUCTION Falconbridge's Raglan Concentrator, located at the northern limit of Quebec's Nunavik Region, was piloted in 1991. Detailed engineering started in January 1995. The plant was constructed in modules in Quebec City, where virtually all of the process equipment was installed. During the summer of 1997, the modules were shipped via barges to Deception Bay, and transported 96 km by land to the concentrator site in Katinniq. The concentrator modules were assembled and connections completed between August and November 1997, with ore first added in December. Located north of the 62"d parallel, the immediate region is typified by shallow topography, little or no vegetation, and permafrost. All electrical energy is generated on-site with diesel generators. There are no road or rail connections to the south, and all of the workers at the site must be transported by air. Workers from the south are flown from Rouyn-Noranda in a company-owned and operated jet, and workers from northern communities in Nunavik are flown to Raglan by a regional airline. The ore is mined from underground and open pit operations, and contains significant quantities of copper, cobalt, and PGM's. The ore occurs in discrete lenses, varying in size from 250,000 tonnes to 1,000,000 tonnes. Characteristics of the ore can vary greatly. Originally designed to process 800,000 tonnes per year of high-grade nickel ore, plant capacity is currently at 1,000,000 tonnes per year.

1

2 3

Senior Metallurgical Engineer - Falconbridge Chile S.A. (formerly Raglan Concentrator Superintendent) Engineering Manager - Koniambo Project, Falconbridge (Australia) Pty Ltd (formerly Raglan Project Metallurgist) Strathcona Mill Senior Process Engineer - Falconbridge Sudbury Operations (formerly Raglan Mill Metallurgist)

2394

STUDY # 1 - GRINDING CIRCUIT Plowsheet Development Early bench scale tests revealed that Raglan ores were very hard, with typical work indices ranging between 18 and 23. At the same time, flotation tests revealed the need to grind to a PSOof 65 pm to achieve acceptable metal recovery. Both observations were a direct result of the finely disseminated occurrence of sulphide minerals throughout the hard ultramafic host rock peridotite. During a 1991 pilot plant program, the following objectives were set: 0

0

Design a circuit to produce a particle size of 80% passing 65 pm to ensure sufficient sulphide mineral liberation for flotation Design a circuit that provides high certainty of reliability and performance Design a circuit with efficient power utilization characteristics since power would be generated on-site Design a circuit that minimized operating costs, primarily by reducing the need for grinding steel

A number of design parameters were investigated during the campaign, including semi versus fully autogenous grinding, the impact of pebble crushing, the opening size on the primary mill discharge screen, and pebble versus ball milling for the secondary mill. For the primary mill, the main focus was on using autogenous grinding to eliminate the cost of shipping SAG mill grinding steel to the high arctic. Conceptually, the design would be typical of the circuits used for milling hard Taconite iron ores in the U.S. The main focus for the secondary mill was achieving the targeted product size while also coping with tonnage swings from the primary circuit. Ore samples for pilot plant grinding testing were taken from one underground zone, which was accessed in the predevelopment period. No single zone could be described as representative of all of the ore at Raglan, however this zone was accessible, and could be described as “average” according to the knowledge of the local mineralogy at that time. Lab testing was used to determine metallurgical differences between ore types. The main observations from the pilot plant were the following: The operating work index of the various ore samples was consistent with previous test data. Values ranged from 19 to 23 with an average value of 20 The net power consumption from the two stages of milling ranged from 23 to 28 kWWtonne The autogenous circuit was marginally less power efficient than the SAG circuit Pebble crushing was shown to be very important with this ore. Crushing pebbles increased throughput by approximately 35% and reduced power consumption by approximately 30% SAG milling with a 6% steel charge increased throughput by as much as 50% Secondary pebble milling consumed 7 kWWtonne less power on average at a pebble wear rate of 1.5%. Pebble milling, however, was less tolerant to variations on feed rate. Based upon the pilot plant test data, an autogenous mill was selected for the primary mill including a mill discharge screen and recycle pebble crusher. For future increases in throughput, the mill would be equipped with a variable speed drive and structurally designed to receive an 8% steel charge. A standard ball mill was selected for the secondary mill. The ball mill was equipped with an oversized 3000 hp motor that was standardised with the primary mill motor.

2395

Selected Flowsheet Description Ore from open pit and underground mining operations is introduced to the underground ore handling system via underground truck dumps. Two underground bins with a combined capacity of 4000 tonnes provide roughly a day of live capacity for ore storage. Ore from the two bins is conveyed to a 150 tonne surge bin immediately ahead of the grinding circuit in the concentrator. The Raglan grinding circuit has a conventional ABC configuration (Figure 1) with a design capacity of 100 mtph (800,000 tonnes per year). The primary 24 ft x 8.5 ft (EGL) autogenous mill is powered by a 3000 hp motor with variable speed drive. Primary mill discharge is sized using an 8 ft x 16 ft single deck vibrating screen, with screen oversize re-circulating to a 5% ft short head cone crusher (400 hp) before returning to the primary mill. Screen undersize is combined with secondary ball mill discharge to feed up to five 15 in. diameter Krebs hydrocyclones. The ball mill is a 14 ft x 21 ft overflow mill driven by a 3000 hp motor at 73% critical speed. Target P80 for the grinding circuit product was 65 pm.

Figure 1 : Grinding circuit flowsheet

2396

Current Circuit Operation The Raglan grinding circuit was commissioned in December 1997. Initial problems with low grinding circuit throughput were overcome by adding many more pebble ports in the AG mill discharge than were originally configured. The target throughput of 100 tph was achieved by the summer of 1998. Table 1 shows some of the key data used in the design of the circuit compared with the actual plant operating data. It can be seen that overall, the plant performance is quite similar to the design data. With the correct pebble port configuration, the range of plant throughput very closely matches design. And at greater than loo%, the crusher recirculating load is very high for AGISAG circuits, as provided for in the design. Most importantly, the circuit produces the design product P80 of 65pm, even at throughputs well over 100 tph. And while the AG mill design had provided for the option to convert to SAG milling if necessary, the AG circuit has demonstrated its costeffectiveness through very low liner wear and reasonable mechanical availability. While the grinding circuit has met its overall objectives, there are a few notable differences between design and actual plant performance that are worthwhile highlighting. Firstly, the overall specific power consumption is somewhat higher in the plant (31.5 kWWt vs. 27.2 kWWt design). This increased energy requirement is the result of inefficient grinding of the coarsest fraction of the circuit feed (+150 mm). Whereas the top-size tested in the pilot plant was 100 mm, the fullscale mill sees feeds with Fsoof 130 to 180 mm. The +lo0 mm rocks have a low grinding rate and tend to build-up in the mill, forming a critical size which limits throughput. The disproportionately-high energy requirement for the feed top-size was not evident at the smaller scale of the original pilot plant. The grinding rate of the coarse rocks is discussed further below under Critical Size and Pebble Crushing. The second important difference between pilot plant and full-scale results is the screen transfer size (T~o), which is rather coarser than design. In order to compensate for the slow grinding rate of the coarse particles, it became necessary in the plant to use a larger screen size than originally intended (9 mm vs. 3.4 mm design) to get the fines out of the AG mill. It was also necessary to run at a much lower density in the AG mill (-5560% vs. 75% design) to keep the mill flushed of fines as much as possible. Both of these operating parameters helped to reduce the mill charge level and increase the energy input to the coarse particles, but resulted in a coarser transfer size to the ball mill. A final key difference is the ball mill percent recirculating load, which is much lower than design (-200% vs. 300% design). This is interesting because the pilot plant very closely predicted the fines content of the AG mill screen undersize (AG mills typically produce a significant amount of fines). Table 1 shows the percent passing 37 pm in the cyclone OF, design and plant. With similar fines content in the feed as estimated in the design, the ball mill is clearly producing more fines than predicted in the piloting, resulting in a reduced recirculating load with the given cyclones.

Variation in Ore Hardness Net energy consumption for the AG mill can range from 17 to 26 kWWt. This variation in power requirement is indicative of the significant variation in ore hardness seen in the plant, and results in throughputs ranging from 80 to 120 tph. Short-term changes in ore hardness frequently cause rapid changes in AG mill charge level and recirculating load, forcing the operator to make quick tonnage changes to compensate.

2397

Table 1 . Design vs. Operating Data

The variation in ore hardness was quantified in 1999 through grinding circuit benchmarking and extensive drill core sampling for MinnovEx SPI testing (SAG Power Index). The MinnovEx SPI test is designed to measure ore hardness and infer power requirement under the prevailing ore breakage mechanisms of SAG milling, and is analogous to the Bond work index for ball milling2. Figure 2 shows 190 SPI test results in a plot of cumulative distribution of required energy for AG milling with crusher to produce a transfer size of 1,200 microns (standard correction factors used)3. The median energy value is 17.6 kWh/t corresponding to a median SPI time of 321 minutes, indicating that the ore is very hard relative to MinnovEx’s database of results from other grinding circuits2. Furthermore, 20% of the material has a value greater than 20.8 kWh/t, nearly a 20% increase in energy requirement, whereas the softer 20% of the feed requires less than 14.2 kWh/t. This range of ore hardness reflects the variation in plant throughput over time. It is noteworthy that 5% of the samples showed extremely high energy requirements. These samples correspond to waste fractions. The results point to the strong influence of ore dilution on the energy consumption in the grinding circuit, and are corroborated by circuit behaviour.

Critical Size and Pebble Crushing As mentioned earlier, the AG mill was observed to fill with large rocks (4 to 8 in.) to the point that throughput was limited. The presence of a very coarse critical size is illustrated using the grinding rate curves derived from JKSimMet modelling. The grinding rate curve represents the breakage rates required for each size fraction to satisfy a steady-state mass balance around the mill (perfect mixing mill model)2,and is derived through model fitting. Figure 3 depicts grinding rate (units of hr-’) as a function of particle size on log-log scales. The Raglan AG mill curve is shown in contrast with the default SAG grinding rate curve, which represents the average grinding rate distribution for the database on which the JK SAG model is based. The difference in shape of these two curves is telling. Whereas the default curve shows increasing grinding rates at particle sizes greater than about 25 mm, the Raglan AG mill curve shows decreasing rates above that size, with the lowest grinding rates in the coarsest size fractions. This feature of the Raglan AG mill model points to the coarse particles as representing a critical size for the mill which serves to limit overall circuit throughput.

2398

0

4

8

12

16 20 24 28 32 36 40 Primary Mill Energy to T80 = 1200 microns, kWhlt

44

48

Figure 2: Cumulative AG Mill Energy Distribution for 1,200 pm Transfer Size3

. ~. ..

0.1

1

10

100

1000

Particle Size (mm) [-Raglan

AG rate curve

-

-Default SAG rate curve [

Figure 3: Grinding Rate Distribution Curves

Increasing Grinding Circuit Throughput - 1 MTPY Project There was, and continues to be, strong economic pressure to push circuit throughput at Raglan, as there is at many milling operations. As the benchmarking and modelling data show, efforts to maximise throughput must focus on the breakage of the coarse particles, and must take into account the variation in ore hardness seen in the process.

2399

In 1999 and 2000, a number of process changes were made in an attempt to increase the rate of breakage of the coarse particles. The intent was to push the AG option to its limits before considering the more expensive conversion to SAG milling. The process changes included: 0

Higher mill speed to increase impact frequency in the mill Larger 4 in. x 5 in. pebble ports to enable discharge of some of the coarser rocks New liner configuration on the pebble crusher to allow tight closed side settings.

These changes had limited impact on circuit throughput. Not much effort has been put into AG mill lifter re-design. This might be an area of focus in the future, given the successes seen elsewhere with SAG operations. In the spring of 2000, two changes were made which together produced a significant increase in throughput. Pebble port size was increased to 5 in. x 5 in., replacing the 4 in. x 5 in. ports, and this allowed significantly larger rocks out of the mill. This change initially produced poor results, as the crusher was unable to adequately reduce the coarser particles. The effect was to build an excessive circulating load of fines and to deplete the mill of coarse media, thereby shifting the circuit to a much more inefficient mode of operation. A grizzly with 4 in. bar-spacing was then added in front of the pebble crusher to bypass some of the largest rocks directly back to the AG mill. The impact of this second change was dramatic, with crusher performance returning to normal and average plant throughput increasing from 115 tph to 130 tph. Effectively, the combination of very large pebble ports coupled with the recirculation of +4 in. material allowed the plant to achieve the 1 MTPY target. While the critical importance of effective pebble crushing in the AG circuit has been demonstrated, the reason for the significantly improved throughput with the large ports and grizzly combination is not obvious. The negative impact of recirculating uncrushed coarse material to the AG feed would seem to be outweighed by the higher rate of production of fines due to improved crusher operation. One possible explanation for this favorable trade-off is that with the bypass of large pebbles back to the mill, the charge is kept much coarser (i.e. medium-sized particles are removed and crushed while the big rocks go back). The presence of more grinding media may result in more efficient grinding of the fine fractions (-25 mm), thereby increasing the rate of production of minus screen size material. Another possibility is that mass transport through the mill has been affected. It has always been observed during inspection of the charge that coarse particles (plus pebble port size) build up in the mill charge next to the discharge grates. There is a very clear size gradient along the axis of the mill, with fines toward the feed end and coarse particles at the discharge. The presence of coarse material at the discharge may present an added restriction to the flow of medium-sized particles through to the grates and out of the mill. By removing more of the large rocks and sending them back to the feed end, this restriction may be diminished, allowing slightly faster discharge to the crusher and lowering the charge level. Though the specific cause of improved performance is unclear, it is evident that a more optimum balance between crushing and grinding has been struck with the addition of large ports and the bypass grizzly. STUDY #2 - CONCENTRATE DEWATERING, STORAGE AND TRANSPORTATION Flowsheet Development Under specific conditions, some sulphide concentrates that contain pyrrhotite are known to exhibit self-heating properties that can lead to a significant hazard for both production personnel and property. Not surprisingly, the majority of pyrrhotite-containing concentrate is therefore handled in slurry form to eliminate this risk. The high arctic location of Raglan eliminated slurry transport as a feasible option due to the prohibitive cost of shipping the water contained in concentrate, and

2400

also due to the problems associated with freezing. All transport options, therefore, needed to minimize the amount water contained in concentrate. Standard testing methods applied to Raglan concentrate samples obtained from the pilot plant program confirmed this self-heating behavior and helped to quantify its severity. The phenomenon of self-heating is discussed in detail in the literature, most recently by Rosenblum et al’. The result of self-heating is the release of heat, the formation of sulphates that bind the concentrate particles together to form solid agglomerates, the smell of sulphur dioxide fumes and the eventual physical transformation of the concentrate filter cake into a sintered mass. Extensive testing at Lakefield Research and the Noranda Technology Centre concluded that to avoid spontaneous heating, Raglan concentrate must be dried to a moisture level less than 0.3% and also cooled to a temperature less than 30°C following drying. The cooling requirement was discovered as a secondary conclusion of the work at Noranda Technology Centre. Using the information described above, several drying and material handling unit processes were examined. Four types of drying unit operations were investigated, most of which required some form of upstream filtration: Fluid bed dryer Flashdryer Steam coil (Myren) dryer Spray dryer. The fluid bed dryer was selected on the basis of proven operating history on sulphide concentrates, a tolerance to variations in feed moisture content in the event of upstream pressure filtration problems, and the ability to easily utilize the exhaust gas from the diesel generators located in the power station. The major disadvantage with this unit was the requirement for a 700 HP blower to boost the pressure of the diesel generator off-gas. The other dryer options were rejected for a variety of technical reasons. The flash dryer was of interest since the diesel generator off-gas could be used directly for drying without a boost in pressure. However, the unit lacked proven operating history and was known to be intolerant to variations in feed moisture content. The steam coil dryer also lacked proven operating history and was not suited to using the diesel generator exhaust in place of steam. The spray dryer was of interest since it combined filtration and drying into one operation. However, it required more heat for drying than was available from the diesel exhaust flow. The cost of installing a separate combustion chamber and burning additional arctic diesel fuel to generate dryer hot-gas more than offset the capital and operating cost of a filtration operation. Following drying, three dry concentrate handling alternatives were considered, as follows:

.

