IP_15_Calculations in Support of IP15-The Area Classification Code for Petroleum Installations_November 2001.pdf

Calculations in Support of IP15: The Area Classification Code for Petroleum Installations

November 2001

By P. T. Roberts OGCH/2 HSE Business Group Shell Global Solutions (UK), Cheshire Innovation Park, P.O. Box 1, Chester CH1 3SH, England Report No. OP.00.47110

Published by The Institute of Petroleum, London A charitable company limited by guarantee

Copyright © 2001 by The Institute of Petroleum, London: A charitable company limited by guarantee. Registered No. 135273, England All rights reserved No part of this book may be reproduced by any means, or transmitted or translated into a machine language without the written permission of the publisher. ISBN 0 85293 339 8 Published by The Institute of Petroleum Further copies can be obtained from Portland Press Ltd. Commerce Way, Whitehall Industrial Estate, Colchester CO2 8HP, UK. Tel: +44 (0) 1206 796 351 email: [email protected]

CONTENTS

Page

Foreword...................................................................................................................... vii Acknowledgements .................................................................................................... viii Executive summary ....................................................................................................

1

1.

Introduction .....................................................................................................

2

2.

Flammability limits for two phase releases ..................................................

3

3.

Shape factors and hazard radii for pressurised releases ...........................

6

4.

Hazard radii from vents .................................................................................. 10 4.1 Vents from the storage of petroleum products ......................................... 11 4.2 Process vents........................................................................................... 20

5.

Evaporation from pools and sumps .............................................................. 22 5.1 Vapour pressure comparisons of some commonly used Category C fluids...................................................................................... 27

6.

Releases into confined areas......................................................................... 29

7.

Discussion and conclusions.......................................................................... 34

8.

References ....................................................................................................... 35

Appendix A: Methodology................................................................................................................ 37 Appendix B: Preliminary investigation of hazard radii and shape factors for the Revision of IP15: the Area classification code for petroleum installations .......... 40

v

vi

FOREWORD

The Institute of Petroleum commissioned this report to address concern over the effect of a release containing droplets or a mist on the dispersion distances determined by the methodology used in IP publication A Risk-Based Approach to hazardous area classification, 1998. The report also reviews flammability limits, evaporation from pools and releases into confined areas. The aim of this publication is to provide a record of the calculations, methodology and assumptions used to calculate dispersion distances. It provides a traceable scientific basis that will be applied to the 2nd edition of IP publication Model Code of Safe Practice Part 15: Area Classification code for petroleum installations 1st edition, 1990. Although it is believed that the adoption of the recommendations of this report will assist the user, the Institute of Petroleum cannot accept any responsibility, of whatsoever kind, for damage or loss, or alleged damage or loss, arising or otherwise occurring as a result of the application of this report.

vii

ACKNOWLEDGEMENTS

This report was prepared by Dr Peter Roberts and Dr Les Shirvill and was reviewed by members of the Institute of Petroleum’s Area Classification Working Group: Phil Cleaver Howard Crowther Kieran Glynn Alan Tyldesley Mick Wansborough

Advantica Technology Consultant (formerly BP) BP Health and Safety Executive Shell

The Institute wishes to record its appreciation of the work carried out by the members of the group. The Institute also wishes to record its thanks to the Health and Safety Executive for cosponsoring this research.

viii

Calculations in Support of IP 15: The Area Classification Code for Petroleum Installations Executive Summary The Area Classification Code for Petroleum Installations published by the Institute of Petroleum (IP 15) offers guidance on the immediate area of hazard associated with the normal processing and handling of petroleum products and is in very wide use. A major revision of IP 15 is being prepared which aims not only to update the guidance based upon best current practice but also to provide a traceable and scientific basis for the guidance given. This latter is not a trivial task and necessarily depends to a great extent on the methodology developed for assessing the consequences of accidental releases on a large scale - much larger than would arise from normal processing and handling. The quantification of hazard necessarily starts with specifying the type of material and the size of release which is very much unknown in the case of small spills and leaks. Material types have been simulated using 5 example fluid compositions coded (A, B, C, G and Gii) following earlier work to update IP 15. Release rate values used here represent the lower end of the “hazardous release” scale. These should be larger than arise in normal handling and certainly should not be taken as indicative of the magnitude of “acceptable” spills. In all circumstances the potential for spills to occur should be rigorously assessed and a full hazard assessment carried out where necessary. This note contributes a methodology and the physical basis for the deriving several guidance parameters relating to: x the characteristics of two-phase releases compared to single phase releases. x the definition of shape factors for pressurised releases of both heavier than air and lighter than air fluids (fluid categories A, B, C, Gi, Gii). x the flammability limits for the fluids used as examples of categories A, B, C, Gi, Gii. x hazards arising from the evaporation of category C fluids. x releases into confined areas. The major findings arising from this work are summarised below. It has generally been possible to defend the key recommendations of IP 15 as conservative. Where revisions are recommended these are strongly dependent on scenario and fluid type. x The hazard radii for pressurised releases of category B and C fluids should be derived assuming a mechanically generated flammable mist; previously gaseous releases were assumed. The Hazard radii for category B and C fluids are increased relative to previous guidance. x Numeric flammability limits published in Annex D of “A Risk-Based Approach to Hazardous Area Classification” for the category B and C fluids have been updated to take account of the composition of the flammable mist; previously low vapour components were assumed to rain-out and not contribute to the lower flammability limit evaluation. x Shape factors for pressurised releases are revised to take better account of the role of initial jet momentum on the jet trajectory. In particular the lighter than air gases (category Gi and Gii fluids) are found to have qualitatively more similar shape factors to the two-

1

x x

x

x

x x

x

phase category A, B and C releases; previously buoyancy was assumed to dominate their dispersion. Hazard radii for discharges from vents are evaluated. The hazard radius varies from slightly smaller to slightly larger than that in the existing guidance depending on the properties of the vented vapour. The composition of vapour from vents on storage facilities maintained at atmospheric pressure may be variable (in composition, density and flammability) and the user of the new guidance should be aware of the effect of this variability because of the consequence for hazard radii. The example range of venting rate and vent sizes used in the guidance are not wholly consistent with the assumption that the discharge takes place at atmospheric pressure. The relationship between venting rate and pressure of discharge is investigated and a value of 300 mb suggested as a threshold above which the consequences of pressure should be assessed. This is of significance for multicomponent fluids where condensation may occur. The existing guidance for liquid spillages is conservative, judged by the volatility of the model category C fluid. The guidance is applicable to materials with approximately twice the vapour generation rate of category C fluid under the specimen conditions. Relative vapour pressures for some common hydrocarbon compounds are listed. The existing guidance for sumps is conservative, judged by the volatility of the model category C. Vapour generation at the source of spillage of category C fluids is a potential hazard dictated by the spill rate and conditions and not the rate of evaporation of the liquid pool. The new guidance should emphasise the role of release conditions in determining the initial vapour generation from spills of category C fluids. For releases into confined areas the relative size of spillage and building are of key importance. The classification “Adequate ventilation” has been assessed with respect to these parameters.

1. Introduction Shell Global Solutions, on behalf of the Institute of Petroleum (IP), has worked to establish a methodology by which certain guidance parameters in the IP 15 document can be calculated from specific scenarios. The benefit is two-fold. Firstly, the existence of a methodology enables the guidance to be independently verified, secondly, it allows an end user to derive specific fluid and process dependent values in a consistent way when required. The methodology closely follows that used for assessing the consequence of hazardous events. Several of the scenarios adopted to illustrate the effect, say on hazard radius, of changing release scenarios were found to produce events that in practice would require a formal assessment of risk; i.e. they fall outwith the definition of normal processing and handling of petroleum products. None of the discharge rates used in this report should be taken as representing normal or acceptable routine practice. This work took place in two stages. A preliminary investigation was carried out for Shell (UK) in order to verify the hazard radii reported in “A Risk-Based Approach to Hazardous Area Classification”, Institute of Petroleum, November 1998 and to see if the shape factors reported in IP 15 were adequate. x the values of the hazard radius for category B and category C fluids in the Risk-Based Approach Document were too small, and the release scenarios unrealistic. x the shape factors for lighter than air gases (category Gi and Gii fluids), and to a lesser extent, the low vapour pressure category B and C fluids needed to be revised.

2

An abridged version of the report of this preliminary investigation is included as Appendix B to this note. The major findings are restated in the body of this note. The implied changes to key values in IP 15 were significant. The Institute of Petroleum requested that Shell Global Solutions pursue 6 work items to verify and quantify the necessary changes for the revision to IP 15 . These were (in short) Work Item 1: To examine results of the AIChE Release modelling program and other recent data and derive, if possible, an improved estimate of the flammability limit for high flash-point releases (e.g. taking account droplet size, rain-out) that would allow the category C hazard radii to be better assessed. Work Item 2: To describe the method used to calculate the hazard radii; define the shape factors for the fluid releases; state the hazard radii. Work Item 3: Cross check the hazard radii for the process vents. Work Item 4: Liquid Pools due to Spillage (section 5.11) To determine the hazard radii/shape factor for shallow liquid pools. Work item 5: Open Sumps and Interceptors (section 5.12) To determine the hazard radii/shape factor for deep liquid pools by taking the steady state evaporation rate and a steady dispersion calculation. Work item 6: Propose a simple ‘low momentum’ calculation method; implement this method and use it to assess the external hazard for releases of category A and category B fluid inside a building and evaluate the hazard radii. Progress on these work items was reviewed. Specific scenarios were discussed and amended in discussion with the Area Classification Committee of the IP to give the results below. These results are in a form suitable for inclusion in the IP 15 revision. A consequence of changes to some scenarios is that numeric values obtained in the original investigation on behalf of Shell (UK)(Appendix B) are changed. Only values from the main body of this note should be transferred to the new guidance. Where possible publicly available and publicly evaluated hazard assessment models have been used in this work.

2. Flammability Limits for Two Phase Releases. A major advance in hazard assessment has been the development of models capable of describing the dispersion of two-phase liquids. The AEROPLUME model is one example, being part of the HGSYSTEM v3.0 (1995) suite of models developed by Shell for industry consortia as publicly available tools subject to peer review and acting as a standard benchmark in model evaluation exercises. Two-phase releases can arise in two ways: x by the atomisation accompanying the expansion and phase change of material that is liquid under storage conditions of high pressure, and gaseous at atmospheric temperature and pressure.

3

x by the mechanical break-up of a (volatile) liquid into small droplets and their subsequent evaporation. Mists of fine droplets and dust clouds of combustible material can be very highly flammable. Unlike a gaseous mixture that is combustible only within narrow flammability limits each droplet can act as a fuel source surrounded by a plentiful air supply. Mists are optically thick and, once ignition has occurred, heat transfer by radiation very effectively preheats droplets/particles distant from the ignition front. In extreme cases and especially for dust clouds, this preheating is sufficient for auto-ignition to take place causing the cloud to burn throughout its volume. This can be a much more vigorous process than a gas cloud fire where a flame-front passes through the mixture. The flammability hazard arising from a two-phase release depends in a complicated way upon the ease of ignition of the fluid, the droplet size distribution and the concentration of droplets and vapour in air. Unfortunately very little is known about precise flammability criteria for mists arising from “real” releases. The summary guidance based upon a review of available literature (Appendix B, Lees, 1998) is that the potential of a mist to burn should be assessed: take all the droplets present in a volume, evaporate them and see if the resulting vapour and air mixture lies within the known vapour phase flammability limits. The easiest way to evaluate this is to use a mass based flammability limit (kg fuel/m3 air) in place of the standard and familiar volume based limit (m3 fuel/m3 air) used for vapours. This definition of a flammability limit: x has NO effect on hazard distances calculated for gaseous mixtures or two-phase mixtures of very volatile componentsi. x has a profound effect on the hazard distances calculated for the category B and category C example fluids resulting in a substantial INCREASE in hazard radius over previous advice based upon volumetric flammability limit values. Not all of the liquid released from a pressurised source might remain airborne. Thus, loss of fluid to the ground through rain-out may mitigate the hazard associated with two-phase releases with a low volatility component. The question of rain-out has been investigated at length by the Center for Chemical Process Safety of the American Institute of Chemical Engineers by means of a series of experiments and an extended modelling exercise. Unfortunately the problem has not been satisfactorily solved. The combined results of this study are reported by Johnson and Woodward (1999) and expressed in software form as a model, called Release. The Release model does not account for all of the physical processes involved in two-phase releases of low volatility materials and is strongly tuned to account for the results of the experiments that were carried out. These experiments aimed to measure the total liquid deposition from pressurised releases of several materials under a limited and non-ideal set of environmental conditions. An obvious concern is that the validity of the model outside of the range of the calibration data is unknown. The model performance is also poor in several respects.

i

This applies to the category A (two-phase), Gi and Gii (gaseous) example fluids used in the IP 15 revision at the reference atmospheric conditions and release conditions therein.

