of Ind. aaiHazardAss_ess-ent Tech.l _u _

5-7~~~~~~~

.-t;; .ir_ t .44FAN1_>.PM -:

1.~~~~~~~~~~J

- - * -~ ~ ~ ~ ~ ~ -- _ _ ' VK _, ''<

__t_____.~~~~~~~~~~~~~~~~~~~~~~~~~ _W _

Office c.f Emr-L-onmental and Scie;ntific Affa:rs _. T 4rd

1-146-1 Bank~~~~~~~~~~~~~~~~~

-t4 L 1E

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

MANUAL OF INDUSTRIAL HAZARDASSESSMENT TECHNIQUES

WORLD BANK

OCTOBER 1985This manual has been prepared incoooeration with the World Bank by:TECHNICA LTD., LONDON, ENGLANDVers: 1.0 Edited by P.J. Kayes

MANUAL OF INDUSTRIAL HAZARD ASSESSMENT TECHNIQUES

FOREWORD:

As an integral part of the appraisal and supervisionfunction for industrial development, the World Bank andInternational Finance Corporation (IFC) are required toevaluate the adequacy and effectiveness of the measures tocontrol major hazard accidents affe^tin.g people and theenvironment outside the plant boundary. For. this purpose,the Office of Environmental and.Scientific Affairs hasdrawn up "The World Bank Guidelines for Identifying,Analysing and Controlling Major Hazard Installations inDeveloping Countries."

In order to implement these guidelines, it is necessaryto carry out a hazard analysis of the proposed installationtvo determine the potential damage of accidental releases oftoxic, flammable or explosive materials from the proposedinstalla.tion. From the results of this haza'rd analysis,potential causes and sources of major hazards are identifiedand ranked according to the degree of severity. For thosefailures that would cause major damage of loss of life on oroff the plant site, the-first objective is to reduce themagnitude of the potential damage through the introductionof alternative processes or process changes, the reductionin hazardous inventories, the provision of robust secondarycontainment systems, the modification of site layouts, theidentification of alternative sites and the optimisation ofcontrol and management techniques.

If it is not possible to reduce the magnitude of thehazards by these methods, a risk analysis may be required todetermine if the probability of the hazardous event can bereduced through process changes, additional or improvedsafety systems, improved training or testing and maintenanceprocedures and soforth. In the final analysis, it may beshown that the particular process, storage requiremerts, orsite that has to be selected "or the plant are incompatibleand that a new site has to be selected. These hazard andrisk assessment methodologies can be applied also toexisting operations as well as to rehabilitation orexpansion projects.

This manual provides as far as possible the latestsimplified techniques used in the chemical industry toassess the consequences of major hazard accidents releasingtoxic, flammable And explosive materials into the atmos-phere. A spreadsheet methodology has been devised tosimplify hand calculations on scientific calculators whereaccess to computers may not be available to the user of thismanual. The spreadsheet methodology also simplifiescomputer applications of the manual., but more complex model-ing procedures may be prepared for programming on microcomputers.

While this manual has been prepared primarily forapplication to World Bank and IFC projects, themethodologies which are presented have wide application inthe chemical industry and its use by others is welcomed andencouraged. Further informaticn concerning the environ-mental and health and safety activities of the World Bankare available by writing to:

Office of Environmental and Scientific AffairsThe World Bank

1818 R Street, N.W.Washington D.C. 20433

U.S. A.

CONTENTS

Page No.

* CHAPTER 1 INRODUCTION

CHAPTER 2 HOW TO USE THIS MANUAL

2.i Background2.2 Description of Steps

CHAPTER 3 FAILUR.E CASES

3.1 Background3.2 Release Cases3.3 Selecting the Event

Sequences

CHAPTER 4 CONSEQUENCE CALCULATIONS

4.1 Outflow Calculations4.1.1 Liquid Outflow4.1.2 Gas Outflow4.1.3 Two Phase Outflow

4.2 Behaviour ImmediatelyAfter Release

4.2.1 Spreading Liquid Release4.2.2 Jet Dispersion4.2.3 Adiabatic Expansion

4.3 Dispersion in the Atmosphere4.3.1 Dense Cloud Dispersion4.3.2 Passive Dispersion4.3.3 Buoyant Surface Release

1.

Page No.

4.4 Fire4.4.1 Pool Fires4.4.2 Jet Fires4.4.3 Fireballs4.4.4 Flash Fires4.4.5 Fire Damage

4.5 Explosions4.5.1 Explosion Correlation

CHAPTER 5 SUMMARY OF EFFECTS

5.1 On-site Hazards5.2 Off-site Hazards

CHAPTER 6 APPLICATION OF THEMETHODOLOGY

6.1 Knock-on Effects

6.2 Reduction of Consequences6.2.1 Reduce Inventories6.2.2 Modify Process or Storage

Conditions6.2.3 Elimination of Hazardous

Material6.2.4 Improve Plant Operability

and Reliability6.2.5 Other Protective Measures

6.3 Reduction of Impacts

CHAPTER 7 HAZARD AND OPERABILITYSTUDIES (RAZOP)

7.1 Background7.2 HAZOP Method7.3 Further Information

ii .

LIST OF FIGURIS

Page No.

Figure 2.1 Guide to the Use of theManual

Figure 3.1 Pipe

Figure 3.2 Flexible Connection

Figure 3.3 Filter

Figure 3.4 Valve

Figure 3.5 Pressure Vessel/ProcessVessel

Figure 3.6 Pump

Figure 3.7 Compressor

Figure 3.8 Storage Tank (Atmospheric)

Figure 3.9 Storage Vessel(Pressurised/Refrigerated)

figure 3.10 Flare/Vent Stack

Figure 3.11 Failure Case Definition

Figure 3.12 Event Tree for FlammableGas Releases

Figure 3.13 Event Tree for LiquidRelease: Flammable orToxic

Figure 3.14 Event Tree for Toxic GasRelease

il.l.

LIST OF TABLES

Page No.

Table 4.1 Atmospheric DispersionParame ters

Table 4.2 Damage Caused at VariousIncident Levels ofThermal Radiation

Table 4.3 Explosion Limit Values forVarious Characteristic Typesof Damage

Table 4.4 Transformation of Percentagesto Probits in ToxicityCalculations

Table 5.1 HAZAN Results Proforma

iv.

LIST OF APPENDICES

Page No.

Appendix 1 Nomenclature and Glossaryof Terms

Appendix 2 Extract of World BankGuidelines for Identifying,Analysing and ControllingMajor Hazard Installationsin Developing Countries

Appendix 3 Summary of PotentialSources of Ignition

Appendix 4 Selected Data and theProperties of SomeHazardous Materials

Appendix 5 Bibliography

V.

:hiapter

Introduction

< tt I~~~~~~~~~~~~V

1. INTRODUCTION

The chemica and energy industries throughout the worldutilise a wide variety of manufarturing, storage and controlprocesses. These processes may. involve many different typesof raw materials or intermediate chemicals as part of theproduction of products required by both industrialised anddeveloping countries alike. Some of the materials used maybe potentially harmful if released into the environment dueio inherent toxic, flammable or explosive properties. Inthe technical processes involved in the modern chemicalindustries and energy, these materials may also be subjectto elevated temperatures and/or pressures; similarly, inorder to facilitate storage active refrigeration ofcondensable gases may also be employed.

Under these circumstances it is essential that highstandards of plant design, management and integrity.can beachieved and maintained. Indeed, in the context of thelarge quantities of t.te potentially hazardous materialswhich are now handled on a routine basis, it is clear thateffective methods have been developed to ensure adequatecontrols and safeguards in the facilities using suchmaterials. Nonetheless, accidents do occur and these mayhave very serious consequences upon the employees, among thepublic and to property. It is therefore a high prioritythat potential hazards are properly identified and accountedfor in the assessment of design and development proposalsfor such plant.

In order to conduct a sensible analysis of thepotential hazards associated with accidental releases oftoxic, flammable or explosive material, it is important tofollow a structured approach applying methods of calculationfor the estimation of such matters as discharge rate,di.spersion and effect distances which will be straight-forward to use and of adequate reliability. In the initialstages of an hazard assessment, it is appropriate to applysimplified techniques which may assist in the ordering andranking of potential impacts; subsequently, if necessary,more refined calculations may be conducted to assist inplant optimisation.

1

This manual has been prepared with the aim o*f providingthe framework necessary for the structured identification ofmajor hazards. The manual employs simplified formulae use-ful in guiding the calculation of potential effect distancesor damage ranges; it is intended to provide the minimumbasis for any initial assessment utilising tested techniqueswhich have been found to be effective when applied to avariety of petrochemical and process plant.

The procedures described in these manual are intendedfor use by the hazard analyst in conjunction with such othermethods of saf'ety assessment as may be appropriate for theolant under exmination. These include such techniques asHazard and Operability (HAZOP) studies and Failure Mode andEffective Analysis (FEMA). In addition, other methods ofhazard ranking based upon the development of indices ofpotential hazard may also be useful, such as the applicationof the Dow Index and/or the ICI Mond Index. These lattertechniques have provided an important role in theestablishment of safecy in the chemical it.'ustries, butdetailed descriptions of the methods involved are notincluded here. the interested reader is referred to thevarious references given elsewhere in this manual.

2.

Chapter 2

How To Use Thus Manual

�

-i

- $PVI �

� .-

- .... �* �

,', �-

E1 .4I � -

I'.

'�1�

�V: �k.

- 4 -

/ '% * -. ".�.--

I t IJ�,49

�U7'.. - A

U

2. HOW TO USE THIS MANUAL

2.1 Background

This manual aims to describe methods which may be usedby engineer and safety analysts to estimate the majoraccident hazards associated with industrial plants. Thebasic procedure requires that potential failures should beidentifed, the quantities of hazardous material releasedshould be calculated and the impact of each release on plantequipment and personnel or the surrounding population andenvironment should be estimated. The methods have beensimplified so that they may be used by an engineer with ahand-held prog'rammable calculator or preferably on a minicomputer.

Large industrial sites may contain several processunits each of which will be made up of vessels, pipes andinstruments. A large part of this manual consists of adviceon how to organise the safety analysis including the sub-division of the plant into manageable units, the avoidanceof unnecessary calculations and the collation of largenumbers of results. The calculational models for estimatingthe discharge rates and consequences of plant failure mayalso be applied outside of the context of a complete plantanalysis. The analyst may conveniently use these models toanswer more specific and less complex questions. For.example, he may estimate whe.ther a fire ata particularhydrogen storage tank would be likely to cause failures onadjacent equ'pment and whether such a fire would renderescapeways unusable.

In order to guide the analyst through the hazardassessment of a complete plant, the method has been dividedinto 14 main steps. Figure 2.1 shows the order in which therelevant sections of this manual where each modellingprocedure is described. A rigorous hazard assessment of aplant is a complex and subtle problem which has been greatlysimplified for this manual. The the analyst responsible fora hazard assessment should be aware of the advantages andlimitations of the methods described here and the completemanual should be read and digested before embarking upon anycomputational work. Although the sections indicated inFigure 2.1 provide the primary source of information on thetechnicalities of the calculational procedures which may beadopted in the assessment of potentially hazardousinstallations, a brief description of the purpose of eachstep in the analysis is given below.

3.

FIGURE 2.1 GUIOE 10 USE OF THIS MANUAL

ANALYS6 OF ONE

JvESSEL

whOL£ snXA^YN15 ~SltPI SlfP 2 SIEP I 1 STIP 4 SltEP 5 TSfP6 |P T1

O>iVISON Of Sif INIo OIVISION OF UNiIS I FIND I i'NvlOfltEs Of RANK VESS(LS BY FINO RIPAMSNIA:It _ cLUST(a AEttASE _CALIULAIE RILEASE_ - UNITS _ VESStLS ASF1 PIPES rVSSfLS A£( PIPtS IrV(NIORY _RALASE CASES FOR CASES RAVES

^ -l z ~~~~~~~~~~~~~~~~~~~~~~VESSELS AND PIPES .OC(sR*St IN SEtTON I 1 3 1 3 J

StPs SltEP 9 SltP to SltP 11 SE SltP ILCLtUSTER REtASE _ t ULL^l _ OblAE RESULTS _ PiOT EffEtCT AvPPtr sESULIs _ REMIEDIAL MEASURESR ATIS (ONStOuutCES DISTNCES

3 * S S A S

STEP 12

fRtOUthtES

2.2 Description of Steps

Step O - Read the whole manual

Step 1 - Divide the Site into Functional Units

The choice of units is often rather arbitrary. It iscustomary to begin the sub-division of the plant into unitsaccording to the process functions which are involved e.g.import, export, storage, distribution and fabrication.Usually, the, releases from any given unit are described ascoming from one particular point which defines the givenco-ordinates of the unit. For very large plants, there maybe a need to split a single process unit into subunits forthe purposes of this analysis if the components of the mainunit are installed over a large area and large separationdistances are involved.

Step 2 - Divide the Units into Vessels and Pipes

Each unit must be split into "building blocks". Theseare pieces of equipment *such as those shown in the Figures3.1 to 3.10. If the analyst is confronted with a novelcomponent not included in this list, he should considerwhich of the listed items most nearly correspond with hisplant so that the analysis may proceed.

Step 3 - Find the Inventories of the Vessels and the Pipes

The inventories of all hazardous materials should befound by consulting relevant process flow and instrumenta-tior. diagrams. The description of each inventory shouldinclude material type, phase, pressure, temperature andvolume or mass.

Step 4 - Rank the Vessels by Inventory

In order to confine the amount of calculation needed tomanageable proportions, it is important to limit theanalysis to o.nly those components containing significantinventories. The minimum significant inventory should bejudged with consideration of the lower flammability limitand/or the toxicity of the material. For a hazardassessment concerned with the potential on-site consequences

5.

of accidents, it is very difficult to quote minimuminventories. However, for hazards bearing the potential tocause serious off-site impacts, reference may be made to theextract from the World Bank Guidelines reproduced inAppendix 2. It should be noted that potentially hazardousquantities may range from hundreds of gr.ams to hundreds oftonnes depending upon the detailed flammability and toxicityof the materials in questions. As a general rule, .however,vapour releases can usually be ignored in the assessment ofacute off-site risks, if the vapour pressure within thevessel is less than 1 bar gauge.

Step 5 - Find Representative Failure Cases for the Vesselsand the Pipes

A small number of major hazard failure cases may bepostulated for each vessel, component and pipe. A guide tothe scale and size of the most commonly used failure casesis shown as part of the "building-blocks" given in Figures3.1 to 3.10. These failure cases were chosen to representthe sizes of failure which may be encountered usingconservative assumptions.

Step 6 - Cluster the Release Cases

Some of the releases which may be postulated in anhazard assessment may involve the same material undersimilar conditions albeit at different locations in theplant. To reduce the amount of calculation needed, thesesimilar releases may be grouped together and only onecalculation for each specified release rate or quantity isthen required.

Step 7 - Calculate the Release Rates

The postulated failures may be followed by an instan-taneous or a continuous release of hazardous material. Figs3.11 to 3 " provide guidance on the selection of thecorrect m;.. .1 to be used, dependent upon the nature of the

material and the assumed discharge condition. The quantityor rate of this release is calculated using the modelsdescribed in Section 4.1. These modules are designed toaccount for whether the material discharge is in the form ofa liquid, or gas or a -ixture of both.

6.

Step 8 - Cluster the Release Rates

Tn order to reduce further the anount of calculationrequired, those releases which involve similar amounts ofany given material at similar temperatures may also be"clustered" together. Dispersion and consequencecalculations may then be carried out just once for eachcluster.

Step 9 - Calculate the Consequences

Guidance on the selection of the correct model to applyfor the estimation of off-site consequences, under thevarious conditions which may be encountered in practice, isgiven in Figures 3. 11 to 3.14. The calculational methodsfor the estimation of spreading/expansion, dispersion, firesexplosions and toxic impacts are given in Section 4.2 to 4.6inclusive.

Step 10 - Collate the Results

A pro forma tabulation has been included in Section 5of this manual in order to assist the analyst in recordingordering and collating the results of an hazard assessmentfor a complex modern plant, where a diverse range ofstorage, distribution and continuous or batch processes maybe involved. By examination of the effect-ranges summarisedas a result of the analysis, it is possible to order or rankpotential hazards for further consideration.

Step 11 - Plot Effect Distances

Ultimately, the results of the hazard assessmentcalculations should be considered in the context of localdemographic, geographic and land use patterns. Since theresults are available for each release case as an "effectdistance", to a first approximation, hazard impacts may be

estimated by superimposing "effect radii" or circles on mapsof the areas under investigation.

7.

St,ep 12 - Estimate Event Frequencies

By examination of reliabilty and other data, eachfailure case may be associated with a frequency ofoccurrence. Estimates of the frequencies of failures ofvarious sizes can be made on the basis of previousexperience. If the analyst has data which relates to theparticular plant under examination, used in preference tofailure statistics available generic information. Nonethe,-less, it should be noted that, at this stage, the analystwill only be able to use the available frequency data in asemi-quantative manner; a full risk analysis is beyond thescope of this manual and would require additional resourcesand data in th.e form of reliability and availabiltyanalyses.

Step 13 - Interpret the Results

Advice is given in Section 6 of this manual on theapplication of the results of the hazard assessmentmethodology to any given plant having a major hazardpotential.

Step 14 - Examine the Need for Remedial Measures

There are established and diverse means in engineeringand management techniques which may be deployed to mitigate,and in some few cases, eliminate hazards in complex processplant. Many remedial measures will be plant specific and .tis not the intention in the preparation of this manual toprovide a comprehensive description of all of the optionswhich may be available. Nonetheless, by way of examples,some suggestions are made in Section 6. Where improvementsto plant design and operation may be identified, the degreeof reduction in hazard potential may be examined by theanalyst by the selective repetition of the consequencecalculations for those failure cases which would be alteredby the design changes. In this way, the benefits of designchanges in terms of possible reductions in hazards may bequantified and compared.

8.