Bulk handling using conventional conveying and loading equipment, similar to many base metal operations Flexible bulk bags of 1 to 2 tonne capacity, similar to those used for matte and other high value products Pneumatic handling using equipment typical to what most producers of bulk cement use.

A simple qualitative comparison was made between the three alternatives. While the bulk handling option presented very low technical risk, the likelihood of exposure to rain and/or snow would surely trigger spontaneous heating. (i.e. high technical risk) In addition, high product loss and environmental contamination through the system would be expected due to the fine particle size of the bone-dry concentrate. (PBo= 20 pm). The use of flexible bags may have addressed the abovementioned problems, however operating costs were expected to be high as a result of transferring

240 1

13,000 bags to the cargo ship, five to six times during a nine month shipping season. The final option, pneumatic handling, was selected for Raglan based on cement industry experience with clean storage and high rate loading of dry, fine bulk cement products, often very near highly populated areas.

Selected Flowsheet Description Concentrate is thickened in a 13-m diameter Supaflo high-rate thickener, and pumped via peristaltic pumps to a filter feed tank. Two Svedala VPA 1530 filters treat the slurry, from which the filter cake is conveyed at 10 to 12% moisture to a dryer feed bin using twin screw-type feeders and a rotary air lock. The 2.4 m diameter by 7 m high Fuller fluid bed dryer is supplied with hot exhaust gas from the power plant generators via a 700 HP blower. The fluid bed media is currently generated from the mill grinding circuit. The inlet gas, tempered by ambient outside air, varies according to the feed rate to a maximum of 330 “C. The outlet temperature is maintained between 100 to 110 “C. Outlet gas and dry concentrate pass through a vertical duct to two parallel pulse jet dust collectors. Clean gas is expelled to atmosphere via a fan and stack, while hot concentrate passes along air slides to a cooler. The cooler consists of a 1.8-m by 1.6-m column, 10-m high, with an extensive network of cooling pipes through which cold water passes. Cool concentrate is discharged into an airlift, which transports the material to a distribution air slide feeding any of three storage silos, with a combined capacity of 4,000 tonnes. The silos discharge into air slides which feed 50-tonne tanker trucks. Trucks travel 96 km to Deception Bay, and are pneumatically discharged into a 50-m diameter storage dome, with 50,000-tonne capacity. A mechanical arm in the dome reclaims concentrate to a central bottom discharge point followed by conveying to pneumatic pods. The pods transfer the concentrate into the MV Arctic for ocean voyage to Quebec City. The ship transports 27,000 tonnes per voyage, five or six times per year, with a cessation of shipping between March 1 and June 1 due to local agreements. Current Circuit Operation The concentrate thickener is operating without any significant problems. There is a persistent froth accumulation on the surface, however a downstream clarifier on the overflow stream has made this problem less important. The concentrate pressure filters are operating at satisfactory levels of availability, throughput and product quality. A number of minor mechanical and programming modifications have been implemented in order to facilitate operation and maintenance. As with any filter system, proper maintenance is critical to preserve performance of the filters. In order to increase plant throughput and peak filter capacity, the concentrate filters were increased from 8 to 12 chambers, with the relatively simple addition of more plates. It is highly recommended that this flexibility be preserved when sizing filters, especially for new plants. There are several reasons for this: 0

0

The plant throughput will inevitably increase within a few years of start-up. Potential for added chambers means potential for “instant” additional capacity. Flowsheet modifications such as additional regrinding may change the size distribution of the concentrate, increasing the filtration requirements. The ore used in pilot plants andor filter selection testing is often not completely representative. An expandable filter is prudent.

The provision to add more plates means having a slightly longer frame, and a few other provisions. The additional cost should not be great, provided that the filter length is reasonable.

2402

Filter cloth selection can play a significant role in the filter performance, as well as cost. At Raglan, the use of a monofilament fabric has been very successful in the concentrate filtering application. However, the same fabric has not yet proven completely successful in the tailings application. A conservative well-proven filter cloth fabric should be used when sizing filters and selecting initial cloths. The client should insist that the filter supplier not use “high-tech” cloths in order to obtain optimistic filtering rates and hence smaller filter areas. Expanding an existing plant with proven experience with such cloths might be an exception to this recommendation. A solid working relationship with a cloth supplier was very important during the first few years of operation at Raglan. A fluid-bed dryer is not a piece of equipment often installed in concentrators, and therefore there was a fairly steep learning curve. Early in the operation, there was also a catastrophic failure of the fluidizing fan, leading to massive equipment damage and some production losses. This event reinforces the need to conduct independent design and fabrication audits on equipment which are of unique design or application, and where the results of failure can be severe. The material handling equipment feeding the filter cake had some difficulties with abrasion and concentrate build-up, despite careful design. A number of modifications were required in this area in order for the dryer sector to operate at full capacity. The dryer itself has undergone only one minor modification. Being shorter than ideal due to building constraints, there is an increased probability of media being picked up by high velocity gas localised at the dryer exit. Therefore, a deflector was installed inside the dryer to reduce the carry-over of bed media into the exhaust duct. The media size was also increased slightly to compensate for this problem. The dryer capacity meets current needs, however, there are opportunities for further increases, if required. One problem that persisted over the first year and a half of operations was the occurrence of concentrate fires. Principally occurring in the dust collectors, the fires were a result of exposing concentrate to the conditions determined in laboratory self-heating test work to be dangerous. The conditions include warm temperatures, some small amount of moisture, and a supply of oxygen. When the dust collector exit became blocked, concentrate would accumulate, containing a small amount of moisture, and the exhaust gadfresh air mixture supplied oxygen and heat. In a surprisingly short period of time, the concentrate would start to oxidize. In order to eliminate the problem, the discharge air slides were modified to have a steeper slope and larger exit, allowing for more efficient transport of concentrate from the dust collectors. This has almost eliminated the potential for build-up of concentrate and greatly reduced the risk of fire. More sensitive level alarms were installed which immediately alert the operator of any build-up. A number of other upgrades have been incorporated in the two main dryer dust collectors. The original inlet design created excessive turbulence, leading to excess dust loading in the air surrounding the bags. This led to pulses in the dust collectors being too frequent, causing excessive bag replacement and limiting the dust collector capacity. A new inlet design was modelled to determine velocity vectors and was able to greatly reduce inlet velocity and turbulence, hence increasing bag life and capacity. The tube sheets in the dust collectors were also changed, partly due to fire damage, and partly to provide a different style of bag that was easier to install and remained in-place more securely. All of the air slides transporting concentrate were also replaced with high-top units, allowing less dust capture, and reducing subsequent plugging of the air slide dust collection system. Plugging of this dust collection system has also led to poor flow conditions in the air slides. Some additional modifications are necessary to reduce dusting while trucks are being loaded.

2403

STUDY #3 - TAILING DEWATERING AND DEPOSITION Piowsheet Development Site conditions under which Raglan tailing are discharged generally are as follows:

0

0 0

Below freezing temperatures for roughly nine months Permanently frozen ground (permafrost) with a temperature of -7 “C An active surface layer of broken rock approximately 1 to 2 metres thick Average wind speeds of 30 km/h from the northwest Virtually no vegetation, and little topographical relief Limited aggregate supply for dam construction Annual precipitation of 650 mm (net) and no nearby lakes.

The challenge was to design an environmentally safe and cost effective tailing system taking into account a number of regulatory trends, such as:

0

Zero effluent discharge Perpetual monitoring Submission of closure plan and posting of insurance bond prior to start-up Requirements for progressive restoration during the operating life of the mine.

The above conditions and challenges resulted in a decision to evaluate tailing discharge design “first principles”. After considering many options, three were selected for more comprehensive study, each being characterized by the pulp density of the watedsolid mixture. Each option was required to safely contain 680,000 tonnes of solids annually for a period of 25 years. The end result of the study was the selection, design and commissioning of a tailing system that worked “hand-in-hand” with the natural permafrost ground conditions of the Arctic. Conventional Option: 55% Solids At 55% solids, the Raglan tailing is typical of most slurry streams from a mineral processing plant. Well proven, simple technology could be used to thicken, pump and contain the slurry. Operating costs could easily be benchmarked and were relatively low. However, due to the characteristic flat terrain and limited supply of construction aggregates at this site, the capital cost to construct tailing dam walls was estimated to be very high. This option would also result in a large tailing site “footprint” and a significant recycle water management system. Advantages and disadvantages of this option are summarised in Table 2.

Advantage

Disadvantage

Proven, simple technology

Recycle water management in extreme arctic conditions Large surface area to manage environmentally No potential for continuous reclamation Negative visual impact on landscape Only suitable site > 5 km from the concentrator High capital cost - $45/annual tonne installed capacity High closure cost - $12/annual tonne installed capacity

Known operatiodmaintenance Low operating cost - $1.34/tonne ore

2404

A net present value cost estimate was prepared, taking into account the initial capital outlay, annual operating costs and post-production phase closure costs. A value of Cdn%53million (1 992 dollars) was estimated for this base option. It was concluded that the technically riskier higher density options should be thoroughly researched in an attempt to mitigate the problems of water management, large open surface area, and high capital plus closure cost.

Paste Option: 70% Solids When the pulp density of the Raglan tailings reaches 70% solids, a typical thixotropic paste is formed. Conceptually, the idea involved utilizing equipment and operating techniques employed by aluminum producers that dewater and contain “red mud” as their tailing. High density deep bed thickeners followed by positive displacement pumps would be used to thicken and transport the tailing paste to a disposal area. At the disposal area, the tailing paste would be placed in layers that gradually would freeze into permafrost. The greatest perceived advantage of this approach was the positive impact on the size of the tailing containment facility. With only 70% of the original slurry volume in the form of a paste, the size and cost of a containment facility would be significantly reduced vs. a conventional design. Also, the high potential for the majority of the tailing water to freeze in-situ would result in minimal recycle and a much simpler, smaller, lower cost water management system. Against these advantages was the overwhelming concern of using thickening technology that was unproven for a sulphide tailing together with difficult placement logistics in an arctic situation where flexibility during upsets is vital to achieving target production rates. Added to this was the unknown long-term behaviour of residual water and percentage of “ice lensing” within the tailings. Laboratory testwork predicted a stable material. However, extrapolation of results to “real world” was considered risky in this situation. Advantages and disadvantages for this option are summarised in Table 3. Table 3- High Density Option

1 Advantage I Disadvantage I I Simplified, lower cost recycle water management I Less proven, more complex technology I system I

Potential for continuous reclamation Less flexibility in upset conditions Less known operatiodmaintenance Smaller surface area to manage environmentally Higher operating cost - $1.75/tonne ore Suitable site < 1 km from the concentrator Lower capital cost - $2 1/annual tonne installed capacity Lower closure cost - $8/annual tonne installed I The net present value cost calculation for this option was Cdn$31 million (1992 dollars). Significantly lower estimated capital and closure costs more than offset the increase in estimated annual operating cost. However, due a much higher perceived technology risk, a further option was considered that would hopefully capture the advantages of these first two options.

Filtering Option: 85% Solids At 85% solids, the Raglan tailings moves from an unstable paste state to a stable solid state in the form of a filter cake. The concept involved the use of pressure filters to produce a cake that could be easily conveyed, stored in silos, transferred to trucks and subsequently moved to a containment area for stockpiling using standard mobile equipment. Many of the previously stated advantages would be achieved but at the expense of higher long term operating costs to maintain pressure filters and to handle tailing filter cake with mobile equipment. Project management decided to accept this higher cost in an effort to minimize technology risk, maximize operational flexibility at a harsh arctic site, maximize environmental protection by utilizing continuous reclamation

2405

practices, minimize water recycle and snowmelt run-off problems and minimize both the initial capital and final closure cost. Advantages and disadvantages for this option are summarised in Table 4. Table 4- Filtered Option

I

1

Advantage

Disadvantage

I

The net present value cost estimate for this option was CdnMO million (1992 dollars), representing a value midway between the previous two options. Selected Flowsheet Description Tailings are thickened in a 30.5 m diameter Supaflo High Rate thickener, and pumped via SRL pumps to a filter feed tank. Three Svedala VPA 2040 filters treat the slurry, with completely independent operation. Three Elliot 1000 HP centrifugal compressors are installed to provide air to the filters and other plant requirements. Filter cake is discharged via individual belt feeders onto a single common conveyor discharging into a 150 tonne silo. The silo is equipped with an airactuated clam gate, operated to fill two Volvo 35 ton articulated trucks. Tailings filter cake is hauled 3 km one-way to the disposal site, where it is dumped, spread, and compacted. Haulage is based on a 24-hour operation, and spreading and compacting is done on dayshift only. Run-off water from the tailings pad is collected in an excavated pond and pumped into the mill recycle water system. Filter operation Operation of the tailings filters has been successful overall, producing good quality filter cake. As with any filter, a rigorous maintenance program is essential in order to maintain availability. Some modifications to controls and mechanical elements have been implemented. The availability of membrane air at the proper pressure and filter cloth condition are both very important factors in the filter performance. The filters have been expanded from the original design of 42 chambers to 46 in order to accommodate increased throughput. In retrospect, it would have been desirable to include the provision for a few additional chambers beyond 46. The tailings filters are occasionally subject to significant swings in performance with different ore types. Some of the more highly altered ore zones have very poor filtering characteristics, sometimes to the point of slowing down plant throughput. These types of problems are difficult to foresee in pilot the preproduction testing and design stages. Constant efforts are made to reduce cycle times for the filters, in order to increase the peak capacity, and lower overall filter utilization. The centrifugal compressors have performed satisfactorily, and their operation has been optimized significantly in order to realize substantial power savings.

2406

Tailing Placement Operation The most important issue related to the tailings operation is that of freezing. So far the tailings have frozen very successfully, and seasonal thawing is limited to minor depths of uncovered tailings. This is despite an accumulation of salts in the process water due to the use of brine for drilling underground.

During the first several years’ tailings deposition, there were a number of operational issues to be resolved. Rainfall patterns were not entirely as expected, with occasional heavy rain for extended periods, particularly in late summer. During these periods, it was almost impossible to drive on the tailings to dump loads. As long as bare terrain within the tailings impoundment area was available, dumping could be done. This option was not possible as the occupied area expanded, and a system of ramps constructed from waste rock had to be prepared in advance of the rainy season. The tailings stack is constantly growing higher, therefore the ramps need to be replaced regularly. Due to freezing conditions during most of the year, large quantities of tailings filter cake built up in the truck boxes, affecting net payloads. Heating the boxes with exhaust did not prevent the problem. Eventually, bolted box liners of PTFE were found to be the best solution. The cold temperatures also mean that tailings freeze fairly quickly after dumping, and therefore the piles must be bulldozed and compacted before this happens. Accumulated snow must be cleared in before applying tailings to an area, adding to the work during the nine or ten months of winter. Another issue relates to dusting of the tails. Dust from the pad is a winter issue, due to sublimation of frozen water in the tailings, essentially desiccating the tailings. Extremely low temperature and low air humidity are experienced during winter at Raglan. If the tailings are compacted with a roller compactor before the water freezes or is sublimated, the dusting is reduced greatly, but not eliminated. Several methods of dust control have been attempted, with varying degrees of success. Water spray is not effective due to rapid sublimation of the ice cover. Trials with a chemical binder have not been effective either. Snow is an effective dust suppressant, however maintaining a snow cover on the pad has proven to be difficult due to persistent and sometimes extreme wind conditions on the exposed top surface. Snow fences were tested, but could not withstand the wind. Trucked and compacted snow remains in place much longer. The most effective solution at this time is a thin cover of gravel, which is about to be applied in “dormant” areas, minimising exposed surface areas while leaving the “active” working surface for tailings deposition. Use of a roller compactor in active working areas is the best defence against dust.