4

Two versions of the model are supplied on CDROM by AIChE; an original model and a corrected model. These have different functionality. The corrected model is intended for end-use and is referred to hereafter in this report. The Release model was reviewed by AEA Technology for the UK Health and Safety Executive (Ramsdale and Tickle (2000)). We conclude from the AEAT report, together with our less detailed investigation of the model, that: x the Release model would over-predict rain-out by a substantial margin for category C fluids. We also believe that the relationship between over-pressure and rain-out is not robustly developed and may not be extrapolated to conditions outside of the tests. With these reservations in mind, but for completeness, we used the Release model to calculate rain-out for cyclo-hexane using the conditions of release temperature, pressure and hole size used in this study for the IP. Cyclo-hexane is the closest to a category C fluid of those tested. We found that: x mass flow rates calculated by the Release model as a function of hole size and pressure were realistic. x rain-out as a fraction of mass-flow rate was independent of hole size and a function of pressure only. The calculated rain-out values were: Pressure (bar) 5 10 50 100

Rain-out (fraction of mass released) 0.98 0.44 0 0

Table 1 Results of the “Release” model for Cyclo-Hexane. x the rain-out fraction did not depend upon the axial location in the jet.(input parameter ZJ) We do not believe that the values in Table 1 are necessarily correct although they confirm our intuition that pressurised releases should become atomised. Further the consensus of the model reviewers is that Release overestimates the amount of rain-out. For the purposes of this work we therefore assume that: x rain-out of liquid from pressurised releases can be neglected in calculating hazard radii. This is in accord with the concluding remarks (5.2) of the AEAT review which suggest that low volatility materials should conservatively be treated as a mist. There must be a lower limit to the drive pressure for which this is true and further work is needed to establish the proper limits. It does seem credible, from intuition and from Table 1, that the 5 bar pressure releases would rain-out. However, if they do not form a flammable mist they would instead form an initially coherent liquid jet. The throw of a liquid jet can be quite substantial and could credibly extend the same distance as we calculate here for dispersed phase jets. A liquid jet of gasoline, say, would present a contact hazard to electrical equipment on exposed surfaces and also form a liquid pool on the ground that will flow away from the source. The result on ignition would be a pool fire rather than a jet-fire or cloud deflagration. For these reasons we retain the hazard radii for Category C fluid down to the 5 bar condition.

5

It is worth noting that, for flammable as opposed to toxic hazards, the Release model prediction of total rain-out does require qualification as to at what distance from the source the rain-out occurs. Clearly to mitigate the hazard this must be smaller than the calculated hazard radius. For toxic hazards (evaluated at a long distance from the release point) only the total released needs to be known. The revision of the IP guidelines is based on model fluids for the five fluid categories. These first appeared in the Addendum to IP 15: A Risk Based Approach to Hazardous Area Classification. The fluid properties are quoted in Table 2. We note: x The vapour phase flammability limits have been re-calculated and differ from those in the risk-based addendum to IP 15. The derivation of those flammability limits is not known. However, we infer that this is because of a changed assumption that all of the released material might participate in a fire. Certainly the previously published values are consistent with the assumption that only the light ends of the mixture would burn and match those calculated for a pseudo-mixture of hydrocarbons smaller than C8. Stream component (mol %) N2 Nitrogen C1 Methane C2 Ethane C3 Propane C4 Butane C5 Pentane C6 Hexane C7 Heptane C8 Octane C9 Nonane C10 Decane H2O Water Carbon Dioxide Hydrogen Average MW LFL (vol %) LFL (kg/m3)

Fluid Cat. A

Fluid Cat. B

Fluid Cat. C

0.00 0.00 0.00 70.00 30.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0.00 4.00 0.00 6.00 7.00 9.00 11.00 16.00 22.00 0.00 25.00 0.00 0.00

0.00 0.00 0.00 1.00 1.00 2.00 3.00 3.00 27.00 25.00 38.00 0.00 0.00

Fluid Cat. G (i) 2.00 88.45 4.50 3.00 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.05 0.00

0.00 48.30

0.00 100.06

0.00 125.03

0.00 18.74

80.00 7.03

2.00 0.039

1.05 0.042

0.86 0.043

4.6 0.034

4.00 0.011

Fluid Cat. G (ii) 2.00 10.00 3.00 3.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00

Comp. LFL (vol %) 5.00 3.00 2.10 1.80 1.40 1.20 1.05 0.95 0.85 0.75 -

28.01 16.04 30.07 44.09 58.12 72.15 86.17 100.20 114.23 128.26 142.28 18.02 44.01

4.00

2.02

MW

Boiling point °C -196 -161 -87 -42 -1 36 69 98 126 151 173 100 -78 (sub) -253

Table 2 Composition of the example category A,B,C, G(i) and G(ii) Fluids and their lower flammability limits (LFL)

3. Shape Factors and Hazard Radii for Pressurised Releases The major findings of this study, compared with earlier guidance are that: x Pressurised releases give rise to an approximately spherical hazard zoneii for all categories of fluid, except where the release comes into ground contact where the hazard distance is extended.

ii

In the context of flammable hazards.

6

x Hazard radii for category B and category C fluids are substantially greater than those quoted in the Addendum to IP 15: A Risk-Based Approach to Hazardous Area Classification. x The increase in hazard radius is due to the redefinition of flammability limit and not to substantial differences in modelling changes. The new shape factors are shown in Figure 1. of the release and the hazard radius.

7

The shape factor depends upon the height

R1

Source

R1

(a) Releases where H > R1 + 1

H

Source

1m

R2

R1

b) Releases where 1 < H d R1 + 1

Source 1m

R2

c) Releases where H d 1 Figure 1 Shape Factors for Pressurised Releases

8

The key features are: x Releases below a height (H) of 1 m are declared to be influenced by the ground and to have a hazard radius R2. x Releases above 1 m, but at heights below the hazard radius R1 + 1 m are declared to be influenced by the ground if the release is directed downward and passes below 1 m. x Releases at a height above the hazard radius R1 +1 m are declared independent of the ground. The Hazard radii are given in Table 3 for the primary radius R1 and in Table 4 for the ground radius R2. For small releases, giving a dimension R1 not substantially larger than 1 m, the radii are similar. The numerical values given in Table 3 and Table 4 are specific to the example fluids. The release rate for these fluids is only weakly dependent upon small variations in the assumed storage temperature about 20 qC, which is chosen to reflect a daily average UK summer temperature. Other fluids may be more sensitive to temperature changes.

Fluid Category

A

B

C

G(i)

G(ii)

Release pressure (bara) 5* 10 50 100 5 10 50 100 5 10 50 100 5 10 50 100 5 10 50 100

1mm 0.01 0.01 0.03 0.05 0.01 0.02 0.04 0.06 0.01 0.02 0.04 0.06 0.001 0.001 0.007 0.015 0.0004 0.001 0.004 0.007

Release flow rate (kg/s)

Hazard radius R1 (metres)

Release hole diameter

Release hole diameter

2mm 0.04 0.06 0.14 0.2 0.04 0.07 0.15 0.2 0.06 0.1 0.2 0.25 0.002 0.005 0.03 0.06 0.001 0.003 0.02 0.03

5mm 0.3 0.4 0.9 1.20 0.30 0.40 1.0 1.4 0.3 0.4 1.0 1.4 0.02 0.03 0.2 0.4 0.01 0.02 0.1 0.2

10mm 1.0 1.5 3.5 5.0 1.0 1.7 4.0 5.5 1.1 1.7 4.0 6 0.06 0.10 0.7 1.5 0.04 0.07 0.4 0.7

1mm 2 2.5 2.5 2.5 2 2 2 2 2 2.5 2.5 2.5 <1 <1 <1 <1 <1 <1 <1 1

2mm 4 4 5 5 4 4 4 4 4 4.5 5 5 <1 <1 1.0 1.5 <1 1 2 2

5mm 8 9 11 11 8 9 10 10. 8 9 11 12 <1.0 1.0 2.5 4.0 1.5 2 4 6

10mm 14 16 20 22 14 16 19 20 14 17 21 22 1.5 2 5 7 3 4 8 11

Table 3 Primary Hazard radius, R1 for example releases. (Rounded figures) *At the fluid storage temperature of 20 oC the nominal discharge pressure of 5 bara is below the saturated vapour pressure of the category A fluid. The saturated vapour pressure (6.8 bara) was used to calculate the discharge rate and dispersion.

9

Fluid Category

A

B

C

G(i)

G(ii)

Release pressure (bara) * 5 10 50 100 5 10 50 100 5 10 50 100 5 10 50 100 5 10 50 100

1mm 0.01 0.01 0.03 0.05 0.01 0.02 0.04 0.06 0.01 0.02 0.04 0.06 0.001 0.001 0.007 0.015 0.0004 0.001 0.004 0.007

Release flow rate (kg/s)

Hazard radius R2 (metres)

Release hole diameter

Release hole diameter

2mm 0.04 0.06 0.14 0.2 0.04 0.07 0.15 0.2 0.06 0.1 0.2 0.25 0.002 0.005 0.03 0.06 0.001 0.003 0.02 0.03

5mm 0.3 0.4 0.9 1.2 0.3 0.4 1.0 1.4 0.3 0.4 1.0 1.4 0.02 0.03 0.2 0.4 0.01 0.02 0.1 0.2

10mm 1.0 1.5 3.5 5.0 1.0 1.7 4.0 5.5 1.1 1.7 4.0 6 0.06 0.10 0.7 1.5 0.04 0.07 0.4 0.7

1mm 2 2.5 3 3 2 2.5 2.5 3 2.5 2.5 3 3 <1 <1 <1 1.0 <1 <1 1 2.0

2mm 4 4.5 5.5 6 4 4 5 5 4 4.5 5.5 6 <1 <1 1.5 2.0 <1 1 2 3

5mm 16 20 20 20 14 16 17 17 20 21 21 21 1.0 1.5 3.5 5 2 2.5 6 8

10mm 40 50 50 50 40 40 40 40 50 50 50 50 2 3 7 11 4 5 11 14

Table 4 Hazard Radius at Ground level, R2, for the example releases For larger releases R2 can be approximated from R1 using Table 5. The ratio decreases as the release pressure increases because mixing improves.

Fluid Category

R2/R1 Low pressure High Pressure A 3.0 2.2 B 2.6 2.0 C 3.5 2.0 G(i) 1.4 1.5 G(ii) 1.3 1.3 Table 5 Quick estimator for the hazard radius at ground level, R2.

4. Hazard Radii from vents In this chapter we address hazard radii for discharges from vents using a standard matrix of conditions and ambient conditions of neutral atmospheric stability and a temperature of 30 °C. A vent is defined as a means of release of vapour at or near to atmospheric pressure. This is distinct from the pressurised releases considered in section 3. The results of the calculations for the storage of petroleum products are given in section 4.1 and for process vents in section 4.2. The main findings, which result from the use of the prescribed matrix of conditions, are that:

10

x low flow rates through the larger orifices implies insufficient exit momentum to disperse the vented material as a jet. The dispersion process then becomes dependent on the detailed flow interaction at the vent tip. Downwash, contacting of material below the vent height and dense vapours in the vicinity and below the vent exit are to be expected in these cases. x some combinations of venting rates and vent sizes are incompatible with the assumption of discharge at atmospheric pressure. x a guideline value of 300 mb for the pressure drop across a vent is suggested as an upper bound to “atmospheric pressure” releases. Above this pressure account needs to be taken of density changes due to pressure differences in calculating discharge rates for ideal gases. The effect of pressure on vapour composition should be screened because components of high flashpoint fluids may condense. x Emissions from the storage of category B and category C fluids necessarily imply that the fluid vapour are in some admixture with air or an inert gas. Hazard radii are derived for a range of possible vapour compositions treated as ideal gases with molecular weights between 48 and 100. The impact of vapour composition on flammability limits is also taken approximately into account. x The shape factor from the existing guidance is retained. The actual shape of the plume from a vent varies from an upright jet, to a bent-over plume, to a plume subsiding below the vent height depending upon the released gas composition, flow rate, vent size and wind speed. With this complexity it is appropriate and conservative to retain a spherical hazard radius around the exit plane. x The hazard radii vary from just smaller to just larger than those in the existing guidance depending upon the vapour composition and flammability limit. x The existing guidance for process vents is in agreement with these results. 4.1. Vents from the Storage of Petroleum products Section 3.2.5 of Revision 6 of IP 15 provides hazard radii for differing emission rates and vent sizes for Class I, II(2) and III(2) materials. Model fluids are used to derive the hazard radii. The composition of these fluids was given Table 2. It is assumed that the vents are remote from any structure and, if attached, are elevated sufficiently for the dispersion to the lower flammability limit to be unaffected by building induced flow. The comments below are specifically addressed for the early stages of dispersion for flammability assessments. Other problems, such as the assessment of odours, assessment of health impacts or environmental impacts, require a more detailed treatment. Work Item 3 is to verify/update the hazard radii given in the original report according to the matrix given in Table 6 for materials in Categories B and C. Vapour Emission Rate (filling rate) m3/h 250 500 1000 2500

Vent diameter (mm) 50 x x x x

80 x x x x

100 x x x x

250 x x x x

Table 6. Matrix of vent flow rates and diameters used in this study Flow from a vent is of vapour only and it is implicit that there is only a small pressure drop across the vent. This contrasts with releases from pressurised containment which may exhibit two-phase behaviour as well as density changes near to the discharge point.