Chmapter 3

Failure Cases

t~~~~~~~~~~~~~~~~' v.. -

|r~~~~~~~~~~~~~~~~~~~~~a .* ~- 2 - ..

.-- f '4'% ' 'tb

@\ s-' ' ' '~~-Ct

'~ b - a ;;} ..

3. FAILURE CASES

3.1 Background

In this chapter, those steps are described which arenecessary to enable a consequence analysis of plant failuresto be cond-ucted. An overview is provided of typical processplant equipment. This overview will help the analyst toident'fy the relevant accidental release cases of the plant.Relevant -cases should include the typical (ie. mostfrequent) leaks and the worst cases which can be postulatedin order to examine the full range of potential impacts.

As part of the release cases, the properties of thereleased media must be determined. This is accomplished bystudying the equipment inventories under normal operationand for abnormal situations. A particular case will be todetermine whether any part of.the process or any reactionmay release a flammable or toxic gas to a vent stack whichnormally does not discharge such gases. Similarly, thepossibilities for reactive accidents or "run-away"exothermic reactions should be considered at tI,e earliestpossible stages in the plant.

In order to assist the analyst in these respects, eventtrees, which describe typical release sequences following anaccident or failure, have been included in this chapter.These trees are provided to guide the analyst in theexploration of possible failure sequences and in thecalculation of consequences. They provide the links to theappropriate consequence models in Chapter 4. It should benoted, however, that the figures and event trees have beenincluded solely as a guide in identifying and analysingreleases and are not intended for derivation of plantreliability and failure frequency estimates.

9

3.2 Release Cases

The first step in establishing a representative set ofrelease cases is to list the components from which the plantis comprised. Only 'a relatively small number of differenttypes of component are of importance in hazard analysis.Most of those c.omponents that th3 analyst is likely toencounter are shown in Figures 3.1 to 3.10, in which both anexample of a component and a typical process flow diagramsymbol are given. While there may be several variants ofparticular process components, the main functions arelimited to the 10 generic cases which are considered here.These are expected to be the important elements inconducting a simplified hazard assessment. These 10,elements include the following:

Pipes Figure 3.1Flexible connections Figure 3.2Filters Figure 3.3Valves Figure 3.4Pressure/process vessels Figure 3.5Pumps Figure 3.6Compressors. Figure 3.7Storage tanks (atmospheric) Figure 3.8Storage vessels Figure 3.9(pressurised/refrigerated)Flare vent stacks Figure 3.10

Representative failures may be propcsed for eachcomponent. The potential failures which should be con-sidered are those that might lead to significant hazards.In figures 3.1 to 3.10 typical failures that might beconsidered are listed. The analyst should consider whetherthese failure modes, are appropriate to the precise designof the component in his plant. For each failure mode, arepresentative range of failure sizes is given. This is theminimum set of failure sizes that would usually be con-sidered to represent the component in a hazard assessment.

10

3.3 .Seleoting the Event Sequences

Having identifed the potential leak/release sources,the analyst can start to estimate the consequences of thepostulated releases. In order to select the appropriateconsequence models, the relevant properties of the fluidbeing discharged muct be identified. These properties are:

Phase (liquid, gas or two-phase)PressureTemperatureFlammabilityToxicity

Once at this point, the analyst may use the "FailureCase Definitioa Tree" (Figure 3.11) to determine which eventtrees or parts of event trees h'e should use to guide theconsequence calculations. Starting at the left and readingthe tree through the relevant branches, the outcome tellsthe analyst which steps to take further into the analysiIf the release is both flammable and toxic, and bothproperties are considered to be significant, beth branchesof the assessment technique should be appli.ed. In additioni,t should be noted that a flammable release may generatetoxic products of combustion. In the case of two-phaserelease, the "Liquid tree" is applied. The appropriatebranches of this tree feed into the 'Gas trees" forcalculation of gas cloud beh,aviour.

In this manual, the inventories have been put into thefollowing categories:

Liquid at ambient pressure and temperature("liquid, ambient pressure)

A liquid at ambient pressure with temperature below itsboiling point is treated as "liquid, ambient pressure".

- Liquefied gas under pressure and at ambient temperature("liquid, pressurised")

- Liquefied gas at ambient pressure and low temperature("liquid, refrigerated")

-1 1

A liquefied gas may be both under pressure and at a lowtemperature. A release of such a fluid would lead to someinitial flashing of vapour followed by a slower evaporation.For simplicity, however, such a fluid should be treated as"liquid, refrigerated" unless the pressure is considerablyhigher than ambient.

- Gas under pressure("gas, pressurised").

Special note should be taken of the possibilities toproduce a BLEVE (Boiling Liquid Expanding Vapour Explosion).Accidents in this category may occur when a pressurised tankwith a flammable liquid is subJect to a fire. the exposedsteel on top of the vessel is not wetted inside by theliquid contents, so the temperature of the steel at thispoint can rise rapidly until the steel no longer has thestrength to withstand the internal pressure. It may thenstretch to such an extent that a ductile rupture occurs,releasing a great amount of gas under high pressure which isignited immediately to form a large fireball. A BLEVE willalso produce large projectiles from the ruptured vessel orcommunities which may cause significant damage inneighbouring plant if safety distances are too small orshielding of sensitive locations is inadequate. .

Rigorous analyses of failure modes and dischargecharacteristics becomes less feasible if postulated releasesinvolve mixtures of hazardous materials. Nontethelesshazard analyses involving such mixtures may be required andmore approximate methods have been developed. For example,if the mixture consists mainly of one component, thephysical properties of the major component may used withoutthe introductions of any significant inaccuracies.Similarly, if only one component of a mixture is toxic, thetoxicity values adjusted for this component may be appliedto the overall mixture. For example, a leak of butane witha trace of hydrogen sulphide may be treated using thephysical properties of butane to calculate release rates butwith the toxicity value of hydrogen sulphide applied inorder to assess the overall toxic impact.

12

FIG 3.1 PIPE

Includes: Pipes, flanges, welds, elbows

Typical failures: Sugaested failure sizes:

U1 Flange leak 20% pipe diameter

© Pipe leak 100% pipe diameter and

20% pipe diameter

© Weld failure 100% pipe diameter and

20% pipe diameter

FIG 3.2 FLEXIBLE CONNECTION

Includes: Hoses, bellows, articulated arms

Typical failures: Suggested failure sizes:

O Rupture leak 100% diameter and

20% diameter

@ Connection leak 20% diameter

O Connection mechanismfailure 100% diameter

FIG 3.3 FILTER

Includes: Filtets, strainers

Typical failures: Suggested failure sizes:

O Body leak 100% pipe diameter and

20% pipe diameter

© Cover leak 20% pipe diameter

FIG 3.4 VALVE

I~ .

Includes: Ball, gate, globe, plug, needle, butterfly,

choke, relief, ESV -valves

Typical failures: Suggested failure sizes:

Q Housing leak 100% pipe diameter and

20% pipe diamater

® Cover leak 20% pipe diameter

© Stem leak 20% pipe diameter

FIG 3.5 PRESSURE VESSLT./PROCESS VESSEL

_ ~ ~~~ I _

Includes: Separators, s,crubbers, contactors, reactors,

heat exchangers, pig launchers/receivers,

fired heaters, columns, reboilers

Typical failures: Suggested failure sizes:

® Vessel rupture,leak Total rupture,

100% pipe diameter

2 Manhole cover leak 20% opening diameter

* ©Nozzle failure 100% pipe diameter

® Instrument line failure 100% line diameter and

20% line diameter

* | Internal explosion Total rupture

FIG 3.6 PrMp

Includes: Centrifugal' pumps, reciprocating pumps

Typical failures: Suggested failure sizes:

C) Casing failure 100% pipe diameterand

20% pipe diameter

O Gland leak 20% pipe diameter

FIG 3.7 COMPRESSOR'

Includes: Centrifugal compressors, akial compressors,

reciprocating compressors

Typical failures: Sucgested failure sizes:

® Casing failure 100% pipe diameter and

20% pipe diameter

( Gland leak 20% pipe diameter

FIG 3.8 STORAGE TANK (ATMOSPHERIC)

Includes: Atmospheric tanks, pipe connections,

bund wall

Typical failures: Suggested failure sizes:

GD Vessel failure Total rupture

© Connection leak 100% pipe diameter and

20% pipe diameter

FIG 3.9 STORAGE VESSEL (PRESSURISED/ REFRIGERATED)

Includes: Pressurised storage/transport vessels,refrigerated storage/transport vessels,

buried or not buried vessels

Typical failures: Suggested failure sizes:

D BLEVE (not buried caseonly) Total rupture (ignited)

© Rupture Total rupture

® Connection leak 100% pipe diameter and

20% pipe diameter

Note: These storage vessels may have bund walls whichshould be caken into consideration in the analysis.

FIG 3.10 FLARE/VENT STACK

0~~~~~~0

Includes: Manifold, vent scrubber, knock-out drum,

flare/vent stack

Typical failures: Suggested failure sizes:

® Manifold/drum leak 100% pipe diameter and

20% pipe diameter

® Discharge beyondspecification Should be estimated

FlGURE 311 *AILIJHE CASE DEFIIIIIION

DETERmtiC tlATULRE OF HAZARD PHASE _ EtAt CALt APPIY EVENT TREE I

GAS |__ . .-- - FFIG 3 12

F t AMM Aft E _"' PHfi (aAIt

_.____ E_IOULL |l t- - FIG 312 BRA4C H L-

I INVENTORlt . .- - . Ot tH tLIOUIOC r

1~~ ~ ~~~~~ ~ ~~~ ~ T OI4fR P _ -_~ftG 313

I TOXC } 0- TWO PHASE ]

G{7~AS -o _ - FIG3 14

3.4 The Event Trees

There are three bas.ic event trees which illustrate thepotential accident sequences which would arise following arelease:

Flammable Gas Figure 3.12Liquids Figure 3.13Toxic Gas Figure 3.14

In the case of a toxic liquid release (non-flashing), amajor cloud of material would not be formed although sometoxic vapour may be produced. In these circumstances,potential impacts are not likely to cause off-sitefatalities and detailed consequence modelling is probablynot needed. However, the properties of the soil and thedistance over which drainage to rivers, to the sea, thewater reservoirs and to other sensitive areas should beevaluated. This evaluation forms a basis for the design ofbund walls, the design of separate drain systems and thedevelopment of contingency plans.

Over each branching point in the event trees, there isa question which defines the sequence of events following arelease. If both the "yes" and the "no" outcomes arerelevant, both branches should be investigated-further.

Under each branch of the event trees, an indication isgiven of the properties of the release which should becalculated and of the events which should be assessed.Where relevant, there is a reference to the appropriatesections in Chapter 4 which provides the details of thecalculatinal methods. The term "Est" indicates that theanalyst should make a quantitative estimate based on hisknowledge of the plant.

Each tree begins with the determination of appropriaterelease rates. For the special case f an instantaneousrelease, the calculations start with -he initial cloudbehaviour. The jet dispersion and jet fire models are notneeded for such cases. There may be, however, a jet releasefollowing an instantaneous release if another high pressurereservoir is not isolated from the ruptured vessel. Thesesituatons should be considered by the analyst in conductingthe initial division of the plant into basic units.

24

If the release point is connected to a large reservoirof hazardous material, it is important that a realisticestimate of the time required for the isolation of thereservoir should be made so that the leak duration may bedetermined with some confidence. In considering theduration of a gaseous or liquid leak, the availability ofshutdown will be dependent on:

- leak detection: the presence of gas and flame detectorsnear the release point or in the affected enclosure shouldbe determined and the expected reliability of thedetectors should be estimated.

- shutdown actuation: the presence of automatic or manualactivation systems for shutdown should be determined. Ifthe system is manual, the methods of informing theoperators should be assessed including identification ofthe personnel responsible for decisions to activate theshutdown mechanisms. From this assessment an estimate ofthe duration of the response time of the systems may bemade.

- shutdown valves: the .availab4lity or reliability of thevalves should be determined and the actual closing timesshould be identified.

As a general rule, it is usual to assume that majorruptures and massive leaks will be detected immediately,either by instruments or by the operating personnel. Formanual activation of shutdown systems, the response timedepends upon the alarm des4gn, the operating procedures andthe adequacy of operator training. Response times of 3, 5and 15 minutes may be typical, taking account of the highlevels of mental stress which would be encountered during anemergency and the possibilities, therefore, for makingmistakes. For automatic shutdown systems, the closing timeof shutdown valves will depend upon the valve size andpressures involved. Typically, for large, high pressurevalves, closing times of 30 seconds are experienced.

25

If the hazardous material released during an accidentor failure is flammable, the ignition possibilities(immediate and delayed) must be assessd by the analyst. ifthe release is both flammable and toxic, it would beessential to analyse both the ignited and non-ignitedevents. For ignited jets or gas clouds (Figure 3.12), thecalculated heat and pressure loads are used:

- directly to assess fatalities and material damage on andoff site,

- indirectly to determine whether the fire/explosion mayresult in knock-on effects, such as those which may damageother equipment also containing-hazardous materialsleading to an escalation of the accident.

The analyst examining any potentially hazardous processmust check th.t releases caused by preceding events due toknock-on effects are analysed by following the appropriateevent trees in Figure 3.12 to 3.14. For the purposes of thehazard analysis, it will usually be bound that a good numberof potential knock-on effects will have alread beenidentified as potential release cases anyway. The analystshould proceed to determine the cloud density to establish%f buoyant, neutral or dense cloud dispersion modes r ouldbe applied. It should be noted that the distance from therelease point to a potential delayed souroe of ignition willform the basis for determining the flammable mass in thecloud and thus the heat and pressure loads which may beencountered in the surrounding environment.

For a flammable or toxic liquid (Figure 3.13), therelease is treated as either a single phase or two-phasedischarge, depending upon the thermophysical conditionsunder which the particular inventory is held. The latterinclude:

- A stabilised liqu'd at ambient pressure and temperaturewould form a pool. In the case where ignition isconsidered relevant, the heat loads from a pool fire maybe calculated. However, if ignition is not considered tobe important, any hazards to water and land pollutionshould be evaluated.

26

- A gas which is liquefied by refrigeration will initiallyform a pool, but, thereafter, evaporation will take place.In the case of early ignition, the fire should be treatedas a pool fire. For other cases, a gas cloud would beformed, and either the flammable gas event tree and or thetoxic gas event tree should be applied.

- A.gas which is liquefied by pressurisation would expandimmediately upon release, and thus one of the appropriategas event tree should be applied directly.

For a toxic gas release (Figure 3.14), the clouddensity governs the sel-ection of appropriate dispersionmodels. in the majority of cases, emphasis will be placedon dense gas dispersion due to the low cloud temperatureswhich are usually encountered. Also, many gases are heavierthan air and can form dense clouds even at ambienttemperatures. The methods for the consequence calculationsdue to toxic impacts are discussed in Section 4.6.

In many plant, safe handling of discharges and upsetconditions is acheived by the design and operation ofadequate vent, scrubbing and flare facilities. Wherereliance is placed upon such systems, the hazard analystmust take particular care to ensure that:

- if any flammable or toxic material may be discharged intothe system during normal operation or during an emergencyat a rate above which the system is designed, the locationof overpressure and ultimate discharge is identified. Athorough check should be conducted of drain systems, ofany additional pressure relief systems dischargingdirectly to the atmosphere and of the possibilities forequipment to rupture.

- flammables may not be discharged into a cold vent systemwhich is not designed to handle flammables.

- in the event of a flame failure at the flare stack,hazardous quantities of flammable (and possibly toxic)material may not be discharged.

- the potential for reactive or runaway reactions with theproduction of excess heat, toxic or flammable materialsfor discharge to the vent or flare system is fullyexamined.

27

- the potential for abnormal reactions to cause thedischarge of hazardous materialc to the vent or flaresystem is fully, examined.

A full examination of accident sequences involving theabove circumstances will require the estimation of releasequantities and the estimation of potential impacts using theevent sequence trees for Flammable Gas, Liquids or Toxic Gasas described above.

28

FIGURE 3 12 ty fit!f_ FOR LAMMAL GAS RELASE

OU611TON FOR * II0AK *I.V | *6Srt10A | C tiO p4 IINf u.OM1 | .R Dt1RAb.(INCM Pot? til IflS CO.fA,0.Sf I40 AOI** VIO

UMUZ tt_ _ . -. I . . . _ _ _ -,_ . _ _._____ ______ lOuIp.E..Is _

c NO to i s

AULEASt .Al

7 isE3~~~~~~~~~~~~~~~~~~~~~~ * lls

C~~~~~~~. W L _

X.( 0LE U AI A ) f - - - - - - . - .- . - -O .. - s F .. : xCA~~~~(U~~~~hf~~~fiS T.. ill S L.0w.. I-.1 *I IS, ll P,A..g .192 0(4.1 C%OUD FLA.4AOL to 40 -O. LdAl AI.IA54 CA.4I J.40 AI '1'0l(AL(LI1,01) IS

t D.SFV.AL(L.10SLA4.O. OdlE' A-~~ ILAlAfOAl (oALul P.D is 00,C 11 I ... oi a¶0 t4wiflt I. I)..* E C -'s '

.. . ..- . - -1 "I' I . so - -. OU.L.PI .4.0 lIh

"51-jc to , -W.