Costs of the tailings disposal method are higher than expected. A summary of costs (from 2000) is presented in Table 5 , based on one million tomes per year operation. Even with some improvements in productivity, vehicle selection, and specific methodology at the pad, the costs are likely to remain high. The conventional disposal of tailings in an abandoned asbestos open pit mine was examined briefly in 2000/2001, however high capital costs and higher than expected operating costs were determined in a preliminary study, and the idea was abandoned.

2407

Table 5- Filtered Tailings Costs OperatinglMaintenance Costs I $/yr Filtering (includes feed system, filters, compressor 1 1,800,000 allowance, and delivery to bin) 1,700,000 Trucking/spreading/compacting(includes costs at load out) Capital Costs Ditching and cover cost Mobile fleet replacement

$/yr 2,000,000 600,000

I $/tonne(ore) 1 1.80 1.70 $/tonne(ore) 2.00 0.60

In the summer of 1999, after 18 months of operation in a demonstration mode, a project was initiated to complete the final detailed design for the water containment ditches and cover, which is based on the “walk-away” approach. AGRA (Amec) was selected from several consultants to perform the engineering. Several other firms are involved for their particular expertise in geochemistry, thermal modelling, and civil engineering. Starting with the original preliminary designs, and with the practical operating experience gained, the designs are now completed and the first phases of construction have begun. Currently a progressive reclaim approach is proposed, and therefore the construction will be completed in phases. This will allow for evaluation of technical and economic factors in the design at each stage of the project. Now that the detailed design of the cover is complete, it is clear that the cover costs are higher than originally thought. It is not the intention of this paper to discuss in detail the design and construction, however a publication focussing on the tailings may be made in the future. OVERALL CONCLUSIONS AND LESSONS LEARNED

The grinding, concentrate dewatering, and tailings disposal sectors of the Raglan concentrator have all performed at or above the original design throughput. The plant is currently operating at 1,000,000 tonnes per year, compared with an original target of 800,000 tonnes per year. The flotation circuit, while not discussed in this paper, is also producing recoveries better than in the plant design. Therefore, while the Raglan concentrator must be judged a success overall, there are some lessons to be learned. The lessons learned from the grinding circuit design are not dramatic, but they are important. While the circuit has met its objectives, the extreme variability of the ore hardness and its impact on throughput was perhaps not fully appreciated. Better tools are widely accepted today that perhaps allow less reliance on pilot plant data as the key design factor. These tools include the SPI approach developed by Starkey and MinnovEx, and simulators such as JKSimMet. While the pilot plant is certainly a key part of any greenfield project, one must always ask whether or not the feed stock for the pilot plant is representative enough to yield a sufficiently robust grinding circuit design. The answer in most cases will probably be no, and therefore the best available tools must be employed to look at the variations, and account for them in the design. The concentrate dewatering circuit design has demonstrated that the concerns and precautions about concentrate self-heating were indeed well founded. Concentrate oxidation does take place under conditions more or less predicted with the lab testing. Carefil design of the process can avoid those conditions and prevent fires on a consistent basis. While a number of the details were not exactly right initially, a good understanding of the situation helped to remedy the problem by identifying the required design modifications.

2408

With respect to tailings disposal, the main lessons learned are essentially about cost. The day-today operating cost of the disposal system is somewhat higher than predicted, and the cover costs are very much higher. With respect to the cover cost, both the quantities and unit construction rates were underestimated. Quantities would have been better predicted by completing a more detailed design of the cover much earlier. These kinds of details are most often determined later in the operating life of a mine. At Raglan, the progressive reclamation aspect of the tailings deposition plan meant that a detailed cover design was required very early in the mine life. If the design had been detailed in parallel with the plant design, then this cost could have been more accurately forecast and used in the decision making process. Having said this, it is quite likely that the choice of design would have remained the same, due to the other factors involved in the decision. There are a couple of other aspects peculiar to Raglan, which are not discussed in the previous sections, but are worth mentioning briefly. The first of these is layout. There probably does not exist an operator who thinks his plant is big enough. In any plant design, there is a struggle between efficiency (i.e. keep it smaller and cheaper) and spaciousness. In the case of a modular plant, the drive for efficient use of space is very strong. After a few years of operation at Raglan, it is probably fair to say that the layout is perhaps too efficient, to the point where maintenance of equipment is more time-consuming and costly. In a few areas, such as the dryer module, building restrictions were such that equipment design was compromised, which has led to costly modifications. There is no easy solution to this eternal struggle, except for a constant quest for balance between the two needs. The second aspect was the heat recovery system associated with the power plant. The diesel generators at Raglan were designed to be extremely energy efficient. Waste heat is captured from the engine cooling systems and hot exhaust gas, and is then transferred to a glycol system, which provides most of the heat for the buildings, and also provides all the hot water. In addition, the exhaust gas is used to dry the concentrate, saving enormous amounts of energy. All of these systems have been very successful, due to a very careful design.

ACKNOWLEDGEMENTS The authors would like to acknowledge all of the people who have contributed to the development and operation of the Raglan concentrator over the past 10 years. Their tremendous efforts and enthusiasm has resulted in a very successful operation in a very inhospitable part of the world. Finally, Falconbridge Limited and SMRQ are thanked for supporting this work and allowing it to be published.

2409

REFERENCES 1.

Hyma, D. and Williams, S., “Falconbridge’s Raglan Proiect: A development Update and Description of the Concentrator Circuit Design”, Proceedings of the 1993 Canadian Mineral Processors Conference, Ottawa, Ontario, Canada.

2.

Starkey, J. and Dobby, G.S., Application of the MinnovEx SAG Power Index at Five Canadian SAG Plants; International Autogenous and Semiautogenous Grinding Technology, Volume 1 of 3, pp.345-360, 1996.

3. MinnovEx Technologies Inc., “SPI Testing and Grinding Circuit Expansion Study: Preliminary Report”; Report to SMRQ, 1999. 4.

Napier-Munn, T.J., Morrell, S., Morrison, R.D., and Kojovic, T., Mineral Comminution Circuits. Their ODeration and ODtimisation, Indooroopilly:JKMRC, 1999.

5.

Rosenblum, F., Nesset, J., and Spira, P., “Evaluation and Control of Self-Heating in Sulphide Concentrates”, Proceedings of the 200 1 Canadian Mineral Processors Conference, Ottawa, Ontario, Canada

2410

The Formal Basis of Design Design Criteria: The Formal Basis of Design J. W.Scott ..............................................................................................................................

3

Design Criteria: The Formal Basis of Design John W. Scott, P. Eng., M I M E , MCIMM’

ABSTRACT Engineering design for a mineral processing plant is formally based on the design criteria developed for the project. The design criteria document evolves through preliminary studies and conceptual designs, and provides the formal documented source of design information, the basis of any assumptions and the governing codes and standards for each discipline. The assembly, content, format and presentation of typical design criteria documents for a mineral processing plant design at both the conceptual and detailed design stage are reviewed and discussed. INTRODUCTION The design criteria for a mineral processing plant will provide the formal basis for design of the process, equipment and facilities. These criteria will specify the life of the mine, annual throughput, design capacities and operating schedules for the equipment, as well as the quality of the feed to be processed and products obtained including tailings. The general climatic and geographical conditions at the site, specific information on soil or rock conditions, and applicable design standards and codes will also be included. The design criteria for a mineral processing plant design are assembled from a variety.of sources, including historical information, site specific soil and geotechnical data, metallurgical testwork on representative samples, pilot plant results, codes and standards and qualified assumptions. The key criteria are normally the process design criteria, as the plants we are concerned with are “Mineral Processing Plants”. In this paper, the major emphasis is on process design criteria, however each engineering design discipline also has a formal document or specification for design criteria. The process design criteria will be based on an interpretation of testwork carried out on the particular ore and site, and will thus parallel this testwork in detail and completeness. As the project moves from the early conceptual phases through to final detailed design, the design criteria will be developed and become more detailed as information is generated and made available for use. The design criteria provide the formal specification to the designer of what and how much is to be processed, what and how much is to be produced, and where and under what conditions. The design criteria document will be a formal tabulation of the design basis information developed for the project in question. For a preliminary study, this may consist of a single page of data; for a feasibility study several pages and for a detailed design up to several hundred pages and multiple volumes. Depending on the complexity of the process and the level of detail included, each project is unique to some degree. The design process for a project progresses from preliminary conceptual phases through to a final detailed design. The criteria used for the initial concepts are generally developed fkom rough data and preliminary testwork on the ore in question. The ore reserve or resource base, notional mine life and products are defined, thus giving a scope and definition to the study phase. In most studies a range of throughputs are analyzed in order to arrive at the most economically attractive size for the proposed operation. This optimum size is then used as the design throughput for the more detailed design phases. Similarly as more testwork results are available and the ore characteristics and process become better defined, a continuous updating of the design criteria is undertaken. I Director Technology, Fluor Daniel Wright Ltd, Vancouver, B. C., Canada

3

At the completion of the preliminary design phase or basic engineering, sufficient information will be available to prepare a formal design criteria document which will form the basis for detailed design work. It is important that at each phase of the development of the project when cost estimates or studies are prepared, the design criteria used to form the basis of design and estimation for this phase are carefully and fully documented. The format of the document produced is usually a standard of the engineering or mining company preparing the design, a key element is the attribution or source of information used, and adequate reference to source documents. The design criteria are used throughout all stages of a design as the primary written source of information for the engineers carrying out the design work, and the source document for further work.

DEVELOPMENT & USE OF DESIGN CRITERIA Conceptual Designs (Scoping Studies) At the outset of the conceptual design phase for a mineral processing plant there will normally be little information available. The first task of the conceptual design group is to gather what information is available and then develop test programs to provide the essentia1,data that are missing. During this phase, the metallurgical response of the ore is most important as it will define the flowsheet and equipment requirements. The ore will be characterized mineralogically and a bench scale test program will be developed and carried out on the available samples (normally split drill cores) to provide basic information for flowsheet development. If the ore is widely variable, the flowsheet will be designed accordingly and a further more-detailed metallurgical test program planned. At this stage the metallurgical testwork will be exclusively carried out on small composite samples representative of particular ore zones or rock types, or in some cases, based on production schedules. The basic metallurgical testwork such as crushing and grinding, work index determination, batch flotation tests, cyanidation bottle rolls, magnetic separation, thickening and filtration tests on products, etc., is carried out to confirm the metallurgical response, identify possible problem areas, and provide a planning basis for more detailed studies. The design criteria for the non-process areas of the design (civil, structural, mechanical, architectural, electrical, etc.) will be mainly generic at this stage with some obvious exceptions, such as seismic conditions, extreme climate, altitude, and environmental constraints. Design Criteria During Conceptual Stage. In the initial conceptual stage, the general criteria with respect to plant capacity, ore characteristics and preliminary metallurgical results are used to develop a flowsheet and preliminary plant layout. The metallurgical and mechanical engineer will rely on the criteria to prepare the flowsheet and size and select the major process equipment. Based on this work and any site constraints a plant layout will be developed to form a basis for cost estimation. The areas where the criteria are poorly defined or missing will be covered by past experience at this stage, and testwork or site investigations planned to obtain the missing information. Table 1 outlines design criteria. Assumptions made will be formalized in a short summary and included in the study report. Table Conceptual Design - Outline of Design Criteria 1. General Criteria Metallurgical Balance 2. Operating Schedule and Throughput 3. Process Criteria 4. Assumptions 5. Recommended Testwork 6.

4

Feasibility Studies A further feasibility study will be undertaken after an order of magnitude or scoping study and preliminary estimate has indicated that the project (or expansion or renovation) is financially attractive enough to merit further work. During the preliminary study scoping tests would have been completed on the metallurgical response of the ore and a generally suitable flowsheet selected. Any problem areas requiring resolution with further testwork would have been identified and suitable laboratory and pilot plant programs initiated. The pilot work in particular, would be aimed at verifying the flowsheet and also providing an opportunity for thickening, filtration, drying and other ancillary tests on samples produced from continuous operation on representative bulk samples. The effects of recycle water streams on the metallurgy will have been noted, with provision made for water treatment or possibly even a once-through system. For physical separations, unwanted slimes buildups, the need for settling ponds or unusual drying requirements, etc., will have been noted. Unusually abrasive or slimy characteristics will have been observed and noted. The small flows typical in pilot plants exaggerate the rheological problems encountered in certain ores and these also will be studied at this phase; particularly for extremely fine grinds and high pulp densities - characteristics perhaps more common in present day ores. The feasibility study will have considered any alternative processing schemes in depth and rejected all except the most logical and profitable via a series of technical-economic studies. Once again, the criteria available will be checked with the selected flowsheet and any deficiencies covered by further testwork. If the ore is difficult to treat, involves a complex or innovative metallurgical process or new technology, then a continuous pilot plant is mandatory. For grinding circuit design, pilot scale testing is still regarded as the most reliable method of selecting a grinding flowsheet and generating design criteria for equipment sizing and selection. Careful sample selection and close coordination with geology, mine planning and production scheduling to ensure representative bulk and composite samples is required. The plant site will now have been selected and the buildings and facilities laid out. The site will have been drilled for geotechnical information and carefully surveyed to provide surface contours and an estimate of cut and fill quantities. An effort will have been made to find local materials suitable for compacted fill, and for concrete aggregate in the case of remote locations. The bearing strength of the soil and underlying bedrock will be tested and criteria for foundation design prepared. Architecturally the climatic conditions will be carefully noted, and a program to develop wind and snow loadings undertaken if necessary. Otherwise IocaVregional records will be used. Prevailing winds and severe climatic conditions will be considered to ensure the best orientation of buildings and openings in the buildings, subject to site constraints. The power supply will have been determined, with a preliminary supply contract providing the characteristics of the supply negotiated. Power distribution within the plant and plant site will also be defined. Water supply and tailings disposal will have been carefully studied, with hydrological and geotechnical studies commissioned as required. It should be noted that for most mineral processing plants, tailings comprise more than 80% of the ore, and in many cases more than 95%. The design of the tailings impoundment, and the stability of the tailings dam structure whether constructed from engineered materials or cycloned sands, wilI depend on information obtained from the geotechnical program. Rainfall and evaporation rates are critical to the overall water balance and need to be quantified from historical records, or during the environmental baseline study. For the detailed feasibility study phase of a project, the general operating and metallurgical design criteria will have been refined and augmented by further laboratory and pilot plant testing. At this stage the plant site will have been carefully surveyed and soils investigations carried out to provide site specific design criteria to the structural and civil engineering groups. From information collected at this point, an expanded “Design Criteria” section, which will be included in the feasibility report is outlined in Table 2.