11

Venting usually takes place in a vertical direction. Material vented upwards will rise above the vent stack as a consequence of vertical momentum. As it mixes with the ambient air it acquires horizontal momentum and bends over toward the horizontal. If the vented gas is less dense than air it will then slowly rise, if it is heavier than air it will slowly subside. The extent to which the plume path is affected by density differences depends upon the gas properties, the venting rate and the prevailing atmospheric conditions. The degree of plume rise in the immediate vicinity of the vent depends mainly upon the exit momentum and hence on the fluid density as well as the exit velocity. However, most of the modelling that has been carried out is for stack gas effluent and uses velocity based criteria for design purposes. The most commonly applied rule is to guard against plume downwash. Experiment shows that, when the exit velocity is less than 1.5 times the wind speed at the vent height, the aerodynamic interaction between the wind and the vent stack causes the pollutant to be drawn down into the near wake of the vent tip. This is called downwash and results in a reduction in the height of release, extra mixing of the pollutant and contacting of the pollutant with the external stack. For releases where the flow rate is sufficient to fully avoid downwash then dilution to below flammable concentrations will occur in the vicinity of the vent and through the mechanism of jet mixing rather than ambient turbulence. In order to be conservative we have neglect the effect of downwash on plume dilution. An ambient wind speed of 2 m/s at the vent top is assumed for these calculations. We observe that: for wind speeds greater than 2 m/s the plume trajectory will be flattened toward the horizontal. The shape factor is thus conservative in the vertical for wind speeds of 2 m/s and higher. For wind speeds less than 2 m/s the hazard will be above the vent point so that the shape factor is conservative in the downwind direction. The hazard radii given in this report take account of the contribution of the 2 m/s wind speed to the downstream extent of the plume. This should render the estimation of hazard radius conservative for lower wind speeds. In critical cases the user should check by calculation. The shape factor given in the existing guidance is retained because it is conservative. For very slow venting rates and under low-wind conditions denser than air gases might flow down the outside of the vent pipe. The situation should not occur for simple vents, designed with an adequate exit velocity. It is more likely to occur for large area vents on a structure. There is no simple model available to treat this problem which needs experimental or computational investigation on a case by case basis. The flow phenomenon is complicated and may involve substantial mixing within the exit of the vent pipe. We believe that it is necessary, but conservative, to retain the zone 2 classification outside of and beneath the zone 1 hazard radius to account for this possibility. Nominal exit velocities for the vent matrix are given in Table 7. These cover a large range. We observe that: x For the range of flow rates shown the 250 mm diameter vent in particular shows nominal exit velocities below and close to 10 m/s. This implies downwash under common wind conditions and so these vent combinations may not be realistic in practice for continuous emissions. x Exit velocities above 100 m/s are indicated for the 50 and 80 mm diameter vents at the higher flow rates. The implications of this are assessed below for the different fluids.

12

Vapour Emission Rate (filling rate) m3/h

Exit Velocity m/s

50 250 500 1000 2500

35 71 141 283

Vent diameter (mm) 80 100 14 28 55 110

9 18 35 71

250 1 3 6 11

Table 7. Nominal exit velocity based on vent area (m/s). The shaded cells represent combinations of vapour rate and vent diameter that need special consideration. Exit velocities below 10 m/s in the top right of the table imply possible downwash effects. Exit velocities above 100 m/s in the bottom left imply a significant pressure forcing. The exit velocity of the vented flow is illustrative but, because it is the exit momentum of the flow that determines the rate of dispersion, the density of the effluent needs to be taken into account as well. We assume that the effluent is an ideal gas with properties evaluated at atmospheric pressure. In reality any rate of discharge requires some driving force and a positive pressure differential is the most common. We need to evaluate what a “negligible” pressure is and how it affects the realistic range of vent sizes and venting rates for the different category fluids. We also need to assess what are the likely properties of the material that is vented. For liquid storage at atmospheric pressure the vapour space will necessarily be occupied by some mixture of fluid vapour and air or fluid vapour and an inert gas. The mixture composition should not be flammable but will be variable according to the filling level and the history of changes in level, ambient conditions etc. Thus for vents : x The vented mixture may vary in composition. x The flammability limit and physical properties of the vapour may vary. x For some conditions (high venting rates and small vents) it may be necessary to take account of the effects of pressure on the release. x If the release does not take place close to atmospheric pressure then the ideal gas assumption may not be appropriate for mixtures with low vapour pressure components. Because the effective molecular weight, and hence density, of the vented vapour is not closely defined it is appropriate to make some example calculations using a range of values. We assumeiii that the average molecular weight of the gas vented from the storage of category B and C fluids will lie in the range 48 - 100. Figure 2 shows how the mixture molecular weight varies with dilution. For no-added air the molecular weight is that of the pure vapour (Table 2) and as more and more air is added then the mixture molecular weight tends to that of air (29) or if mixed with nitrogen (28).

iii

Note: because the model fluids are illustrative we have made no attempt to reconcile the example ideal gas properties with the actual equilibrium air-vapour-liquid compositions of the model fluids at atmospheric pressure and the reference temperature of 30 °C used throughout this guidance

13

Mean Molecular Weight for Mixtures of Category A, B, C fluids with Air, g/mol MW 48 - A MW 100 - B MW 125 - C

140

Mixture Molecular Weight

120 100 80 60 40 20 0

0.2

0

1

0.8

0.6

0.4

Mol fraction of Air in Mixture

Figure 2 The effect on mixture molecular weight of mixing category C (top), category B (middle) and category A vapour with air assuming volatilisation of all components. The upper flammability limits of the Category B and C fluids lie far to the right of Figure 2 as shown in Table 8. Fluid B C

Upper Flammability Limit of vapour % 6.84 6.08

Mol fraction Air in mixture 0.93 0.94

Table 8. Upper Flammability limits of Category B and C vapour. The mol fraction of air in the vapour in a storage tank must be substantially less than these values to avoid flammable/explosive mixtures. The mixture molecular weight of 48 corresponds to the following mixtures, Table 9: Fluid Type A B C

Mol Fraction Air 0 0.72 0.81

Mol Fraction Fluid Vapour 1 0.28 0.19

Table 9: Equivalent mixtures for a vapour molecular wt of 48. The Category B and C fluids Table 2 have a lower flammability limit falling into a narrow range of 0.039-0.043 kg/m3, 0.86-2.0 %v. We also included Category A vapour as an additional example in this sensitivity analysis although, of course, for practical reasons it is not stored at atmospheric pressure and temperature.

14

As the pure vapour becomes diluted then the lower flammability limit of the mixture increases. The effect of dilution on the flammability limit of the mixture is shown qualitatively in Figure 3 where we have used Le Chatelier’s law and treated air as inert fuel component. This is not strictly accurate for high dilutions but it does illustrate that lower flammability limit increases as air is added. Only mixtures initially richer than the upper flammability limit are considered which is why the mol fraction of air is terminated at 0.9. Lower Flammability Limit for Mixtures of Category A, B, C fluids with air, %

Mixture Lower Flammability Limit, %

100

MW 48 - A MW 100 - B MW 125 - C

10

1

0.1

0

0.1

0.2

0.3

0.5

0.4

0.6

0.7

0.8

0.9

Mol fraction of Air in Mixture

Figure 3. Lower Flammability limits of the three model fluids used in the Area Classification guidance as a function of dilution with air. Only mixtures originally above the upper flammability limit are considered.

We now verify and quantify the range of venting rates and vent diameters that qualify as “negligible pressure”. Figure 4 shows the variation in mass-flow rate from a short pipe as the applied pressure, expressed in millibar gauge, increases. Specific calculations (points) were calculated for an ideal gas of molecular weight 48 and for a pipe diameter of 50 mm. At low drive pressure the mass flow rate is proportional to the square root of the applied pressure as would be expected from the Bernoulli equation for an incompressible fluid i.e. keeping the density unchanged from its base value. A correlation lineiv is drawn to indicate this relationship. At high pressure the mass flow rate (points) increases more quickly with increasing pressure than indicated by the correlation line. This is because the increase in fluid density in the vent is significant. To identify a suitable cut-off we compared the goodness of fit of the Bernoulli equation to the flow calculations. We found that the fit became progressively worse as we included results for pressures above 300 mb. Therefore we adopt 300 mb as marking the upper limit of “negligible” pressure. iv

A correlation line was used for convenience. Equally the slope could be derived from the discharge model for an ideal gas through a round hole using a discharge coefficient of 0.8. This was the basis for the model calculations to which the regression line was fitted.

15

To see the consequence of this for the flow conditions in Table 6 we take the maximum flow rate of 1.34 kg/s corresponding to the volume flow-rate of 2500 m3/h (evaluated at 1 atm. pressure and 30 °C) and plot this as a horizontal line on Figure 4. We then construct the (incompressible) flow rates for the additional hole sizes of 80, 100 and 250 mm as parallel lines to the 50 mm pipe calculations. Figure 4 shows that for pipes smaller than 80 mm diameter the 2500 m3/h flow rate implies a drive pressure greater than 300 mb. Mass flow rate as a function of applied pressure 10

mass flow rate in kg/s

1

0.1

Mol. Wt 48 50 mm pipe 80 mm pipe 100 mm pipe 250 mm pipe 2500 m3/hr

0.01

0.001 0.1

1

100

10

1000

10000

over-pressure, mb

Figure 4. Mass Flow rate through pipes of different diameter for an ideal gas of molecular weight 48. Points denote calculations with a discharge model. Lines denote a correlation based on drive pressures less than 300 mb. The horizontal line denotes the mass flow rate consistent with a volume flow rate of 2500 m3/h at atmospheric pressure and 30 °C. Figure 5 shows the equivalent graph for a vapour of molecular weight 100. For the larger molecular weight greater pressures are needed to achieve a given volumetric flow. We find that pipe diameters smaller than 100 mm require a drive pressure greater than 300 mb to achieve a flow rate of 2500 m3/h.

16

Mass flow rate as a function of applied pressure 10

mass flow rate in kg/s

1

0.1

Mol. Wt. 100 50 mm pipe 80 mm pipe 100 mm pipe 250 mm pipe 2500 m3/hr

0.01

0.001 0.1

1

100

10

1000

10000

over-pressure, mb

Figure 5. Mass Flow rate through pipes of different diameter for an ideal gas of molecular weight 100. Points denote calculations with a discharge model. Lines denote a correlation based on drive pressures less than 300 mb. The horizontal line denotes the mass flow rate consistent with a volume flow rate of 2500 m3/h at atmospheric pressure and 30 °C These calculations suggest that for ideal gases a threshold of 300 mb is a suitable choice to distinguish between “vent” flows where difference in the physical properties of the vapour within the vent and at atmospheric pressure can be neglected. If the vent design implies higher over-pressures then the implications for the possible change in properties of mixtures with low vapour pressure components needs to be assessed. Hazard radii were obtained for vertical discharges of an ideal gas with molecular weight in the range 48 - 100 fluid. The effect of changing flammability limit was tested using values appropriate to mixtures of category B and category C vapour with air and including category A vapour as a worse case. Results are given in Table 10 and have been rounded. Values marked with a double asterisks(**) are conservative, at such low exit velocities mixing is dependent on the detail flow in the vent exit.

17

Vapour Emission Rate (filling rate) m3/h

Hazard radius, m

50 250 500 1000 2500

2 2.5 3 4

Vent diameter (mm) 80 100 2.0 2.5 3.5 5

2.5 2.5 3.5 5

250 3** 4 6. 7

a) Assuming category B vapour mixed with air. The shaded cells require a drive pressure greater than 300 mb to achieve Vapour Emission Rate (filling rate) m3/h

Hazard radius, m

50 250 500 1000 2500

2 2.5 3.0 4.0

Vent diameter (mm) 80 100 2.0 2.5 3.5 5.0

2 2.5 3.5 5

250 3** 4.0 5.5 6

b) Assuming category C vapour mixed with air. The shaded cells require a drive pressure greater than 300 mb to achieve Vapour Emission Rate (filling rate) m3/h

Hazard radius, m

50 250 500 1000 2500

2.5 3.5 4.5 6.5

Vent diameter (mm) 80 100 4.0 3.5 4.5 7

6.0 4.5 5.0 7

250 6** 6.5 9 13

c) Assuming pure category A vapour. The shaded cells require a drive pressure greater than 300 mb to achieve Table 10. Hazard radii for a fluid of molecular weight 48 g/mol treated as an ideal gas and three example flammability limits The results show a trend for the hazard radius to increase as the lower flammability limit of the mixture decreases as would be expected. The hazard radius values for the category B and category C vapour/air mixtures are quite similar but substantially smaller than those for the “pure vapour” category A simulation. This suggests that the dilution with air that characterises an open venting system is key to reducing hazard distances and should be taken account of in any specific site assessment. Table 11 below gives results for a molecular weight 100 fluid, evaluated as a mixture of category C fluid and air and as pure category B fluid. This is almost certainly an extreme example but, when compared to Table 10, illustrates that as the molecular weight of the

18

mixture increases the hazard radius also increases for a fixed volume flow rate. We also notice in the detailed calculations and showing in Table 11 that, for a fixed venting rate and increasing vent diameter, the trajectory of the vent gases alters as the exit velocity decreases. At low exit velocities the vent gases subside quite quickly after leaving the vent and the downward path of the gases passes relatively close to the vent. The hazard radius, measured from the exit plane to the point of dilution to LFL starts to decrease because of the trajectory whereas, for lighter (but still dense) gases the vented material passes away from the stack and the hazard radius increases as the exit velocity and hence the jet mixing component of entrainment decreases. When the molecular weight is as large as 100 g/mol then discharge from a vent sized 250 mm or larger has insufficient bulk velocity to disperse the gases according to the assumptions made here. In practice, and especially for substantially larger vents, such as hatches, complicated flow interactions can take place in the actual opening with a dense gas flow preferentially around the edges and a degree of inflow and mixing taking place inside the centre of the vent opening. A different methodology, such as computational fluid dynamics or experimental measurement must be used if the mixing within the geometric features of the holding vessel is to be understood and quantified. Vapour Emission Rate (filling rate) m3/h 7

Hazard radius, m

50 250 500 1000 2500

Vent diameter (mm) 80 100

4 5 6.5 9

6.5 5.5 7 10

6 8 7 10

250 n/a n/a n/a 11

a) Assuming category C vapour mixed with air Vapour Emission Rate (filling rate) m3/h 7

Hazard radius, m

50 250 500 1000 2500

3.5 5 6.5 9.5

Vent diameter (mm) 80 100 6.5 6 7 10.5

6.0 8.5 7.5 10.5

250 n/a n/a n/a 11.5**

b) Assuming pure category B vapour Table 11. Hazard radii for a fluid of molecular weight 100 g/mol, treated as ideal gas. For venting rates less than 2500 m3/h the vent of 250 mm diameter gives too small an exit velocity to assure dilution of the gas and a down flow in the vicinity of the vent stack is to be expected. Affected combinations of vent rate and vent size are marked as n/a. In Table 10 and Table 11 we note that, for the highest flow rates and the smaller vent sizes, the hazard radius is shaded. This denotes that more than 300 mb pressure drop is needed to achieve this flowrate and a check on the vapour composition should be carried out to see if condensation of any components is implied.