CICLAL 0. .,.V. I II

xts-hS Rlt^S *^ Sl*_ r _ __ __ _ __ ___ ______ ___ OSUAIDPS LALf.hf ... Salo t4F. B-;sVIt{RtAt(U^teh 0".b10X tSlj $"1 0w U." tSI If wf¢^i Slf Otht CiOV F1 if ".$ 1 CR ~O Ij I1II AC ICI .O0. fL,ALIL1 SJ-w"TnRtittNt ¢ "[;S U". Oh ^1 #9l ADI s C"C -R 14 UZS hfXT r r. s$' a..^ s "0 .,ll vl...iA r?-iLL

*ao hSthl ILOLL OhA.1I LII ^c 1 wN

FIGURE 313 EVENT TREE FOR LIOUIU HELEAS£, fLAMMAELE OR TOXJC

THERMODYONAMIC OUESTIONS FOR |mSUIOwON AN DNill[ON RUPIURE ANO SI8O,IU,,fEhEMD ONtS BRANCHING ISOLAtION REILASE fkW Olt4Ch

POINTS. SUltESSfUL EQU=PM-NT

G r 0 rA YF -- -- -- -

NO

NoL .S NoNO ~ ~ ~ ~ N

S R~~~~~~~~~~~~~~ENO ~~~~NO

EIIEASE RATE .. I 2 Al. f.f C t LOW_; I,EUUVN(fS kfFER TO RELEfANIA'TIONS ',1 s SHUIUOWN taM EST POut SIZ5 ' 2 1 IW Ot A,U Off SItE RitEASE CASES ANUCALCULATION WITlH jNVNIORlT (SI POOL FIRE SltZf A R t I 04NNELt A!S4Sf POS!.ISLitVCALCULATION WITH ~~~~~REL[ASEf DURATION (ST FlEA! LOAD RADiI RR46 tNia444tfItdT AND) FOR Et,AIA1IONREFERENCE TO Ot RAIl I INVINtURY DAKAGE TO OTHER fAtIJlALt VAtUESCHAPTER A IMMESSURISEO) EDUIPIENT ASS 5)

AOIABATII EXMPNSION CSlT8.f,D L__ Y X Y Z 5 _ _INVENTORY (ST EVAPORAIION RATE INPUIT INTO II, II7

RElfASt DURATIONfS 2 tFtAHtAEIIl,IBY RAtf & INVENTORY ANDIOR FI1 It 14tPUt SSdRISEOI IIU .AOlAIABAt EXPANSION

FIG 3 14 EVEN1 TREE FOR TOXIC GAS RELEASE

QUESTION FOR INSIANlAN(OUS S S1UT00WN AND ISOLATION DENSE (lOUOBRANCHING POINTS IELM tINIINUOUSI SUCCESSFUL? _ J

NO F___ f_-

AtC I KNS: RtLIAst RAilt _ _ __ E G, A:StS: IuuirItCALCULATIONS WITH INVINILJ tst SbIJtUWsN *Int ESI D(Nzt tUCD OKJlAS4. 1r, At,O Of f'3REFERENCE TO (tOULO OENStsl '2 1IL1 "IL(AJITVY RNAQUS .. 1CHAPTER 4.

INV[NTORY (St

RELEASt OURATlON (StIBASfDDN RATE AND

IttVfNICTY I

PRtSSURIStG * t 2)IT Om.PEso2N

1U)D DWNSITY J II

REtIAGRAIlD__ _ __ _ __ _ __ ____ ________ _ ___________*_31 ____.___ ____J(LOUD LXN5JTY '. Ii

AS ABOVE XlCEPI fOR N UTRA ItBUOlANI ttOUDSHUTDOWN TIMlE jtYtRLIIY 4 J2i. I II j

_ _ _ _ _ | tt~~~~~~~~~~~~~ITKALtYl RACIUS ia

Chapter 4

Consequence Calculations

_L~-7 ?

a..~~~~~~~~~~~~~~~~* -

4. CONSEQUENCE CALCULATIONS

In this section, the theoretical basis is given for theestimation of discharge rates, dispersion rates and toxic orflammable impacts resulting from postulated plant-failurecases. Wherever possible, simplifications to the equationsneeded for the corresponding calculations have been in-troduced in order to assist the analyst in conducting aninitial hazard assessment. The nomenclature and notationused in the. derivation of these modelling procedures issummarised in Appendix 1.

4.1 Outflow Calculations

Most incidents involve the initial escape from con-finement of hazardous material, be it through a hole or acrack or an open ended pipe rupture. There are well knownequations which can be used to calculate the rate ofrelease, provided the size of the hole and the physicalproperties of the material contained within the system areknown. Judgement may be required on occasions when non-uniform holes or non-uniform conditions are involved. Theseproblems are usually solved by defining an equivalent holediameter for the net open area.

It is important to appreciate, however, that thepressure within the container will fall during leakage and.hence the rate of release will also fall during the periodof discharge. Additionally, if the material was originallyheld as a pressurised liquid at temperatures above itsnormal (atmospheric pressure) boiling point, the drop inpressure following a release to atmosphere will cause partof the liquid to "flash off" or vaporise instantly. Theremainder will be present as liquid but its temperature willbe greatly reduced relative to its normal (atmosphericpressure) boiling point. These effects may be representedby equations, such that the amount o' gas and liquidreleased over the period of a leakage may be estimated.

32

When conducting an hazard assessment, three main typesof flow or discharge phenomena may be encountered. These

are:

Liquid Release the release of a liquid fromcontainment; the fluid remains liquidduring the process

Gas Release the discharge release of a gas fromcontainment

Two Phase Release the release of a mixture of gas andliquid (usually resulting from boilingof the liquid under dischargeconditions).

33

4.l.1 LIQUID OUTFLOW

METHOD: Use of Bernoulli flow equation to establishrelease rate of hazardous liquified gas fromrefrigerated and pressurised containment.

OUTPUT: Discharge rate and fraction of liquid whichflashes off immediatel.y after discharge.

CONSTRAINTS: None.

The Method Discharges of liquid are treated by twomethods depending on whether or not theliquid will flash when released toatmospheric pressure. For non-flashingliquids, standard liquid discharge formulaeare used, e.g. the Bernoulli flow equation.The discharge rate in kg/s for a liquid out-flow from a pressurised containment is givenby:

Q - Cd Ar D1 [2 (PI - Pa) + ghl 1/2

Dlj

Although the driving force is primarily theliquid head, h, for releases of refrigeratedliquid, the model can also be used to analysebottom pipe failures close to vessels con-taining saturated liquid under pressure, P1.

The fraction of the liquid which may flashimmediately to form a vapour, Fvap, is givenby Fauske (1965) as:

Fvap = Cpl (T1 - Tb)/Hvap

If this fraction is less than one, furtherdilution of the liquid spray with ambient airis necessary to complete vaporisation of theremaining cold liquid. The temperature andvelocity of the resulting vapour/air mixturecan be found. However, if Fvap is greaterthan one, the liquid has evaporated com-pletely before reaching atmospheric pressure.

34

The temperature and corresponding-saturatedvapour pressure are calculated, and vapourrelease equations are used to estimate thedischarge rate under the remaining pressuredifference.

Outputs This method gives the discharge rate of aliquid from refrigerated or pressurisedcontainment. It is also used to calculate,where appropriate, the fraction of liquidwhich flashes to atmosphere immediacely afterdischarge.

Input (i) Choice of discharge coefficient, Cd;this may be calculated using standardchemical engineering methods (seeHandbook of Chemical Engineering -

Perry); alternatively, a 'standard'figure of 0.6 (conservative) may beused for a hole in a vessel.

(ii) Thermophysical data (Perry).

(iii) Initial Pressure and/or hydrostatichead.

(iv) Selection of an effectivenes openarea, Art for the hole size character-istic of the failure case underconsideration.

Assumptions This method is based on the Bernoulli equation

and which is a standard general differentialConstraints equation expressing the behaviour of flow.

The main assumptions are the selection of thevalue'of Cd and the assumption of incompres-sibility of flow. This method yieldsinstantaneous discharge rates and no allowanceis made for the time dependency of thedischarge as the pressure or liquid headfalls.

Accuracy This method for instantaneous discharge ratesis considered to be as accurate as necessaryfor the type of study involved (i.e. betterthan + 5%). It should be noted that whenapplied to a discharge over a finite period oftime, the results computed are conservative.

35

Application This method may be applied to releasesofliquids from enclosed systems, which may besubject to an hydrostatic liquid head, h,and/or an internal pressure, P1, in excess ofambient atmospheric pressure.

36

4.1.2 GAS OUTFLOW

METHOD: Use of the gas flow equation for caiculatingdischarge rates for gases from sources underpressure.

OUTPUT: Gas discharge rate.CONSTRAINTS: The simple equations assume ideal gas

behaviour which is probably reasonable for allbut very high (near critical) pressures.

The Method Discharges from vessels and pipes.containingonly gas or vapour under pressure are normallyreadily calculated using standard equationsfor gas flow. Such equations exist for thedischarge rate of both critical andsub-critical flows. The distinction is madeas follows, assuming reversible adiabaticexpansion and ideal gas behaviour.

If Pa < P1 {2 G/(G-1) flow is critical

(G+l

If Pa > Pi 2 GI(G-1) flow is

G+1

sub-critical

Having established whether the flow iscritical, the formula giving the dischargerate from a pressurised gas release is:

Q = Y Cd Ar P1 [(MG/RTl)(2/G+l)(G+l)/(G-l li2

For critical outflow, Y = 1.0.

37

For sub-critical outflow,

[(2) (] [ ) (G+11(G-1 3 1/2

It may also be necessary to take account ofthe size of the vessel from which the materialis emitting to decide whether a time dependentoutflow should be calculated. If the durationof the release is short compared with the timerequired to empty a given vessel, it isnormally adequate to utilise the initial dis-charge rate calculation in subsequent analysesof consequences.

Outputs This method gives the initial discharge ratefor a pressur.ised gas release.

Inputs (i) A discharge coefficient, Cd, of 0.8 maybe used when calculating the rate ofleakage from pipes connected tovessels.

(ii) Thermophysical data for the gas(Perry).

(iii) Initial pressure of release.

(iv) Selection of an effective open area Ar,for the hole size characteristic of thefailure case under consideration.

Assumptions This method assumes reversiblc adiabaticand expansion and ideal gas behaviour.Constraints

38

Accurac_ This method is sufficiently accurate forHazard Assessment provided ideal gas behaviouris a reasonable assumption.

Application This method may be applied to the discharge oftoxic and flammable gases from large storagevessels or pipes.

39

4.1.3 TWO PHASE OUTFLOW

METUOD: Fauske/Cude method to calculate rate ofdischarge from two phase critical flows.

OUTPUT: Discharge rate and fraction of liquid whichflashes.

INPUTS: Discharge coefficient and thermophysical data.CONSTRAINTS: Only valid if the calculated fraction of

liquid flashing is less than 1. Alternativemethods are required for more complexsituations; in this guide suggestions areprovided to yield approximate solutions.

The Method Two phase critical flows can occur in failuresof connections to the vapour space of vesselscontaining superheated liquids under pressure.They also occur in failures of pipeworkcontaining superheated liquids remote from thevessel, where a fully developed critical flowwould be established. The calculations usedto find the discharge rate are based on theFauske/Cude (1975) method. Fauske worked onsteam-water critical discharges, and thismethod makes the assumption that the pressureratio between, the throat and the upstreampressure for water systems (0.55) can beapplied to other materials, with the assump-tion that the two phases are homogeneous andin mutual equilibrium. Thus,

Pc = 0.55 Pl

The liquid fraction Fvap which flashes at Pcis given by:

Fvap = Cpl (T 1 - Tc)/Hvap

40

The mean density of the two phase mixture isthen:

DmFvap/Dv + (l-Fvap)/Dl

and the discharge rate Q is given by:

Q = Ar Cd (2Dm(Pl-Pc)) 1 /2

If Fvap is greater than one, the method ofcalculation is not appropriate. Thecalculations may be repeated treating therelease as a gas in which case some under-estimation of the release rate may beencountered. Alternatively, the ratio ofPC/Pl may be adjusted to satisfy the conditionwhen Fvap'l and the discharge calculationsconducted using associated thermophysicalproperties for these conditions.

Outputs This method provides a means for estimatingthe fraction of a liquid discharge which mayflash to a vapour/gas and enables thedischarge rate from a two phase flow to becalculated.

Inputs (i) Discharge coefficient, Cd, using astandard value of 0.8

(ii) Thermophysical data for the fluids

(iii) Initial pressure of release

(iv) Selection of an effective open area,Ars for the hole size characteristic ofthe failure case under consideration.

41

Assumptions The mathod assumes that the two phases areand -homogeneous and in mutual equilibrium. It isConsLraints a simple empirical method, suitable for hand

calculations, which gives reasonably accuratetesults for simple systems commonly ofinterest in hazard assessments.

AccuracSy The accuracy of this method is questionablefer discharges involving lorng lengths of pipeline where two phase flow may develop withinthe line.

Application This method may be applied to releases toatmosphere of saturated liquids stored underpressure, at a temperature above the normalboiling point.

42

4.2 Behaviour Immediately after Release

Once discharged from its containment, a hazardousliquid material will tend to spread out in the surroundingenvironment. This may involve the spreading of a liquidpool on the land or on water, which will affect the rate ofevaporation and dispersion in the atmosphere. Such spread-ing will also be affected by any secondary containment (e.g.bunds, walled areas) and by the topography of the land. Inthe case of volatile material, should a spreading processoccur, it will be quickly followed by evaporation.

Spreading of a release on water is a -special case ofthe general approach to be taken in estimating the degree ofspreading of spilled materials. However, the output of aspreading model or other relevant calculations, is usuallyused as an input to more complex hydrological models,particularly if currents and drift rates are expected tocontribute significantly to the spatial distribution of aspilled pool of material.

Evaporation of spilled material may be transporteddownwind to sensitive areas thereby contributing to hazards.However, the evaporation process requires heat which isextracted from the surroundings (air, land and or water).It is. necessary, therefore,. to examine the evaporationprocess in order to determine the rate of vapour evolutiongiven the type of material spilled and the properties of thesurrounding environment.

43

4.2.1 SPREADING LIQUID RELEASE

METHOD: Spreading liquid model for land or watertaking into account the buoyancy factor forthis liquid.

OUTPUT: Spreading rate providing a pool diameter anddepth after a given period of time.

CONSTRAINTS: This method may be applied to spills on landor water and calculates the "idealised"spreading (i.e. on a flat surface or on calmwater).

The Method Evaporation of the material is a necessaryprocess if significant concentrations of thematerial are to be air borne to the "receptor"i.e. the object(s) at risk. A liquid releasemay be hazardous if there is a risk of fire orif a flammable or toxic vapour evaporates.Evaporation is considered in this sectionwhile liquid fires are considered elsewhere in

Section 4.4.

As a general rule, significant quantities ofvapour evaporate only if the boiling point ofthe spilled liquid is below ambient temper-ature.

Evaporation requires heat, which in the caseof refrigerated liquids is extracted from thesurroundings. The evaporation process is

analysed to determine the rate of vapourevolution. After a release, it is assumedthat the pool may be represented by a cylinderof radius, r, and uniform height, h, at time,t. For spills both on land and on water, the

spreading rate is given by:

dr - (egFh)1/ 2

dt

44

constant base where F is the buoyancy factor,and e is based.on emperical measurements(values r.ange from 1 to 6). If the liquidspill is spreading on land, the buoyancyfactor F can be taken to be unity and thevalue of e can be taken as 2.

On the other hand, if the liquid spill is onwater, the buoyancy factor is a function ofthe relative density, given by the equation.

F = I - Dl/Dw

and the value of e can be taken as 1.6.obviously, this model can only be applied tospills of liquids which are lighter than waterand which are neither soluble in nor reactwith water.

A pool will contirue to spread if the spillrate exceeds the vaporisation rate. In theabsence of restrictions such as bund walls orcomplex topography, the size of a liquid poolis limited ultimately by the affect on thevaporisation rate of the heat transfer fromthe contacting surfaces.

Once a spill has reached its limit of spread-ing along a surface, either by meeting anartificial boundary or by reaching a minimumdepth at which it no longer spreads, heattransfer continues until the pool vaporisescompletely. The pool will expand until itreaches the sides of a deliberately con-structed bund or some obstruction. For thepurposes of calculating evaporation in aninitial risk analysis, it is often sufficientto assume that the period of outflow andexpansion are negligible. This can be checkedby using the above description of expansion.The evaporation rate from the pool in kg/s isthen estimated by:

Q = Ce Ap t-1/2

45

where A is the surface area of the pool andCe is a constant. The value of Ce depends onthe nature of the substrate which is'moistened. If no data exist for thecombination of substrate and material underconsideration, an estimate of Ce may be madeby:

0e =.Cs (Ts - Tb)/Hv ( (As)1 /2

where C5 is the thermal conductivity of thesubstrate and as is the corresponding thermaldiffusivity..

Outputs This method provides a way of estimating thespread of a non-reactive liquid either on landor water. 'It gives the radius of a spillafter a given time has elapsed, assuming thatthe liquid has reached its characteristiclimiting thickness.

Inputs (i) Parameters in the equations above

(ii) Rate of liquid spillage from anappropriate discharge model.

Assumptions This model makes a simplifying assumptionand about the spreading process by hypothesisingConstraints that the initial release takes the form of a

tall cylinder which then spreads under gravitywith no addition of liquid. More complexmodels have been suggested but there is littlejustification for their use. Complicationsoccur for spills on *water if the liquid issoluble in, or reacts with water, or if it isheavier than water. Unless heats of dilutionor reaction are significant in raising thebulk temperature of che spilled material, themodels indicated here would tend to yieldconservative estimates in hazard assessmentsdue to reduced evaporation rates.

Accuracy Accurate data on the behaviour of largespreading pools is sparse but suggests thatthis simple model is sufficiently accurate foruse in hazard assessment.

Application This method may be applied to the initialspreading of refrigerated liquid releases.

46

4.2.2 JET DISPERSION

METHOD: Simple Jet Model.OUTPUT: Plume dimensions and concentratior of

hazardous material along the plume.CONSTRAINTS: Plume characteristics can be calculated whece

jet momentum is dominating the mixing process.

The Method The jet shape for vapour produced by materialescaping under pressure can be treated eithersimply or with a complex model. Existingcomplex models such as that of Ooms can in-volve large computational loads and simplermodels are preferred where possible.