5

Table 2 Feasibility Study - Outline of Design Criteria Section 1 1.1 Introduction 1.2 General Criteria 1.3 Operating Schedule and Capacity 1.4 Metallurgical Balance Section 2. 2.1 Process Description 2.2 Equipment Description FlotatiodMagnetic Separation, etc. Thickening and Filtering Drying and packing 2.3 Facilities Description Buildings Services Site and access 2.4 Required testwork

Based on the design criteria and the selected flowsheet, the mechanical equipment will have been specified and quotations solicited from suppliers to obtain sufficiently accurate cost data for the report. The plant layout will have been refined and developed to provide a good basis for civil and structural takeoffs and preliminary design work. The power supply and electrical requirements will be defined by the equipment selection, and yard and building lighting, heating requirements, etc. The architectural finishes will be selected on the basis of the duty, weather and plant areas. The criteria used for this selection will be based on experience in similar installations, specified codes and any specific chemical resistance required. A typical format for the process design criteria document prepared for a bankable feasibility study is provided in appendix 1A; Appendix 1B has a table of contents and format for a document for a master structural design criteria. Basic Engineering and Detailed Design Following a positive feasibility study, the next phase of a project involves basic engineering, followed by detailed design. The basic engineering for a project involves developing the feasibility study to a point where the major equipment has been tendered, a supplier recommended and design advanced to a point where detailed engineering can commence. The end product is the detailed design criteria document, serving as a specification to the detailing engineer and a basis for the project budget estimate. At the commencement of the basic engineering, a draft design criteria document will be prepared, based on a format used by the particular engineering company doing the design work. The preparation of this draft will normally bring to light a number of areas where specific design criteria are missing, so that further testwork can be carried out or the necessary information obtained. As the engineering progresses and major pieces of equipment are committed to purchase, actual weights and dimensions of the equipment in question will be available for inclusion in the appropriate sections of the criteria. At the same time, the flowsheets and general arrangement drawings will be updated to show the recommended equipment. The mechanical, structural and civil groups will work very closely with each other during this period to ensure that the basic criteria are available for such problem areas in the design as large rotating equipment or severe vibration. Detailed engineering will be critically dependent on design criteria, and in fact the first phase of detailed design or basic engineering is carried out to provide a detailed design criteria to be used as the basis for final detailing. During the basic engineering the major and long lead equipment is tendered and will be selected, and the flowsheet, equipment list and general arrangement drawings are updated to near final form. The design criteria developed for the detailed feasibility study will also be updated, and any further information included as it is made available. At this stage, actual equipment loads will

6

be available so structural and foundation design may be advanced. Normally, the first detailed process and instrumentation diagrams are developed and will form the basis for piping design in conjunction with the flowsheet. Final selection of such equipment as filters, dryers and thickeners will enable a detailed water balance to be prepared, providing the flows of flesh and reclaim water as the design criteria for the water system. The heating and ventilation requirements will have been defined by the climatic conditions and by any local codes and ordinances. Similarly codes governing noise levels and personnel protection will be incorporated into the criteria to ensure compliance. As the basic engineering proceeds, areas will be identified which require further specific tests for design to proceed. These would be carried out and the criteria updated in those areas. The final document produced at this stage is known by various names: design criteria, design basis memorandum, etc. However it will consist of a formal document including all design criteria used, a description of the process and the set of basic engineering drawings. These documents then form the basis for the detailed engineering design. A generalized table of contents for such a document is given in Table 3. Excerpts from typical design criteria documents for the structural discipline is shown in Appendix 1B. Other discipline design criteria are similar in format and level of detail. Table 3 Index for a Design Criteria I. General Criteria 1.O Introduction 2.0 Scope of Work 3.0 Criteria Summary 4.0 Mechanical Design Criteria 4.1 Conveyors 4.2 Chutes 4.3 Fabricated Items 4.4 Process Pumps 4.5 Piping 4.6 Linings 4.7 Codes and Standards 5.0 Electrical Design Criteria 5.1 General 5.2 Codes and Standards 5.3 Power Supply System 5.4 Main Substation 5.5 System Voltages 5.6 Emergency Power 5.7 Metering 5.8 Distribution Switchgear 5.9 600 Volt Power 5.1OMotors 5.1 1Power Factor Correction 5.12Motor Controllers 5.13Controls & Pilot Devices 5.14Lightning and Surge Protection 5.15Grounding 5.160verhead Transmission Line 5.17Lightning and Receptacles 5.18Wiring Methods 5.19Wire and Cable 5.20Conduit 5.21Enclosures 5.22Cable Tray

7

Table 3 Index for a Design Criteria 5.23Special Equipment 5.24Communications 6.0 Structural Design Criteria 6.1 General 6.2 Design Loads 6.3 Unit Stresses and Limiting Deflections 6.4 Load Factors 6.5 Load Combinations 7.0 Mechanical Services Criteria 7.1 Design and Drafting Standards 7.2 Materials 7.3 Design Conditions 7.4 HVAC 7.5 Building System 7.6 Plumbing 7.7 Fire Protection 8.0 Air Pollution Control Criteria 8.1 Codes and Standards 8.2 Materials 8.3 Design 9.0 Architectural Design Criteria 9.1 General 9.2 Washrooms 9.3 Change Rooms 9.4 Lunch Rooms 9.5 Offices and Labs 9.6 Building Enclosures 10.OInstrumentationDesign Criteria 10.I Purpose 10.2References 10.3General 10.4Instruments 10.5Control Panels 11. Process Area Criteria 11.OOperating Schedule and Capacity 11.1General 1 1.2Schedules and capacity 12.0Process Metallurgy 12.1Ore Characteristics 12.2Metallurgical Balance 12.3Material Balance 13.O Water Supply 13.1Operating Criteria 13.2Process Description 13.3Equipment 14.0Tailing Disposal 14.1Operating Criteria 14.2Process Description 14.3Equipment 15.OAncillary Buildings 15.1Offices 15.2Change House

8

Table 3 Index for a Design Criteria 15.3Laboratory 15.4Shop & Warehouse 16.0Conditioningand Flotation 16.1Operating Criteria 16.2Process Description 16.3Equipment 17.0Reagents 17.1Operating Data Summary 17.2Reagent Preparation and Equipment Description 17.3Reagent Fume and Dust Control 18.0Hydrosizing & Tabling 18.1Operating Criteria 18.2Process Description 18.3Equipment 19.0Filtration and Drying 19.1Operating Criteria 19.2Process Description 19.3Equipment 20.0Concentrate Packing & Storage 20.1 Operating Criteria 20.2Process Description 20.3 Equipment REQUIREMENTS FOR DESIGN CRITERIA The requirements for design criteria for a specific process plant are naturally dependent on the particular ore, the flowsheet chosen and the site constraints. Normally, the design criteria will be developed in two complementary sections; the first including a description of the ore, process, plant site and geographical location along with the general design criteria developed by each engineering discipline. The second section will present design criteria for each area of the plant and process including an operating data summary and detailed process and equipment descriptions. The general arrangement drawings, flowsheets and process and instrument diagrams form an integral part of the design criteria documentation and are included and referenced extensively for clarity and convenience. General The design criteria consist of the basic project description plus the general criteria for each engineering discipline. The outline for a complete design criteria manual as required for detailed design is provided by the index referenced in Table 3. The summary of basic design criteria will include: Description of the project Scope of work 0 Location 0 Meteorological data Site and soils description 0 Utilities 0 Transportation 0 Applicable laws and codes. Operating Schedule and Capacity The heart of any design is the required plant capacity and the operating schedule. These define the size of the equipment and the operating duty. The operating schedule is normally twenty-four 0 0

9

hours per day, three shifts per day, 365 days per year for a typical concentrator. In some cases the plant may be designed to operate fewer days per week, but continuous twenty-four hour operation is usual. In some dewatering sections filters may be operated on a single shift basis; this will be clearly set out in the design criteria for the area. In addition to the operating schedule and throughput, some estimation of the availability of the plant is used for setting the actual hourly throughput. This availability is a function of the maintenance requirements of each individual piece of equipment and also the operating strategy of the plant. For instance, a weekly maintenance shift may be planned for certain areas due to required preventive maintenance on certain critical items of equipment. These items in the design criteria are normally tabulated in a format as shown in Table 4. Table 4 Typical Operating Schedule PLANT AREA Flotation 365 7 3 8 168 94

Period Daydyear Daydweek ShifVday Hours/shift Hourdweek Availability ("!)

Process Metallurgy The design criteria for process metallurgy will include a description of the ore treated, the products to be made, the physical and chemical characteristics of the ore which affect the process, an estimated metallurgical balance, reagent additions, nominal equipment residence times and product specifications. The metallurgical criteria for each area of the plant or unit process are summarized in an operating data summary for these areas and are utilized with an accompanying process and equipment description to provide a complete criteria for detailed design. The design criteria required by each of these disciplines is outlined in the typical index given in Table 3. As can be seen, each discipline requires a general compilation of design codes and standards as well as specific criteria for the various areas covered by the discipline. The full details of the general criteria would require more space than is available at this time, however, a typical excerpt from the general structural design criteria is presented in Appendix 1B. The references to codes and standards normally refer to Canadian Standards; for plants built in other countries the applicable national, state and local codes would apply. For each discipline involved in the detailed design there are a number of basic criteria based on fundamental design practice for that discipline and other points which are specifically related to mineral processing plant design. These have been developed through experience and are particularly required when the detailed design is being carried out by a firm less involved in the mineral industry, perhaps off-shore in an under developed country or in a country with few mineral resources. Metallurgical (Process): The metallurgical criteria required for each of the unit processes is listed in a previous section. In addition to the process criteria, an operating philosophy for the plant is developed which reflects the plant location and the human resources available to operate it. This operating criteria should be carehlly considered and documented in order to be of use to the design team and for the training of operating personnel. Mechanical: The mechanical engineering for a concentrator, once the major equipment has been selected, is mostly involved with material handling, pumping and design for minimum spillage and easy maintenance of the equipment. In this respect the design criteria are particularly specific to concentrator, with such points as launder slopes, pipe connection details, wear liners, conveyor skirting and pump selection being critically important. With the use of finer grinds in some flotation plants, the behaviour of thickened slurries and other pulp streams has become less

10

predictable, therefore slurry viscosity testwork for pump and piping selection has become more common. Electrical: The electrical design criteria for a plant involves an extremely wide scope possibly fiom site power generation to wiring details. Particular emphasis must be given to energy conservation through such methods as power factor correction, high efficiency motors and lights and correct cable sizing. CiviYStructural: The structural design requires particular emphasis on rotating equipment, vibration analysis, and recognition of plug loads in such vessels as flotation tanks and conditioners. In addition, careful investigation of the underlying soil or rock characteristics are required for successful foundation design. Care should be taken at the earliest opportunity to obtain as much information as possible over the entire mill site, as buildings may be relocated several times during the progress of the design. A general structural design criteria specification is provided in Appendix 1B as an example. Mechanical Services: The services required in a modem concentrator complex for heating, ventilation, fire protection, plumbing, sewer and water supply, etc., can be a major design area. It is essential to develop the design criteria for these areas in accordance with local codes and regulations and also bearing in mind the number of workers in the plant. Sanitary services must also consider the provision of separate facilities for men and women as well as for supervision and labour. This set of criteria must be carefully discussed with operations personnel who are involved at the design stage. Architectural: The architectural criteria reflect the style of building exterior and interior finishes desired for the plant. In addition, the provision of specific criteria for heavy use areas, particular chemical resistance requirements and climatic conditions are important. The development of criteria for offices, washrooms, drys and laboratories requires a detailed description of the proposed work force and operating schedules for the various areas. As for the structural discipline, codes such as the UBC (Uniform Building Code) provide definition to the design. Occupancy rules, fire separations, sprinkling, entrance and egress requirements are all used to define the requirements. Instrumentation: The instrumentation design criteria are developed in conjunction with the process and mechanical engineering groups, and in accordance with the proposed operating criteria. The specific instruments chosen for each application will reflect an overall control philosophy for the plant whether it is a DCS, PLC, or local start-stop switches for simple circuits. The particular design criteria referred to in Appendix 1A refer to a gold plant. The criteria developed for a large base metals flotation concentrator or an iron ore concentrator utilizing spirals or magnetic separation would be similar in nature but reflect the specific ore and process under consideration. PLANTPROCESS AREA For each area of the plant and stand alone unit processes, such as flotation, the design basis document will provide the operating criteria, a process description and a detailed equipment description. The operating criteria include the operating schedule for the particular area along with the availability predicted for the equipment, and the nominal flowsheet and design throughputs. The relationship between nominal and design flows is often misunderstood and should be clearly explained and specified in the design criteria, and if possible noted on the flowsheets as well. In the initial operation of the plant this may not be a major point, but at the stage where debottlenecking, production limiting factors and possibly expansion become a fact of life, there is a critical need for such information to be well defined and available. Once again, space limitations prevent including a complete design criteria for all process areas of a typical mineral processing plant. The areas covered are applicable to most plants but would be augmented and modified specifically for the plant and process being designed. The equipment descriptions in this particular design criteria are not specific as to supplier as the equipment has not yet been purchased. In a fully detailed design criteria for a project in

11

detailed engineering, the major equipment would be committed to and described specifically. As this equipment is purchased the design criteria document will be updated to reflect the particular machinery to be installed.

SUMMARY AND CONCLUSIONS The compilation and use of a formal design criteria document is essential to the engineering design of a mineral processing plant. Careful preparation of this document prior to commencing detailed design along with continuous updating ensure a consistent basis for design in all areas and for all disciplines involved in the design. ACKNOWLEDGEMENTS This chapter is a revision and update of a previous paper: Scott, John W., 1986. Design Criteria: The Formal Basis of Design in Design and Installation of Concentration and Dewatering Circuits, Eds. Andrew L. Mular and Mark A. Anderson, SME-AIME, Little CO. The support of Fluor Mining and Minerals is gratefully acknowledged. LIST OF APPENDICES Format for Process Design Criteria Appendix 1A Appendix 1B

Structural Design Criteria - Excerpt from Master Document

12

APPENDIX 1A FORMAT FOR PROCESS DESIGN CRITERIA

13

Project Design Criteria

MINING GROUP LTD Project Feasibility Study

Client Name:

