AUTOMATED MINERALOGY FOR PROCESS PLANNING AND OPTIMISATION, QUALITY CONTROL, AUDIT STUDIES AND TROUBLE SHOOTING Patricio Jaime, Paul Gottlieb, Alan Butcher & René Dobbe FEI Australia
ABSTRACT This paper reviews the relevance and impact of scanning electron beam-based Automated Mineralogy technologies on understanding, problem-solving and improved efficiency at different stages of the mining cycle. Two main technologies dominate the market, namely MLA and QEMSCAN®. These analysers have advanced rapidly with improved SEM and X-ray detector hardware, and the development of automated processing and data presentation software has had a revolutionary effect on mineral processing. Previously, using manual methods, it was not feasible to attempt this work because the large data sets required could not be assembled in a realistic timescale. The speed, reliability, and repeatability of the present automated measurements have now made this type of analysis routine. Automated Mineralogy, linked to geometallurgy, aids in quantifying the variability of an ore deposit in terms of process parameters such as hardness, liberation, flotability, and leach response. The combined geological and metallurgical data are applied to individual block models and mine plans. Geometallurgical model outputs are forecasts on economic parameters like target grind, throughput, concentrate grade and metal recovery. Optimization of plant performance with respect to ore variability and effective mining and processing over the entire mine life are now achievable. From a practical point of view, the improved technology has reduced Automated Mineralogy analysis time considerably and allows quick evaluation of composite feed, concentrate and tail samples with regard to grain sizes and liberation characteristics of Base Metal-bearing minerals, their associations with Fe-sulphides, and silicate gangue, and characterisation of losses to the tails, enabling necessary future adjustments of key plant parameters.
CHAPTER 5
INTRODUCTION Automated Mineral Analysers were originally developed as diagnostic metallurgical tools in mining to aid in improving the efficiency of mineral processing plants using samples drawn from plant surveys and pilot-scale tests. They were used mainly for the assessment and auditing of size fractions of feed, concentrates and tailings from base and precious metal ores [1]. Subsequently, they began to be used for concentrator design and optimisation where ore characterisation was used to help understand the relationship between the feed ore and its subsequent behaviour in the plant (see Figure 1).
Figure 1: Relationship between the feed ore and the plant.
In recent years, Scanning Electron Beam-based quantitative mineralogy tools have advanced rapidly with improved SEM and X-ray detector hardware, and the development of sophisticated and automated image analysis methods. Automated Mineralogy has now established itself as an essential enabling technology for the reliable acquisition of statistically valid mineralogical data from particulate samples, sections of rock and drill core. This has had a revolutionary effect on the industrial use of such data in the study of geology, mining and mineral processing. Previously, using manual methods, it was not feasible to attempt this work because the large data sets required could not be assembled in a realistic timescale. The speed, reliability, and repeatability of the modern automated measurements have now made this type of analysis routine. More than 150 Automated Mineral Analysers have been installed around the world and consist of MLA (Mineral Liberation Analyser) and QEMSCAN® (Quantitative Evaluation of Minerals by Scanning Electron Microscopy), now both part of FEI Company.
BASIC PRINCIPLES OF AUTOMATED MINERALOGY System configuration
An Automated Mineral Analyser typically consists of a Scanning Electron Microscope (SEM) equipped with multiple Energy Dispersive X-ray (EDX) detectors. The analyser is typically used as a complementary technique to optical and X-ray Diffraction and the Electron Probe Micro Analyser. An Electron Backscatter Diffraction (EBSD) detector and a Wavelength Dispersive Spectrometer (WDS) may be added for specific requirements. Automation software controls the SEM hardware to quantitatively analyze mineral and material samples. The system has the ability to measure up to 16 sample blocks without the need for operator assistance. Different approaches to mineral identification have been used in electron microscope-based systems, ranging from BSE (Backscattered Electrons)
314
automated mineralogy for process planning...