For flammable materials, the free area withinwhich the vapour is above its lower flammablelimit (LFL) is represented by an ellipse. Thelength of the major semi-axis is calculatedby:

A=[((bl+b 2 )/bijc> + Dvy-1 dm(Do) 1 / 2 / 0.32Dv

where jc is the LFL concentration. The lengthof the minor -semi-axis is calculated by:

B - A (-log [l/2(l+jc(l-Dv)bl/(bl+b 2 ))})1/2

If the je.t occurs in a volume where thebackground concentration, CA, is not zero thenthe true concentration, CT, will be given by:

CT = CA + C.CV (I-CA)

where C is the cloud concentration assuming nogas in the original air and Cv is the volumefraction of vapour in the original mixture.

For toxic releases, hazard ranges willnormally be greater than those calculated forflammable limits. To cater for thesecircumstances, a second stage in the calcula-tion may be required to enable downwinddispersion of the toxic gas cloud to becontinued until hazardous concentrations areno longer encountered.

47

Outputs This method calculates the characteristic jetplume dimensions as a function of length andwidth to the lower flammable limits of tkeplume or any other pre-defined plumeconcentration.

Inputs (i) Discharge rate and size of dischargeorifice produced by a sonic gas releaseor a two phase outflow model, andthermophysical properties.

(ii) Thermophysical properties for thematerials in question.

Assumptions The model described above is approximate andand is based on a simplified approach. MoreConstraints accurate methods are available but require the

use of coniputer solutions, such as thosedeveloped by Ooms.

Accuracy This simple method is considered to besufficiently accurate for most hazard analysiscalculations, particularly those required inan initial assessment.

Application This method may be applied to estimate thebehaviour and dispersion of a high velocityjet of vapour, down to the LFL. Toxicreleases may require further calculation ofdownwind dispersion characteristics.

48

4.2.3 ADIABATIC EXPANSION

METHOD: Two-stage expansion model for instantaneousrelease of flashing liquid or pressurisedvapour.

OUTPUT: Concentration and radius of expanded cloud.Mass of air in the cloud. Volume of thecloud.

CONSTRAINTS: The model applies to rapid expansion with noheat exchange between the expanding cloud andthe surrounding air.

The Method For the estimation of the rapid adiabaticexpansion experienced during the instantaneousrelease of a flashing liquid or pressurisedvapour, a simple two-zone, hemispherical modelmay be applied. In this model, it is assumedthat a core of a uniform concentration con-taining 50%e of the released mass, issurrounded by a peripheral zone characterisedby a Gaussian distribution of concentration.

The expansion process is idealised intostages. In the first, the gas or liquidaerosal expards down to atmospheric pressureand the liquid flashes, as appropriate. Thekinetic energy, developed by the initialexpansion, drives the turbulent mixing of airinto the cloud in the second stage as thecloud spreads outwards. The end of thisspreading phase is taken to be when thespreading velocity of the core falls below agiven velocity.

During the first phase, the expanding gas orliquid does work against the atmosphere andsome internal energy goes to increasing thebulk energy of the substance. If it isassumed that the increase in kinetic energy isgiven by (P-Pa)dV, the initial expansion toatmospheric pressure may be treated as areversible adiabatic process. The energy ofexpansion becomes the difference between theinitial and final energies, minus the workdone on the atmosphere. This first phase,idealised in this way, is isentropic.

49

For a gas, the energy can be determined bycalculating what the final state is after areversible adiabatic expansion to P1 and T1with internal energy U1, to the state definedby Pa and T2 with i-nternal energy U2. Thechange in internal energy is then:

U1 - U2 = CV (T 1 - T 2 )

and the energy of expansion is:

E = Cv(Tl - T2) - Pa(V2 V v1)-

Similarly, for a liquid, the fraction, Fvap,that.flashes is calculated by assuming thatthe entropy is constant during this idealisedinitial phase, i.e.

SI(M) (1 - Fvap)Sl(2) + F 3apSv(2)

Fvap =.[Sl(l) - Sl(2)]/[Sv(2)'- S1l(2)]

Tb[Sl(l) - Sl( 2)]/Hvap (2)

As before, the energy of expansion is thedifference in initial and final internal.energies minus the work done against theatmosphere.

E = Ui(l)-[(l-Fvap)Ui( 2 ) +FvapUv(2)] -Pa(V2-Vl)

=.Ul(1)+PlVl - [(l-Fvap)Ul( 2 )+FvapUv( 2 ) +PaV2] - (Pl Pa)Vl

= Hl(l) - [(l-Fvap)Hi(2)+FvapHv(2)]

(PI-Pa) Vl

= Hl(l)-H1(2)-(PI-Pa)vl-Fvap[Hv(2)-H1(2)I

= Hl(l)-Hl(2)-(Pl-Pa)VI-Tb[Sl(l) S1(2)I

50

The differences in enthalpy and entropy aremade consistent by using the heat capacityalong the saturation curve to calculate both.For simplicity the heat capacity is taken tovary linearly between states (1) and (2).

As a result of the impulse developed in theexpansion, extensive turbulence is generated.This turbulence is the determining factor forfurther mixing of the gas cloud with air fromthe environment. Once the expansion energy,E, has been obtained, the expression for theturbulent diffusion coefficient is:

Kd = .O0l37El/2 (Vgo)l/ 3 [t(E)l/ 2 /(Vgo)l/ 3]-1/4

and the expression for the core radius as afunction of time is:

rc = (4Dt) 1 /2

A useful criterion to be used for decidingwhen the spreading is complete is to set thelimit condition to drc/dt = lm/s. Combiningthis with the other two equations gives thefollowing expressions for the core concen-trations and radius as at the end of themixing.

jc - 172.95 E-0 -9rc - 0.08837 EO. 3 (Vgo)l/ 3

It is, however, necessary to invoke a furtherassumption to determine the effective edge ofthe cloud in the Guassian region. This wasachieved by using the model to match the Freonspill tests carried out by van Ulden (1974).It was found that the end of the initialexpansion, as reported by van Ulden, occurredat a cloud radius r such that r/rc = 1.456.At this value of r/rc, 91% of the releasedmaterial is still within the cloud. The valueof ic, or of jgl, can be greater than one

51

since a large part of the material released isin liquid form. Jc is the vol./vol concentra-tion, where the volume of the material is.taken to be the volume if it were all vapourat the appropriate pressure and temperature.The amount of air mixed into the cloud may becalculated iteratively, if there is stillliquid present. For a given mass of air mixedinto the cloud, which has a vapour fractionf2, the temperature T3 is given by:

M CplTl + MairCaTa = M f2Hvap + MairCaT3

while there is still the requirement that:

Mair + f 2 M + (l-f 2 )MVcloud '____

Da Dv Dl

This temperature and vapour fraction must beconsistent with the requirement that thepartial pressure should equal the saturatedvapour pressure, (s.v.p) at T3, i.e.

s.v.p (T3) =-Pa(f2M/(DvVcloud)]I

The mass of air is adjusted until theseconditions are satisfied.

Having determined the initial amount of airmixed into the cloud, the subsequent dis-persion calculation can-be made using themodels described in Section 4.3.

Outputs The method gives the concentration and radiusof the expanded cloud, the mass of airentrained in the cloud and the volume of thecloud.

Inputs (i) Initial volume, pressure andtemperature

(ii) thermophysical properties.

52

Assu-otions The model assumes an instantaneous release ofand material and such a rapid expansion that noCovistra5nts heat exchange takes place between the

expanding cloud and the surrounding air.

Accuracy The model is less accurate than the jetdispersion model because it is less wellsupported by experiments and observation.However, it is considered the best modelcurrently available for lar7e instantaneousreleases and it is suitable for most hazardanalyses.

Application This method describes the initial behavicur ofan instantaneous pressurised release. Theoutput may be used for subsequent dispersioncalculations

53

4.3 Dispersion in the Atmosphere

Dispersion of hazardous and pollutant materials in theatmosphere has been the subject of intensive interest forsome decades and has resulted in the development of manydifferent models. In the first instance, models of neutral(or tracer) plumes were of interest because these wererelevant in examining the behaviour of pollutants dischargefrom vents and stacks. W*ith the growth of interest inhazard analysis, the behaviour of clouds of material inwhich the vapour density is significantly different Zromthat of the surrounding air became of interest. Usually, itis the denser than air behaviour which is of most interestbecause a lighter than air gas will 'float' upwards and maytherefore disperse harmlessly. The main interest is in socalled 'dense gas dispersion' and this type of dispersion isusually central to hazard analysis of major chemical plant.

Dispersion of pollutants or small particles in theatmosphere is a function of the stability of the air, thewind speed and the atmospheric turbulence. 'The standardcategorisation system'normally adopted for air stability isbased on that derived'by Pasquill. This scheme employs 6 or7 categories to cover unstable, neutral and stableconditions; the latter are designated as A to F or sometimesA-G. Neutral stability is.designated category D and is thecondition expected when there is total cloud cover with orwithout precipitation. If the sun is shining, atmosphericturbulence is increased due to the insolation or radiation.This g'ives rise to the less stable conditions classed as,categories A to C inclusive. The most stable atmosphericconditions occur on clear, calm, cloudless nights, and isdesignated category F (or sometimes G-is used as well).

Besides the thermal effects on the atmosphere turbu-lence is increased by higher wind speeds. These lattereffects depend upon the surface roughness of the ground overwhich the atmosphere flows. These terms occur in thevarious dispers'ion models which are available. Some )f themore recent and more advanced models introduce more complexdescriptions of turbulence based upon eddy diffusivity.These latter methods, however, have rarely been encounteredin hazard assessment studies because of their complexity andhigh computer resource requirements. Instead, most hazardassessments employ the simpler, top hat or "box" models.

54

The models described in this section of the manual are:

(i) A typical sin.ple "top-hat" model for use inanalysing dense gas behaviour.

(ii) The Gaussian plume model which is used, eitherin conjunction with a dense gas model todescribe the behaviour of the plume when itbecomes "neutral", or by itself if the plume isneutral in the first olace.

55

4.3.1 DENSE CLOUD DISPERSION

METHOD: Cox and Carpenter dense gas dispersion model.-OUTPUT: Gives cloud dimensions after a time interval

for both continuous and instantaneousreleases.

CONSTRAINTS: Applicable only while the cloud is spreadingdue-to gravitational forces.

The Method The model described here for dense clouddispersion is thac of Cox and Carpenter (1980)but there are several other "top-hat" or "box"models in existence. An instantaneous releaseis represented as a cylindrical cloud, or"top-hat" which is assumed to adopt a"pancake" shape and spread radially relativeto its centre while advecting with the wind.Continuous release, is represented by a plumeof rectangular cross-section, which in thepresence of wind is r.elative.ly narrow andspreads laterally down-wind because ofgravity. In both types of release, air isentrained through exposed surfaces.

For an instantaneous release, the initialamount of air mixed into the cloud isestablished using methods which make allowancefor the dilution occurring with the surround-ing air. The mixed cloud is taken to be acylinder of uniform density, with a chosenheight to radius ratio, which is usually takento be unity. Thus, for a continuous release,lateral spreading may be expressed by:

dR - [kHg (Dp/Da-1)] 1 /2

d t

where the constant k is given various valuesby different authors.

56

For both continuous and instantaneousreleases, there are two sources of mixing asthe cloud spreads: mixing in the edge vortexand mixing through the top surface, determinedby atmospheric turbulence and the densitydifference between the cloud and air. Foredge mixing we have

Qe =<J 2RH dRdt

For atmospheric turbulence giving entrainmentthrough the top of the cloud, the entrainmentvelocity is given by:

Ue = Aul/(Ri)

where the Richardson number, Ri, is given by

glRi = . (Dp/Da-l)

The values of ul and 1 depend on surfaceroughness, weather conditions and cloudheight. The rate of air entrainment is foundby multiplying by the area of the top surfaceof the cloud. Combining the two expressionsgives overall dilution of the cloud, thetemperature of which is correspondinglyaltered. This model uses the heat flowequations for forced or natural convection,whichever gives the greater value.

When the Iateral spreading rate due toturbulence starts to exceed that due to gravi-tational spreading, a transition is made tothe neutral density model. The criterion usedis:

dR dOy

dt dt

where is the corresponding atmosphericcoefficient.

57

This model provides a reasonable estimate ofthe behaviour of those hazardous gases whichmay .be significantly heavier than air,provid,ed that the gas cloud is in a relativelyflat environment free from any significantobstructions.

Outputs This method gives the spreading and downwinddisplacement of the cloud.

Inputs Initial cloud dimensions and concentrations,.wind speed, atmospheric stability,, materialinput rate or quantity, thermophysical data,characterisation of ground condition.

Assumptions Notes on the application of these models areand given in the detailed descriptions.Constraints

Accuracy Many comparisons and some tests have beenconducted to validate these-models. Usually,they can be calibrated to fit the test data.These methods are regarded as sufficientlyaccurate for modelling a gas cloud inrelatively 'flat' environment with no,significant' obstruction. Care should betaken with situations where these conditionsare not met.

Application These methods may be applied to examine thebehaviour of hazardous gases or clouds whichare heavier than air.

58

4.3.2 PASSIVE DISPERSION

METHOD: Gaussian model for neutral clouds.OUTPUT: Cloud correlations for both instantaneous and

continuous releases.CONSTRAINTS: Only applicable to those phases of gas

dispersion dominated by atmosphericturbulence.

The Method Neutral cloud dispersion occurs when thespreading rate of a dense cloud due toturbulence starts to exceed that due togravitational spreading. A Gaussian form ischosen, and the uniform cross-sectionalconcentration distribution of the dense phasemodel has to be matched to the non-uniformGaussian spread in the neutral phase. It isassumed that the crcsswind width at theatching point is equal to the crosswind width

at the virtual source of the Gaussian plumeplus (2 y) where y is the standarddeviation measured at the matching point. Thecentreline concentration is also matched tothe cylindrical cloud or rectangular plume.

Instantaneous For a release at ground levelCase

C(x,y,z) =

2Q 1 2r y 2 Z2\)(2'I)/ 2zuixdYGzexP - (.+ r2+*,I

where C is the cloud concentration andand az are the standard dispersion parameters,dependent on stability category and distancefrom the source as described by Pasquill(1961). These parameters are given in a con-venient form by McMullen (1975). It is alsonormally assumed that Ex = C'y. The x, y and zco-ordinates are local to the cloud centre atground level, which itself advects with thewind.

59

For ground level concentrations at the cloudcentre-line

2Q

(2,K;)l/Zul I-x My C'Z

Continuous For a release at ground level:Case

Q y2 ZC(x,y,z) = exp -+

s uIVy Crz _2 y. z

where the origin is now the source position, Qis the rate of release and ul the windspeed.

On the plume centreline, at the surface, thisreduces to:

Qc (x) r_

iYR u I6-y a-Z

In the above equations, it should be notedthat absorption at the surface has beenassumed zero, such that the plume is"reflected" at the surface; This may bepessimistic for some cloud materials.

The standard atmospheric dispersion parametersdescribe the increase in the cloud radius asthe cloud drifts downwind. Simple formulae ofthe type

5 = axb and

CZ = cxd

have been proposed to describe the increase instandard deviation. Table 4.1 overleaf showssuggested values proposed by TNO for theparameters a,b,c and d valid for conditionswhere the downwind distance exceeds lOOm.

60

TABLE 4.1 ATMOSPHERIC DISPERSION PARAMETERS

Parameter a b c d/Category

very un-stable (A) 0.527 0.865 0.28 0.90unstable (B) 0.371 0.866 0.23 0.85slightlyunstable (C) 0.209 0.897 0.22 0.80neutral (D) 0.128 0.905 0.20 0.76stable (E) 0.098 0.902 0.15 0.73verystable (F) 0.065 0.902 0.12 0.67

To match the initial source conditions eitherthe cloud size or the initial concentrationmay be'used by selecting a suitable point onthe x-axis as the "virtual source" locationfor the purposes of determining the standarddispersion parameters, since the latter aredefined as functions of downwind distance instandard formulations of the Pasquillstability categories.

Outputs The model gives cloud concentrations, both forinstantaneous and continuous releases.

Inputs (i) Released mass(ii) Dispersion parametersCiii) Windspeed.

Assumptions The dispersion is based upon the assumption ofand Gaussian distributions of turbulence in theConstraints atmospheric boundary layer.

61

Accuracy Subject to a relevant selection of wind speedand stability category short term averageconcentrations of pollutants may be estimatedwithin at least a factor of two of the ex-periment. This degree of precision isgenerally satisfactory for the purposes of anhazard assessment. However, due to limits inthe applicabilitv of dispersion coefficients,the methods are generallv applied to downwinddistances in excess of lOOm and less thanlOkm from the point of discharge.

Application The method determines the concentrationdistribution of clouds of neutral density.

62

4.3.3 BUOYANT SURFACE RELEASE

METHOD: Briggs Plume Rise Model.OUTPUT: Release height estimate.CONSTRAINTS: Based upon empirical observations.

The Method For a surface release of buoyant material, itis not necessarily clear that the materialwill lift off the surface under the action ofbuoyancy forces. The effects of turbulence,which may be intense near the ground due tofriction effects and obstacles may ',edominant. For the cloud to lift o- , thelower parts of the cloud edges must move in-wards due to the external hydrostatic pressureacting against the spreading influence of theturbulent dispersion. Briggs (1976) hassuggested that a criterion may be developed bycomparing a characteristic lateral turbulentspreading velocity with a characteristic in-ward movement associated with buoyancy. Thisarises from considerations of the drawing inof the cloud sides near the surface as thebulk of the cloud starts to rise.

Briggs takes [G H (Dp - Da)/Da ]1/2 as thelatter value and the friction velacity Uf, asthe former, but other criteria could bedevised. Briggs therefore defines a parameter

G H (D p -Da)

Lp =Da UfL

and he finds that lift off occurs when this isgreater than about 2.5 for instantaneous.releases of roughly hemispherical shape andfor continuous releases of roughly semi-cylindrical shape at the ground. It should benoted, however, that the critical value couldbe different, but in the absence of otherinformation to the contrary, the same value of2.5 is generally adopted for both cases.

63

If the cloud lifts off, the trajectory anddispersion may be predicted using the passivedispersion model or a conventional plume riseformula (e.g. Briggs, 1969) applied with aGaussian dispersion model such as describedabove.