19

Comparison with section 3.2.5 of revision 6 of IP 15 shows that hazard radii range from just below (molecular weight 48) to just above (molecular weight 100) the existing guidance using flammability limits for category B and category C vapour diluted with air. 4.2. Process vents Section 5.7 of Revision 6 describes releases from process vents to atmosphere. It is again assumed that venting is restricted to vapour phase releases. Release rates in the guidance and required for work item 3 follow the matrix shown in Table 12 within which the implied exit velocities are included. We note that the range of values is extremely small compared with those used above for vents from storage. This seems counter intuitive as one might expect venting from a process to involve quite large flow rates.

Vapour Emission Rate m3/h 50 10 100 250

1.5 14 35

Exit Velocity m/s Vent diameter (mm) 100 250 0.5 3.5 9

0.1 1 1.5

Table 12. Matrix of conditions for assessing hazard radii from process vents and their associated exit velocity. Shading denotes combinations prone to downwash. The exit velocity from a vent should exceed the wind speed by a factor 1.5 if downwash is to be avoided and exit velocities less than about 10 m/s may be assumed to give rise to downwash under common meteorological conditions. Downwash is to be avoided as it can lead to soiling and corrosion of the vent pipe and to low level exposure of structures and personnel to the emitted gases. Table 12 shows exit velocities very substantially less than 10 m/s for a majority of cases. Where the exit velocity is less than 1 m/s some very complicated interactions between the external flow and the pipe flow can occur. It is known that for lighter than air gases the external flow can enter the vent pipe from above and that mixing can take place within the vent pipe itself to a good degree. Experimental and/or computational fluid calculations are needed to make accurate calculations for these cases. Hazard radii are calculated in the same manner as for the vents from storage. The vented fluid is assumed to be an ideal gas of molecular weight, 7, 19 and 48 corresponding to the category G(ii), G(i), and A fluids used in the guidance. The hazard radii are given in Tables 13-15 below. In view of our comments above with respect to the overall low flow rates in the guidance we have included larger values in the tables.

20

Vapour Emission Rate m3/h

Hazard radius, m

50 10 100 250 500 1000 2500

< 2** < 2** 2 2 3 4.5

Vent diameter (mm) 100 250 < 2.5** < 2.5** 3 3 3.5 4.5

< 5** < 5** < 5** < 5** < 5** 5.5

Table 13. Hazard radii for Fluid category G(ii) as ideal gas of Molecular weight 7 g/mol. Vapour Emission Rate m3/h

Hazard radius, m

50 10 100 250 500 1000 2500

< 1** 1 1.5 2 3 4

Vent diameter (mm) 100 250 < 2** < 2** 2 2 3 4

< 4** < 4** < 4** < 4** 4 5

Table 14. Hazard radii for Fluid category G(i) as ideal gas of Molecular weight 19 g/mol. Vapour Emission Rate m3/h

Hazard radius, m

50 10 100 250 500 1000 2500

2.5 2 2.5 3.5 4.5 6.5

Vent diameter (mm) 100 250 < 4.5** 5 6 4.5 5 7

< 6** < 6** < 6** 7 9 13

Table 15. Hazard radii for Fluid category A as ideal gas of Molecular weight 48 g/mol. No solutions were obtainable for the smallest releases. The dilution is entirely dominated by the flow interactions at the vent tip. We suggest that, to be conservative, the largest calculated hazard radius for a given vent diameter is used. These values are indicated by a double asterisk(**) in the tables. The results are in agreement with those in the existing guidelines.

21

5. Evaporation from pools and sumps. IP 15 provides good qualitative guidance on the evaporation and dispersion of vapour from category B and C fluids discharged to pools and from sumps. This work sets out to derive the relevant hazard radii as requested in work items 4 and 5. The transient nature of a spill problem and the multicomponent nature of category C and B fluids leads to a semantic problem when providing a reference document for the IP guidance. The hazard radius defined for a pool of specific size should reflect the hazard from the pool itself and be independent of the manner of release of the fluid if it is to be of generic use. Investigation of several scenarios leads to the conclusion that the evaporation rate of the heavier components of category C fluid is very low under the reference conditions. The evaporation rate of the lighter components on the other hand is rapid such that the maximum vapour generation rate is actually dictated by amount and rate of release of the liquid. Consequently we find that: x the existing guidance for liquid spillages and for sumps is conservative in so far as it applies to the hazard posed by the residual liquid in a spill of category C fluid. The guidance is appropriate to a spill with an average double the evaporation rate of category C fluid. x the hazard arising early in a spill arises from the more volatile components and is dictated by the conditions of the release rather than the size of the pool itself. x the existing guidance may still be conservative because the transient nature of the vapour release results in a rapidly dispersing “puff” of vapour rather than a large cloud. Some scenario investigation suggests that releases would have to be large ones to present substantial hazards and thus be more typical of incidents as opposed to more commonly occurring events. x evaporation from sumps will be less than from pools and therefore the existing guidance is conservative for sumps as well. This is verified for releases of temperature 50oC. In this section the basic phenomenology of liquid spills is described and the hazards quantified. We define: x a pool to contain a shallow layer of liquid e.g. arising from the spillage of a category C fluid onto a surface such as concrete. x a sump to contain a much deeper layer of liquid; possibly floating upon a sub-layer of water. This distinction is necessary because the physical process of evaporation requires the supply of energy equal to the latent heat of evaporation of the fluid. For a shallow layer this energy can be obtained from the substrate by conduction. Clearly the availability of energy depends on the actual depth of the pool, the temperature and physical properties of the substrate and the external conditions but, for very shallow pools, evaporation will take place at a rate determined by the ground temperature. For a deep layer this energy is obtained from the internal energy of the spilled fluid. The temperature of the pool will drop until a heat balance is reached between the evaporation rate and the total energy transfer to the liquid by conduction and insolation. As the liquid cools then the evaporation rate decreases.

22

Clearly these are extremes and spills can behave as “shallow” or “deep” or have mixed behaviour depending upon the fluid properties and spill conditions and whether or not the spill is physically contained. For a given fluid we would generally expect a greater evaporation rate from a shallow spill onto a warm substrate than from a deep pool. Implicit in this expectation is some consideration as to how the fluid enters the pool or sump, how long it remains there and what the time history of the event is. A pool will originate with a breach in a tank or pipe containing liquid. As the spill commences, liquid spreads out over the ground until it meets a restriction or until the pool thickness has reached a limiting value determined by surface tension. While the spill is spreading over warm ground the lighter components evaporate preferentially. There may also be some vapour released at the early in the spill which adds to the overall vapour generation rate. The rate of total evaporation increases until the pool stops spreading. If the pool is unconfined then the vapour generation rate then falls quickly because the pool dries out. If the spread of liquid is confined by a bund or similar then a finite depth of liquid remains. The vapour generation rate will then remain constant or decline gradually in time depending on the overall heat balance on the pool. When the pool reaches the minimum depth then the evaporation rate will fall rapidly as the pool dries. A sump may behave as a deep pool or may contain a residual amount of material that is refreshed with new releases. If the fill-up rate exceeds the potential evaporation rate of the fluid components then the sump will behave as a deep pool. If the fill up rate is smaller than the evaporation rate of the lighter components then the sump will behave more like a shallow pool. For work items 4 and 5, hazard radii are sought for pools of 5, 10, 15 and 20 m diameterv formed by the retention of spilled materials by planned bunds, naturally occurring blockages such as kerbs, or emergency measures. There are a number of ways in which such pools can occur: x A fixed volume of fluid could be spilled. If this is captured in bunds of the different sizes above then the smallest pool will be deeper than the largest pool. x A variable volume of fluid could be spilled. This could result in the different sized pools having more similar depth. Each of these scenarios will result in a different evaporation history. Quantifying all of the possible scenarios is a large task. Before investigating some actual releases we consider the simpler problem of steady state evaporation of a category C fluid. Using the steady state modelling assumptions for the model category C fluid and for pools of size 5 - 15 m it can be shown that the evaporation flux must exceed certain limits if the existing guidance is correct. Some example hazard distances are given in Table 16. The pool size is the length of the pool in the wind direction and the hazard distance is the distance to LFL measured from the downward edge of the pool. Two values are given for the hazard distance. The minimum value uses a lower flammability limit appropriate to the lightest vapour components of a category C fluid. The maximum uses the lower flammability limit appropriate to the heaviest components of a category C fluid, c.f. Table 2,

v For practical reasons the hazard radii are calculated for square sources of equivalent area as described in Appendix A: Methodology.

23

Evaporation Flux (kg/m2/s)

Evaporation Rate (kg/s)

Pool Size L (m)

Min. Hazard Distance D (m)

Max. Hazard Distance D (m)

0.002

0.05 0.45 0.04 0.36 0.03 0.23

5 15 5 15 5 15

2.9 13.0 0 7.5 0 0

5.2 21 3.2 14.5 0 0

0.0015 0.001

Table 16. Hazard distance ranges for category C vapour from area sources The existing guidance gives hazard distances for ranges of pool sizes, Table 17: Pool Size L (m) Less than 5 Between 5 and 10 Greater than 10

Hazard distance D (m) 3 7.5 15

Table 17. Original Hazard distances from the Guidance. Comparing Table 16 and Table 17 and taking the maximum hazard distance we see that the existing guideline is equivalent to a source with an evaporation flux of between 0.0015 and 0.002 kg/m2/s depending upon the flammability limit that is used. Indeed, considering that the more volatile components evaporate first, the existing guidance is equivalent of a source decreasing in strength from a flux of 0.002 to 0.0015 kg/m2/s over the life of the spill. These points are illustrated with some examples below: Figure 6 shows the time dependent evaporation flux for 4.5 m3 of category C fluid rapidly dumped into a 15 m pool. This volume is chosen to adequately fill the pool and leave fluid still present after a time of 30 minutes (1800 s). Details are given in the methodology section, Appendix A. The horizontal lines show the evaporation flux range 0.0015 - 0.002 kg/m2/s relevant to the existing guidance. For the release the evaporation flux decreases from a value of 0.0012 kg/m2/s to a value of 0.0005 kg/m2/s after a time of 300 s. This marks the end of the period in which the lighter components are evolved. Subsequently the evaporation rate is nearly constant. In fact it shows a slight increase because the simulations assume noon-time insolation and the pool is in fact warming slightly over the later period of the modelled spill. Thus a rapid spill of category C fluid gives evaporation fluxes a fraction 0.6-0.4 smaller than accounted for by the existing guidance.

24

Spill rate 0.45 m3/s

evaporation flux kg/m2/s

0.01

15m pool 10 s 2.e-3 1.5e-3

0.001

0.0001 -200

0

200 400 600 800 1000 1200 1400 1600 1800 time from spill start, s

Figure 6. Evaporation rate of a rapid dump of category C fluid Figure 7 shows the effect of spilling the same volume over the longer period of 300 s. Again there is no special significance to the choice of a volume of 4.5 m3 other than that this is sufficient to adequately fill a pool of 15 m diameter. Here three stages of the spill can be seen. There is an initial high rate as the liquid spreads to fill the pool and is passing over a hot concrete substrate. The liquid quickly reaches the bund walls and the pool depth starts to increase. The vapour generation rate falls because the substrate and pool are cooled by the evaporation. The evaporation of the light components of the newly spilled fluid is visible as a plateau lasting just longer than the 300 s of the actual spill. Thereafter the evaporation rate falls again to give results similar to those for the rapid spill.

25

Spill rate 0.015 m3/s

evaporation flux kg/m2/s

0.01

15m pool 300 s 2.e-3 1.5e-3

0.001

0.0001 -200

0

200 400 600 800 1000 1200 1400 1600 1800 time from spill start, s

Figure 7. Time history of a five minute spill of 4.5 m3 of category C fluid.