based systems to X-ray dominated systems [2, 3, 4]. Although both MLA and QEMSCAN® (formerly QEM*SEM) technologies have uniquely evolved from rather distinct philosophies, the principles behind them are essentially the same in that they use backscattered electron image analysis and X-ray mineral identification to provide automated quantitative mineral characterization. Automated stage control and image acquisition allows for BSE imaging and subsequent X-ray analysis of several thousand particles within the time span of around one hour, depending on sample-type and mineral texture. The MLA exemplifies a BSE based system where the information obtained from the acquisition of a backscattered electron image is fundamental to the nature of the follow-up X-ray analysis [5]. The QEMSCAN® is optimized for X-ray throughput and proves to be beneficial on samples that require full X-ray mapping [6, 7]; also see Figure 2 for the process of data acquisition.
Figure 2: Data acquisition schematic for QEMSCAN.
Automated Mineral Analysers – Data Acquisition Technologies
Automated Mineral Analysis involves the setting of a representative set of particles into a mould (typically 30 mm diameter) with epoxy resin to form a hardened block. Typical particle sizes range from 5 μm to 3 mm and should preferably be of a defined narrow size fraction. The block is then ground down to expose a representative cross-section of particles, which is subsequently polished, and then coated with carbon before being presented to the SEM (see Figure 3).
Figure 3: Sample presentation. Multiple sample holders, thin section holders, custom holders.
P RO C E MI N 2009. Santiago, Chile | 315
CHAPTER 5
Various factors come into play when deciding on the measurement parameters for a particular analytical run using an Automated Mineral Analyser. In virtually all cases, fast time-todata requirements have to be counterbalanced against the need for high image resolution from BSE acquisition, as well as chemical energy resolution from X-ray collection. The effects of pixel resolution to mineral classification results are shown in Figure 4. The choice of image and pixel X-ray resolution is user-defined and depends on application requirements.
Figure 4: QEMSCAN measurement of a single particle at a range of pixel spacing settings.
The modal data are reproducible across all pixel settings when the entire sample is measured (117mm2); however run-time is drastically reduced at larger pixel spacings. Other textural features such as grain size are better represented at finer pixel spacings.
EXAMPLES OF THE USE OF AUTOMATED MINERALOGY IN THE MINING INDUSTRY Ore Characterization of Base Metal Ores
The mineral and textural variability of an ore deposit, and hence the feed to a mineral processing plant, has implications for concentrate grades and recoveries produced by that plant. The deportment of deleterious and precious elements in that feed is also important for minimization of environmental impacts and the maximization of profits. Quantitatively capturing the variations of important ore characteristics is the aim of ore characterisation and can be effectively captured using Automated Mineralogy. Gu and Burrows [8] describe a coarse particle ore characterisation measurement technique. Here the key to effective ore characterization is the selection of a particle size for analysis that preserves the original textures of the sampled material. Figure 5 illustrates the coarse particle analysis at a particle size around 600 microns. The purpose of the measurement is to characterize the ore using values such as, modal mineralogy, elemental deportment, phase size and association of the ore samples (e.g. Cu). Verification that the particle size being analysed is representative of the entire ore sample is essential and can be done by comparing external chemical assay data with the chemistry calculated by the coarse particle analysis.
316
automated mineralogy for process planning...
Figure 5: Ore characterisation – Coarse-particle measurements.
Precious Metal Search / Trace Mineral Search
Mineralogical examinations of tail samples can simply have the aim of characterising only grains of the metals ores of interest. Searching for very few bright grains within a very large population of dark particles is ideal task for automated mineral analysers. The Rare Phase Search (RPS) analysis mode employed by the MLA searches the BSE images for bright phases of interest (e.g. Cu Sulphides, Au, PGM) using a BSE or spectral trigger and collects a corresponding characteristic X-ray spectrum. For each grain found, the system saves the image of the particle containing the grain, the stage location and its X-ray spectrum. The operator can subsequently move to the SEM stage location where the grain was located for further investigation. RPS is designed to efficiently locate very fine (sub-micron) components in large particle populations, such as gold in tailings and deliver data such as grain size and associated minerals (see Figure 6). Quantitative information is obtained for liberation and associations. The ability to classify off-line allows the operator to automatically eliminate other bright phases, such as galena, from the phases of interest. The overall analytical technique for precious metal search has been described by Gu for MLA and by Gottlieb et al. for QEMSCAN® [5, 6].