Project Name: Project Number: FLUOR

~~~

Mining and Minerals

PROJECT DESIGN CRITERIA

-

SECTION 2 PROCESS DEStGN CRITERIA 2.1

Process Description

2.2

Design Criteria The following design criteria establishes the design parameters that are to be followed in project design and execution. The criteria have been compiled from the various sources listed below. The source for a particular criterion is identified per the code in the following List of Sources: MG DER FDW TL

2.3

-

IND

- Industrial practice

PL VEN

- Process Licensor - Vendor information

UNITS

VALUE

SOURCE

tonnesld days years

20,000 365 10

-Mining Group Ltd. Derived or calculated result -Fluor -Testing Lab

Ore Characteristics Mine Operations: - Ore production rate - Operating days per year - Mine life - Schedule: shifts per day hours per shift Plant Performance: - Availability - Daily throughput (design) Specific Gravity: - Average, design

.L.

hours

YO tonnes

MG

12

MG MG

93 2 1,500

MG DER

2.4

MG

2.6 - Bulk density ROM:- ore size - moisture content - angle of repose - draw down angle Low Impact Crushing Index Autogenous Work Index Bond Work Index (rod mill) Bond Work Index (ball mill) Abrasion Index Average Feed Assay: - gold - copper - cyanide soluble Cu

tonnedm’ mm degrees degrees

2.4 1,000 nom. 3Imax. 5 37 60

kW/t kWh/t kWJt kWt

TBD 14 16 TBD

Yo wiw

g/t

I .9

YO

0.1 0.04

%

14

COMMENTS

MG MG MG FDW FDW TL FDW TL TL TL MG MG MG

In-situ, wax immersion; use for crushing and milling. Pychnometer, use for rest of the process. Based on in-situ SG.

FDW to specify FDW to specify FDW to specify FDW to specify To be checked

Client Name: Project Name: Project Number:

MINING GROUP LTD Project Feasibility Study

Project Design Criteria

FLUOR Mining and Minerals PROJECT DESIGN CRITERIA

UNITS UNITS 2.4

2.5

Primary Crushing Circuit Operating Schedule Utilization Crusher availability Crushing time Processing rate, design Product 80% passing Crusher size Crusher open side setting: - maximum - minimum Maximum crusher capacity Crusher discharge feeder type Dump pocket capacity Discharge hopper capacity Coarse Ore Stockpile - Live capacity

- Total capacity - Angle of repose, outside - Drawdown angle

Wshift shiftsld Wyear %

Yo Nd t/h

mm in. rnm mm t/h

h tonnes tonnes degrees degrees

VALUE VALUE

SOURCE SOURCE

2 7,300 83.3 80% 16 1,344 I50 48 x 74

MG MG MG MG MG MG DER DDR FDW

175 140 TBD Apron TBD TBD

FDW FDW VEN MG FDW FDW

24 2 1,500 TBD 37 55

MGlFDW

10

COMMENTS COMMENTS

to be verified

FDW recommends belt

DER FDW FDW

Reclaim:

2.6

- No. of conveyors - Reclaim feeder type - No. of feeders Grinding Circuit - General Operating Schedule

dJY Nd

Plant availability ?h Processing rate, design t/h 2.6.1 Primary Grinding Circuit- SALG - Mill size ftxft -Commercial a'genous W.I. kWt - Scale-up factor - Design operating power kWh1t -New feed size (bo) microns - Product size (kso) microns -New feed rate t/h - Selected power kW - Nominal steel charge YO - Design steel charge % YOsolids -Mill slurry density UNITS

I Belt 4

MG

365 24 93 896

MG MG MG DER

32~13.25 TBD TBD 8.2 150,000 800 896 7,500 6 10 70 VALUE

FDW

15

MG FDW to specify

FDW FDW FDW DER FDW FDW FDW FD W SOURCE

FDW to specify FDW to specify FDW to specify FDW to specify FDW to specify FDW to specify FDW to specify FDW to specify FDW to specify FDW to specify COMMENTS

Client Name: Project Name: Project Number:

MINMG GROUP LTD Project Feasibility Study

Project Design Criteria

FLUOR Mining and Minerals PROJ~CTDESIGNCRITERIA

UNITS

VALUE

2.6.2 SAG Mill Discharge Screen - Type horz.,vibrt - Aperture mm - Unit area m2/m3/h - No. of screens - Screen o/size, % SAG % discharge - Screen o/size flowrate uh - O/size solids conc. % - Size ftXft 2.6.3 Pebble Crushing Circuit 2.6.4 Secondary Grinding Circuit: - Grinding mill type Ball - Bond work index kWh/t - Ball mill circuit feed (kso) microns - Ball mill circuit product microns @SO) - Design power per mill kW - Ball mill size, dia.xlength ftXft -Number of mills - Ball mill density Yo cw 2.6.5 Product Classification - Type - Cyclone O/F product Pg0 microns - Cyclone underflow density % solids - Cyclone overflow density YOsolids % - Circulating load - Cyclone diameter in. - Number of cyclones operatinghall mill - Number of cyclones installeclhall mill 2.7 Leaching Circuit: 2.7.1 Trash Screens vibratinf - Type m3/h/m - Screen unit area mesh - Aperture mxm - Screen size -Number of screens kW - Screen power t/h - Required water m3/h - Required air

1.7 TBD 2 30

270 90 8x24

SOURCE

COMMENTS

FDW FDW FDW FDW FDW

1 op., 1 standby FDW to verify

FDW FDW FDW FDW to determine if required

16 800

FDW TL FDW

FDW to specify FDW to specify FDW to specify

80 4,500 18x30 2 65

MG FDW FDW FDW FDW

FDW to specify FDW to specify FDW to specify FDW to specify FDW to specify

Wcyclone 80 75 40 250 15

IND MG IND FDW VEN VEN

FDW to verify FDW to verify FDW to verify FDW to verify

8

VEN

FDW to verify 10

TBD 20 2 TBD TBD TBD

16

VEN FDW MG VEN IND VEN DER VEN VEN VEN

FDW to specify

APPENDIX 1B STRUCTURAL DESIGN CRITERIA EXCERPT FROM MASTER DOCUMENT

17

Client Name: Project Name: Project Number:

Specification 160.215.00910 Page 2 of I Attachment 4 Rev. A

FLUOR DANIEL Mining & Minerals

STRUCTURAL ENGINEERING DESIGN CRITERIA

1.

GENERAL

1.1

Summary NOTE!!! Notes, including this one, otherwise indicated, are to be removed or reconciled when this is issued as a Project specification. This specification prescribes a uniform format for Project level presentation of structural engineering design criteria for a Project. The Lead Structural Engineer has the responsibility to modify, add, or delete requirements to meet other governing criteria, including Client requirements, regional or local codes, or insurance underwriters' requirements. A.

B.

Scope of Specification I.

This specification establishes the criteria for structural engineering and design on a Project. The specification defines applicable codes and standards, design loads, materials of construction, and foundation requirements.

2.

Whenever this specification differs from the referenced codes, standards, or specifications, this specification shall govern.

Related Specifications The following specifications prescribe items of related work: 1.

000.215.03300

2.

000.21 5.03600

Structural Concrete and Reinforcing Grouts

3.

000.215.05120

Structural Steel

4.

000.220.04220

Concrete Unit Masonry

Coordinate work prescribed by this specification with work prescribed by the above listed specifications. C.

Related Technical Requirements The following specifications describe items of related work: 1.

Practice 000.215.1207

Anchor Bolt Design Criteria

2.

Practice 000.215.1215

Wind Load Calculation

3.

Practice 000.215.1216

Earthquake Engineering

NOTE!!! Include related calculations and documents. Where FD Procurement systems and PAR forms are not used, related drawings may be listed here. 1.2

References The publications listed below form part of this specification, and shall be used for the purposes listed. Each publication shall be the latest revision and addendum in effect on the date this specification is issued for construction unless noted otherwise. Except as modified by the requirements specified herein or the details of the drawings. Work included in this specification shall conform to the applicable provisions of these publications. NOTE!!! Engineer to verify latest dates of referenced documents before issuing specification. A.

-

AASHTO American Association of State Highway and Traffic Officials

18

Specification 160.215.00910 Page 3 of i Attachment 4 Rev. A

Client Name: Project Name: Project Number:

FLUOR DANIEL Mining 8 Minerals STRUCTURAL ENGINEERING DESIGN CRITERIA 1.

Standard Specification for Highway Bridges, Sixteenth Edition, 1996. a.

B.

C.

ACI - American Concrete Institute I.

318/318R-95

Building Code Requirements for Reinforced Concrete

2.

301-96

Specifications for Structural Concrete for Buildings

3.

530/530.1-95

Building Code Requirements for Masonry Structures, Specifcations for Masonry Structures and Related commentaries NOTE!!! If UBC is specified as the governing code, delete item 8.3 above and add masonry structures under UBC heading.

AlSC -American Institute of Steel Construction 1.

D.

Truck Loads

Ninth Edition, 1989. a.

Steel Design

b.

Steel Erection

API -American Petroleum Institute 1.

STD 650-93 - Ninth Edition a.

Welded Steel Tanks for Oil Storage (including seismic design)

E.

AREA - American Railroad Engineering Association

F.

ASCE - American Society of Civil Engineers I.

7-88 a.

Load Definitions

b.

Live Loads

c.

SnowLoads

d.

Impact Loads

e.

Wind Loads

NOTE!!! If UBC is specified as sthe governing code rather than ASCE, delete Live Loads and Wind Loads here and add the same under UBC heading. G.

AWS - American Welding Society 1.

H.

NFPA - National Fire Protection Association

1.

1.

D1.l-96 Structural Welding

NFPA 101-97

Safety to Life from Fire in Buildings and Structures

OSHA Occupational Safety and Health Act 1.

Part 1910, Subpart D

19

Client Name: Project Name: Project Number:

Specification 160.215.00910 Page 4 of I Attachment 4 Rev. A

FLUOR DANIEL Mining 8, Minerals STRUCTURAL ENGINEERING DESIGN CRITERIA

J.

Guardrails

b.

Handrails

c.

Stair Treads

d.

Ladders

UBC - Uniform Building Code 1.

2.

a.

1997 Edition.

a.

Earthquake Loads

b.

Rain Loads (Ponding)

c.

Building Design

PRODUCTS

Not Applicable 3.

EXECUTION 3.1

Design Loads Loads and forces used for design shall be as defined in ASCE 7-88, hereafter referred to as ASCE, and as specified below. Buildings and structures, unless noted otherwise, shall be classified as Category I according to ASCE, Table 1, for the purpose of determining wind and snow loads. A.

D (Dead Loads) These are the vertical loads due to the weight of permanent structural and nonstructural components of a building or structure, including empty vessels, equipment, built-in partitions, fireproofing, insulation, piping and ducting, electrical conduits, and permanent fixtures.

B.

0 (Operating Loads) These are the dead loads plus the weight of any liquids or solids present within the vessels, equipment, or piping during normal operation. Also consider unusual loading such as that occurring during regeneration or upset conditions when fluid levels could be higher than normal operating levels. Operating load shall have the same load factor as dead load.

C.

Te (Test Loads) These are the dead loads plus the weight of liquids necessary to pressure test vessels, equipment, or piping. Test load shall have the same load factor as dead load.

D.

L (Live Loads) 1.

These are the loads produced by the use and occupancy of the building or structure. They include the weight of all movable loads, including personnel, tools, miscellaneous equipment, movable partitions, cranes, hoists, parts of dismantled equipment, and stored material.

2.

Live loads and reduction of live loads shall be according to ASCE Section 4; OSHA Part 1910, Subpart D; and as specified in Attachment 01.

20

Client Name: Project Name: Project Number:

Specification 160.215.00910 Page 5 of 1 Attachment 4 Rev. A

FLUOR DANIEL Mining & Minerals

STRUCTURAL ENGINEERINGDESIGN CRITERIA

3. E.

Minimum Lr (roof live loads) shall conform to ASCE, Section 4.11.

Ve (Vehicular Loads) Bridges, trenches, and underground installations accessible to truck loading shall be designed to withstand HS20 load as defined by AASHTO Standard Specifications-for Highway Bridges. Maintenance or construction crane or bridge crane loads shall be considered also, where applicable. Truck or crane load shall have the same load factor as live load.

F.

S (Snow Loads) 1.

Snow loading on roofs or other exposed surfaces (platforms) shall be considered according to ASCE Section 7, and the following parameters: Ground snow load, Pg

=

[::

::1 psf

Occupancy Category=

I (ASCE Table 1)

Exposure Factor, C(e)

=

Importance Factor (I)=

1 .O(ASCE, Table 20)

Thermal Factor, C(t) =

1.2 (ASCE Table 19)

1.0 (ASCE Table 18)

NOTE!!! The occupancy category and other factors listed here shall be adjusted for each Project or building requirement.

G.

2.

In no case shall the minimum design snow load be less than 20 psf. The effects of snow drifts at valleys, parapets, and offsets in roofs shall be considered.

3.

Live load shall not be assumed to act concurrently with snow load upon the same surface. Use the larger of either load. Snow load shall have the same load factor as live load.

W (Wind Loads) 1.

Every building, structure, component, and cladding shall be designed to resist wind effects according to Attachment [:: ::I. NOTE!!! Include Attachment 02 when ASCE is selected as the governing code and Attachment 03 when UBC is selected as the governing codes. NOTE!!! Detailed procedures for wind load calculations for buildings, structure, and equipment are given in Structural Engineering Practice 000.215.1215.

2.

Wind load shall be separately computed for all supported equipment, ladders, and stairs except for vessels whose wind increase factors have already accounted for these items.

3.

No reduction shall be made for the shielding effect of vessels or structures adjacent to the structure being designed.

4.

The overturning moment due to wind shall not exceed 2/3 of the resisting moment of the structure during its lightest possible condition after plant construction is complete.

5.

Flexible wind-sensitive structures, including vessels, are defined as having at least 1 of the following conditions: Height exceeds 5 times the least horizontal dimension Fundamental frequency less than 1 Hertz (continued ....)

21

1 Sampling Sectlon Co-Editors: Graham Farquharson and Art Winckers

Sampling a Mineral Deposit for Feasibility Studies K.J. Ashley ............................................................................................................................

25

Sampling in Mineral Processing

J. W. Merks ............................................................................................................................

37

Sampling High Throughput Grinding and Flotation Circuits J. Mosher, D. Alexander .......................................................................................................

63

Practical and Theoretical Difflcuities When Sampling Gold F.F. Pitard ............................................................................................................................

77

Sampling a Mineral Deposit for Metallurgical Testing and the Design of Comminution and Mineral Separation Processes J. Hanks, D. Barratt .............................................................................................................

99

Bench Scale and Pilot Plant Testwork Section Co-Editors: JeffAustin and John Mosher

Overview of Metallurgical Testlng Procedures and Flowsheet Development

T.P. McNulty .........................................................................................................................

119

Bench-Scale and Pilot Plant Tests for Commlnution Clrcult Design

J. Mosher, T. Bigg ................................................................................................................

123

The Selection of Flotation Reagents vla Batch Flotatlon Tests

P. Thompson .........................................................................................................................

136

Bench and Pilot Plant Programs for Flotation Circuit Design S. R. Williams, M.0. Ounpuu, K. W. Sarbutt ..........................................................................

145

Bench-Scale and Pilot Plant Testwork for Gravlty Concentratlon Clrcuit Design A.R. Laplante, D.E. Spiller ...................................................................................................

160

Bench Scale and Pilot Plant Tests for Magnetic Concentration Circult Design D.A. Norrgran, M.J. Mankosa ..............................................................................................

176

Bench-Scale and Pilot Plant Tests for Thickening and Clarlflcatlon Clrcult Design B.K. Pocock, C.B. Smith, G.D. Welch ..................................................................................

201

Bench-Scale and Pilot Plant Tests for Flitration Clrcuit Deslgn T. Kram .................................................................................................................................

207

Gold Roasting, Autociaving, or Blo-Oxidatlon Process Selection Based on Bench-Scale and Pilot Plant Test Work and Costs

J. McMullen, K.G. Thomas ...................................................................................................

211

Bench-Scale and Pilot Plant Tests for Cyanide Leach Circuit Design G.E. McClelland, J.S. McPartland .......................................................................................

251

Bench-Scale and Pilot Plant Work for Gold- and Copper-RecoveryCircuit Design

D. Thompson ........................................................................................................................

264

Guiding Process Developments by Uslng Automated Mineralogical Analysls D. Sutherland, Y.Gu.............................................................................................................

270

Financial and Feasibility Studies Section Co-Editors: Brian H. Johnson and John W. Scott

Guidelines to Feasibility Studies

J. Scott, B. Johnston .............................................................................................................

281

Major Mineral Processing Equlpment Costs and Preliminary Capital Cost Estimations

A.L. Mular ............................................................................................................................

310

Process Operating Costs with Applications in Mine Planningand Risk Analysis

D. Halbe, T.J. Smolik............................................................................................................

326

Financlal Analysts and Economic Optlmization

L.D. Smith .............................................................................................................................

346

Mining Project Finance Explained

R. Halupka ............................................................................................................................

371

4 Models and Simulators for Selection, Sizing, and Design Section Co-Editors: Dr. Brian Flintoff and Dr. John Herbst

Mineral ProcessingPiant/Circuit Simulators: An Overview

J. Herbst, R.K. Rajamni, A. Mular, B. Flintofs...................................................................

383

BRUNO Metso Minerals' Crushing Plant Simulator