For sumps there is a concern that hot liquid may be introduced. Figure 8 shows that preheating the fluid to 50 qC from 20 qC enhances the vapour rate in the early spill stages by a factor of about two. Even so it only reaches the 0.002 kg/m2/s value for the few seconds in which the liquid spreads to cover the pool. For a sump, where liquid is being added to existing material, this initial transient would be absent. It is notable that over a halfhour period the effect of the initial temperature difference on evaporation rate has disappeared.

26

Spill rate 0.015 m3/s Storage Temperature 50 C

evaporation flux kg/m2/s

0.01

15m pool 300 s 2.e-3 1.5e-3

0.001

0.0001 -200

0

200 400 600 800 1000 1200 1400 1600 1800 time from spill start, s

Figure 8. Time history of evaporation for liquid initially at 50 C. The liquid in the pool rapidly cools so that the residual evaporation rate is similar to the other cases. These calculations show that in so far as the evaporation of the model category C fluid is concerned the existing guidance is conservative with respect to the evaporation of the bulk of the fluid. They also show that the manner of discharge itself (rate, amount) coupled to the initial volatilisation of light components determines the early peak in vapour rate. This cannot be characterised simply in terms of a pool dimension. For the example we have used here - the release of 4.5 m3 of fluid into a 15 m bund over time scales of 10 to 300 s which perhaps is not an unreasonable scenario, the guidance is conservative even for this initial spill period. Larger spill possibilities would likely occur as incidents and thus require explicit modelling. 5.1. Vapour pressure comparisons of some commonly used Category C fluids. Table 18 shows the vapour pressures of some hydrocarbons (C7+) relative to n-octane. Liquid Spill Hazard for all of these compounds exceeds that of n-octane and specific account of volatility must be taken into account when the ratio exceeds 2.

Name

Formula

n-octane cis 1,2-dimethylcyclohexane 3-vinylcyclohexene 2,2,4-trimethylhexane isopropyl cyclopentane di-sec-butyl ether 2,2,5-trimethylhexane trans-2-octene

C8H18 C8H16 C8H12 C9H20 C8H16 C8H18O C9H20 C8H16

27

Vapour Pressure relative to n-octane at 30 qC 1 1.02 1.09 1.13 1.14 1.15 1.17 1.19

octene-1 tra 1,3-dimethylcyclohexane cis 1,4-dimethylcyclohexane 2-methyl bicyclo 221 heptane tra 1,2-dimethylcyclohexane 3-methyl heptane 2-methyl-1-heptene 2,4-heptadiene 3-ethyl hexane 1,7-octadiyne 2-methyl heptane 1,7-octadiene cis 1,3-dimethylcyclohexane cycloheptane 3,4-dimethylhexane 1,1-dimethylcyclohexane 3-methyl 3-ethyl pentane 2,3-dimethylhexane 2-methyl 3-ethyl pentane di-tert-butyl peroxide 2,3,3-trimethylpentane 2,3,4-trimethylpentane 1,1,2-trimethyl cyclopentane toluene 3,3-dimethylhexane 2,4-dimethylhexane 2,5-dimethylhexane 2,2,3-trimethylpentane 1,3-heptadiene 2,2-dimethylhexane 1,6-heptadiyne 2,4,4-trimethyl pentene-2 heptyne-1 2,2,3,3-tetramethylbutane ethyl cyclopentane 2,4,4-trime-1-pentene n-heptane cis-2-heptene methyl cyclohexane trans-2-heptene cis 1,2-dimethylcyclopentane 2,2,4-trimethylpentane trans-3-heptene cis-3-heptene 1,4-heptadiene 1,5-heptadiene heptene-1 3-ethyl pentane 3-methyl hexane 2-methyl-1-hexene 1,6-heptadiene tra 1,2-dimethylcyclopentane

C8H16 C8H16 C8H16 C8H14 C8H16 C8H18 C8H16 C7H12 C8H18 C8H10 C8H18 C8H14 C8H16 C7H14 C8H18 C8H16 C8H18 C8H18 C8H18 C8H18O2 C8H18 C8H18 C8H16 C7H8 C8H18 C8H18 C8H18 C8H18 C7H12 C8H18 C7H8 C8H16 C7H12 C8H18 C7H14 C8H16 C7H16 C7H14 C7H14 C7H14 C7H14 C8H18 C7H14 C7H14 C7H12 C7H12 C7H14 C7H16 C7H16 C7H14 C7H12 C7H14

28

1.23 1.24 1.26 1.29 1.35 1.38 1.39 1.41 1.42 1.43 1.45 1.49 1.5 1.52 1.52 1.58 1.6 1.64 1.67 1.78 1.87 1.88 1.93 1.98 1.99 2.11 2.12 2.22 2.24 2.36 2.38 2.5 2.73 2.73 2.75 3.06 3.15 3.15 3.17 3.21 3.24 3.37 3.5 3.52 3.67 3.75 3.86 3.97 4.21 4.22 4.31 4.35

tra 1,3-dimethylcyclopentane 2-methyl hexane cis 1,3-dimethylcyclopentane 2,3-dimethylpentane 1,1-dimethylcyclopentane 3,3-dimethylpentane 2,4-dimethylpentane 2,2,3-trimethylbutane 2,2-dimethylpentane

C7H14 C7H16 C7H14 C7H16 C7H14 C7H16 C7H16 C7H16 C7H16

4.39 4.49 4.56 4.68 5.12 5.58 6.63 6.85 7.06

Table 18 vapour pressures of some hydrocarbons (C7+) relative to n-octane

6. Releases into confined areas. A release of flammable material into a confined space, such as a building, is potentially an extremely hazardous event. Ignition may lead to the development of over-pressure causing structural damage to the enclosure and neighbouring buildings. Consequently, any events leading to a sustained release of flammable material should not be considered as normal operation but be subject to a detailed consequence analysis. For the purposes of area classification there is a need to quantify the difference between an event definitely leading to a hazardous condition and an event that is potentially hazardous but which might be managed by precautionary action such as the active control of ignition sources near to the point of handling of flammable material. A key concern is how to assess the conditions under which a release might escape a building at a flammable concentration and require external ignition prevention precautions. General guidance for the safe ventilation of building enclosures given in IP 15 is, quite soundly given by dividing activities that may lead to releases into categories requiring different grades of ventilation. Grades of ventilation are then parameterised by the air change rate, D, expressed as the number of times per hour that the air in the building is changed. Four categories are identified:

Adequate Ventilation:

To quickly reduce possibly flammable concentrations to safe concentrations in the event of a leak or spillage, and following action to stop the fluid source, 12 air changes per hour is recommended.

Dilution Ventilation:

Forced ventilation at sufficient rate to limit the formation of a gas volume at a concentration of 20% of the Lower Flammability limit is recommended. Typically dilution ventilation will be vigorous (30 - 90 air changes per hour) and the output diverted to vents.

Local Artificial Ventilation:

The use of either small scale dilution ventilation (use of extractors etc.) or an enhancement of flow in obstructed areas to attain adequate ventilation is recommended.

29

Overpressure Ventilation:

Prevention of the ingress of flammable material to a confined area by maintaining an over-pressure within it is recommended for cases where buildings are close to potential sources but do not contain sources themselves.

Of these categories we need only to try to quantify the first, “Adequate Ventilation” as applying generally to small spills. The Dilution Ventilation rate needs to be specially designed for each application. Local Artificial Ventilation is essentially the same as Adequate Ventilation for small enclosures within a larger confined workplace, unless the effluent is ducted to a vent in which case it is an example of Dilution Ventilation. Overpressure ventilation precludes the need for assessment. What does “adequate ventilation” mean in terms of a release? Let V be the ventilated volume. Consider the volume to be well mixed then the concentration within the enclosure x

of a gaseous flammable material released at time 0 at a constant rate m kg/s is given by a simple mass balance:

UV

dC dt

x

m  DUVC

where D s-1, is the air change rate, U kg/m3, is the density of the flammable gas and C m3/m3 is the volumetric (molar) concentration. This equation has the solution: x

C

m 1  e  Dt  DUV

for initial conditions C = 0 when t = 0. The maximum concentration (volume fraction) that can occur throughout the well mixed volume is given by: x

m DUV

C max

To estimate some values for the leak rate than would give rise to flammable concentrations in an enclosure we can take some typical values: For a dense gas: V = 1000 m3 U = 2 kg/m3 D = 12/3600 = 0.0033 s-1

C max  >0.02,010 . @

corresponding to a building with typical dimension 10 m. density of the flammable gas e.g Cat A. an adequate ventilation rate of 12 changes per hour: a typical flammable limit range, [LFL,UFL] by volume fraction

.

30

x

>

@

from which we find that m  013 . , 0.66 kg/s. A factor 2 variation on these end range values would be a reasonable uncertainty to apply. The time taken to reach 90% of maximum concentration is ~700 s. Thus, to round figures, if the gas is well mixed, a release of # 0.1 kg/s could give rise to potentially flammable conditions through an adequately ventilated building, of volume 1,000 m3, on a time scale of ~700 s. This would be an extremely hazardous condition. Larger releases might lead to potential flammable regions outside of the building because the material leaving would be above the lower flammability limit for the gas. Certainly a release as large as # I kg/s, capable of filling the building to the upper flammability limit should not be tolerated. The basic premise of the ventilation calculation is that the volume should be well mixed. Mixedness will not occur through the ventilation flow itself as the implied ventilation wind speed is very small, of order 33 mm/s assuming a typical building length-scale of 10 m. Some other “stirring” mechanism is needed. If we consider the pressurised releases treated in section 3, then a release rate of # 0.1 kg/s for a category A or B fluid could arise for a hole size as small as 2 mm for pressures below 50 bar. The hazard radius for an unconfined release of this magnitude is # 5 m which is half the characteristic length scale of the example building and the order of magnitude of the free path of a centrally placed release. The hazard radius is also the distance over which the unconfined jet entrains enough air to dilute the mass flow to the LFL. This entrainment is comparable to the circulation in the building and it is thus reasonable to propose that the volume of air passing through the building can be well mixed by a pressurised release. For a lighter than air gas: We can rework the above example for the category G(i) and G(ii) gases. We have: V = 1000 m3 U = 0.78, 0.29 kg/m3 D = 12/3600 = 0.0033 s-1 CLFL = 0.046, 0.04 m3/m3 .

corresponding to a building with typical dimension 10 m. corresponding to a category G(i) and G(ii) gases respectively. an adequate ventilation rate of 12 changes per hour: LFL for category G(i) and G(ii) gases respectively.

x

giving m

012 . , 0.04 kg/s for category G(i) and G(ii) fluids respectively.

From these simple arguments we can deduce that quite small pressurised releases of the type used to underpin the hazard radius recommendations for unconfined releases and of size # 0.1 kg/s could result in the establishment of a flammable mixture within a 1000 m3 building having a ventilation rate of 12 air changes per hour for Category A and G(i) fluids, and a somewhat lower value of # 0.04 kg/s for category G(ii) fluid.

31

External Hazards The implication of this is that there may be an efflux of flammable material from the buildings for releases of order 01 .

V V kg/s for category A and G(i); 0.04 kg/s for category G(ii) 1000 1000

gases where V is the volume adequately ventilated. assessed.

The fate of this material needs to be

Flow and mixing around buildings Flow and dispersion are strongly influenced by the building shape and orientation to the wind. Buildings are classified as bluff rather than streamlined bodies and this greatly complicates the description of the flow. Consequently here is a large body of work reporting the study of flow and dispersion in building wakes.(e.g. Castro & Robins, 1977; Hosker, 1979; Britter, Hunt & Puttock, 1976; Snyder and Lawson, 1994). The main feature of a bluff rather than a streamlined body is that the wind flow cannot pass smoothly around the body but instead separates from the upwind edges of the body. This forms a region along the sides and behind the body in which pressure is reduced. The pressure difference across this region causes the separated flow from each edge and side of the body to curve towards each other and eventually intercept to form a recirculation region behind the body in which the mean flow is actually reversing as fluid is returned towards the low pressure regions. Depending on the shape of the body this region can extend back up the building sides to the building front. The flow within the recirculation region is highly time varying and unsteady. It was originally thought that the recirculation region was a closed "bubble" bounded by a separation streamline and that material was transported in and out of the region by turbulent mixing across this boundary. This remains a useful simplification of the flow but it is now known that transfer of material in and out of the recirculation region is not limited to turbulent transport across the boundary. Material is also advected into the region along entering mean streamlines and advected out via vertical spiral vortices or through entrainment by horseshoe vortices. The entire recirculation region may even collapse intermittently causing all the contents to be flushed downstream. There is no simple and accurate way of assessing the fate of material released into the wake of a building. Dispersion models in regulatory use (ADMS, ISC, AERMOD) recognise that significant mixing takes place in a building wake and this can reduce concentrations downwind of the building by up to an order of magnitude compared with an unconfined release. The models do not attempt to describe events in the near wake and treat this as a well-stirred region of constant concentration i.e. any material leaving the building is predicted to undergo a step change in concentration. The models should not be used closer to the building than 3-5 times the extent of the recirculation zone. Typically the extent of the recirculation zone for squat shaped buildings scales as the building height and for tall thin buildings it scales as the building width. For a building of characteristic dimension 10 m this imposes an a region of modelling uncertainty of 30-50 m. This is of greater extent than the present guidelines for the external hazard radii. An alternative to regulatory models is to use experimentation or Computational Fluid Dynamics to explore some release scenarios. The disadvantage, in addition to that of cost, is that generalisation of the results is difficult. This particular problem is also difficult to solve with computational methods because of the physical complexity of the problem and the requirement for validation.