Figure 6: X-ray image of very small grains of Au, Platinum Group Mineral (PGM).
Modern Mining
Recently, mining leader and Automated Mineralogy pioneer Anglo Platinum, South Africa, was the first to submit even faster but well contextualized time-to-data requirements for automated mineralogy instruments at their operational sites for concentrator process management applications [9]. The central laboratory at Anglo Platinum routinely
P RO C E MI N 2009. Santiago, Chile | 317
CHAPTER 5
monitors plant performance of various operations and rapidly assists in the detection of potential inefficiencies. There is great time-to-data benefit if this work can be performed at the operational mining site. Acquired data at Anglo Platinum shows that the mode of the mineral occurrence varies with geological setting and reef type. Knowledge of the complex mineralogy and liberation characteristics of the plant feed makes it possible to operate the plant more efficiently. The faster the analytical turnaround, the quicker the response time to changes in feed mineralogy and thereby improvement of overall recovery. Frequent analysis of the feed as part of concentrator process management may also help to better resolve the background noise of the plant performance from actual variations in mineralogy. For operational mining site requirements, fast time-to-data results from Automated Mineral Analyzers make it possible to set targets for the optimization of processes at the concentrator plant within the time span of a single 8-hour shift on the mining operations [10] (see Figure7). In this figure, instrument set up and data reporting included. Sample preparation time excluded.
Figure 7: Generalized simulation of analysis possibilities during an 8 hour shift.
TWO CASE STUDIES Case 1: Candelaria Concentrator 2000
A concentrator survey combining QEMSCAN® data and metallurgical test work was performed by Phelps Dodge (now Freeport McMoRan) to quantify the influence of copper and gangue mineralogy during normal plant operation, identify specific optimization areas of the circuit and establish a size by size mineralogy of the circuit streams. [11] The QEMSCAN findings showed that Chalcopyrite Liberation characteristics exhibit differences between the different ore types having also an impact in the comminution circuit, defining middling generation, in this case insoluble content in the final concentrate and losses of ultrafine chalcopyrite to the rougher tails. Ore types rich in quartz and magnetite reduce the efficiency of the chalcopyrite liberation. Silicate gangue (chain and phyllosilicates) and carbonates occur in the fine to ultrafine grain size range and represent the major causes for sliming during primary grinding and re-grinding. The Candelaria concentrator exhibited typical metallurgy during the survey period and averaged 95.3% Cu recovery. Losses to tailings averaged 4.7%, of which 0.5% was lost from the scavenger circuit and 4.2% from the rougher circuit. The degree of liberation of chalcopyrite and gangue, mainly in the -75 µm size fraction, provides quantitative evidence that further upgrading of the final concentrate is possible 318
automated mineralogy for process planning...
(see Figure 8A). The improvements in Cu recovery then need to target the 4.2% lost from the rougher circuit. The size by size liberation of chalcopyrite in the rougher tailings shows 2 populations (see Figure 8B); in coarse (>75 µm) particles containing fine (25 to 30 µm) disseminated chalcopyrite (58% of the Cu losses) and fine (<20 µm) liberated chalcopyrite. Liberated particles account 31% of the Cu lost to the rougher tailings and represent a recovery target of 1.3%.
Figure 8: Size by size liberation of chalcopyrite in the rougher Feed and rougher tailing stream [11].