D.M. Kaja .............................................................................................................................

404

PiantDesigner": A Crushing and Screening Modeling Tool

P . Hedvall, M. Nordin ..........................................................................................................

421

JKSimMet: A Simulator for Analysis, Optimlsation, and Design of Comminution Circuits

R.D. Morrison, J.M. Richardson ..........................................................................................

442

JKSimFloat as a Practical Tool for Flotation Process Design and Optimization

M.C. Harris, K.C. Runge, W.J. Whiten, R.D. Morrison .......................................................

461

USiM PAC 3: Design and Optimizatlon of Mineral ProcessingPlants from Crushing to Refining

S. Brochot, J. Villeneuve, J.C. Guillaneau, M.V. Durance, F. Bourgeois ............................

479

Emergence of HFS as a Design Tool In Mineral Processing

J.A. Herbst, L.K. Nordell ......................................................................................................

495

Reducing Maintenance Costs Using Process and Equipment Event Management

O.A. Bascur, J.P. Kennedy ...................................................................................................

507

Enterprise Dynamic Simulation Models

D. W. Ginsberg ......................................................................................................................

528

Comminution (Crushing and Grinding) Section Co-Editors: R.E. Mclvor, Tony Moon, and James Vanderbeek

Factors Which Influencethe Selectlon of Commlnutlon Clrcults D. Barratt, M. Sherman ........................................................................................................

539

Types and Characteristicsof Crushing Equlpment and Clrcult Flowsheets

K. Major ...............................................................................................................................

566

Selection and Slzlng of Primary Crushers R. W. Utley.............................................................................................................................

584

In-Pit Crushing Design and Layout Conslderatlons

K. Boyd, R. W. Utley..............................................................................................................

606

Selection and Slzlng of Secondary and Tertlary Cone Crushers G. Beerkircher, K. O’Bryan, K. Lim .....................................................................................

621

Selection, Sizing, and Speclal Conslderatlonsfor Pebble Crushers

K, O’Bryan, K. Lim...............................................................................................................

628

Selection and Sizing of Hlgh Pressure Grlndlng Rolls R. Klymowsky, N. Patzelt, J. Knecht, E. Burchardt ..............................................................

636

Crushing Plant Design and Layout Conslderatlons K. Boyd .................................................................................................................................

669

Types and Characteristics of Grlndlng Equlpment and Clrcult Flowsheets

M.I. Callow, A.G. Moon .......................................................................................................

698

Selection of Rod Mills, Ball Mills, and Regrlnd Mills C.A. Rowland Jr. ..................................................................................................................

710

Selection and Slzlng of Autogenous and SemlAutogenous Mills D. Barratt, M . Sherman ........................................................................................................

755

Selection and Slzlng of Ultraflne and Stlrred Grlndlng Mills

J. K.H. Lichter, G. Davey ......................................................................................................

783

Grlndlng Plant Design and Layout Conslderatlons

M.I. Callow, D.G. Meadows .................................................................................................

801

Selection and Evaluation of Grinding Mill Drives G.A. Grandy. C.D. Danecki. P.F. Thomas ...........................................................................

819

The Design of Grinding Mills V. Svalbonas .........................................................................................................................

840

538

Size Separation Section Co-Edltors: Patrick Turner and James E. Wennen

Slzlng and Appilcatlon of Gravity Ciasslflers W.M. Reed ............................................................................................................................

867

HydrocycloneSelection for Plant Design T.J. Olson, P.A. Turner.........................................................................................................

880

Coarse Screenlng M.A. Bothwell, A.L. Mular ...................................................................................................

894

Fine Screenlng In Mlneral Processlng Operations

S.B. Valine, J.E. Wennen ......................................................................................................

917

The Use of Hindered Settlers to Improve iron Ore Gravity Concentration Clrcults S. Hearn ................................................................................................................................

929

7 Solid-Solid Separation Sectlon Co-Editors: George H. Hope and Donovan F. Symonds

Types and Characteristics of Gravity separation and Flowsheets

R.O. Burt ..............................................................................................................................

947

Types and Characteristics of Heavy-Media Separators and Flowsheets

R.A. Reeves ...........................................................................................................................

962

Types and Characteristics of Non-Heavy Medium Separators and Flowsheets

J.K. Alderman .......................................................................................................................

978

The Selection and Sizing of Centrifugal Concentration Equipment: Plant Design and Layout

A.R. Laplante ........................................................................................................................

995

Slzing and Selection of Heavy Media Equipment: Design and Layout

D. F. Symonds, S. Malbon .....................................................................................................

1011

Photometric Ore Sorting B. Arvidson ...........................................................................................................................

1033

Electrical Methods of Separation

A.L. Mular ............................................................................................................................

1049

Selection and Sizing of Magnetic Concentrating Equipment: Plant Design/Layout

D.A. Norrgran, M.J. Mankosa ..............................................................................................

1069

Flotation SectDon Co-Editors: M. Ian Callow and Glenn Dobby

Overview of Flotation Technology and Plant Practice for Complex Sulphlde Ores N.W. Johnson, P.D. Munro...................................................................................................

1097

Overview of Recent Developments In Flotation Technology and Plant Practice for Copper Gold Ores A. Winckers...........................................................................................................................

ll24

An Overview of Recent Developments In Flotation Technology and Plant Practlce for Nickel Ores A. Kerr ..................................................................................................................................

1142

Nonsulflde Flotation Technology and Plant Practlce J. Miller, B. Tippin, R. Pruett ...............................................................................................

1159

Design of Mechanical Flotation Machines M.G. Nelson, F.P. Traczyk, D. Lelinski ................................................................................

1179

Flotation Equipment Selection and Plant Layout

K.R. Wood.............................................................................................................................

1204

Column Flotation

G. Dobby ..............................................................................................................................

1239

Solid-Liquid Separation Section Co-Editors: Frank Baczek and Benjamin K.Pocock

Characterization of Process Objectives and (General) Approach to Equipment Selection C.E. Silverblatt, J.H. Easton.................................................................................................

1255

Centrifugal Sedimentation and Filtration for Mineral Processing

W. Leung ...............................................................................................................................

1262

Characterization of Equipment Baged on Filtration Prlnclpais and Theory

G.D.Welch ...........................................................................................................................

1289

Testing, Sizing, and Specifying Sedimentation Equipment T. Laros, S, Slottee, F. Baczek ..............................................................................................

1295

Testing, Sizing, and Specifying of Flltratlon Equipment

C.B. Smith, I.G. Townsend ...................................................................................................

1313

Design Features and Types of Sedlmentatlon Equipment

F. Schoenbrunn, T. bras .....................................................................................................

1331

Design Features and Types of Filtration Equipment C. Cox,

F. Traczyk................................................................................................................

1342

Plant Design, Layout, and Economic Considerations

M. Erickson, M.Blois ...........................................................................................................

1358

10 Pumping, Material Transport, Drying, and Storage Section Co-Editors: Ken Boyd and William E. Norquisr

Selection and Sizing of Slurry Pumps

M.J. Bootie ...........................................................................................................................

2373

Selection and Sizing of Slurry Llnes, Pumpboxes, and Launders

B. Abulnaga, K. Major, P . Wells ..........................................................................................

1403

Slurry Pipeline Transportatlon

B.L. Ricks ..............................................................................................................................

1422

The Selection and Sizing of Conveyors, Stackers, and Reclalmers G. Barfoot, D. Bennett, M. Col .............................................................................................

1446

Selectlon and Sizlng of Concentrate Drying, Handling, and Storage Equipment

M.E. Prokesch, G. Graber....................................................................................................

1463

The Selection and Slzlng of Bins, Hopper Outlets, and Feeders

J. Carson, T.Holmes ............................................................................................................

1478

Pre-Oxidation Section Co-Editors: Dr. Ralph P. Hack1 and Dr. Kenneth Thomas

Design of Barrick Goldstrlke’s Two-Stage Roaster

D. Wamica, A. Cole, S. Bunk ...............................................................................................

1493

Selection of Materials and Mechanical Deslgn of Pressure Leaching Equipment

K. Lamb, J. Gulyas ...............................................................................................................

1510

Barrlck Gold-Autoclaving and Roasting of Refractory Ores

K.G. Thomas, A. Cole, R.A. Williams...................................................................................

1530

Selectlon and Sizing of Biooxldatlon Equlpment and Clrcults

C.L. Brierley, A.P. Briggs.....................................................................................................

1540

12 Leaching and Adsorption Circuits Section Co-Editors: Dr. Chris Fleming and Michael R. SchafSner

Copper Heap Leach Design and Practice

R.E. Scheffel .........................................................................................................................

1571

Precious Metal Heap Leach Design and Practice

D. W. Kappes .........................................................................................................................

1606

Agitated lank Leaching Selection and Design

K.A. Altman, M. SchafSner, S. McTavish ..............................................................................

1631

CiP/CiL/CiC Adsorption Circuit Process Selection

C.A. Fleming .........................................................................................................................

1644

CiP/CiL/CiC Adsorption Circuit Equipment Selectlon and Design

K.A. Altman, S. McTavish .....................................................................................................

1652

13 Extraction Section Co-Editors: Paul G. Semple and Ron Bradburn

Zinc Cementatlon-The Merrlll Crowe Process A.P. Hampton .......................................................................................................................

1663

Selection and Design of Carbon Reactivatlon Clrcults J. yon Beckmann, P.G. Semple .............................................................................................

1680

Selectlon and Slzlng of Elutlon and Eiectrowlnnlng Clrcults P. Hosford, J. Wells..............................................................................................................

1694

Selection and Sizing of Copper Solvent Extractlon and Electrowinning Equlpment and Circuits C.G. Anderson, M.A. Giralico, T.A. Post, T.G. Robinson, O.S. Tinkler ...............................

1709

14 Bullion Production and Refining Section Editor: Dr. Corby Anderson

Bullion Productlon and Reflnlng C.O. Gale, T.A. Weldon........................................................................................................

1747

Platlnum Group Metal Bulllon Production and Reflnlng C.G. Anderson, L.C. Newman, G.K.Roset ...........................................................................

1760

Fundamentals of the Analysls of Gold, Sllver, and Platlnum Group Metals

C.G.Anderson ......................................................................................................................

1778

15 Tailings Disposal, Wastewater Disposal, and t h e Environment Section Co-Editors: James R. Arnold and Dr. George W. Poling

Management of Tailings Disposal on Land B.S. Brown ............................................................................................................................

1809

Design of Tailings Dams and impoundments P.C. Lighthall, M.P. Davies, S.Rice, T.E. Martin................................................................

1828

Hazardous Constituent Removal from Waste and Process Water

L. Twidwell, J. McCloskey, M. Gale-Lee..............................................................................

1847

Treatment of Solutions and Slurries for Cyanlde Removal

M.M. Botz, TI.Mudder ........................................................................................................

1866

Strategies for Minimization and Management of Acid Rock Drainage and Other Mlnlng-InfluencedWaters

R.L. Schmiermund ................................................................................................................

1886

Environmental and Social Considerations in Facility Siting

B.A. Filas, R. W. Reisinger, C.C. Parnow .............................................................................

1902

16 Construction Materials for Equipment and Plants Sectlon Co-Edltors: Dr. Rod McElroy and Wesley Young

Selectlon of Metalllc Materlals for the Mlnlng/Metallurglcal Industry G. Coares ..............................................................................................................................

1911

Elastomers In the Mlneral Processlng Industry P. Schnarr, LE. Schaeffer, H.J. Weinand ............................................................................

1932

Plastlcsfor Process Plants and Equipment

G. W. McCuaig ......................................................................................................................

1953

Cornmerclal Acceptance and Appllcatlons of Masonry and Membrane Systems for the Process Industries R.E. Aliasso Jr., T.E. Crandall, D.M. Malone, R.J. Storms..................................................

1962

17 Power, Water, and Support Facilities Section Co-Editors: M.N. Brodie and Charles R. Edwards

The Development of an Electric Power Dlstributlon System M.N. Brodie ..........................................................................................................................

1973

Selection of Motors and Drlve Systems for Comminution Clrcults P.F. Thomas .........................................................................................................................

1983

Selectlon of Metallurgical Laboratory and Assay Equipment: Laboratory Designs and Layouts P.F. Wells .............................................................................................................................

2011

On-Line Composition Analysis of Mineral Slurries T.F. Braden, M. Kongas, K. Saloheimo ................................................................................

2020

18 Process Control and Instrumentation Section Co-Editors: Robert Edwards and Aundra Nix

introduction to Process Control

B. Flintoff .............................................................................................................................

2051

Well Balanced Control Systems T. Stuffco, K. Sunna ..............................................................................................................

2066

The Selection of Control Hardware for Mineral Processing R.A. Medower, R.E. Cook .....................................................................................................

2077

Basic Field Instrumentation and Control System Maintenance In Mineral Processing Circuits J.R. Sienkiewicz ....................................................................................................................

2104

Strategies for Instrumentation and Control of Crushing Circuits S.D. Parsons, S.J.Parker, J. W. Craven, R.P. Sloan.............................................................

2114