32

What can we say about the external hazard zone? i) An external hazard zone will only exist if the release within the building is sufficiently large to raise concentrations within the building above the LFL. For a building of size ~ 1000 m3 this implies pressurised releases have to exceed ~ 0.1 kg/s and assumes that the releases take place within the body of the building. ii) The efflux from the building will comprise a flow of volume DV ~ 3.33 m3/s for a building of size ~ 1000 m3 and with adequate ventilation and take place through the normal ventilation openings in the building fabric. iii) The efflux is a relatively small flow and will be mixed into the recirculation zone behind the building in the presence of wind. This mixing process is a result of highly unsteady flow and is wind speed and wind direction dependent but is highly effective. iv) Hazardous areas will be restricted to the immediate vicinity of vent openings and the building fabric. If releases are significantly larger than 0.1 kg/s then a greater hazard will pertain. For slightly higher flow rates then, if the efflux buoyancy (positive or negative) is not negligibly small then it will influence the dynamics of the ventilation flow. Specifically it will preferentially direct hazards to roof or floor and enhance ventilation and if the buoyancy driven outflow exceeds the ventilation flow a counter flow (additional ventilation) will be induced. For much larger flow rates then specific cases need to be considered. Containment and building effects will keep external concentrations low but events that create an opening in the fabric of the building, such as opening a door, that make greater ventilation possible would lead to the outflow of flammable material presenting a localised hazard.

Recommendations: Enclosure of facilities handling flammable materials should be subject to a hazard assessment to determine the level of adequate ventilation. The adequate ventilation recommended in the existing guidance is appropriate to pressurised releases of Cat A, G(i) fluid of up to ~ 0.1 kg/s for a 1000 m3 building provided that the reaction time to respond to a leak is less than ~ 700s. For an adequately ventilated building the hazard zone is confined to the immediate building fabric. Escaping vapour will be rapidly entrained and mixed into the recirculation zone behind the building. This is a region of unsteady but average reverse flow toward the building at ground level and can be considered to extend to a dimension given by the smaller of the building height or width. Conventional dispersion models are only accurate at about 3 -5 times this distance for neutrally buoyant releases. If ventilation is not adequate then hazard zones will extend at ground level or roof level depending on the buoyancy of the release. Buoyancy driven flow around vents/openings will probably enhance the ventilation flow.

33

If ventilation is far from adequate then changing the open area of the building may result in a large volume of potentially flammable material. The present guidance is conservative as far as shape factors are concerned but should relate the size of release and the volume of the building in some more explicit way. Notes: The case where a release takes place in a doorway arose in discussion and, by implication, exists in the present guidelines. The ventilation rate for structures with open doorways will probably be higher than the 12 changes per hour considered as adequate ventilation.

7. Discussion and Conclusions In this work we have used tools, developed to investigate the consequence of major releases, to quantify appropriate guidance for the Area Classification of Petroleum Installations. The quantification of hazard necessarily starts with specifying the type of material and the size of release which is very much unknown in the case of small spills and leaks. Material types have been simulated using 5 example fluid compositions coded (A, B, C, G(i) and G(ii) following earlier work to update IP 15 and published as an addendum a “Risk Based Approach to Area Classification”. That work utilised some very large flow rates that certainly exceeded any definition of normal spillage. The release rate values used here necessarily represent the lower end of the “hazardous release” scale. These should be larger than arise in normal handling and certainly should not be taken as indicative of the magnitude of “acceptable” spills. In all circumstances the potential for spills to occur should be rigorously assessed and a full hazard assessment carried out where necessary. It has generally been possible to defend the key recommendations of IP 15 as conservative. Where revisions are recommended these are strongly dependent on scenario and fluid type. The major findings of the study are that: x The hazard radii for pressurised releases of category B and C fluids should be derived assuming a mechanically generated flammable mist; previously gaseous releases were assumed. The Hazard radii for category B and C fluids are increased relative to previous guidance. x Numeric flammability limits published in Annex D of “A Risk-Based Approach to Hazardous Area Classification” for the category B and C fluids have been updated to take account of the composition of the flammable mist; previously low vapour components were assumed to rain-out and not contribute to the lower flammability limit evaluation. x Shape factors for pressurised releases are revised to take better account of the role of initial jet momentum on the jet trajectory. In particular the lighter than air gases (category G(i) and G(ii) fluids) are found to have qualitatively more similar shape factors to the twophase category A, B and C releases; previously buoyancy was assumed to dominate their dispersion. x Hazard radii for discharges from vents are evaluated. The hazard radius varies from slightly smaller to slightly larger than that in the existing guidance depending on the properties of the vented vapour. x The composition of vapour from vents on storage facilities maintained at atmospheric pressure may be variable (in composition, density and flammability) and the user of the new guidance should be aware of the effect of this variability because of the consequence for hazard radii.

34

x The example range of venting rate and vent sizes used in the guidance are not wholly consistent with the assumption that the discharge takes place at atmospheric pressure. The relationship between venting rate and pressure of discharge is investigated and a value of 300 mb suggested as a threshold above which the consequences of pressure should be assessed. This is of significance for multicomponent fluids where condensation may occur. x The existing guidance for liquid spillages is conservative, judged by the volatility of the model category C fluid. The guidance is applicable to materials with approximately twice the vapour generation rate of category C fluid under the specimen conditions. Relative vapour pressures for some common hydrocarbon compounds are listed. x The existing guidance for sumps is conservative, judged by the volatility of the model category C fluid. x Vapour generation at the source of spillage of Category C fluids is a potential hazard dictated by the spill rate and conditions and not the rate of evaporation of the liquid pool. The new guidance should emphasise the role of release conditions in determining the initial vapour generation from spills of category C fluids. x For releases into confined areas the relative size of spillage and building are of key importance. The classification “Adequate Ventilation” has been assessed with respect to these parameters. The major uncertainties remaining are: x The role of liquid rain-out in influencing flammability hazards for releases from low (~ 5 bar) processes. x How properly to advise the effect of different fluid compositions in hazards. x How to properly balance guidance for area classification and necessary procedures for hazard assessment.

8. References A Risk-Based Approach to Hazardous Area Classification, Institute of Petroleum, November 1998. Shell FRED 3.1 User and Technical Manual, Shell Global Solutions, (2000). HGSYSTEM v3.0; Technical Reference. Shell Global Solutions (1996) available in electronic form (pdf) at http://www.hgsystem.com/. Johnson D.W. and Woodward J.L.; Release: a model with data to predict Aerosol Rainout in Accidental Releases. AIChE. (1999). Ramsdale S., Tickle G. Review of Release rain-out model and the Center for Chemical Process Safety (CCPS) data, Contract Research report 277/2000, Publ. HSE Books. Lees F. Loss prevention in the Process Industries, ed 2. Publ. Butterworth/Heinemann, 1996. Bearman P.W., 1972, "Some Recent Measurements of the Flow Around Bluff Bodies in Smooth and Turbulent Streams", Paper presented at a synopsium on external flows, University of Bristol.

35

Britter, R.E. Hunt, J.C.R. and Puttock, J.S., 1976, "Predicting Pollution Concentrations Near Buildings and Hills." Presented at the Institution of Measurements and Control Conference on Systems and Models in Air and Water Pollution, London, Sep 22-24, pp 7-1 to 7-15. Castro, I.P., & Robins, A.G., 1977, "The Flow Around a Surface-Mounted Cube in Uniform and Turbulent Streams", J. Fluid Mech., vol. 79, part 2, pp. 307-335. Hosker, R.P., 1979, "Empirical Estimation of Wake Cavity Size Behind Block Type Structures." In the reprints of Fourth Symposium on Turbulence, Diffusion and Air Pollution, Reno, NV, Jan. 15-18, pp 603-609. Snyder, W.H. and Lawson, R.E., Jr., 1994, "Wind tunnel measurements of flow fields in the vicinity of buildings", Reprint vol.: 8th AMS Conf. on Appl. Air Poll. Meteor., with AWMA, Jan. 23-28, Nashville, TN.

36

Appendix A: Methodology A1. Calculating Flammability limits. The flammability limit of a mixture, L (%v), should therefore be computed by first using Le Chatelier’s law to establish the volumetric flammability limit of a vapour only mixture:

1 L

yi

¦L

i

Where yi denotes the mol fraction of the species i and Li the species vapour phase flammability limit. This flammability limit can then be converted to a mass-based criteria most simply by the assumption of an ideal gas whereby the mixture density is given by:

U mix

¦y M i

i

P kg/m3 RT

where Mi are the species molecular weights, P the pressure, R the ideal gas constant and T the temperature. The mass based limit is.

Lm

U mix

L kg/m3 100

The calculation procedure for flammability limits is embodied in the Shell FRED 3.0 (and higher) software tool available from Shell Global Solutions.

A2. Calculating hazard radii for pressurised releases. Hazard radii need to be calculated using a multicomponent and multi-phase dispersion model. Shell Global Solutions has developed such a capability as part of HGSYSTEM which is a set of dispersion models, developed at the specification of industry consortia, and made publicly available via the internet. HGSYSTEM has been used as a reference model in model evaluation exercises and the model components set a standard against which commercial models are tested. Model vendors should be able to relate the performance and physical basis of their models to HGSYSTEM. The calculations used the fluid properties given in Table 2 of the main report and the following conditions: Storage Temperature - 20 qC Ambient Temperature - 30 qC Wind Speed - 2 m/s Stability Class - D Roughness - 0.03 m Horizontal Release @ 5 m for R1 Horizontal Release @ 1 m for R2

37

Of these the surface roughness length 0.03 m has been increased from the value of 0.003 m used in the original guidance. The 2-phase jet calculations used for the hazard radii are not sensitive to this parameter so the change is semantic rather than critical. A value of 0.003 m is appropriate to extremely flat terrain with even and low lying vegetation. This is not characteristic of process sites where values as large as 0.3 m could in some circumstances be justified. The choice of 0.03 m is at the lower end of realistic values for process sites and is conservative. The role of roughness length in dispersion models is to characterise the ambient turbulence. This is the dominant mixing mechanism for low-momentum releases. For convenience, and following the methodology in the previous update to IP 15, the discharge rates were calculated using the Shell Global Solutions Generalised Release Model (GENREL) incorporated in FRED 3.1 and setting pipe friction to be negligible. The HGSYSTEM model AEROPLUME has a simpler release model and, for two phase releases, generates a range of possible mass flow rates for a set of input conditions rather than a single value. The GENREL and AEROPLUME suggested rates are, for most mixtures, comparable. The hazard radii were calculated using the two-phase jet dispersion model AEROPLUME. AEROPLUME calculates cross-sectional average concentrations. The hazard radius R1 should be taken as the distance at which the average concentration has fallen to a value of 0.7 LFL. The hazard radius R1 should be taken as the distance at which the average concentration has fallen to a value of 0.5 LFL. This arises because the hazard radius is evaluated with respect to the maximum concentration in the jet and the relationship between the maximum and average jet concentrations depends upon the shape of the concentration profile. Within FRED 3.1 the AEROPLUME results are automatically processed to take account of the concentration profiles and give hazard radii directly and in a manner transparent to the user. We note that a key result of this study, that a mass-based flammability limit should be used for two-phase flow calculations; is not implemented in the post-processor tools supplied with HGSYSTEM 3.0. The post-processor PROFILE from v 3.0 should not be used for high flash-point fluids. A3. Calculating Hazard Radii for vents. Ambient conditions for calculating the hazard radii for vents were the same as for the pressurised releases. The key feature of the vent flows is that they are assumed to be ideal gases, i.e. no account is taken of the potential for two phase flow. The dispersion calculations were carried out using AEROPLUME assuming a vertical release. The hazard radius was defined as:

R

 Z  H

2

 X2



where H was the assumed vent height of 10 m and Z, X the vertical and downstream coordinates of the point of maximum concentration in the jet equal to the flammability limit. This is the centroid position of the jet calculated by AEROPLUME with an average concentration equal to 0.7 LFL.

38

A4. Calculating pool evaporation and dispersion The pool evaporation calculations were carried out using the liquid spills model LPOOL which is part of HGSYSTEM. LPOOL is an implementation of the ExxonMobil Liquid Spills Model LSM90. It takes account of the multicomponent evaporation of material either spilled into a bund or onto an unconstrained surface. The basis of pool evaporation models is set out in Lees (1996). The evaporation rate is based on a forced convection analogy and increases with wind speed and with pool area. For non-boiling pools the evaporation rate increases with the vapour pressure of the spilled liquid. A pool model adds a liquid spread, a heat balance and a component balance to the base evaporation correlation. For consistency with the other hazard radii a wind speed of 2 m/s wind speed and an ambient temperature of 30 qC was used. To enhance the evaporation rate further, maximum insolation (spill at mid-day, cloudless sky) was assumed. The insolation factors in LPOOL are conservative (high) for mid-latitudes. As noted in the text, pool evaporation is a transient process that gives rise to a time-varying gas cloud. In particular we found that the highest evaporation rates came during the first moments of the spill when the lightest components were evaporating. Calculations made using the HGSYSTEM unsteady dense gas dispersion model HEGADAST showed that the vapour cloud dispersed extremely quickly and, with the target conditions, it was not possible to attain the hazard radii in the previous guidance. We then carried out a sensitivity test with the HGSYSTEM steady state dense gas dispersion model HEGADAS-S to determine what continuous release rates of dense gas would be consistent with the existing guidance. We found, as given in the main text, that these rates are higher than the vapour rates for category C spills.