Assumptions If the cloud does not lift off, it can only be

and treated as a passive tracer using the 3ppro-Constraints priate dispersion models, e.g. the Gaussian

model, for which relevant equations are givenabove.

64

4.4 Fires

The radiation effect of fires is normally limited to

areas close to the source of the hydrocarbon (say within200m). In many cases, this means that surrounding com-munities are not affected. However, there are some types offire which would have a more pervasive effect. The means bywhich the various types of fire may be analysed is described

below,

Fires may be categorised as follows:

- Pool fire (e.g. a tank fire or fire from a pool of fuelspread over the ground or water).

- Jet fire from the ignition of jets of hydrocarbons or

other flammable materials.

- BLEVE (Boiling Liquid Expanding Vapour Explosion)resulting from the overheating of a pressurised vesselby more minor primary fire which causes the vessel toexplode and a large and very intense fireball to be

produced.

- Flash Fire involving the ignition of a vapour cloudwhich does not explode. That is, the flame speed isnot high, but the fire spreads quickly throughout theflammable zone of the cloud.

Fires affect the'surrounding environment primarilythrougn the radiated heat which is emitted. If the level of

heat radiation is high enough other objects which areflammable may themselves be ignited. Alternatively, living

organisms may be burned by heat radiation and thereby suffer

either injury or death.

The damage associated with heat radiation may beassessed on the basis of the dose of radiation received. A

measure of the received dose is the energy per unit area of

the surface exposed to the radiation and the exposure dura-

tion. Alternatively, the likely effect of radiation may beestimated by using the power per unit area received. Thislatter approach is particularly relevant where the equili-

brium between the power received and the power absorbed

dictate.s the degree of damage that may encountered.

Simplified models for the assessment of pool, jet,BLEVE and flash fires are given in the following sections.

A summary of some potential sources of ignition is given in

Appendix 3.

65

4.4.1 POOL FIRES

METHOD: Use of classical empirical equations todetermine burning rates, heat radiation andincident heat. The intensity is matched tolikely damage levels by reference to historic

and other data.OUTPUT: Heat intensity and thereby an indication of

the potential to cause damage/casualties.CONSTRAINTS: This empirical model for calculating the form

of the fire has only been validated forrelatively small fires.

The Method The model employed in the estimation of poolfires is derived from those indicated by TNO

and involves the use of classical empiricalequations to determine burning rates, heat

radiation and incident heat. Some of thelatter equations-are related to hypotheticalpooi fires of infinitely large radius.

The rate of burning of the liquid surface perunit area for liquids having boiling points

above ambient temperature is given by:

dm 0.001Hc

dt cp(Tb - Ta) + Hvap

On the other hand, for liquids with boilingpoints below room temperature, the expressioni s:

dm 0.001 Hc

dt Hvap

The heat flux is given by:

dmq = Hc - XE

dt

where H. is the heat of combustion and XE is

the fraction of heat produced as radiation.

66

XE normally takes a value in the range 0.13 to0.35; in the absence of better data, the value0.35 may be used to prQvide conservativeestimates of heat flux for pool fires. Whenthe heat flux at the surface of the pool firehas been calculated, the heat incident uponnearby objects may then be determined. Asimplified method assumes that all the heat isradiated from a small vertical surface at thecentre of the pool. For a ground pool, theheat incidence at a distance, r, rom the poolcentre is given by:

T qI =

where T is the transmissivity of the air pathand the other parameters are as defined else-where. In the absence of good data to thecontrary, the transmissivity is set to unity;this gives conservative results.

Outputs The method provides reasonable estimates ofradiative flux in the event of a pool fire.

Inputs (i) Thermophysical properties for thematerials of interest

(ii) Fraction of heat liberated as radiation

(iii) Transmissivity of the air path to areceiver.

Assumptions These methods are based upon empiricaland correlations.Constraints

Accuracy Generally considered valid to be for the typesof pool fires which might give rise tooff-site impacts.

Application This method may be applied to the estimationof effects of pools of fuel which igniteincluding tank fires and spreading pools onland or water.

67

4.4.2 JET FIRES

METHOD: Use of jet dispersion model to determine flamesize and radiation formula to determineintensity.

OUTPUT: Heat intensities and potential to causedamage/casualties.

CONSTRAINTS: Estimated radiation levels close to the baseof the flame may be subject to error due toflame-lift-off at the source.

The Method The model employed in estimating thermalradiation effects from jet flames is anextension, developed by the American PetroleumInstitute, of the model used for jetdispersion including wind effects (API RP521,"Flare Radiation"). The flame is consideredto be in the form of a series of point sourcesspaced along the centre line of the jet withall sources radiating equal quantities of heat

q.

The radiation I from a particular point in theflame to a receptor at distance, r., can betaken as:

Xgq'I = --- …-----

4 <r 2

where r is the distance from the point and Xgis an emissivity factor dependent on thenature of the combustible material involved inthe flame. A value of X of 0.2 is suggestedfor jet fires. To calculate the powerradiated to a receptor point, the flame isrepresented by selecting a number of locationsalong its central axis.

These are assumed to be point sources ofradiation whose total power output equals thatof the flame. The power incident at a

68

receptor point is then evaluated as the sum ofpower from all of these points within theflame. A more complex jet dispersion modelnaturally involves more complex jet shapes andrequires the use of a computer program tocalculate appropriate "view factors". Thelatter are the integrals of the visiblesurface.of a particular flame with respect tothe various receptors of radiation). Tablesof view factors are available but it is notusually considered necessary to use them in aninitial hazard analysis because the simplerexpressions outlined here generally are

- adequ-ate for these purposes.

Outputs The calculations produce an estimate of theradiative heat flux which is received at aplane with a defined inclination to the flame.

Inputs (i) Fuel input rate(ii) Length of jet(iii) Distance and orientation of receptor

Assumptions It is generally assumed for convenience thatand the flame will have approximately the sameConstraints length as an unignited jet. If detailed

radiation envelopes are required, theiterative procedures necessary to accomplishthe calculations are best conducted bycomputer.

Accuracy The method is not accurate at the base of aflame if lift-off of the flame occurs. Theseconditions are likely to arise when highpressure jets are under consideration. Forlarge flames an additional allowance for a-nextended flame shape due to flame 'thrust maybe needed over and above the estimates basedon unignited jets. Conservative estimates aregenerally detained if a flame thrust factor of1.5 times the distance to LFL is used.

Application These simplified metheds may be used toestimate radiation levels from jet releases offlammable material.

69

4.4.3 FIREBALLS

METHOD: Empirical Correlation of fireball radius basedupon work by the American Petroleum Institute.

OUTPUT: Fireball radius and heat flux.CONSTRAINTS: Applicable to fire-balls occurring in the

outdoor environment.

The Method both the radiation intensity at a distancefrom the fireball centre and the duration ofthe fireball can be determined using a verysimple calculation. The maximum radius of thefireball is given by:

Rf = 2.665 M 0O 3 2 7 (in m.)

where M is the flammable release mass inkilogrammes.

The fireball has a durati-on of tf secondswhere

tf = 1.089 M0O 3 2 7 (in seconds).

The release o-f energy by combustion is tho-igiven by:

K = Hc M 0 .6 3 7

1.089

So the radiation flux, I, at a distance r isgiven by:

I KXET

The fireball duration and diameter expressionsused above are those proposea in a recentreview of fireball models. (API RP521 "FlareRadiation").

70

Output This method gives the radiation intensity atspecified distances from the centre of thefireball and permits an estimation to be madeof the fireball duration.

Input Mass of release and heat of combustion.

Accuracy These methods are based upon empirical corre-lations which may be updated from time to timein the light of new evidence. The methods areconsidered to be-adequate for an initialhazard assessment using current state of the

- art techniques.

Assumptions As indicated elsewhere.andConstraints

Application These methods may be used to estimate theeffect distance and range of,fireball impactsi;ithin the ranges inherent in the API reviewAPI RP 521.

71

4.4.4 FLASH FIRES

METHOD: The gas dispersion models described previouslyare applied directly

OUTPUT: Extent of flash fireCONSTRAINTS: As for dispersion models

The Method It is generally assumed that a flash firespreads throughout that part of the vapourcloud which is above LFL. There is, however,little information on the effects of a flashfire on people. The dispersion calculationspresented previously may be used to establishUFL and LFL contours. Subject to ignition ofthe cloud, the conservative assumpLion isgenerally made that all of the people.outsidebuildings, but in areas between these contoursare considered, to be killed; of the peopleinside buildings, a fraction, f, may beassumed to be killed. In an initial hazardassessment and in the absence qf otherinformati.on, f is usually taken. to be zero.

For a more detailed analysis, the effects ofthe vapour cl.oud igniting taking into accountthe various atmospheric conditions which maybe encountered at several different timesafter release should be examined.

Outputs This method yields the extent of potentialflash fires and provides broadly conservativeestimates of the expected fatalities.

Inputs (i) Cloud density profile from previous dis-persion calculations.

(ii) Flammability limits.

Assumptions The method assumes that significant over-and pressures do not occur and includes only anConstraints appraximate assumption about the magnitude of

* the potential impacts.

72

Accuracy Not known but considered to be adequate forthe purpose of an initial hazard analysis.

Application This method is anplicable to flash firescenarios which may be encountered in openterrain.

73

4.4.5 FIRE DAMAGE

METHOD: Fire damage estimates are based upon correla-tions with recorded incident radiation fluxand damage levels

OUTPUT: Indication of damage as a function of incidentradiation

CONSTRAINTS: Since damage estimates are based uponempirical evidence the damage responsecharacteristics should be updated as and whennew evidence comes to light.

The Method Various tables have been created to set upcriteria for damage to people and propertyfrom fire. Sometimes they are expressed interms of radiation intensity and sometimes asa power dosage. The effect on buildings,natural surroundings and equipment is measuredin terms of the likelihood of ignition, parti-cularly if wooden structures or buildings arein the vicinity. Spontaneous and flame-induced ignition values can be considered forvarious levels of radiation. The radiative orincident fluxes recorded i-n Tables 4.2overleaf are related to the levels of damageand impact upon people, including plantpersonnel, based upon observations arisingfrom actual incidents and- large fires.

Outputs The method provides estimates of fire damage,fatalities and injuries.

Inputs (i) Estimates of thermal flux at selectedreceptor points using appropriate firemodels described in previous sections.

74

TABLE 4.2 DAMAGE CAUSED AT VARIOUS INCIDENT LEVELS OFTHERMAL RADIATION

INCIDENT FLUX TYPE OF DAMAGE CAUSED(kW/m 2 )

37.5 Sufficient to cause damage to processequipment. 100% lethality.

25 Minimum energy required to ignite woodat infintely long exposures. (nonpiloted). 100% lethality.

12.5 Minimum energy required for pilotedignition of wood, melting plastictubing. 100% lethality.

4 Sufficient to cause pain to personnel ifunable to reach cover within 20s;however, blistering of skin (ist degreeburns) is likely. 0% lethality.

1.6 Will cause no discomfort for longexposure

Assumptions At the lower levels, where time isand required to cause serious injury toConstraints people, there is often the possibility to

escape or take shelter.

AccuracY The accuracy of the incident flux damagerelationships is considered to be adequatefor initial hazard assessments and within theestimation of hazardous incidents.

Application The correlations of thermally induced damageor injury may be applied to hazardassessment.

75

4.5 Explosions

An explosion is a sudden release of energy in a

material, its violence depending on the rate of energy re-

lease. Considering explosions in the most general sense the

types involved in hazardous installations may include the

release of both physical or chemical energy.

Where chemical energy is involved, there are two main

types of explosions which are of importance to the process

industries. These are deflagrations, in which the flamefront travels relatively slowly through the combustiblematerial, and detonations, in which the flame front becomes

coupled with a shock wave which is travelling faster than

the speed of sound into the combustible gases. A deflagra-tion may turn into a deronation, if the source of ignition

is strong, or if the combustible gases are in a confined or

semi-confined area and a sufficient run up distance is

available.

It should be noted that an explosion is one of the twopossible results of ignition in a flammable release, theother being a flash fire. The probability split between the

two events is a matter for these analyst's judgement.Typically, for those situations where a delayed ignition -could occur a reasonable assumption, in the absence of dataof the contrary, is that approximately 15% of the releases

could result in explosion characteristics, with theremainder being in the form of flash fires.

One major consideration in analysing explosions is to

consider whether the explosion is essentially confined orwhether it is unconfined. An explosion within a vessel is

obviously confined and the effects of these are treated

separately below.

By definition, an unconfined explosion is one in which

a gas cloud is formed on flat ground with no significantstructures or obstructions which would tend to restrict the

expansion of the burning cloud. An explosion of a vapour

cloud in this manner is referred to as an Unconfined Vapour

Cloud Explosion (UVCE).

76

A confined explosion may also be encountered whenthere is a significant amount to 'obstruction' of theexpansion of the burning gas or vapour cloud in more thanone dimension (i.e. more than by the surface on which thegas has spread). Typically, a confined explosion may occurin built up areas particularly where buildings or structuresare present.

A number of attempts have been made to analyse thebehaviour of flammable mixtures under explosion conditionsto provide a general theoretical interpretation of such pro-cesses. These have not been very successful largely becauseof the uncertainties in adequately describing the conditionof the explosible gas mixtures prior to ignition and thecomplex coupling between turbulence, shock waves and flamespeed. A more pragmatic approarh is to adopt correlationmethods based on field studies with well defined explosiblematerials. The correlation methods are adopted in this.manual.

An explosion within a vessel manifests its effects bythe properties and fragments which it produces. An approxi-mate method for estimating the sizes, velocities and dis-tances reached by such projectiles has been included in thissection for completeness.

77

4.5.1 EXPLOSION CORRELATION

METHOD: Correlation of damage produced with energy ofexplosion.

OUTPUT: Distances to various levels of damage causedby a vapour cloud explosion;

CONSTRAINTS: Should not be extrapolated for very large orvery small clouds.

The Method Hemispherical cloud explosions arise fromground level ignition in unconfined areas, andthere are two methods of estimating theeffects.. The first is to estimate the damagelevels directly and the second to find theoverpressure and. other parameters, and esti-mate damage from them.

The damage radius R(S) is given by

R(S) - C(S)(NEe) 1 /3

where C(S) is an experimentally derived con-stant defining the level of damage based uponwork by the Dutch State Mines (DSM) Company.It varies from 0.03 for heavy damage to 0.4for light glass damage and light injury. Eeis the total energy of the explosion and N isthe yield factor or the proportion of theenergy, Ee, which is available for pressurewave propagation.

Outputs The method gives the distances to variousdefined levels of damage. Using the notationabove, these are typically:

R(1) for C(l) - 0.03 Heavy damage tobuildings and toprocessing equipment

R'3) for C(3) = 0.15 Glass damage causinginjury

R(4) for C(4) - 0.4 10% glass damage

78

TABLE 4.3 EXPLOSION LIMIT VALUES FOR VARIOUSCHARACTERISTIC TYPES OF DAMAGE

C(S) LIMIT VALUE CHARACTERISTIC DAMAGE(mJ-1/3)

C(1) 0.03 Heavy damage to buildings and toprocessing equipment

C(2) 0.06 Repairable damage to buildings andfacade damage to dwellings

C(3) 0.15 Glass damage causing injury

C(4) 0.4 Glass damage about 10% of panes

Inputs In finding damage levels directly, it is assu-med that the total amount of combustiblematerial in the explosive part of the cloudand the heat release per unit mass of thematerial are known. The product of these'givesthe total energy Ee of the explosion.

Assumptions If the total energy available for explosion,and Eet is greate'r than 5 x 1012 joules, there isConstraints almost no information on damage effects. For

lower values Ee there is sufficient data tomake an estimate of damage. It is assumedthat a proportion N of the energy Ee isavailable. This is called the yield factorand is further assumed to be the product oftwo terms designated by Nc and Nm. Nc is theproportion of yield loss due to the con-tinuous development of fuel concentration andNm represents the mechanical yield of the com-bustion. Nc and Nm can be chosen according tothe analyst's judgement; correspondingly, thevalues usually taken for Nm are 33%. Typical-ly, N. is taken to be 30%Q for isochoriccombustion and 18% for isobaric.

Accuracy It is generally considered that the corre-lation methods when applied to vapour cloudexplosions will tend to yield a conservativeestimate of damage.

79

Application This method may be used to provide an estimateof effect distances for a range of explosionseverities for flammable clouds ofhydrocarbons containing upto 5 x 1012 joulesof energy or approximately 100 tonnes ofmaterial.

80

4.6 Effects of Toxic Releases

For an assessment of major hazards, we are concernedwith the acute toxic effects of short-term exposure at highconcentrations and not the long-term chronic effects, whicharise from long-term exposure at lower concentrations.Toxic gases and vapours cause damage to living organisms bya variety of physical and chemical mechanisms, many of whichare not fully understood. Much of the data on the acuteeffects of toxic materials have been derived from controlledexperiments on anim3ls. The effect on humans has only beencorroborated through the fortunately very limited experiencewe have in this matter, usually from occupational exposuresresulting from inplant accidents.

The effects of acute exposure to toxic gases orvapours includes: mild irritation, severe irritation,injury, irreversible injury and lethal effects. Some ofthese are manifest immediately in the exposed person, whilesome may be manifested as a delayed response to anaccidental release. Both the exposure concentration andtime of exposure are important in determining the acutetoxicity effects. This may be repressed in terms of adose/response relationship.

An approximate quantitative measure of the ability ofa chemicals to produce the most acute toxic manifestation,ie. death (through inhalation), is the lethal concentration(LC50) at which 50% of the exposed population would not beexpected to survive over the exposure period. For anatmospheric release of hazardous gases and vapours,inhalation is the main route of exposure causing acute toxiceffects. In some hazard analysis cases, it may beappropriate to specify a non-lethal injury level of exposurefor a given period as the limiting exposure criteria. Or,alternatively, a useful exposure limit to adopt may be inthe NIOSH/OSHA, Immediately Dangerous to Life of Health(IDLH) value (1978) for a 30 minute exposure.