Strategies for the instrumentation and Control of Grinding Circuits R. Edwards, A. Vien, R. Perry ..............................................................................................

2130

strategies for the instrumentation and Control of Solid-Solid Separation Processes G.H. Luttrell, M.J. Mankosa .................................................................................................

2152

Strategies for Instrumentation and Control of Thickeners and Other Solid-Liquid Separation Circuits F. Schoenbrunn, L. Hales, D. Bedell ....................................................................................

2164

Strategies for Instrumentation and Control of Flotation Circuits H . Laurila, J. Karesvuori, 0. Tiili ........................................................................................

2174

Pressure Oxidation Control Strategies J. Cole, J. Rust ......................................................................................................................

2196

Engineering, Procurement, Construction, and Management Section Co-Editors: Roger M.Nendick and Robert C. Schenk

Development of a Mineral Processing Flowsheet-Case History, Batu Hljau T. de Mull, S. Saich, K. Sobel ...............................................................................................

2211

Speclflcatlon and Purchase of Equipment for Mineral ProcessingPlants C. Hunker, S. Maldonado .....................................................................................................

2223

The Management and Control of Costs of Capital Mineral Processing Plants D. W. Stewart .........................................................................................................................

2230

Schedule Development and Schedule of Control of Mineral Processing Plants P. Kumar ..............................................................................................................................

2238

The Risks and Rewards Associated with Dlfferent Contractual Approaches P.J. Gard ..............................................................................................................................

2245

Success Strategies for Building New Mining Projects R.J. Hickson ..........................................................................................................................

2250

20 Start-up, Commissioning, and Training Section Co-Editors: Ken Major and Mike Mular

Pre-Commissioning,Commissioning, and Training

T.Watson..............................................................................................................................

2277

Plant Ramp Up and Performance Testing

R.M.Nendick ........................................................................................................................

2285

Preparation of Effective Operating Manuals to Support Operator Training for Metallurgical Plant Start-ups S.R. Brown ............................................................................................................................

2290

Planning and Staffing for a Successful Project Start-up

K.A. Brunk, L.J. Buter, K.M. Levier .....................................................................................

2299

Maintenance Scheduling, Management, and Tralnlng at Start-up: A Case Study

P. Vujic.................................................................................................................................

2315

Operator Training

A. Vien ..................................................................................................................................

2328

Safety and Health Considerations and Procedures During Plant Start-up

L.A. Schack ...........................................................................................................................

2337

21 Case Studies Section Co-Editors: Dr. Martin C. Kuhn and Donald C. Gale

Sunrise Dam Gold Mine-Concept to Production W.R. Lethlean, P.J. Banovich ...............................................................................................

2345

A Case Study in SAG Concentrator Design and Operations at P.T. Freeport Indonesia R. Coleman, A. Neale, P. Staples..........................................................................................

2367

High Pressure Grindlng Roll Utlllzatlon at the Empire Mine D.J. Rose, P.A. Korpi, E.C. Dowling, R.E. Mclvor ..............................................................

2380

The Raglan Concentrator-Technology Development In the Arctlc J. Holmes, D. Hyma, P. Lunglois .........................................................................................

2394

AUTHOR INDEX

Index Terms

Links

A Abulnaga, B.

1403

Alderman, J.K.

978

Alexander, D.

63

Aliasso Jr., R.E.

1962

Altman, K.A.

1631

1652

Anderson, C.G.

1709

1760

1778

Arvidson, B.

1033

Ashley, K.J.

25

539

755

B Baczek, F.

1295

Banovich, P.J.

2345

Barfoot, G.

1446

Barratt, D.

99

Bascur, O.A.

507

Bedell, D.

2164

Beerkircher, G. Bennett, D.

621 1446

Bigg, T.

123

Blois, M.

1358

Bootle, M.J.

1373

Bothwell, M.A. Botz, M.M.

894 1866

Bourgeois, F.

479

Boyd, K.

606

Braden, T.F.

2020

Brierley, C.L.

1540

Briggs, A.P.

1540

669

This page has been reformatted by Knovel to provide easier navigation.

Index Terms Brochot, S.

Links 479

Brodie, M.N.

1973

Brown, B.S.

1809

Brown, S.R.

2290

Brunk, K.A.

2299

Bunk, S.

1493

Burchardt, E.

636

Burt, R.O.

947

Buter, L.J.

2299

C Callow, M.I.

698

Carson, J.

1478

Coats, G.

1911

Col, M.

1446

Cole, A.

1493

Cole, J.

2196

Coleman, R.

2367

Cook, R.E.

2077

Cox, C.

1342

Crandall, T.E.

1962

Craven, J.W.

2114

801

1530

D Danecki, C.D.

819

Davey, G.

783

Davies, M.P.

1828

de Mull, T.

2211

Dobby, G.

1239

Dowling, E.C.

2380

Durance, M.V.

479

E Easton, J.H.

1255

Edwards, R.

2130

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Erickson, M.

1358

F Filas, B.A.

1902

Fleming, C.A.

1644

Flintoff, B.

383

2051

G Gale, C.O.

1747

Gale-Lee, M.

1847

Gard, P.J.

2245

Ginsberg, D.W.

528

Giralico, M.A.

1709

Graber, G.

1463

Grandy, G.A.

819

Gu, Y.

270

Guillaneau, J.C.

479

Gulyas, J.

1510

H Halbe, D.

326

Hales, L.

2164

Halupka, R.

371

Hampton, A.P. Hanks, J.

1663 99

Harris, M.C.

461

Hearn, S.

929

Hedvall, P.

421

Herbst, J.A.

383

Hickson, R.J.

2250

Holmes, J.

2394

Holmes, T.

1478

Hosford, P.

1694

Hunker, C.

2223

Hyma, D.

2394

495

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

J Johnson, N.W. Johnston, B.

1097 281

K Kaja, D.M.

404

Kappes, D.W.

1606

Karesvuori, J.

2174

Kennedy, J.P.

507

Kerr; A.

1142

Klymowsky, R.

636

Knecht, J.

636

Kongas, M.

2020

Korpi, P.A.

2380

Kram, T.

207

Kumar, P.

2238

L Lamb, K.

1510

Langlois, P.

2394

Laplante. A.R.

160

995

Laros, T.

1295

1331

Laurila, H.

2174

Lelinski, D.

1179

Lethlean, W.R.

2345

Leung, W.

1262

Levier, K.M.

2299

Lichter, J.K.H.

783

Lighthall, P.C.

1828

Lim, K.

621

Luttrell, G.H.

628

2152

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

M Major, K.

566

Malbon, S.

1011

Maldonado, S.

2223

Malone, D.M.

1962

Mankosa, M.J.

176

Martin, T.E.

1403

1069

2152

1828

McClelland, G.E.

251

McCloskey, J.

1847

McCuaig, G.W.

1953

McIvor, R.E.

2380

McMullen, J.

211

McNulty, T.P.

119

McPartland, J.S.

251

McTavish, S.

1631

Meadows, D.G.

801

Medower, R.A.

2077

Merks, J.W.

1652

37

Miller, J.

1159

Moon, A.G.

698

Morrison, R.D.

442

461

63

123

Mosher, J. Mudder, T.I.

1866

Mular, A.L.

310

383

1049 Munro, P.D.

1097

N Neale, A.

2367

Nelson, M.G.

1179

Nendick, R.M.

2285

Newman, L.C.

1760

Nordell, L.K.

495

This page has been reformatted by Knovel to provide easier navigation.

894

Index Terms

Links

Nordin, M.

421

Norrgran, D.A.

176

1069

O’Bryan, K.

621

628

Olson, T.J.

880

Ounpuu, M.O.

145

O

P Parker, S.J.

2114

Parnow, C.C.

1902

Parsons, S.D.

2114

Patzelt, N.

636

Perry, R.

2130

Pitard, F.F.

77

Pocock, B.K.

201

Post, T.A.

1709

Prokesch, M.E.

1463

Pruett, R.

1159

R Rajamani, R.K.

383

Reed, W.M.

867

Reeves, R.A.

962

Reisinger, R.W.

1902

Rice, S.

1828

Richardson, J.M.

442

Ricks, B.L.

1422

Robinson, T.G.

1709

Rose, D.J.

2380

Roset, G.K.

1760

Rowland Jr., C.A.

710

Runge, K.C.

461

Rust, J.

2196

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

S Saich, S.

2211

Saloheimo, K.

2020

Sarbutt, K.W.

145

Schack, L.A.

2337

Schaeffer, L.E.

1932

Schaffner, M.

1631

Scheffel, R.E.

1571

Schmiermund, R.L.

1886

Schnarr, P.

1932

Schoenbrunn, F.

1331

2164

3

281

Scott, J.W. Semple, P.G.

1680

Sherman, M.

539

Sienkiewicz, J.R.

2104

Silverblatt, C.E.

1255

Sloan, R.P.

2114

Slottee, S.

1295

Smith, C.B.

201

Smith, L.D.

346

Smolik, T.J.

326

Sobel, K.

755

1313

2211

Spiller, D.E.

160

Staples, P.

2367

Stewart, D.W.

2230

Storms, R.J.

1962

Stuffco, T.

2066

Sunna, K.

2066

Sutherland, D.

270

Svalbonas, V.

840

Symonds, D.F.

1011

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

T Thomas, K.G.

211

1530

Thomas, P.F.

819

1983

Thompson, D.

264

Thompson, P.

136

Tiili, O.

2174

Tinkler, O.S.

1709

Tippin, B.

1159

Townsend, I.G.

1313

Traczyk, F.P.

1179

Turner, P.A.

880

Twidwell, L.

1847

1342

U Utley, R.W.

584

606

V Valine, S.B.

917

Vien, A.

2130

Villeneuve, J.

2328

479

von Beckmann, J.

1680

Vujic, P.

2315

W Warnica. D.

1493

Watson, T.

2277

Weinand, H.J.

1932

Welch, G.D.

201

Weldon, T.A.

1747

Wells, J.

1694

Wells, P.F.

2011

Wells, P.J.

1403

Wennen, J.E.

917

Whiten, W.J.

461

1289

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Williams, R.A.

1530

Williams, S.R.

145

Winckers, A.

1124

Wood, K.R.

1204

This page has been reformatted by Knovel to provide easier navigation.

SUBJECT INDEX

Index Terms

Links

A Acid rock drainage prediction techniques Acid wash circuits Activated carbon adsorption in gold recovery

1886

1899

1894 1680 264

AG mills bench-scale and pilot plant testing for circuit design motor and drive system selection selection and sizing

123 1983 755

Agitated tank leaching compared with heap leaching

1634

and counter current decantation (CCD) thickeners history

1631 1632

and Merrill-Crowe zinc precipitation recovery process process selection and design

1631 1631

Air jigs

990

Air tables

989

Aluminum

1924

AMMTEC

2211

Anglo American Research (AARL) elution process

1694

Arctic grinding and dewatering studies (Falconbridge Raglan Concentrator) Arsenic removal from wastewater

2394 1847

1857

Assay fire assaying

1778

laboratories

2011

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Australian Minerals Industry Research Association

462

Autoclaves Barrick Goldstrike acidic autoclave

1530

1538

pressure oxidation and pressure leaching autoclaves

1510

Autogenous grinding mills. See AG mills Automated mineralogical analysis

270

B BacTech Environment Corporation

1547

1548

Ball mills characteristics

698

fine and ultrafine grinding

783

following SAG mills

801

motor and drive system selection selection factors Barrick Goldstrike Mines

1983 721 1530

acidic autoclave

1530

oxygenated roaster

1535

two-stage roaster

1493

Batu Hijau case study

745

1538

2211

Belt conveyor systems calculations

1449

components

1447

types and selection criteria

1446

Bench-scale and pilot plant testing automated mineralogical analysis

270

in cyanide leach circuit design

251

in comminution circuit design

123

in filtration circuit design

207

in flotation circuit design

145

in gold- and copper-recovery circuit design

264

in gravity concentration circuit design

160

locked cycle testing

150

This page has been reformatted by Knovel to provide easier navigation.

1550

Index Terms

Links

Bench-scale and pilot plant testing (Cont.) in magnetic concentration circuit design

176

overview

119

in selection of flotation reagents

136

in selection of pre-oxidation process to enhance gold recovery in thickening and clarification circuit design

211 201

BHPBilliton

2315

Bidding

2223

Bin selection

1478

Biooxidation

1540

aerated, stirred-tank

1547

in heap leaching of sulfidic-refractory gold ore

1562

principles

1541

BIOX

1547

Bond Work Index BRGM

Bubble-surface-area flux

1548

1550

1548

1550

544 1547

BRUNO Crushing Plant Simulator

1565

404 1187

Bullion determination by spectroscopy and other instrumental methods

1797

fire assaying

1778

platinum group metals

1760

production and refining

1747

C Cadmium removal in bullion production

1755

Carbon reactivation

1680

Carbon-in-columns processing. See CIC, CIL, and CIP processing Carbon-in-leach processing. See CIC, CIL, and CIP processing

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Carbon-in-pulp processing. See CIC, CIL, and CIP processing Case studies Batu Hijau

2211

coal preparation plant control

2158

Empire Mine high pressure grinding roll

2380

Falconbridge Raglan Concentrator arctic grinding and dewatering studies

2394

PT Freeport Indonesia SAG mill

2367

Solid–solid separation process control

2158

Sunrise Dam Gold Mine start-up

2345

Centrifugal separation applications

1275

calcium carbonate

1285

mills

783

separators

953

terminology

1266

theory of

1261

types of centrifuges

1267

CIC, CIL, and CIP processing circuit design and equipment selection

1652

compared with agitated tank leaching

1631

selecting among CIP, CIL, and CIC processing

1644

Circuit surveys Clarification

1635

63 201

Coal centrifugal separation

1275

flotation separation

1171

and heavy-media cyclones

962

heavy-media separator sizing and selection

1011

preparation plant control case study

2158

Comminution autogenous and semi-autogenous grinding mill selection and sizing

755

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Comminution (Cont.) bench-scale and pilot plant testing for circuit design

123

Bond Work Index

544

circuit flowsheets

579

circuit sampling

63

crusher types

566

crushing plant design and layout

669

factors in circuit selection

539

grinding equipment types and circuit flowsheets

698

grinding mill design

840

grinding mill drive selection and evaluation

819

grinding mill selection

710

grinding plant design and layout

801

high pressure grinding roll selection and sizing

636

in-pit crushing design and layout

606

JKSimMet comminution circuit simulator

442

pebble crusher selection and sizing

628

primary crusher selection and sizing

584

sampling for process design

99

secondary and tertiary cone crusher selection and sizing

621

ultrafine and stirred grinding mill selection and sizing

783

Compartment mills

731

Computational fluid dynamics

495

in design of mechanical flotation equipment Computer Integrated Manufacturing model

1730

1189 2052

Concentrate Falconbridge Raglan Concentrator dewatering, storage, and transportation studies handling

2400 1474

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Cone separators

987

Construction materials aluminum

1924

copper and copper alloys

1925

elastomers

1932

irons

1913

masonry and membrane linings for process vessels

1962

metallic

1911

nickel alloys

1923

plastics

1953

stainless steels

1914

steels

1912

titanium

1926

Construction projects

2250

Contracts

2245

bonuses and penalties

2262

Fixed Price

2246

Lump sum

2261

Lump Sum Turn Key

2246

and performance tests

2286

and ramp up

2286

Reimbursable Cost Plus Fixed Fee

2246

Reimbursable Cost Plus Incentive Fee

2246

Reimbursable Cost Plus Percentage Fee

2246

types

2262

2249

2248

2248

Copper as construction material

1925

flotation separation of copper gold ores

1124

heap leach design and practice

1571

recovery by solvent extraction and electrowinning Corrosion control

267

1709

1962

This page has been reformatted by Knovel to provide easier navigation.

1927

Index Terms

Links

Cost estimations capital costs

314

2254

Class I (order of magnitude)

310

315

Class II (preliminary)

315

Class III (definitive)

315

cost indexes

310

equipment costs

310

O’Hara method

317

process operating costs

326

terminology

314

Costs heap leaching

1627

management and control

2230

weighted average cost of capital

351

Counter current decantation (CCD) thickeners

1631

Crowe, T.B.

1665

Crushers and crushing

566

578

2114

2122

crusher types

571

588

factors in crusher selection

567

in-pit crushing design and layout

606

control strategies

instrumentation strategies

2114

plant design and layout

669

selection and sizing of pebble crushers

628

selection and sizing of primary crushers

584

selection and sizing of secondary and tertiary cone crushers Crystal Ball software

621 341

Cyanide leaching bench-scale and pilot plant testing in circuit design

251

bench-scale and pilot plant testing in gold recovery from cyanide solutions

264

and Merrill-Crowe process

1663

process chemistry

1666

1666

This page has been reformatted by Knovel to provide easier navigation.

327

Index Terms

Links

Cyanide leaching (Cont.) solution characteristics Cyanide removal from solutions and slurries

1671 1866

D Davis-Ritchie equations

1889

Davy BB

1716

Demonstration plants

121

Design criteria

3

format

14

requirements sample excerpt sample index

9 18 7

Discounted cash flow method

346

Discrete element methods

495

Discrete grain breakage

495

Drill mud

1285

Dry separation

954

989

Drying equipment selection criteria

1463

storage

1476

types

1466

Dual laminate

1953

Dutch State Mines

963

Dynawhirlpool

962

E Elastomers

1932

defined

1932

types

1934

uses

1941

Electric power distribution systems

1973

Electrical separation methods

1049

1947

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Electrowinning in copper heap leaching

1571

gold

1699

plant sizing and selection for copper EW

1734

and refining

1748

1755

267

1709

and solvent extraction in copper recovery Elution Anglo American Research (AARL) process

1694

in carbon reactivation

1683

gold

1694

Zadra process

1697

Empire Mine high pressure grinding roll case study

2380

Engineering company selection

2306

engineering and construction (E) firms

2250

Enhanced gravity separation

988

Enterprise dynamic simulation models

528

Enterprise resource management systems

507

Environmental factors in facility siting

1902

Equipment specification and bidding

2223

F Facilities and support systems. See also Construction materials electric power distribution systems

1973

environmental and social factors in siting

1902

metallurgical and assay laboratory equipment design, and layout

2011

motor and drive system selection for comminution circuits

1983

on-line composition analysis of mineral slurries research and development laboratories