39

Appendix B: Preliminary Investigation of Hazard Radii and Shape Factors for the revision of IP15: the Area Classification Code for Petroleum Installations. Summary As part of the revision of the Institute of Petroleum publication “Area Classification Code for Petroleum Installations”, more commonly known as IP 15, dispersion calculations were carried out to cross check tables D3.2 and D3.3 of the publication “A risk-based Approach to Hazardous Area Classification”. The objectives of revisiting the calculations were twofold: x To provide a subset of results more in line with petroleum installation needs. x To verify shape factors for releases - especially those representing a refinery hydrogen stream. In repeating the calculations it was noted that the usual practice of specifying flammability limits in terms of volumetric concentrations for gases leads to possibly incorrect and too low estimates of hazard distances for class B and class C liquids. These are materials that have high flash points at ambient conditions but may be placed under high pressure when pumped from one location to another. The leak of high pressure liquid may give rise to a flammable mist or, in any event, lead to the jetting of potentially flammable liquid over distances much greater than those given in the present IP 15 guidance. This note describes the revised calculations and discusses the problem of how to assess the hazard range for high flash-point fluids. The calculations in this report were carried out with the HGSYSTEM models DATAPROP, AEROPLUME and PROFILE to facilitate the presentation of the many results. These models are equivalent to the components of the FRED model version 2.3. As a result of this work the definition of flammable limit used in deriving contours with the software package FRED has been changed. FRED versions 3.0 and higher are compatible with these results. B1. Introduction Hazardous area classification requires that an assessment be made of the extent of the zone enclosing an operational area where flammable materials are handled and where small leaks might occur unnoticed for a short period of time. With this knowledge precautionary steps can be taken to reduce the potential for ignition, for example by placing constraints on electrical installations. The Institute of Petroleum (IP) publication: “A risk-based Approach to Hazardous Area Classification” adopted (in Annex D) tables of hazardous distances from an internal Shell source, which contained calculations made with the FRED dispersion model, version 2.1, that were appropriate to an exploration and production environment where high pressures and hence high flow rates are the norm. These tables included events that would only arise as a result of mechanical damage on a petroleum refinery and which would in any case demand an emergency response. Including these events as part of advice on Area

40

Classification gives a misleading impression of the hazard distance to be associated with normal operation. This work was intended to rework the content of current IP publication advice and verify that the shape factors (Figure 6.2 and Figure 6.3 of IP 15) for dense and buoyant gas releases were realistic. It was expected that the advice would be essentially unchanged. However, It was noticed that the published data gave very small (more than ten times smaller) hazard distances for the Class Cvi fluid, compared with the Class A fluid; for example: Release Conditions Pressure (bar) Diameter (mm) 100 10

Hazard radius (m) category A category C 39 3

Table B.1 Extract from Annex D of “A Risk Based Approach to Hazardous Area Classification” Physical intuition suggests that, to take the example above, a the hazard radius of 3 m is far too small compared with the likely “throw” of a non-volatile but flammable liquid jet driven by a 100 bar pressure through a 10 mm hole. There have been discussions and project proposals to investigate the “throw” of water jets for area clarification and electrical safety purposes and unpublished work carried out for ICI has shown that small holes and low drive pressures can effectively transport material well beyond the present category C fluid guideline distances. Pressurised releases can also atomise material with a high flash-point to give rise to mists that are flammable. The flammability limits for fine aerosols are akin to those for vapours although there is a paucity of informational data in the literature. It is believed that, in general, the flammability limit of fine aerosols, when expressed as the mass of fuel per unit volume, is similar to the flammability of the vapour alone (Lees, F.J. Loss Prevention in the Process Industries, 2nd ed., vol. 2, section 16.4.3). This is significant because it is common practice is to use volumetric units to express flammability limits for gases and for mixtures of gases. For a given mass concentration the equivalent volume fraction is substantially reduced in the presence of a small liquid component because of the very large density difference between the two phases. The use of a volumetric criterion for evaluating the hazard distance may therefore be misleading when two-phase mixtures are involved and, worse, give too small an estimate of the hazard distance. In this note we compare the use of mass and volume based flammability criteria for establishing hazard zones on the assumption that releases of high flashpoint material will form a flammable mist and hence that their hazard can be addressed in a comparable manner to gaseous and low flash-point 2-phase releases. The issue remains of whether more coherent jets exist and are able to deliver flammable liquid to an even more distant target but we do not attempt to quantify this here. We should also comment upon the mitigating effect of liquid droplet rainout from jet releases. For high flashpoint liquids that are atomised and ejected as jets the process of droplet collision will lead to a growth in the droplet size with time and the subsequent deposition of liquid. An extensive body of work has been carried out for single component fluids by the American Institute of Chemical Engineers (AIChE) and a release model developed to predict the loss of liquid from such events (Johnson, D.W., Woodward, J.L., Release: A model with data to predict Aerosol Rainout in Accidental Releases, AIChE, 1999). For area classification purposes we are considering material transport over very short distances and with transit times of a few seconds and over trajectories that are vi

Model fluid compositions are given in Table B2,

41

determined almost entirely by the initial jet momentum and orientation. To give conservative estimates of the hazard zone it is reasonable to assume that liquid droplet rainout in the vicinity of the source is a second order effect. For the larger releases that appear in risk assessment scenarios where fluid transport times of several minutes may occur, the AIChE work should be reviewed to see if accounting for droplet rainout makes a substantial difference to model predictions. B2. Calculation Procedure. Of necessity only a limited number of variables can be considered in this investigation. The area classification process is NOT a full hazard assessment which takes account of the range of meteorological and process conditions, release events etc. Accordingly a single set of release conditions was chosen and these are the same as those used in the IP addendum “A risk-based Approach to Hazardous Area Classification” with the exception that the surface roughness length has been increased by a factor of 10 from an unreasonable overland value to a value that is still low (by a factor of 10) compared to that used for process sites. This has no significant effect on any of the calculations but makes for a more physically realistic input data set. The fluid compositions used are given in Table B2 and the base calculation conditions in Table B3. The two parameters in Table B3 that were varied to investigate the shape of the hazardous zone were the release orientation and the release height. A reduced set of operating conditions were assumed. Four drive pressures (100, 50, 10, 5) bara. and five hole sizes (1, 2, 5, 10 , 20) mm (as diameters). The shape factors for the releases were calculated using the largest hole size and the greatest drive pressure.

42

Stream Component (mol perc) N2 Nitrogen C1 Methane C2 Ethane C3 Propane C4 Butane C5 Ethane C6 Hexane C7 Heptane C8 Octane C9 Nonane C10 Decane H2O Water CO2 Hydrogen Average MW LFL (vol%)

Cat. A

Cat. B

Cat. C

Cat. G Cat. G Comp. (I) (ii) LFL

MW

0 0 0 70 30 0 0 0 0 0 0 0 0 0 48.3

0 4 0 6 7 9 11 16 22 0 25 0 0 0 100.06

0 0 0 1 1 2 3 3 27 25 38 0 0 0 125.03

2 88.45 4.5 3 1 1 0 0 0 0 0 0.05 0 0 18.74

2 10 3 3 1 0 0 0 0 0 0 0 1 80 7.03

28.01 16.04 30.07 44.09 58.12 72.15 86.17 100.20 114.23 128.26 142.28 18.02 44.01 2.02

2.09

1.70

1.52

4.82

4.03

5.3 3 2.2 1.86 1.50 1.2 1.2 0.95 0.83 0.77 4.1

Boiling point o C -196 -161 -87 -42 -1 36 69 98 126 151 173 100 -78(sub) -253

Table B.2 - Definition of category Fluids and their properties. The vapour phase mixture flammability limits are taken from the IP addendum publication and were evaluated using Le Chatelier’s Law. Standard conditions Ambient temperature Relative humidity Wind speed Reference Height Stability class Surface roughness Sample time Release height Reservoir Temperature Release angle (relative to horizontal)

Base Case values 20 o C 70 % 2 m/s 10.0 m D 0.03 m 18.75 s (~ instantaneous) 1.0 m 20 o C 0o

Range of values

0 to 15 m -30 o to 90 o

Table B.3 - Base parameters for the calculation Release rates for the five test fluids were evaluated using the Generalised Release Model from FRED 2.3 and assuming negligible friction losses from any pipe runs. The release rates calculated in this way have proved more consistent than the default release rate used by the Aeroplume model which is a literature correlation for two phase flow. Aeroplume also has a built in two phase discharge mode, used for advice purposes under normal program execution, and this usually agrees to within a few percent of the Generalised Release Model predictions. As expected it was found that the release mass flux per unit area was hole size independent for the release conditions and the release rates are plotted as a function of hole size and pressure for each of the fluids in Figure B2. The results are straightforward except for category A fluid at the lowest drive pressure. Here the standardised 5 bar drive pressure at ambient temperature is below the saturated vapour pressure of propane and above the 43

bubble point of propanevii. This means that in a vessel the category A vessel would exist as discrete two phases giving three discharge scenarios: a vapour only release of a propane rich vapour; a liquid only release of a butane rich butane/propane mix and a two phase release; all depending on the location of the release point. Any significant length of piping between the storage vessel and the release would favour a two phase release because dissolved propane would vaporise in the pipe as the pressure decreases. We have therefore carried out two sets of calculations for this low pressure scenario. A high mass rate discharge of a 45:55 propane/butane mixture corresponding to the liquid phase composition of the 70:30 propane/butane total mixture and a low mass rate discharge of the original 70:30 propane/butane mixture as a 2 phase release. Figure B1 shows mass flow rates for both circumstances.

Release Rate as a function of hole size and drive pressure Category A Fluid

Release Rate kg/s

10

100 bar 50 bar 10 bar 5 bar - liquid 5 bar - 2 phase

1

0.1

0

2

4

6

10

8

12

14

16

18

20

Hole size (dia.), mm

Figure B.1: Release rate as a function of hole size and drive pressure for each of the five conditions, category A fluid.

vii In the Addendum to IP 15 it was recognised but not explicitly stated that the 5 bar scenario was ambiguous. The approach there was to decrease the ambient temperature for this run and hence the saturation vapour pressure of propane.

44

Release Rate as a function of hole size and drive pressure Category B Fluid

Release Rate kg/s

10

100 bar 50 bar 10 bar 5 bar

1

0.1

0

2

4

6

8

10

12

14

16

18

20

18

20

Hole size (dia.), mm

Release Rate as a function of hole size and drive pressure Category C Fluid

Release Rate kg/s

10

100 bar 50 bar 10 bar 5 bar

1

0.1

0

2

4

6

8

10

12

14

16

Hole size (dia.), mm

Figure B.2: Release rate as a function of hole size and drive pressure for each of the conditions, categories B and C fluids.

45

Release Rate as a function of hole size and drive pressure Category G(i) Fluid

100 bar 50 bar 10 bar 5 bar

Release Rate kg/s

10

1

0.1

0

4

2

6

10

8

12

14

16

18

20

18

20

Hole size (dia.), mm

Release Rate as a function of hole size and drive pressure Category G(ii) Fluid

100 bar 50 bar 10 bar 5 bar

Release Rate kg/s

10

1

0.1

0

2

4

6

10

8

12

14

16

Hole size (dia.), mm

Figure B.3 - Release rate as a function of hole size and drive pressure for each of the conditions, categories G(i) and G(ii) fluids.

46

The AEROPLUME model was run using automatically generated input files for the different operating conditions. AEROPLUME outputs a cross-sectional average jet concentration. To derive concentration contours the output of the AEROPLUME model can be post-processed using the PROFILE program. This superposes a Gaussian profile upon the AEROPLUME results and constructs the required contour. This is the normal operating procedure in the FRED model where the results are automatically displayed as a graphic and the underlying calculations are not visible to the user. It is important to note that PROFILE assumes a ground-reflected Gaussian distribution which means that the relationship between the maximum plume concentration and the average plume concentration depends upon the height of the plume above ground. Where we are just interested in the furthest extent of the maximum LFL concentration from the source we can estimate this directly from AEROPLUME by calculating the distance to a slightly smaller concentration. The ratios are shown in Table B.4.

Elevated Plume Grounded Plume

Peak Concentration LFL LFL

Averaged Concentration 0.7 LFL 0.5 LFL

Table B.4 : Equivalence between peak and averaged concentrations in the HGSYSTEM and FRED models.