The effect of a toxic gas is dependent upon theconcentration of the toxic compound in the atmosphere andthe time for which individals may be exposed to thatconcentration. Both of these parameters are dependent uponthe nature of the release and the dispersion of gasdownwind.

81

For an instantaneous, or near instantaneous release of

hazardous material, a cloud may pass over a populationrelatively quickly. However, acutely toxic concentrationscould be encountered even for this relatively short period.

In the case of a continuous release, relief from theexposure would not be experienced for some time unless-someform of safeguard action is taken; as a-result, lowerconcentrations may give rise to detrimental effects. Forthis reason, it is essential to consider toxic affects byusing the predictions of concentration/time profilesprovided by the dispersion models given in the recentsections of this manual.

For toxic vapour cloud calculations, there is nodefinitive lower limit of concentration such as the LFL inthe case of a flammable gas, because toxic effects depend

upon the time of exposure as well as the concentrationexperienced. To avoid calculations at negligible concen-trations, it is usefu'l to specify a level of interest orconcern, based upon the toxic parameters of the materialinvolved and the anticipated duration of the exposure. The

parameters may be adjusted by the analyst to suit thespecific situation under examination. Care should be takento adjust the time step used in assessing toxic impacts to

be compatible with the size of the cloud in order to avoidunnecessary calculations.

82

4.6.1 EFFECTS OF VAPOUR CLOUD OF TOXIC GAS

METHOD: Combining the dispersion models, frequencydata, population data and dose responserelationship to determine expected toxicimpacts.

OUTPUT: Proportion of population affected in variousweather conditiosn.

CONSTRAINTS: The validity of this approach has not beenfullv developed for all toxic materials due tothe paucity of relevant experimental data.

The Method The ear " effects during the rapid dischargeof material are not usually included in thetoxicity calculations because ehese regionsshould be confine-- within the plant boundariesfor most cases. Toxic effects during thesestages are assumed to have little significancefor the overall toxic impact. The only toxiceffects which are usually calculated are forthe subsequent dense vapour cloud dispersionphase or Gaussian dispersion as appropriate.By calculating the concentration profile inrelation to the development of the cloud, thetoxic load ca.n be estimated. The results areintegrated across the cloud to find the toxiceffect at distances, d, from the releasepoint.

where sufficient information is available, aconvenient way of expressing the effect of anexposure to toxic gases is to relate the con-centration of exposure to the duration of thatexposure using what is known as a probitfunction. A probit is a probability unit, Pr,and has the form

Pr = At + Bt In (Cnte)

83

where At, Bt and n are parameters which aredependent upon the nature of the toxicmaterial 1C is the concentration of exposurete is the duration of exposureAt, Bt and n are chosen so that thevalue of Pr is a Gaussian distributedrandom variable w.ith a mean value of 5and a variance of 1. It provides ameasure of the percentage of thepopulation who would be expected to beadversley affected in a toxic releaseof known concentration and duration ofexposure. The probit may be related tothe percentage death probability byusing the transformation given in Table4.4 overleaf.

In theory, the method of probit calculationsis applicable to all toxic materials, butdifficulties may arise in practice. Thederivation of probit functions for humanlethality is restricted by the shortage ofappropriate toxicity data upon which to base areliable judgement for the values of the.parameters forming the probit expression.

Normally, the exposure time should be setequal to the release duration. However, evenin emergency situations, where considerableconfusion may prevail, an exposure duration ofmore than 30 minutes would probably be un-realistic since potential victims would tendto take avoiding or mitigating actions withinthis time.

To evaluate the probit, Cnte must becalculated at positions of interest. This caninvolve lengthy computation and the followingsimple approaches are suggested. The objectof these procedures is to evaluate thedistance from the release at which the toxiceffect would have a value of 50%. The probitwill then have a value of 5 and would equateto the LC(50) for the defined exposure time;

84

For a continuous release:

exp (5.0 -At) Cnte

Bt

where te is the exposure time

A value of C is then obtained from which theradius to 50% fatality may be estimated usingthe models described in Section 4.3 of thismanual.

For an instantaneous release:

As the cloud passes over the population, theconcentration at any g:ven point will vary.In order to make the calculation tractable,the average concentration may b.e calculcatedalong the centre-line of the cloud. Assumingthat the cloud radius does not change duringthe passage of the cloud over the locations ofinterest the average centre-line concentra-tion is given by:

C - 0.585 C(x,o,o,t)

where C(x,o,o,t) may be calculated using themodels in Section 4.3.

The duration of exposure, te may then be givenby:

te = (R 2 - X 2 ) 1 /2

uj

where R is the cloud radius at the location ofinterest.

Toyicity data for some commonly used chemicalsare given in Appendix 4.

85

TABLE 4.4 TRANSFORMATION OF PERCENTAGES TO PROBITSIN TOXICITY CALCULATIONS (Finney, 1971)

% .0 1 2 3 4 5 6 7 8 9

0 - 2.67 2.95 3.12 3.25 3.36 3.45 3.52 3.59 3.6610 3.72 3.77 3.82 3.87 3.92 3.96 4.01 4.05 4.08 4.1220 4.16 4.19 4.23 4.26 4.29 4.33 4.26 4.39 4.42 4.4530 4.48 4.50 4.53 4.56 4.59 4.61 4.64 4.67 4.69 4.7240 4.75 4.77 4.80 4.82. 4.85 4.87 4.90 4.92 4.95 4.9750 5.00 5.03 5.05 5.08 5.10 5.13. 5.15 5.18 5.20 5.2360 5.25 5.28 5.31 5.33 5.36 5.39 5.41 5.44 5.47 5.5070 5.52 5.55 5.58 5.61 5.64 5.67 5.71 5.74 5.77 5.8180 5.84 5.88 5.92 .5.95 5.99 6.04 6.08 6,13 6.18 6.2390 6.28 6.34 6.41 6.48 6.55 6.64 6.75 6.88 7.05 7.33

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.999 7.33 7.37 7.41 7.46 7.51 7.58 7.58 7.65 7.88 0.09

Outputs This method gives the probability of exper-iencing a lethal dose (or injurious dose) ofthe substance in question at a given distancefrom a source-, taking dispersion conditionsinto account.

Input (i) Toxicity parameters for the materialunder consideration.

(ii) Concentration/time profiles of releasedmaterial.

Assumptions Many assumptions are required for this modeland and it should be noted that the resultingConstraints estimates are intended to yield an indication

of the effect distances associated withspecified toxic hazards. The calculations ofdispersion and toxic effect are interlinked;for complex situations access to suLtable com-puting facilities would be required.

Accuracy This method is considered to have an accuracyno better than a factor 2.

86

Application This method may be used to estimate approxi-mate effect distances in the event of a toxicgas or vapour release. These calculations maybe based upon Probit relationships, LC(50),IDLH or other relevant dose criterion for thetoxic pollutant of interest.

87

* . Chapter 5

Summary Of Effects

4~~~~~~4

5. SUMMARY OF EFFECTS

The results from the models are in the form of a listof effect'distances. These are distances at which in theevent of an accident or failure, consequences involvinginjury, death or structural damage.would arise. In general,the analyst should repeat the calculations for a few typicalweather conditions. It should be noted that stable weatherleads to larger effect distances for :hose cases in whichthe effect is determined by gas/vapour cloud dispersion.The analyst should therefore include the most stable weatherconditions that occur at the site under investigation in theoverall analysis in additioa to those weather conditionswhich are more frequently encountered.

To assist in organising-the results of the hazardanalysis calculations, Table 5.1 has been provided forguidance. The analyst should use one sheet for eachclustered failure case. These tables provide a convenientmedium to collate and store the results. The results maythen be plotted as- circles centred on the unit x-yco-ordinates on suitable maps. It-is likely that two suchmaps will be required. One should be of a sufficientlylarge scale so that the site may be shown in detail. Thiswould permit on-site hazards to be displayed. A second map,typically covering an area of a few kilometers around thesite, may be needed to illustrate the off-site hazards. Inaddition to hazard distance estimates, the calculationsprovide toxic gas concentrations within the vapour cloud atvarious times following any accidental release. Ifrequired, these may be plotted on the site maps asconcentration isopleths for the guidance of plant managementand designers, local and state/federal governmentauthorities and emergency planners.

5.1 On-site Hazard

The analyst should plot circles corresponding to theeffect distances shown in the proformas on relevant siteplans. This may assist in identifying knock-onpossibilities. These will exist when large vulnerableinventories and vital control equipment lie within"structural damage circles". Areas with high employeedensities which fall within "damage circles" and theadequacy of escape routes may also be examined in thismanner. Such an assessment may lead to plant design and

88

layout changes as discussed in Section 6 of this manual.

For toxic gas releases gas/vapour dispersion models enable

concentrations cofitours to be plotted at various times

during cloud travel, which may prove helpful in designing

emergency response procedures.

5.2 Off-site Hazard

The chi: concern in examining the safety of i.ndustrial

installationb is frequently the hazard posed to nearby

residential communities and other commercial industrial

installations. The analyst should identify the potential

hazards by locating the effect distance contours on a map of

the surrounding area. The safety assessment will be parti-

cularly concerned with toxic and explosion risks, since

these have the potential to kill or injure at greater

distances from the site than do on-site fires.

The results and suggestions for reducing or eliminating

risks of accidential releases'are discussed more fully in

Section 6.

89

TABLE 5.1 IiAZAN RESULTS PROFORMA

PLANT NAME _

FAILURE CASE _ _ RELEASE MA`CRIAI. _ __ PRESSURE - __

UNIT NAME _ _ RELEASE TYPE CONTfINST TFMPERATIJRE _

UNIT CORDINATES _ RELEASE PHASE (S)

VESSEL/LINE EVENT TREE FOR FLAMMABLE GAS?

FAILURE TYPE LIOUID RELEASE?

AUTHOR TOXIC GAS?

RtLEASE AMOUNT Kg_

MODEL NUMBERS

0 --- ----_ _

FLAMMABLE TOXIC

FLASH OVERPRESSURE BLEVE POOL FIRES FRAGMENTS PROBIT IDLH 1 LC5O

EkFECT DISTANCE(M) R Dl 02 03 D4 R(12 5) R(37 5) A R(SO) R RlIDNi IC50

NO S1RIJ(lURAL HEAVY SIRWUtUAt 51I(,li * ikl SLIliHI NU SkIIT(IkAL SlRUt lURAI '.ltkt IURAL NU kUHUI UkAI

100% DfATHtS 'ARUt lIRAt S0.:)1 AHS StRUt TURAL STIPUt TURAt NO DEAI IIS I0 4 bAN - I A_TMs10f' % LI EATH'% Ni) DEAllt NOJ lif A)))

WEATHERWIND SPEFn STABIUTY

I I I I .I I t * * I I .I

* ' I I ; ' ' I ' ' I

DELETE AS APPLICABLE

' AT SPECIFIED EXPOSURE TIME

Chapter 6

Applhcation Of The M;ethodology

~~ _ ~~ ~

6. APPL-ICATION OF METHODOLOGY

By following the procedures in Section 3 and 4 theanalyst will have identified which off-siLe areas are atrisk from on-site failures. He also will have identifiedfailures, which-could potentially lead to knock-on effectsin the plant. The items of plant equipment responsible forthe most severe of these consequences will have also beenidentified.

In this chapter some corrective measures which theanalyst might consider to reduce the consequences,frequencies and impacts of the failures are proposed. Thesesuggestions are necessarily of a general nature and theanalyst must decide which are suitable for the process andsite under study. It is not possible to be comprehensiveand the emphasis is placed on basic design modificationsrather than secondary "add-on" remedial measures.

6.1 Knock-on Effects

The results from the consquence calculations are in theform of£effect distances which are particularly suitable forthe investigation of knock-on possibilities.. The analystshould-examine plans of the site looking for the vulnerableplant within the effect radii of failures leading tostructural damage, as well as possible effects on theoperability of the plant, safe shut-down, etc. Ofparticular importance are:

- large inventories in vulnerable vessels

The analyst should check to see if inventories thatappear to be at risk are in fact protected by blastwalls, firewalls or some other system. He will normallyfind that large storage vessels are protected fromradiatioin damage by water drer.ching. He may considerwhether the specification and reliability of this systemis sufficient for the calculated hazard.

- load-bearing structures

It is of course necessary to examine qualitatively theconsequences of the failures of the load-bearingstructure due to fires and explosions. The analyst willbe interested in vessels and employees on the structureor positioned under the structure.

91

control or shutdown devices

These devices may be automatic or manual. In the lattercase the analyst will be able to check if the con-sequences of a release prevent an approach to the controlposition for emergency shutdown.

projectile effects

Projectiles from an explosion may damage vessels orpiping nearby and cause secondary releases of flammableand toxic materials. In addition the analyst mustdetermine if operators or the public are at risk. Blastwalls may be indicated in certain cases.

design and siting of control rooms

Fires, explosions and toxic gas releases are importantconsid#rations in the design and siting of-contro] rooms,as this is the critical area because of the concentrationof operators and the need to ensure safe shut-down ofcritical plant areas.

- modify plant layout and/or siting.

92

6.2 Reduction of Consequences

The evaluation method outlined in Sections 2 to 5identifies pieces of plant that have potential of causingsevere daAiiage both on and off-site. A list of measure,leading to reduced effect distances are proposed here. Theanalyst should repeat the effect calculations to estimatethe benefits of the proposed measures.

A hazard analysis is best carried out at the designstage of the plant, where des:gn, layout and sitingmodifications can be made Lo reduce or eliminate potentialhazards. Even at the operational stage a hazard analysiswill reveal critical items of equipment, piping, etc., wheremodifications, improved-containment and/or operation andcontrol will effectively reduce the hazard. Some hazardreduction possibilities itclude: reduction of inventories,modification of process or storage conditions, eliminationof hazardous material, imprcvement in plant operability.

6.2.1 Reduce Inventories

The primary object should be to reduce the inventory ofhazardous material, so that the potp"tial hazard outside cheplant boundary is greatly reduced oL even eliminated. Forexample:

- reduce the inventory of hazardous materials in storageand in the process. Many instances can be sited where ithas been possible to operate plants with considerablylower quantities of raw materials and intermediateproducts than originally designed;

- change the process to produce the hazardous material as asmall quantity of intermediate rather than store largequantities of this material;

- change from batch to continuous reaction system withlower inventories better mixing etc.;

- use a low inventory high efficiency process, eg. indistillation and evaporation systems.

93

6.2.2 Modify Process or Storage Conditions

If it is not possible to reduce the inventory ofhazardous material, it may be possible to change the processor storage conditi ns to reduce the hazard potential of anaccidental release.

- store and process toxic gases ir. a suitable solventrather than in large volumes;

- store and handle all materials in small discretequantities rather than in large volumes;

- prooess hazardous reactive materials in a large volume ofrecycle carrier material containing the catalyst in acontinuous reactor and thus prevent runaway reactions;

- process hazardous -aterial as a gas rather than as aliquid in a flammable solvent;

- store hazardous gas as a refrigerated liquid rather thanunder pressure;

- reduce process temperatures and pressures through processmodifications.

6.2.3 Elimination of Hazardous Material

Should the first two aiternatives prove ineffective, itmay be possible to use material or alternative processroutes to eliminate the hazardous material.

6.2.4 Improve Plant Operability and Reliability

A hazard analysis may identify critical plant itemswhere it is not possible to reduce significantly the hazardby means of the above measures. Other methods such as HASOP(Section ') or reliability studies etc. may be used todetermine corrective design, operation and control measureswhich will reduce the risk of accidents and provide adequateprotection for plant operators as well as the surroundingpopulation.

94

6.2.5 -Other Protective Measures

Other protective measures may include:

- automatic shutdown to reduce release duration;

- the provision of bunkering or blast walls;

- it is occassionally considered necessary to protectemployees or nearby plant from shrapnel from explodingvessels or rotating machinery;

- firewalls/fire proofing of struccures;

- increased bunding (diking);

Using the release calculations, the analyst can checkthat the bunds have sufficient volume for the predictedrelease scenarios. Bunds are more effirient if they havevertical walls. Since this may prevent liquid surgingout after a major leak, bunds with sloping floors ma)reduce the surface area of pool and thus the rate ofevaporation. The analyst should also check that jetreleases are contained, if thesa are considered to belikely. Evaporation rates are reduced if bund areas aredecreased and bund heights are increased. In the caseswhere the calculations indicate that knock-on effects arepossible, the analyst may wish to check that those bundswhich are common to adjacent vessels provide for anadequate containment capacity.

- :'ter curtains which may restrict gas release.

6.3 Reduction of Impacts

A range of measures are available to reduce the impactsof major hazard accidents. The analyst should at leastconsider the following:

- provision of escape routes;

The format of the output as effect distances makes theexamination of the effect or failure on escape routesvery convenient;

- evacuation planning;

- public alert systems;

95

The response of the public to an alert is every difficultto predict. Nevertheless the installation of suchsystems must be considered where significant inventoriesof toxic materials are held.

emergency procedures on and off-site;

maintenance and inspection of plant;

These are normally determined by statutory codes.

safety and emergency training;

provision of safety buffer zones around the plantboundary;

modify siting of proposed plant.

96

Chapter 7

Hazard And Operability Studies (HAZOP)

-4.~

7. HAZARD AND OPERABILITY STUDIES (HAZOP)

8.1 Background

The 'results of an hazard assessment, as described inprevious sectinns, may show that a more detailed safetystudy is required of a particular section of the plant whichpresents a significant hazard to operators and/or thesurrounding community. For this purpose, a HAZOP study maybe required to identify systematically the possible ways inwhich the system could fail.' The studies are carried out bya team including the principal design engineers, HAZOPchairman and an external HAZOP expert.