2020 121

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Falconbridge Raglan Concentrator arctic grinding and dewatering studies Feasibility studies final

2394 281 295

and prefeasibility studies in operating cost estimates

328

preliminary (intermediate economic evaluations) preliminary evaluations Feeder selection

288 282 1486

Filtration bench-scale and pilot plant testing in circuit design

207

and centrifugal separation

1262

equipment testing, sizing, and specifying

1313

equipment types and design features

1342

filter media

1355

pressure

1348

principles and theory for equipment characterization

1289

theory

1314

vacuum

1343

Finance. See also Contracts discounted cash flow method

346

project financing

371

risk-reward balance

371

Fine particle separators

952

Fire assaying Fisher’s F-test Fixed Price contracts Flocculation

1778 54 2246

2248

201

Flotation

2174

Batu Hijau kinetic modeling

2211

bench and pilot plant programs in circuit design

145

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Flotation (Cont.) circuit sampling

63

column flotation

1239

complex sulphide ores

1097

control strategies

2177

copper gold ores

1124

design of mechanical flotation equipment

1179

equipment selection

1204

nickel ores

1142

nonsulfide minerals

1159

plant layout

1218

reagent selection via batch flotation tests

136

Fluor Corporation

2211

FOXCOM protocol

2086

G Gates

1489

GENCOR S.A. Ltd.

1547

Gold Barrick Goldstrike acidic autoclave

1530

Barrick Goldstrike oxygenated roaster

1535

Barrick Goldstrike two-stage roaster

1493

1538

bench-scale and pilot plant testing in selection of pre-oxidation process biooxidation chemistry and tests

211 238

biooxidation in heap leaching of sulfidic-refractory gold ore

1562

bullion production and refining

1747

determination by spectroscopy and other instrumental methods

1797

electrowinning

1699

elution

1694

fire assaying

1778

flotation separation of copper gold ores

1124

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Gold (Cont.) gravity recovery heap leach design and practice pressure oxidation chemistry and tests pressure oxidation control strategies

995 1606 231 2196

recovery by activated carbon adsorption (bench-scale and pilot plant testing)

264

recovery by zinc dust cementation (bench-scale and pilot plant testing)

266

roasting chemistry and tests

219

sampling theory and practice

77

selecting among CIP, CIL, and CIC processing

1644

Gravity separation Air jigs

990

air tables

989

applications

955

basic technology

947

bench-scale and pilot plant testing in circuit design

160

centrifugal separators

953

cone separators

987

density separators

932

dry separation

954

enhanced

988

fine particle separators

952

in gold recovery

995

jigs

983

non-heavy media

978

rising current washers

986

selection of gravity classifiers

867

shaking tables

952

sluices

951

spiral concentrators

867

989

985

929

985 wet separation

983

This page has been reformatted by Knovel to provide easier navigation.

952

Index Terms

Links

Grinding

2130

control objectives and elements

2131

control strategies

2147

Empire Mine high pressure grinding roll case study

2380

equipment types and circuit flowsheets

698

expert systems and advanced controllers

2146

Falconbridge Raglan Concentrator studies

2395

high pressure grinding roll selection and sizing

636

mill design

840

plant design and layout

801

regulatory control loops

2140

selection and evaluation of mill drives

819

selection and sizing of autogenous and semi-autogenous mills

755

selection and sizing of ultrafine and stirred grinding mills supervisory control Gy, Pierre

783 2143 77

sampling constant

47

Sampling Theory

77

H HART protocol

2085

Heap leaching with biooxidation for sulfidic-refractory gold ore

1562

compared with agitated tank leaching

1634

copper heap leach design and practice

1571

costs

1627

defined and described

1607

engineering design

1584

history

1571

1606

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Heap leaching (Cont.) irrigation (sprinkling) systems

1596

and ore types

1610

pads and liners

1585

1614

1618

1619

precious metal heap leach design and practice

1606

preliminary evaluations and testing

1572

reasons for using

1609

stacking and reclaiming

1459

1623

1847

1857

Heavy metal removal from wastewater Heavy-media cyclones

962

Heavy-media separators

962

sizing and selection of Heterogeneity Test High fidelity simulation

1612

1011 77 397

495

High pressure grinding rolls Empire Mine case study selection and sizing of High-gradient magnetic separator (HGMS) Hindered settlers

2380 636 1087

1091

929

HIOX

1550

Holmes & Narver pumper

1716

Homestake Mine

1644

Hopper outlet sizing

1483

HPGR. See High pressure grinding rolls Hydraulic classifiers. See Hindered settlers Hydrocyclones

880

I In-pit crushing and conveying INCO Canada

606 1760

1767

1868 Instrumentation. See also Process control control devices

2109

crusher instrumentation strategies

2114

This page has been reformatted by Knovel to provide easier navigation.

1770

Index Terms

Links

Instrumentation. See also Process control (Cont.) measurement devices Internal rate of return

2106 348

IPCC. See In-pit crushing and conveying Iron Ore Company (Canada) Irons

939 1913

J Jet mills

783

Jigs

983

JKSimFloat flotation simulator

461

JKSimMet comminution circuit simulator

442

Julius Krutschnitt Mineral Research Centre

442

462

K Kaolin centrifugal separation

1281

flotation separation

1168

L Laboratories metallurgical and assay (equipment, design and layout) research and development

2011 121

LARCODEMS. See Large Coal Heavy-media Separator Large Coal Heavy-media Separator

962

Leaching agitated tank leaching selection and design

1631

bench-scale and pilot plant testing in cyanide leach circuit design

251

biooxidation in heap leaching of sulfidic-refractory gold ore copper heap leach design and practice

1562 1571

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Leaching (Cont.) heap leach stacking and reclaiming

1459

precious metal heap leach design and practice

1606

pressure leaching autoclaves

1510

Lead

1097

LIGHTNIN hydrofoil

1716

LKAB Iron Ore Mine (Sweden)

1623

936

Lonrho Refinery

1774

Lump Sum Turn Key contracts

2246

2248

1069

1093

M Magnetic separation bench-scale and pilot plant testing in circuit design

176

high-gradient magnetic separator (HGMS)

1087

integral

1074

tramp metal removal

1071

1091

wet high-intensity magnetic separator (WHIMS)

1087

Maintenance predictive

2080

process control

2073

and start-up

2315

2104

Masonry and membrane linings for process vessels

1962

Matthey Rustenburg Refinery

1772

MEMS technology

2080

Mercury retorting

1750

Merrill, C.W.

1665

Merrill Crowe process

1663

and agitated tank leaching

1631

compared with CIP, CIL, and CIC processing

1644

in gold recovery

1646

266

history

1663

and refining

1748

1756

This page has been reformatted by Knovel to provide easier navigation.

1665

Index Terms

Links

Metso CBT software

2330

Metso Minerals Mica

404 1162

MicroElectro Mechanical systems. See MEMS technology Mine Environment Neutral Drainage

1887

MINEWALL 2.0

1889

Mining-influenced waters (MIWs)

1886

1889

1899. See also Wastewater prediction techniques

1894

Model Predictive Control

2147

Modeling and simulation Batu Hijau flotation model

2211

BRUNO Crushing Plant Simulator

404

computational fluid dynamics

495

Computer Integrated Manufacturing model

2052

in design of mechanical flotation equipment

1179

discrete element methods

495

discrete grain breakage

495

empirical

383

enterprise dynamic simulation models

528

enterprise resource management systems

507

fundamental

383

hierarchy

383

high fidelity simulation

397

history of mineral processing simulation

384

JKSimFloat flotation simulator

461

JKSimMet comminution circuit simulator

442

mining-influenced waters

1888

Model Predictive Control

2147

overview

383

phenomenological

383

PlantDesigner crushing and screening program

421

population balance models

393

USIM PAC 3.0 simulator

479

389

495

495

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

ModL programming language

423

Mount Isa Inlier (Australia)

1097

N National Institute of Metallurgy (South Africa)

1774

NEMA 4 and 4X enclosures

2105

Net present value

348

NET technology

2080

2081

Newmont Mining Corporation

2196

2211

Nickel

1142

Nickel alloys

1923

Nonsulfide minerals

1159

O OCA. See On-line composition analysis of mineral slurries OCA-XRF. See X-ray fluorescence On-line composition analysis of mineral slurries methods

2020

2046

2021

Operating manuals

2292

Operator training

2290

computer based Overflow

2328

2334 868

P Pebble mills Performance testing

733

748

2286

2289

PGMs. See Platinum group metals Phosphate

1172

Photometric ore sorting

1033

Pilot plant testing. See Bench-scale and pilot-plant testing PlantDesigner crushing and screening modeling program Plastic construction materials

421 1953

This page has been reformatted by Knovel to provide easier navigation.

2337

Index Terms

Links

Platinum group metals bullion production and refining

1760

defined

1760

determination by spectroscopy and other instrumental methods fire assaying Population balance models

1797 1778 393

495

Potash centrifugal separation

1278

flotation separation

1175

Pre-commissioning

2277

Pre-oxidation Barrick Goldstrike acidic autoclave

1530

Barrick Goldstrike oxygenated roaster

1535

1538

Barrick Goldstrike two-stage gold ore roaster

1493

biooxidation (bioleaching) equipment and circuits

1540

pressure oxidation and pressure leaching autoclaves

1510

Precious metals. See Gold, Platinum group metals, Silver Pressure leaching

1510

Pressure oxidation autoclaves chemistry and tests process control strategies Process control

1510 231 2196 2051

2063

See also Instrumentation asset management

2080

control devices

2109

crusher control strategies

2114

2122

DCS

2090

2093

design

2066

2075

This page has been reformatted by Knovel to provide easier navigation.

2078

Index Terms

Links

Process control (Cont.) flotation instrumentation and control strategies

2174

FOXCOM protocol

2086

grinding instrumentation and control strategies

2130

hardware

2060

hardware selection

2077

HART protocol

2085

hybrid

2090

instrumentation

2057

level 1 communications

2084

level 2 communications

2087

level 3 and 4 communications

2088

maintenance

2073

measurement devices

2106

MEMS technology

2080

Model Predictive Control

2147

NEMA 4 and 4X enclosures

2105

.NET technology

2080

PLC

2090

predictive maintenance

2080

pressure oxidation control strategies

2196

process

2056

process management

2053

production management

2053

SEVA technology

2080

2096

2104

2081

2067

2081

Solid–solid separation instrumentation and control strategies

2152

strategies

2060

system selection

2070

system types

209

thickening instrumentation and control strategies users

2164 2063

PT Freeport Indonesia SAG mill case study

2367

This page has been reformatted by Knovel to provide easier navigation.

13

Index Terms

Links

PT Newmont Nusa Tenggara

2211

PT Pukuafu Indah

2211

PYROX

1889

Q QEM*SEM

28

Qualitative Evaluation of Materials by Scanning Electron Microscopy. See QEM*SEM Quartz/feldspar Quebec Cartier Mine

1164 929

R Raglan Concentrator arctic grinding and dewatering studies Ramp up

2394 2285

Regrind mills

735

Reimbursable Cost Plus Fixed Fee contracts

2246

Reimbursable Cost Plus Incentive Fee contracts

2246

Reimbursable Cost Plus Percentage Fee contracts

2246

Research and development laboratories

121

Rising current washers

986

748

Risk analysis conventional sensitivity analysis

339

in feasibility studies

288

Monte Carlo technique (Crystal Ball software)

341

risk components in mineral project

351

risk factors in project financing

373

risk-reward balance

371

Risk and contracts Rod mills

306

2245 712

743

S SAG mills bench-scale and pilot plant testing for circuit design

123

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

SAG mills (Cont.) equipment types and characteristics

698

fOllOWed by ball mills

801

motor and drive system selection

1983

PT Freeport Indonesia case study

2367

selection and sizing

755

Sampling central values

40

for comminution processes

99

for feasibility studies

25

Fisher’s F-test

54

gold

77

Gy’s sampling constant

47

Heterogeneity Test

77

interleaving sampling protocol

56

large comminution and flotation circuits

63

in mineral processing

37

for mineral separation processes

99

presentation of results

35

spatial dependence

43

Student’s t-test

50

terminology

38

types of

26

variance

41

61

45

Sand centrifugal separation of tar sand

1281

process control for magnetic separation of mineral sands

2160

Sandvik Rock Processing AB

421

Schedule development and control

2238

Screening coarse

894

fine

917

This page has been reformatted by Knovel to provide easier navigation.

58

Index Terms

Links

Sedimentation bench-scale and pilot plant testing

201

centrifugal

1262

equipment testing, sizing, and specifying

1295

equipment types and design features

1331

theory

1295

Selenium removal in bullion production

1755

removal from wastewater

1847

1853

1857

Semi-autogenous grinding mills. See SAG mills Separation processes Settling rate

99 868

SEVA technology Shaking tables

2080

2081

952

985

Silver bullion production and refining

1747

determination by spectroscopy and other instrumental methods

1797

fire assaying

1778

flotation separation of zinc–silver–lead ores

1097

heap leach design and practice

1606

Simulation. See Modeling and simulation Size separation coarse screening

894

fine screening

917

gravity classifiers

867

hydrocyclones

880

over flow

868

settling rate

868

spiral concentrators

867

929

929

985 underflow Sluices

868 951

This page has been reformatted by Knovel to provide easier navigation.

952

Index Terms

Links

Slurry Bingham plastic fluids

1392

1399

Classifications

1392

1427

conventional

1430

conventional tailings

1431

cyanide removal

1866

heterogeneous

1392

launders

1409

Newtonian viscous

1392

non-conventional

1431

non-conventional steam coal slurries

1431

on-line composition analysis

2020

pipeline systems

1422

piping system sizing and design

1403

properties

1426

pumpboxes

1416

thickened tailings disposal

1431

upcomers

1415

valve selection

1418

1395

1417

Slurry pumps applications

1399

bearing assembly

1375

casings

1379

construction materials

1380

hydraulics

1385

impellers

1375

pumpboxes

1416

selection and sizing

1373

settling velocity

1388

specific speed

1378

wear and cavitation

1383

Social factors in facility siting

1902

Soda ash

1279

This page has been reformatted by Knovel to provide easier navigation.

1427

1427

Index Terms

Links

Solid–liquid separation available equipment designs

1256

centrifugal sedimentation and filtration

1262

control architecture and equipment

2169

filtration equipment characterization

1289

filtration equipment testing, sizing, and specifying

1313

filtration equipment types and design features

1342

instrumentation

2167

instrumentation and control strategies

2164

plant design, layout, and economic considerations process characterization

1358 1255

sedimentation equipment testing, sizing, and specifying

1295

sedimentation equipment types and design features

1331

Solid–solid separation

2152

electrical methods

1049

gravity recovery of gold

995

gravity separation

947

heavy-media cyclone

962

heavy-media separator sizing and selection

1011

incremental grade concept

2153

magnetic methods

1069

non-heavy media

978

photometric ore sorting

1033

plant control strategies

2156

process control case studies

2158

Solvent extraction in copper heap leaching and electrowinning in copper recovery

1571 267

McCabe-Thiele diagrams

1709

mixing concepts

1716

settler concepts

1726

1709

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

Specifications

2223

Spiral concentrators

867

929

985 Stacking systems cascading conveyor (“grasshopper”) system

1456

fixed stackers

1455

heap leach stacking and reclaiming

1459

overhead tripper or shuttle

1455

radial stacker

1456

silo storage

1460

stockpile calculations

1461

stockpile types and selection criteria

1454

traveling stackers (stacker/reclaimers)

1457

1623

Start-up and commissioning maintenance scheduling, management, and training

2315

operator training and manuals

2290

2328

and performance testing

2286

2289

planning and staffing

2299

and precommissioning

2277

and ramp up

2285

safety and health considerations

2337

successful start-up defined

2299

Sunrise Dam Gold Mine case study

2345

Steel

1912

stainless

1914

Stillwater Mining

1761

Stirred media mills

783

Student’s t-test

50

Sumitomo Corporation

2211

Sunrise Dam Gold Mine start-up case study

2345

61

This page has been reformatted by Knovel to provide easier navigation.

952

Index Terms

Links

T Tailings co-disposal with other wastes

1817

conventional

1431

dams and impoundments

1828

dewatering

1837

dry

1817

dry cake disposal

1838

1815

Falconbridge Raglan Concentrator dewatering and deposition studies

2404

land disposal (site development and operations)

1809

new developments in disposal

1833

paste

1816

subaqueous disposal

1835

thickened tailings disposal

1431

Tar sand

1281

Thermal reactivation

1683

Thermoplastic

1953

Thermoset

1953

Thickened tailings disposal

1431

1839

1815

1839

1815

1839

Thickening bench-scale and pilot plant testing in circuit design

201

control architecture and equipment

2169

counter current decantation (CCD)

1631

instrumentation

2167

instrumentation and control strategies

2164

Titanium

1926

U Under flow

868

USIM PAC 3.0 simulator

479

This page has been reformatted by Knovel to provide easier navigation.

Index Terms

Links

W Wasp, Edward J.

1422

Wastewater. See also Mining-influenced waters (MIWs) arsenic removal

1847

1857

heavy metal removal

1847

1857

selenium removal

1847

1853

Weighted average cost of capital Wet high-intensity magnetic separator (WHIMS) Wet separation

351 1087 983

X X-ray fluorescence

2023

Z Zadra elution process

1697

Zinc cementation or precipitation. See Merrill Crowe process flotation separation of zinc–silver–lead ores

1097

This page has been reformatted by Knovel to provide easier navigation.

1857