B3. Effect of using mass and concentration based LFL criteria: The mass equivalent LFL was derived from the vapour properties in Table B.2 by multiplying the vapour phase volume fraction by the vapour phase material density assuming an ideal gas, one atmosphere pressure and the ambient temperature. The downwind LFL position on a mass and on a volume basis was determined by linear interpolation of distance between bounding entries in the AEROPLUME output file and using the ratios given in Table B.4. This is adequate for the purposes of this study. Figure B.4 shows a composite of all results for a 1 m horizontal release. The distance to LFL calculated using the mass concentration and the distance to LFL using the volume fraction criteria are plotted against each other. x For the two gas mixtures (category G(i) and G(ii) ) the flammability limits equivalent and an identity results. In the presence of a liquid phase, volumetric concentrations underestimate the amount of flammable material present because the contribution of liquid to volume is negligible. We see: x For the low flashpoint material (fluid A) sufficient liquid has evaporated for the jet to be wholly gaseous at LFL so the two flammability definitions and calculated LFL distances are equivalent x For fluid category B the volume based definition noticeably understates the hazard distance by ~ 30%. x For fluid category C there is almost an order of magnitude difference in the hazard radius.

47

Influence of fluid type on LFL definition

Distance to LFL using volume fraction

100

10

A B C Gi Gii

1

0.1 0.1

1

10

100

Distance to LFL using mass concentration

Figure B.4 : Effect of using mass concentration and volume concentration definitions on distance to LFL for the five fluids. This is a significant finding. It suggests that using volume based flammability criteria to derive distances to LFL may lead to underestimating the hazard of two-phase releases. To illustrate this more clearly on a linear scale. Figure B.5 shows some example shape factor calculations for the category B fluid. An elevated release (15 m height) is assumed and calculations carried out for different release directions. The resulting overlapping cigarshaped contours are results that would be obtained by following normal practice and using volumetric limits (e.g. with the FRED 2.3 model). The points show where the maximum distance to LFL occurs using the mass based criterion.

48

IsoConcentration contours for CATB fluid, 100 bara, 20 mm hole Releases at different inclinations 50

Values from Aeroplume

40

Height, m

30

20

10

0

0

10

20

40

30

50

Downwind distance, m

Figure B.5 - Visual Example of the difference in using volumetric (contours) and mass concentration definitions (points) for the LFL of a category B fluid.

B.4 Shape factors Shape factors for the hazardous area are based on calculations of jet dispersion in different directions and the existing classification advice implicitly assumes that, because of the effects of body forces, the shape factor for buoyant releases is different to that for dense gas releases. This work shows that the hazard radius for the different fluids is determined by jet momentum rather than body forces, but that there is an effect of the ground on entrainment which is different for dense and for buoyant releases. Figure B.6 shows the shape factor for a category A fluid released from a 1 m height (maximum flow rates are used in all of these calculations). It is assumed that the wind direction is from the left of the picture. For a category A fluid there is no effect of liquid load on the calculation procedure and we have included both the full iso-concentration contours as would be obtained using the FRED model and the points derived from the AEROPLUME model using the mass concentration definition of LFL. For releases that do not touch the ground the hazard range is essentially a flattened circular locus. An increasing wind speed would flatten the shape further by decreasing the vertical penetration of the jet because the contribution to entrainment from cross-wind mixing is increased.

49

For releases that do touch the ground the dispersion distance is increased on two counts: x Within the assumptions of the jet model, the overall entrainment is reduced. Initially the perimeter area of the jet is reduced by the physical presence of the ground and then, as it spreads out over the ground and slows down, the entrainment is inhibited by the density difference between the plume and the air. x The reflected Gaussian shape of the concentration distribution means that the ratio of the peak concentration to the average concentration within the jet changes when the jet hits the ground. For an elevated plume the peak concentration is ~1.42 times greater than the average concentration across the plume and occurs on the plume centroid. For a grounded plume the peak concentration is ~ 2 times the averaged concentration and occurs at the ground. The locus of points from AEROPLUME shown in Figure B.6 reflect this assumption.

IsoConcentration contours for CATA fluid, 100 bara, 20 mm hole Releases at different inclinations 50

Centroid Values from Aeroplume Peak Values from Aeroplume

40

Height, m

30

20

10

0

0

20

60

40

80

100

Downwind distance, m

Figure B.6 : LFL isopleths for a category A fluid released at angles from -20 to 90 degrees to the horizontal and centroid position of LFL from Aeroplume (points). The ground level LFL concentration position from Aeroplume for the three lowest trajectory releases are shown as vertical bars. These two factors substantially increase the hazard distance for horizontal releases near to the ground. This methodology is conservative because extra mixing caused by the friction between the jet and the ground is neglected. Also it results in a step change increase in hazard radius for releases that only just touch the ground as shown in Figure B.7 below.

50

IsoConcentration contours for CATA fluid, 100 bara, 20 mm hole Releases at different inclinations 50

Values from Aeroplume Peak Values from Aeroplume

40

Height, m

30

20

10

0

0

20

60

40

80

100

Downwind distance, m

Figure B.7 : Shape factor for an elevated release of category A fluid Figure B.7 shows the effect of increasing the release height to 15 m. Calculations to the required concentrations can only be made for release angles greater than -20 degrees with the present version of the AEROPLUME model which does not model steep impacts of high momentum jets. We do not expect these steeper impacts to lead to greater dispersion distances as the impact will lead to the generation of additional turbulence which will increase the mixing rate. Figure B.8 shows the shape factor for the release of category G(i) fluid from a height of 1 m We find that the distance to LFL along the plume centroid is nearly independent of release angle. The distance to LFL at the ground for those releases that hit the ground is again greater than that of the elevated jet so that the overall shape factor is similar to that of a dense gas release. The effect of the ground is less dramatic because the spreading fluid is less dense than the air and is convectively unstable. This means that entrainment into the grounded lighter than air gas is greater than it would be for a dense gas. Consequently the hazard distance at the ground is not as great.

51

IsoConcentration contours for CATG(i) fluid, 100 bara, 20 mm Releases at 90, 70, 50, 30, 10, 0,-10 degrees from the horizontal 20

Centroid Values from Aeroplume Ground level Values from Aeroplume

15

10m Height,

5

0 0

5

10

15

20

Downwind distance, m Figure B.8 : Shape factor for the category G(i) Fluid. Points are centroid distance to LFL and the vertical bars show the ground level distance to LFL for the two lowest trajectory releases. The category G(ii) fluid is markedly more buoyant than the category G(i) fluid and to investigate whether the extra buoyancy significantly modifies the shape factor these have been derived for release heights of 1m, 5 m, 10 m and 15 m. Figure B.9 shows that, for a 1 m release height the results are qualitatively very similar to those for a category G(i) fluid. The grounded jets entrain air more efficiently because the flow is more unstable and there is less enhancement of the distance to LFL at the ground.. Increasing the release height to 5 m barely alters the shape factor. As the release height is increased to 10 m the basic shape is still maintained with a downward directed jet at -30 degrees hitting the ground before diluting to LFL. The ground footprint is very similar to that for the 1 m release. Increasing the release height to 15 m again results in a fan shaped hazard range that is almost symmetric about the release height showing that, in the absence of ground contact, the hazard radius is determined by the jet orientation. This contrasts with Figure 6.3 of the IP 15 publication that suggests that there is a smaller hazard zone below the release point than above it.

52

LFL Concentration contours for CATG(ii) fluid 100 bara, 20 mm hole 35 30

Centroid LFL position from Aeroplume Groundlevel LFL from Aeroplume Release height: 1 m

25 20 Height, m 15 10 5 0 0

5

10

15

20

25

30

35

Downwind distance, m LFL Concentration contours for CATG(ii) fluid 100 bara, 20 mm hole 35 30

Centroid LFL position from Aeroplume Groundlevel LFL from Aeroplume Release height: 5 m

25 20 Height, m 15 10 5 0 0

5

10

15

20

25

Downwind distance, m

53

30

35

LFL Concentration contours for CATG(ii) fluid 100 bara, 20 mm hole 35 30

Centroid LFL position from Aeroplume Groundlevel LFL from Aeroplume

25 Height, m

Release height: 10 m

20 15 10 5 0 0

5

10

15

20

25

30

35

Downwind distance, m LFL Concentration contours for CATG(ii) fluid 100 bara, 20 mm hole 35

Centroid LFL position from Aeroplume

30 25 Height, m

Release height: 15 m

20 15 10 5 0 0

5

10

15

20

25

30

35

Downwind distance, m

Figure B.9 : Shape Factors for a release of Cat. G(ii) fluid released from a height of 1 m, 5m , 10 m and 15

m above the ground. Points indicate the plume centroid position at LFL and the vertical bars the ground level distance to LFL for the three lowest trajectories all derived from the Aeroplume model. The isopleths are calculated with the Profile model and are equivalent to the output from the FRED 2.3 model.

54

B.5 Hazard Radii The calculations carried out for each of the releases suggest that for elevated jets the shape factor should comprise a semi-spherical hazard zone evaluated as the distance to LFL on the plume centroid for a horizontal jet in the downwind direction. The calculations given in Tables B.5 & B.6 show that the effect of the ground is significant for low level releases of all materials and therefore the shape factor should be extended at the ground.

Distance to LFL at the plume centroid height for a 1 m high horizontal release, m Fluid Category

Pressure

20 mm

10 mm

5mm

2 mm

1mm

100

59.6

27.8

11.1

4.0

2.0

50

59.8

28.0

10.9

3.9

2.0

10

60.2

28.4

10.8

3.4

1.8

5

63.5

29.2

10.9

3.1

1.7

100

39.2

17.4

7.0

2.9

1.5

50

39.0

17.2

6.7

2.8

1.4

10

30.0

12.5

5.0

2.1

1.1

5

23.4

8.6

3.9

1.7

0.9

100

36.6

16.0

6.6

2.7

1.4

50

36.7

16.0

6.4

2.7

1.4

10

37.0

16.4

5.8

2.5

1.3

5

37.2

16.6

5.3

2.4

1.3

100

13.2

6.4

3.3

1.3

0.7

50

8.4

4.2

2.1

0.9

0.4

10

3.5

1.8

0.9

0.4

0.2

5

2.6

1.3

0.7

0.3

0.2

(bara) A

B

C

G(i)

G(ii)

100

15.6

9.3

4.9

2.0

1.0

50

12.3

6.8

3.5

1.4

0.7

10

6.3

3.3

1.6

0.7

0.3

5

4.7

2.4

1.2

0.5

0.3

Table B.5 : Summary of Distances to LFL at the plume centroid height for the horizontal release from a height of 1 m.

55

Distance to the LFL at Ground Level for a 1 m high horizontal release, m. Fluid Category

Pressure

20 mm

10 mm

5mm

2 mm

1mm

100

83.0

39.8

17.1

5.1

2.7

50

83.1

40.2

17.1

4.9

2.6

10

85.5

41.0

17.2

4.1

2.3

5

93.5

42.7

17.3

3.7

2.1

100

54.6

25.2

9.9

3.8

1.9

50

54.2

25.0

9.6

3.6

1.9

10

42.0

18.8

6.3

2.7

1.4

5

33.1

13.8

4.8

2.1

1.1

100

50.7

23.2

9.1

3.5

1.8

50

50.9

23.4

8.8

3.4

1.8

10

51.0

23.8

8.4

3.1

1.6

5

52.3

24.0

8.3

2.9

1.6

100

18.7

9.2

4.4

1.8

0.9

50

12.0

5.7

2.9

1.2

0.6

10

4.9

2.5

1.2

0.5

0.3

5

3.6

1.8

0.9

0.4

0.2

100

21.3

12.8

6.6

2.8

1.4

50

16.7

9.3

4.7

2.0

1.0

10

8.6

4.4

2.3

0.9

0.5

5

6.3

3.2

1.7

0.7

0.4

(bara) A

B

C

G(i)

G(ii)

Table B.6 : Distance to LFL at ground level for the horizontal release from a height of 1 m.

B6. Conclusions The purpose of this study was to verify shape factors and to cross check tables D3.2 and D3.3. of the publication “ A risk-based Approach to Hazardous Area Classification” with a set of conditions more in line with petroleum installation needs. Dispersion calculations were carried out for 5 types of fluids (Categories A, B, C, G(i) and G(ii)) under a range of pressure and release rate conditions. We found that: The usual practice of specifying flammability limits in terms of volumetric concentrations for gases leads to possibly incorrect and too low estimates of hazard distances for class B and class C liquids. These are materials that have high flash points at ambient conditions but may be placed under high pressure when pumped from one location to another. The leak of high pressure liquid may give rise to a flammable mist or, in any event, lead to the jetting of potentially flammable liquid over distances much greater than those given in the present IP 15 guidance. Shape factors for the hazardous area, based on calculations of jet dispersion in different directions, show that the hazard radius for the different fluids is determined by jet momentum rather than body forces. Consequently, the shape factor for buoyant releases is qualitatively similar to that for dense gas releases.

56

For releases that do not touch the ground, the hazard radius is nearly independent of release angle and the hazard range is essentially a circular locus, slightly flattened at the top owing to the wind. For a buoyant elevated release, the hazard radius is almost symmetric about the release height. This contrasts with Figure 6.3 of the IP15 publication that suggests that there is a smaller hazard zone below the release point than above it. We recommend that: Mass based flammability criteria should be used for multiphase releases. The Shell Global Solutions FRED model (version 3.0 and higher) has been modified on this basis. The shape factor should comprise a semi-spherical hazard zone evaluated as the distance to LFL on the plume centroid height for a horizontal jet in the downwind direction. There should be an extra margin of allowance for grounded jets.

57