7.2 HAZOP Method

HAZOP involves the scrutiny of a large number ofpossible deviations from normal operationing conditions,which are generated by applying guide words such as MORE,LESS, REVERSE etc., to each of the parameters describingconditions in each component or pipeline in the plant.Often there is no realistic cause, or the effects areunimportant (e.g. REVERSE temperature is meaningless, LOWpressure, unless a vaccuum might be created, may not beharmful) such cases can be quickly passed over. Sometimes,the causes are credible and the effects are significanteither for the correct functioning of the process or forsafety and possibly both. In such cases, there may be aneed for design changes to eliminate the identified cause,or alternatively a more detailed reliability study may berecommended, to determine whether the probability of theevent is high enough to justify action. The team may assessthe consequences and probability subjectively as 'large' or,small' and rank the actions accordingly.

At the stage described above, major changes to thedesign may be made relatively cheaply, whereas a detailedHAZOP takes place after the design has been frozen. Thethird type of HAZOP is to consider plant alterations thatmay be made during service. This type of HAZOP provides aformal method of ensuring that any such alterations cannotcontribute to excessive hazards. Usually, a coarse HAZOP isapplied just before the design has been completed.

97

The objectives and the method of HAZOP Studies are:

- To identify those areas of the design which may possess asignificant hazard potential

- To identify and study features of the design whichinfluence the probability of a hazardous incidentsoccurring

- To familiarise the study team with the design informationavailable

- To ensure that a systematic study is made of the areas ofsignificant hazard potential

- To-identify pertinent design information not currentlyavailable to the team

- To provide a mechanism for feedback to the client of thestudy team's detailed comments.

The method of study is as follows:

(i) Identification of drawings relating to equipmentcontaining hazardous materials, based on thedrawing list and on a brief examination of thedrawings themselves.

(ii) Study of the select-d drawings following asimplified HAZOP checklist. A line-by-linestudy is not normally attempted, but in mostcases attention is concentrated on vessels andmajor equipment items. The checklist containssuch questions as -how can the vessel be overpressured?', 'what happens on low or highlevel?' etc.

(iii) The results of the study are recorded as a setof comments to be followed up. Some requirefurther investigation or calculation whereasother are passed to the project design team asrecommendations for design change.

(iv) JAZOPS are recorded in a fixed format althoughthe format varies from study to study.

98

HAZOP methods, as prac-tised to date, are onlyapplicable to process hazards but there is no doubt that itcould be developed to apply to structures, managementprocedures and many other systems that relate to safety.This would, however, probably involve the development of newguide words.

The technique can be laborious but the efficiency ofthe studv team increases rapidly with experience, as trivialcases can be more quickly identified and resolved. It isunwise, however, to take too many short-cuts because thisundermines the main advantage of the method, which is itst..oroughness and comprehensiveness in failure caseidentification.

7.3 Further Information

For further reading, the analyst is referred to thebooklet "A Guide to Hazard and Operability Studies"published by the Chemical Industries Association (1977) andRoach and Lees (1981). Other systematic hazard assessmenttechniques are being used by the chemical industry which arewell presented in the U.S.A Chemical ManufacturersAssociation Publication "Pr.ocess Safety Management" (1985).

99

Appendices

7~~~~~~~* 4

-a,aa,, , - '

*~~~~ =--5^-;.

-

1'y.~ ' __=

APPENDIX 1

Nomenclature and Glossary of Terms

NOTATION USED IN SECTION 4 UNITS

a. thermal diffusivity of substrate m2 /sAr area of release m2A constant in entrainment equation -

A major semi-axis of vapour cloud mApI area of a pool of spilled liquid m2At material dependent toxicit.y

parameterB minor semi-axis of vapour cloud mbl, b2 concentration profile shape

parameterBt material dependent toxicity

parametersCpl specific heat of liquid at

constant pressure J/kg/KCd discharge coefficientCe liquid pool evaporation factor -

CA ambient concentration (volumefraction)

C0 theoretical concentrationCT true concentration kg/m3CV volume fraction of vapour in

initial mixture kg/m3Cv heat capacity at constant volume J/kgKCs thermal conductivity substrate W/mKC local concentration of pollutant kg/m3C(S) damage coefficient in DSM model

for a damage level of S definedin categories 1,2 and 3

D diffusion coefficient m2/sdm equivalent diameter of initial

vapour/air mixture mDa density of air kg/m3

Di initial density of explosivemixture kG/M3

Dl liquid density kg/m3

Dm mean density of liquid/vapourmixture kg/m3

Dp local plume density k/m3

Dv vapour density kg/m3

Dw water density kg/m3Do vapour density under initial

conditions kg/m3e liquid pool spreading factor

having a value of between 1 and 6 -

E energy of the explosion joules

Ee energy of expansion J/kgF buoyancy factorFvap fraction of liquid flashed to vapour -

G Cp/Cv (ratio of specific heats forgas or mixture)

g gravitional acceleration m/s2

h height of liquid in tank abovedischarge point m

H cloud depth mHI enthalpy of liquid per unit mass J/kgHv enthalpy of vapour per unit mass J/kgHvap enthalpy of evaporation J/kgJc core concentration kg/m3Hc heat of combustion J/kgI incident heat kW/m2

ic LFL (Lower Flammable Limit)concentration kg/m3

J edge mixing entrainment coefficient -

k constant in spreading equation -

K rate of energy by released due tocombustion J/s

Kd turbulent deffusion coefficient m2sl1 characteristic tufbulence length

scale mM molecular weightdmdt mass burning rate kg/m2 sn material dependendent toxicity

parameterN yield factor for vapour cloud

explosionP pressure N/m2-a ambient pressure N/m2P1 pressure of reservoir or process

pressure N/M2PC pressure at release plane N/m2Pr toxicity probitQ mass released or release rate as kg or

appropriate ks/sQe air entrainment rate m/sq heat flux kW/m2R radius mdrdt rate of spreading of pool m/sdRdt rate of lateral spreading of a

dense gas cloudrc radius of core of cloud m

/J.

Rf fireball radius mR(S) distance from cloud centre to

damage level S defined in- categories 1,2 and 3 m

r distance from fire centre ma XlyjZ dispersiop standard deviation

parameter (Gaussian dispersion modelfor plume sp:ead) m

S damage level, DSM model -

S1 entropy of liquid per unit mass '/K kgSv entropy of gas per unit mass J/K kgte exposure time to a toxic gas

or vapour minutesT transmissivity of air patht time sTb boi.ling point KTc saturation temperature of liquid

at release plane pressure KTj initial temperature KTs temperature of substrate KU1 internal energy. of the liquid per

unit mass J/kgUV internal energy of the vapour per

unit mass J/kgu1 mean wind velocity m/sUe entrainment velocity m/sUf friction velocity m/sv volume m3

Vgo volume of the gas at standardtemperature and pressure m3

x,y,z space co-ordinates mXE fr'npe o of combustion heat

emitted as radiation.ff¢ . ..... jet-flame emissivity factor,

typically taken to be 0.2e liquid spreading rate constant -

Ce evaporation constant k5/m2sl/2Ap surface area of pool mpi constant 3.14U1 initial internal energy j/kgU2 internal energy after expansion j/kgT2 . ..... temperature after expansion kf2 . ..... vapour fractionT3 . ..... cloud temperature after air

entrainment kMair mass of air kgCa heat capacity of air j/kg ks.v.p. saturated vapour pressure N/mr

Ri Richardson numbera,b,c,d dispersion deviation parameters -

Lp Brig-s parameter s2 /m

tf fireball duration s

Pr probit function

APPENDIX 2

Extract of World Bank Guidelines forIdentifying, Analysing and Controlling

Major Hazard Installation-s inDeveloping Contries

"VORLD BANK GUIDELINES FOR DENTTFYfIG,ANRL7ZING AND CONTROLLING MAJOR HAZARD INSTALLATIONS

IN DEVELOPING COUNTRIES-

September 1985

Office of Environment and Scientific AffairsProjects Policy Department

WORLD BANKWashington, D.C.

~ - i - September 1995

WORLD BANK GUIDPLINES FOR IDENTIFPBG, AALYZING, ANDCONTROLLING MAJOR HAZARD INSTALLATIONS aI DEVELOPING COUNTRIES

Preamble

The European Economic Community have taken a lead indeveloping guidelines controlling maj.r accident hazards of certain.industrial activities. The Environmental Council of the EconomicCommunity met on June 24, 1982 and adopted such a directive whichmember states were required to comply with by January 8, 1984.

Impetue was given to the European Community to considerthe need to control major hazards by, in particular, four seriousindustrial accidents; the Flixborough explosion in 1974 killed 28workers, injured 89 people and caused widespread damage to housing inthe vicinity of the plant; the disaster at Beek in Holland in 1975,an explosion and fire killed 14 people on site following the releaseof propylene at the refinery. The two other cases were at Sevtso andManfredonia in Italy in 1976, where highly toxic substances werereleased contaminating the surrounding districts, and raisingimplications regarding the health of people exposed to the toxicreleases.

Recently the explosion of natural gas in Mexico Citykilling some 450 people anu the toxic gas release at Bhopal in Indiakilling more than 2,500 people has iighlighted the urgent need for theWorld Bank to adopt similar guidelines to those developed by the EEC.These latter two incidents illustrate the even greater risks that mustbe controlled in installations producing hazardous substances indeveloping countries.

TABLE OF CONTENTS

Page No.

1.0 Introduction ...... ............ 1

2.0 Potential Industrial Hazards ............................... 1

3.0 Identification System for Major Hazards ... ...... .3

3.1 Introduction .... ........ ............................. 33.2 Threshold Quantities for a Major Hazard Assessment .....5

4.0 Implementation of the Guidelines ..................... too..6

4.1 Requirements for a Major Hazard Assessment ............. 6

4.1.1 Objectives ................................ .64.1.2 The Contint -f the Major Hazard Assessment ......74.1.3 Informat.'ci to be Included in Major Hazard

Assessm w:. .............................. 8

4.1.3.1 Infor=a:ian Relating to Sub-stances Listed in Appendix II & III ........8

4.1.3.2 Informatic-n Relating to the in-stalltion ...... .................. 9

4.1.3.3 Informa-.:n Relating to theManage"net System forControl)Bng the Activity ................. 9

4.1.3.4 Information Relating to thePotential Major Accidents ............. 9

5.0 Emergency Plans ........................................... 12

5.1 On-Site Emergency Plan ................................ 125.2 Off-Site Emergency Plan ..... .12

6.0 Restrictions on Development in the Vicinityof.Ma'or Hazard Installations ........................... 13

Bili ography .. . .... .. .. ............ 14

Appendix I ............................ 15Appendix II .................................. 17Appendix TIII.......... o........ .... 21Appe-ndix IV .................................. 29Appendix V................... ....... .. 38

1.0 Introduction

1. These guidelines are based substantially on the EECdirective on the major accident hazards of certain industrial activi-ties and regulations promulgated under the United Kingdom Health andSafety at Work Act. Parts of this guidelines have been quoteddirectly from "A Guide to Control of Industrial Major Accident HazardsRegulations 1984", UK Health and Safety Booklet HS (R)-21. (publishedby HMSO, London).

2. Industrial activities involving certain dangerous sub-stances have the potential to give rise to serious injury or damagebeyond the immediate vicinity of the work place. These activitieshave commonly come to be known as "Major Hazards". These guidelinesare concerned with the protection of the health'and safety of personsin the workplace and persons outside the plant boundary, as well asthe protection of the environment. Furthermore, they apply generallyto industrial processes, storage and transport of hazardous material,but do not apply to nuclear or tc extraction or mining operations orto licensed hazardous waste disposal sites. According to the guide-lines, persons in control of activities involving certain dangerous,explosive, flammable and toxic substances must demonstrate that majoraccident hazards have been recognized, and that measures have beentaken to prevent accidents and to control and minimize theconsequences of those that do occur.

3. It is the object of these guidelines to provide a frame-work in which a developer can supply evidence and justification forthe safe operation of the proposed industrial activity. It is not theobjective of these guidelines to provide details or specific methodsof analysis, safe operating procedures, etc., which are the contentsof the World Bank Manual of Industrial Hazard Assessment Techniques.

4. In summary, these guidelines provide the criteria foridentifying acutely toxic, flammable, explosive and reactive hazards,as well as providing an indicative list of these hazardous chemicals.in addition threshold quantities are specified, which require thedeveloper to undertake a major hazard assessment and to implementmeasures to control major hazards that are identified in such anassessment.

2.0 Potential Industrial Eazards

5. Although 'major hazard (or major accident)' is defined inthe guidelines and includes the phrase 'a major emission, fire orexplosion', the definition uses a number of phrases which need to beinterpreted. An occurrence will be a major accident if it meets thefollowing conditions.

2-

(a) that it leads to a serious danger to people or theenvironment;

(b) that it results from uncontrolled events in the courseof an 'industrial activity'; and

(c) that'it involves one or more 'dangerous substances'.

6. 'Serious danger to persons' should be taken to mean deathor serious injury including to health, or the threat of death orserious injury, whether caused immediately by the accident (e.g., col-lapse of a populated building caused by an explosion) or as a delayedeffect (e.g., pulmonary oedema following some hours after exposure to.a toxic gas),.and affecting or potentially affecting people inside andoutside the installation. It is emphasized that the accidents, actualor potential, should be major ones distinguished from other seriousaccidents not onlv by the severity of the casualties but by thenumbers of them, or by the physical extent of the damage.

7. The reference to delayed effects is not intended toinclude the cumulative effects of frequent exposure to small amountsof the dangerous substance and, therefore, brief excursions slightlyabove the routine control limits for toxic substances shculd not beconsidered as major accidents.

8. 'Serious danger to the environment' should be taken tomean a significant, relatively'long-lasting (but not necessarily irre-versible) effect on plants or animals on land, in the air or in thewater which has the potential to lead to a serious danger to man. Forexample, serious pollution by a toxic substance of a water course usedfor drinking water could pose a threat to man.

9. 'Uncontrolled developments' should be taken to mean thatthe occurrences of concern are likely to develop quickly; to beoutside the normally expected range of operating problems; to presentonly limited opportunity for preventive action; and to require anysuch action to be in the nature of an emergency response. It alsoserves to indicite that the guidelines are concerned with acute ratherthan chronic events, i.e., uncovenanted or unusual rather thancovenanted or regular releases of the dangerous substance. Similarly,'a major emission' refers to a relatively large, sudden andunconvenanted release of the dangerous substance from its normalcontainment.

10. It is clearly possible to identify, using a pragmaticapproach, the installations and activities that pose the main threatof a major accident. It is also relatively easy to decide whether anevent was a major accident after it has occurred. It is much lesseasy to define 'major accident' for the purpose of making predictions,as the developer is required to do in his major hazard assessment as

_3-

an essential step in demonstrating the adequacy of the measures takento prevent such accidents. The following examples outline eventswhich may be taken, prima facie, as major accidents:

(a) any major fire giving rise to thermal radiation at thesite or plant boundary exceeding 5 kw/m2 for severalseconds;

(b) any release actual or potential, of a hazardous sub-stance where the total quantity released is a signifi-cant proportion of the quancity which invokes theguidelines, e.g., releases of kilogram quantities ofGroup A toxic substances; ton quantities of othertoxic substances; ton quantities of pressurized orrefrigerated flammable gases; or tens of tons of flam-mable liquid;

(c) any vapour or gas explosion which could give rise to.blast overpressures at the site or plant boundaryexceeding 0.5 bar; and/or

(d) any explosion of a reactive or explosive substancewhich could cause damage to buildings or plant outsidethe immediate vicinity sufficient to render them or itinoperative for weeks.

11. These estimates should include all the quantities of eachsubstance present whether it is in pure form or part of a mixture.However, the substances should be in a form capable of giving rise tomajor accident hazards, e.g., no account should be taken of ammoniaunless it is anhydrous or is in water solution containing more than50% by weight, nor should account be taken of stored chlorinatedpotable water.

12. For the purpose of these guidelines a "developer' isdefined as a manufacturer, distributor, transporter or end-user, whomanufactures, processes, stores, or transports hazardous chemicals inthreshold quantities greater than those identified in Appendix II andIII of this document.

3.0 Identification System for Major lazards

3.1 Introduction

13. The proposed system for identifying major hazards is basedon the quantity (or inventory) of hazardous substance stored or pro-cessed at an industrial site or in transit. In this context the term'installation' is used to descrtibe the general activity which mayresult in a major accident hazard. The term 'installation' is definedfurther in Appendix I.

-4-

14. - The threshold quantities specified in Section 3.2 relateto the total quantity of substance held on site. A developer may beinvolved in an activity in which the quantity of a hazardous substancevaries over a period, due either to seasonal demand, or because thesite is complex.and includes a number of processes each of which hasan inventory which varies from day eo day or even hour to hour. Insuch cases the manufacturer should make an estimate of the quantity ofeach substance liable to be on site and a decision as to whether therequirements of Sections 4, 5, and 6, apply should be based on themaximum anticipatee auantity.

1S. The estimation of the quantity of a substance at a siteshould include all the amounts which are likely to be on the siteunder the control of the same manufacturer. In the case of productionand process activities, this will include quantitites in manufacture,use or processing, and associated storage (i.e., the storage that isused in connection with the process). Account should also be taken ofany quantities in pipelines on the site and in internal transportoperations. These estimates should include all quantities of anydangerous substance whether the substance is in pure form or part of amixture or present as a by-product. For example, if a chemical plantmanufactures a hazardous substance, the estimate of the total quantityshould take account of the quantities which are present in reactionmixtures and purification processes together with the quantities whichare present in storage. In certain processes there may be circum-stances where a significant quantity of a dangerous substance can onlybe produced if abnormal coaditions develop in the plant. If such anevent ca t reasonably be predicted, then this should be taken intoaccount in estimating the overall quantity of the substance on a site.An example of this type of situation was the production of a signifi-cant quantity of TCDD (dioxin) when conditions of excess temperatureand pressure developed in a plant producing 2,4,5-trichlorophenol atSeveso.