It was clear for this analysis that the fine fractions offer the best target for improved recovery because these fractions are dominated by liberated chalcopyrite particles. Since flotation kinetics are responsible for the losses of liberated particles, mineral processing solutions need to address over grinding in the primary grinding circuit and improved fine particle flotation circuits. The results of the QEMSCAN analysis combined with metallurgical tests presented an opportunity to identified the loss of grained chalcopyrite in the rougher tailings, demonstrated opportunities for better control liberation in specific rock types, clarified the need to improve circuit performance and offered the chance to reduce operating costs by decreasing lime consumption. Kendrik et al., (2003) reported that, after adjustments made based on the quantitative assessment of the circuit’s, size by size mineralogy combined with metallurgical testing, an improvement in gold recovery by 10% was achieved, together with reduced losses to copper tailings by 16%, and finally reduced lime consumption by 72%, all of which resulted in improved cash flow.[11] Case 2: The Falconbridge Montcalm Concentrator 2005
The Montcalm Ni/Cu ore is processed at the Kidd Creek concentrator in Northern Ontario. The data generated from automated mineralogy had direct input into plant design, audit and recovery forecasting. In the early stages of feasibility and pre feasibility studies carried out by Falconbridge (now Xstrata) an Ore Characterization involving QEMSCAN (2002) measurements of selected drill core and sized feed samples provided textural information that detected Pyrite dilution of the Ni concentrate. Some problematic textures were flagged by QEMSCAN like Pyrite form and content (see Figure 9A). Pentlandite occurs as coarse grains and flames locked with pyrrhotite (see Figure 9B). Chalcopyrite presents average grain sizes finer than pentlandite and is highly associated with silicate gangue which would result in Cu losses to tails (see Figure 9C). These issues were later substantiated during the early stages of operations and during a Benchmark Survey conducted in 2005. [13]
P RO C E MI N 2009. Santiago, Chile | 319
Figure 9: Typical problematic textures; pyrite content; bi modal population of pentlandite grain sizes and fine chalcopyrite in silicates.
The Figure 9 legend includes pyrite (Py, gray), pentlandite (Pt, black), pyrrhotite(Pyrr, dark gray), chalcopyrite (Cpy, white) and silicate gangue (G, light grays). This figure was modified after Charland et al., 2006 [13]. The principal objective of the 2005 Survey was to produce information that would lead to improved performance. A second objective was to predict the impact of coarsening the primary grind. From the Chemical Mass Balance it was suggested that the plant was performing above the project target for Ni and Cu grades and Ni recoveries. However, a loss of 2.5% in Cu recovery has resulted as the primary grind was coarsened from P80 39 µm to P80 53 µm as expected [13]. The QEMSCAN liberation data demonstrated that Ni and Cu losses to final tails are 16% and 14%, respectively (see Figure 10). Final tails were characterized by two populations of metal–hosted particles. They include an ultrafine (CS6 and CS7) population created from over-grinding a well liberated feed and a second population where both pentlandite and chalcopyrite occurred as textures (middlings and locked particles), which suggested an opportunity to improve recovery.
Figure 10: Size by size mineral analysis of re-grind stream. Modified after Charland et al, 2006.
The recommendations were made in order to optimize the grinding circuit performance to minimize fine losses. A suggested regrind or stage of grind approach would be required to tackle the locked particles. All these are currently being implemented with an overall Cu recovery increasing by 1-2%. [13] Xstrata continues use Automated Mineralogy to audit and keep control of the feed and circuit process.
CONCLUSIONS Mineralogy is clearly useful when applied in the correct way. Rocks contain information that traditionally only geologists and mineralogists have been able to unlock. With the
automated mineralogy for process planning...
advent of more companies adopting automated mineralogical techniques, it is clear that mineralogy is experiencing a new phase, that gives the engineers insights that were previously just not possible. Historically, the principal application for automated mineralogy has been in the area of mineral processing plant optimisation by examining plant streams, but today the spread of applications is much broader. Nowadays SEM-based automated mineralogy analysers implement prudent use of both BSE and X-ray signals, in conjunction with advanced image and pattern recognition analysis and metallurgical testwork to successfully provide quantitative mineralogical data for the mining sector, automatically giving fast time-todata results. The use of Automated Mineralogy technology has proven to be a powerful complimentary tool in flowsheet development stages as well as Plant optimization stage. The Candelaria and Montcalm plant surveys represent 2 very successful case studies of this size-bysize circuit audits application. Several revenue and optimization targets can be identified in timing with the operations in weekly, monthly or quarterly basis by using automated mineralogy work making this type of analysis a routine and definitively making this equipment, mine site or central lab based, an essential process mineralogy tool for ore characterization, control and optimisation of grinding and flotation circuits.
REFERENCES Gottlieb, P., Adair, B. J. & Wilkie, G. J. (1994) QEM*SEM Liberation Indices For Grinding Classification and Flotation. Fifth Mill Operators’ Conference. Roxby Downs. [1] Reid, A. F. & Zuiderwyk, M. A. (1975) Qem*Sem: An Interface System for Minicomputer Control of Instruments And Devices. CSIRO. Investigation Report 115, 1975. [2] Jones, M. P. (1982) Designing an X-ray Image analyser for Measuring Mineralogical Data. XIV International Mineral Processing Congress, 1982, Toronto, Canada. [3] Petruk, W. (1984) Image Analysis Measurements and Data Presentation for Mineral Dressing Applications. Proceedings of the Second International Congress on Applied Mineralogy, Los Angeles, California, pp. 127-140. [4] Gu, Y. (2003) Automated Scanning Electron Microscope Based Mineral Liberation Analysis. An Introduction to JKMRC/FEI Mineral Liberation Analyser. Journal of Minerals & Materials Characterization & Engineering, Vol. 2, No.1, pp33-41, 2003. [5] Gottlieb, P., Wilkie, G., Sutherland, D., Ho-Tun, E., Suthers, S., Perera, K. Jenkins, B. Spencer, S. Butcher, A. & Rayner, J. (2000) Using Quantitative Electron Microscopy For Process Mineralogy Applications. JOM. [6] Fandrich, R., Gu, Y., Burrows, D. & Moeller, K. (2007) “Modern SEM-based mineral liberation analysis” - International Journal of Mineral Processing, Vol. 84: 310-320. [7] Gu, Y. & Burrows, D. (2006) Quantitative Ore characterisation, Proceedings Mineral Process Modelling , Simulation and Control. Sudbury, 2006, pp. 217- 232. [8] Dobbe, R, Moeller, K. & Schouwstra, R. (2008) The New Mla 600f: A Breakthrough Feg-Sem Solution for Ultra-High Throughput Applications in Mining. FEI Application Note. [9] Dobbe, R. Gottlieb, P. Gu, Y. Butcher, A. R., Fandrich, R. & Lemmens, H. (2009) Scanning Electron Beam-Based Automated Mineralogy: Outline of Technology And Selected Applications In The Natural Resources Industry. [10] Kendrick, M. Thompson, P. Baum, W. Wilkie, G. & Gottlieb, P. (2003) The Use of QEMSCAN P RO C E MI N 2009. Santiago, Chile | 321
CHAPTER 5
Automated Mineral Analyser at the Candelaria Concentrator. Copper ’03. Santiago, Chile. [11] Baum, W. Lotter, N. O. & Whittaker. (2004) A New Generation for Ore Characterization and Plant Optimisation. SME Annual meeting’03. Denver Colorado, U.S. [12] Charland, A. Kormos, L. Whittaker, P. Arrue-Canales, C. Fragomeni, D. Lotter, N. O., Mackey, P. & Anes, J. (2006) A Case Study for Integrated Use of Automated Mineralogy in Plant Optimization: The Falconbridge Montcalm Concentrator. Proc. Automated Mineralogy 06, MEI Brisbane, Australia. [13]
322