16. It should be noted that when estimating the quantity ofhazardous substance to assess whether the site becomes subject tothese requirements, it is necessary to add the quantity of substancein process to the quantity in associated storage together with anyamounts of the same substance in any installation within 500 metresowned by the same manufacturer. If the nearby installation is morethan 500 metres away then it is only necessary to add the quantitiesif these are such that there could be, in foreseeable circumstances,an aggravation of the major accident hazard. It is realized that insome cases interpretation of this 500 meter requirement might be dif-ficult but reasons for inclusion or exclusion should be clearly speci-fied.

17. Although an industrial activity involves, or is liable toinvolve, hazardous substances, this does not of itself make the activ-ity subject to these World Bank guidelines. The hazardous substance

must be present under circumstances which could give rise to a majoraccident. Thus, it may be possible for a developer to argue that amajor accident cannot, in fact, arise. It may be that the physicalstate of a subr-ance or the way in which it is distributed round thesite may avoid the possibility of some types of accidents. This maybe particularly relevant to many of the toxic substances which areinvolatile liquids or solids. These may not have the potential tocause a major accident unless some special factor, such as energycontained in a pressurized system, is present, although spillage intowater courses may still remain a problem.

3.2 Threshold Quautities for a Major Hazard Assessment

18. The criteria for hazardous substances and the quantitiesabove which a major hazard assessment is required by the World Bankare presented in Appendix II. This Appendix is divided into foursections, namely:

(A) very toxic substances;

(B) other toxic substances;

(C) highly reactive substances and explosives; and

(D) flammable substances.

19. The criteria for the two groups of toxic substances, name-ly groups (A) and (B) of the Schedule are given in terms of the toxiceffects on populations of specified experimental animals, though intwo cases in addition to fulfilling these criteria the substances musthave physical and chemical properties capable of entailing major acc4-dent hazards. This is taken to mean that the properties are such thatthe toxic substance could be easily distributed throughout the envi-ronment if containment is breached, for example, a gas or highly vola-tile liquid or a solid which might be ejected from a pressurizedreactor. A substance has to satisfy only one of the three criteriafor ingestion, percutaneous (i.e., through the skin) or inhalationtoxicity in the tables to qualify. In the case of Class (A) sub-stances a major hazard assessment is required irrespective of thequantity involved.

20. In the case of Class (B) toxic substances quantities havebeen specified for some of the more common substances. However, forthose unnamed substances which fall into the indicative criteria givenin Appendix II, quantities exceeding 1 ton would require a majorhazard assessment.

21. A major hazard assessment also is required for any processusing plant at a pressure greater than 50 bars when the product of thevolume of the pressure system in cubic meters and the pressure in barsexceeds 10,000. Likewise for highly reactive chemicals (Class (C))

-6-

and'flammable substances. (Class (D)), threshold quantities are givenin Appendix II which would require a major hazard assessment for thepurpose of these World Bank Guidelines.

22. To assist in identifying .major hazard chemical's anindicative list of acutely toxic and reactive chemicals in Classes(A), (B) and (C) have been listed in Appendix III.

4.0 Implementation of the Guidelines

23. The guidelines require that proof of safe operation beavailable at any time. Developers must show that they have identifiedthe major accident hazards arising from their activities and havetaken adequate steps to prevent such major accidents in design, lay-out and siting, that they will provide adequate steps'to prevent suchmajor accidents during operations and will provide people on-site withthe information, training and equipment to ensure their safety.

4.1 Requirements for a Major Hazard Assessment

24. For major hazard installations handling dangerousmaterials in excess of the quantities listed in Appendix II a majorhazard assessment is required. This study must show that the activitywill be carried on safely; it includes a description of the majoraccident hazards that could arise from a manufacturer's activities andthe controls that are exercised to prevent them or to limit theirconsequences. 'Major accident' is defined in Section 2 and guidanceis given on the definition in this section. The guidance that followsdiscusses some of the general issues that bear on the major hazardassessment.

4.1.1 Objectives

25. The objectives of the major hazard assessment are:

(a) to identify the nature and scale of the use ofdangerous substances at the installation;

(b) to give an account of the arrangements for safe opera-tion of the installation, for control of seriousdeviations that could lead to a major accident and foremergency procedures at the site;

(c) to identify the type, relative likelihood, and broadconsequences of major accidents that might occur; and

(d) to demonstrate that the developer has appreciated themajor hazard potential of the company's activities andhas considered whether the controls are adequate.

-7-

26. In addition, the work that the developer does in preparinghis assessment should enable him to provide the competent authorityresponsible for making emergency plans outside. the installation withan estimate of the scale and consequences of the realization of thehazards, in accordance with the requirements of these guidelines(see Section 5).

4.1.2 The Content of the Major Hazard Assessment

27. A major hazard assessment is essentially an abstract ofrelevant information about the major hazard aspects of the activitiesfrom a much more extensive body of information. this body ofinformation will include plant design specifications, operatingdocuments, maintenance procedures, and information derived from theexamination of the major hazard potential by means, of techniques suchas discribed in the World Bank Manual of Industrial Hazard AssessmentTechniques.

28. The information required in the major hazard assessmentfalls into two broad categories: firstly, factual information aboutthe site, its activities and surroundings, and secondly, (the core ofthe assessment) estimates of the scale of potential major accidents.which may occur at the installation and the means to prevent thesehazards being realized.

29. It is not possible to specify precisely what the secondpart of the major hazard assessment should contain because thecomplexity of the potential hazards will vary greatly from site tosite. Eowever, the World Bank manual provides detailed guideance onthe methodology for carrying out such an assessment. The essence ofthe major hazard assessment, and the reason behind the choice of thatterm, is that the onus lies on the developer to assess his ownhazards, take measures to control them adequately, and then to presenthis conclusions to the World Bank.

30. The major hazard assessment should, therefore, contaiasufficient information about the major accident potential of thedeveloper's activities to enable judgment to be made whether thesignificant hazards have been identified and are being properlymanaged. In some instances it may be necessary to ask for informationin the assessment to be supplemented by further information, butideally the aim should be to provide an analysis which stands on itsown as a demonstration that major accident hazards are beingadequately controlled.

31. The major hazard assessment should provide adequatejustification for ics conclusions, either by setting out the sourcesof the evidence for a particular argument, or by recording theprincipal assumptions in sufficient detail to enable them to be

-8-

challenged if it emerges that they are critical to the conclusions ofthe assessment. For example, a major hazard assessment may state thatthe integrity of pressure vessels has been assured by the strictapplication of appropriate design codes, operating duties, maintenanceand inspection procedures, in support of an assumption that the suddenfailure of pressure vessels has been dismissed as a possible cause ofa major accident. A major hazard assessment may also perhaps indicatethat the risk of an aircraft crashing on the installation isinsignificant in comparison with other causes of a major accident,because the site is well separated fom the nearest airport and airtraffic lanes. Clearly, the amount of evidence required on eachaspect of the maior hazard assessment will vary according to theimportance of that aspect and in particular the consequences of theparticular accident being considered.

4.1.3 Information to be Included in a Major Hazard Assessment

32. The report shall contain the following:

4.1.3.1. Information relating to substances listedin Appendix II and III:

(i) The name of the substance as given in Appendix tIIIor for substances included in Appendix II under ageneral designation, the name corresponding to thechemical formula of the substance;

(ii) A general description of the analytical methodsavailable to the developer for determining thepresence of the substance, or references to suchmethods in the scientific literature;

(iii) A brief description of the hazards from the sub-stance.

(iv) In cases where the substance may be isolated fromprocess vessels, its percentage concentration, andthe main impurities and their percentages.

(v) Where there is a potential for runaway reactions afull discription is required such as given in theexample in Appendix V, and the consequences ofrunaway reaction determined as part of the maiorhazard assessment. This may involve computermodelling, as well as data from bench scale testinge.g. as detailed in references (8) and (10).

_9_

4.1.3.2 Information relating to the installation:

(i) A map of the site and its surrounding area to ascale large enough to show any features that may besignificant in the assessment of the hazard or riskassociated with the site.

(ii) A scale plan of the site showing the locations andquantities of all significant inventories of thehazardous substances.

(iii) A description of the processes or storage involving.the hazardous substance and an indication of theconditions under which it is normally held.

(iv) The maximum number of persons likely to be presenton site.

tv) Information about the nature of the land use andthe size and distribution of the population in thevicinity of the activity to which the major hazardassessment relates.

4.1.3.3 Information relating to the management system forcontrolling the activity:

(i) The staffing arrangements for controlling the acti-vity with the name(s) of the person(s), and ifappropriate his (their) deputies or the competentbody responsible for safety and authorized to setemergency procedures in motion and to inform out-side authorities.

(ii) The arrangements made to ensure that the means pro-vided for the safe opera:ion of the activity areproperly designed, const.acted, tested, operatedand maintained.

(iii) The arrangements for training of persons working onthe site.

4.1.3.4 Information relating to the potential major hazardaccidents:

(i) A description of the potential sources of a majoraccident and the conditions or events which couldbe significant in bringing one about.

- 10 -

(ii) A diagram of any plant or plants in which theactivities are carried on sufficient to show thefeatures which are significant as regards thepotential for a major accident or its prevention orcontrol.

(iii) A description of the measures taken to prevent,control or minimize the consequences of any majoraccident.

(iv) Methods of sizing and capacities of emcrgancyrelief systems, blow-down canks, emergency scrubbersystems, vent flares, etc., especially to handletwo-phase flow conditions (i.e. the subject of aforthcoming American Institute of ChemicalEngineers publication).

(v) Information about the emerger.cy procedures laiddown for dealing with a major accident occuring atthe site.

(vi) Information about prevailing meteorologicalconditions in the vicinity of the site.

(vii) An estimate of the number of people on site who maybe particularly exposed to the hazards consideredin the written report.

33. Further details on these items are given in Appendix IV,and the World Bank Manual of Industrial Hazard Assessment Techniques.A flow chart (Figure 1) summarizes the procedures outlined above inSection 4.

- 11 -

Figure 1: Procedures for Major Hazard Assessment-and ControlOf Major Hazard Installation

START HERE

HAVE I A SUBSTANCE DEFINEDIN APPENDIX II OR III?

YES NO

IS ANY SF3STANCE PRESENT ATOR ABOVE THE THRESHOLDQUANTITY IN APPENDIX II?

YES NO

t | | ~~~~~~~~NO ACTIO7N|

PREPARE MAJOR HAZARD ASSESSMENT INCLUDING:

1. IDENTIFICATIOI OF MAJOR ACCIDENT HAZARDS;2. PREVENTATIVE MEASURES EMPLOYED;3. EMERGENCY PROCEDURES TO BE ADOPTED; AND4. SAFETY MANAGEMENT.

(SEE SECTION 4.1)

REVIEW BY W.B.

- 12 -

5.0 Emergency Plans

34. These guidelines require developers to prepare an adequateemergency plan for dealing with major accidents that may occur ontheir sites.

5.1 On-Site Emergeucy Plan

35. It is not the intention of these guidance notes to explainin detail how to prepare an on-site emergency plan. The detail andscope of the emergency plan will vary according to the complexity ofthe site and it is, therefore, not appropriate to prescribe here pre-cisely what the plan should cover. The developer will need to con-sider the potential major accidents which are identified in the majorhazard assessment (Section 4) to ensure that the plan takes account ofthem. Useful guidance in preparing the emergency plan may be found inthe booklet 'Recommended Procedures for 8andling Major Emergencies'published by the UK Chemical Industries Association, and the US EPA"Community Preparedness for Chemical Hazards, Part 3: A Guide forContingency Planning' (1985).

36. The developer should ensure that the on-site emergencyplan is compatible with the off-site emergency plan which should bedrawn up by the local authority. The on-site and off-site plansshould be interlocked to ensure that they provide a comprehensive andeffective response to emergencies.

37. The plan should include the name of the person responsiblefor safety on the site (usually the site or plant manager) and, ifdifferent, the name of the person who is authorized to set the plan inaction.

38. The developer should keep the on-site emergency plan up-to-date, and to ensure that it takes account of any changes in opera-tions on the site that might have a significant effect on the plan.The developer is also required to make sure that people on the sitewho are affected by the plan are informed of its relevant provisions.This should include not only those people who may have duties underthe plan, but also those who may need to be evacuated from the site inan emergency, including contractors and visitors.

5.2 Off-Site Eaergency Plan

39. The intention is that emergency plans should be drawn upor amended by the local authority after consultation with bodies whomight be able to contribute information or advice. Such consultationis seen as an important aspect in the preparation of adequateemergency plans - this has been well demonstrated in the case of planswhich are in operation in many areas of the world. Obviously thedeveloper must be consulted about the major accident hazards and the

- 13 -

possible consequences, and any special emergency measures. Theresults of major hazard assessment discussed in Section 4 will provideuseful data for drawing up these emergency plans.

40. A two-way flow of information is required between thedeveloper and the local authority. Information from the developer isneeded to enable the authority to draw up the off-site emergency plan;information from the authority should be available to the developerwhen he prepares the on-site emergency plan.

6.0 Restrictions on Development in the Vicinity of Major HazardInstallations

41. The extent of the safety buffer zone or restrictivedevelopment zone that may be required for a major hazard installationshould be determined on a "case-by-case" basis. The importance ofmaintaining a restricted development, safety buffer zone is clearlyshown by the experience in Mexico City and Bhopal, as well as at otherhazardous installations around the world.

42. The results of the major hazard assessment may indicatecertain critical areas around the plant boundary where restrictions'should be imposed on further development, taking into account localfactors, as well as site storage, process, and management factorsetc. Planning authorities may restrict the land use in these safetybuffer zones to warehousing, light industry or agricultural use, butexclude residential, shanty towns, hospitals, schools and commercialdevelopment. Some developers are purchasing safety buffer zones andare planting a dense tree cover as a safety screen, as well as toprevent squatting and shanty town developments.

- 14 -

Bibliography

1. European.Community Directive, 1982. "On the MajorAccident Hazards of Certain Industrial Activities".82/501/EEC. Official Jcurnal of the European Community,.L230, June 1982.

2. "A Guide to the Control of Industrial Major AccidentHazards Regulations 1984." UK Health and SafeEy ExecutiveHS (R) 21, EMSO (1985).

3. "UK Notification of Installations Handling HazardousSubstances Regulations, 1982." SI No. '357 (and UK Healthand Safety Executive Guide, Booklet HS(R) 16).

4. 'A Guide to Razard and Operability Studies". ChemicalIndustries Association, London, 1977.

S. "Recommended Procedures for Handling Major Emergencies",Chemical Industries Association, Alembic House, 93 AlbertEmbankment, London SE1

6. "Codes of Practice for Chemicals With Major Hazards;Chlorine", Chemical Industries Association, 1975.

7. 'Community Preparedness for Chemical Hazards, Part 3; AGuide for Contingency Planning," US EPA (draft), August1985, Document 6291H.

8. 'Guidelines for a Reactive Chemicals Program," DowChemical Co., 1981.

9. "Manual of Industrial Hazard Assessment Techniques, iJorldBank, 1985.

10. 'Daniel Stull, 'Fundamentals of Fire and Explosion," AICHEMonograph Series No 10, 73 , 1977.

- 15 -

APPENDIX I

Defiuition of the Term Installation'

~ -1. Installations for the production or processing of organic or.inorganic chemicals using for this purpose, in particular:

- alkylation

- amination by ammonolysis

- carbonylation

- condensation

- dehydrogenation

- esterification

- halogenation and manufacture of halogens

- hydrogenation

- hydrolysis

- oxidation

- polymerization

- sulphonation

- desulphurization, manufacture and transformation ofsulphur-containing compounds

- nitration and manufacture of nitrogen-containing compounds

- manufacture of phosphorus-containing compounds

- formulation of pesticides and of pharmaceutical products

2. Installations for the processing or organic and inorganicchemical substances, using for this purpose, in particular:

- distillation

- extraction

- soivation

- 16 -

- mixing

- drying

3. Installations for distillation, refining or other processingof petroleum or petroleum products.

4. Installations for the total or partial disposal of solid orliquid substances by incineration or chemical decomposition.

5. Installations for the production or processing of energygases, for example, LPG, LNG, SNG.

6. Installations for the-dry distillation of. coal or lignite.

7. Installations for the production of metals or non--me:als bythe wet process or by means of electrical energy.

8. Storage of dangerous materials identified in Appendices II andIII.

9. Transportation Distribution Systems:l/

- Pipelines (quantities between block valves)- Shipping and terminal facilities (including in-land

waterways)- Road- Rail

1/ There may be an overlap between these guidelines and many nationaland international regulations, and guidelines concerning transferof hazardous substances. When a national or internationalregulation applies to a particular installation, the World Bankguidelines should be used only as a check to ensure that all safetyaspects have been identified and controlled.

- 17 -

APENDItX II

List of Hazardous Substances RequiringA hijor Hazard Assessment

(A) -Very Acutely Toxic- Substances

The following indicative criteria are used to identify anyvery toxic" substance requiring a major hazard assessment. Thesecriteria are independent of the quantities of the substance stored, orprocessed, or that are formed by an unwanted by-product reaction.

Very toxic substances are defined as:

- substances which correspond to the first line of the tablebelow;

- substances which correspond to the second line of thetable below and which, owing to their physical and chemi-cal properties, are capable of entailing major-accidenthazards similar to those caused by the substance mentionedin the first line:

LD 50 (oral) (1) | LD 50 (cutaneous) (2) LC 50 (3)mg/kg body weight mg/kg bodv weight mg/l (inhalalation)

1 LD 50 <5 LD 50 <10 LC 50 <0.1

2 5<LD 50<25 10(LD 50<50 0.laLC 50<0.5

Note: (1) LD 50 oral in rats.Note: (2) LD 50 cutaneous in rats or rabbits.Note: (3) LC 50 by inhalation (four hours) in rats.

If an LC 50 v lue is available for a shorter exposure time 't" theLC 50 (4 hr) may be estimated as follows:

LC 50 (4 hr) - LC 50 (t hr) x t4

(B) Other Acutely Toxic Substances

(1) The following quantities of toxic substances represent thethreshold above which compliance with Section 4.1 is required.

Named Substances Quantity Tonnes

Phosgene 2Chlorine 10Hydrogen fluoride 10Sulphur trioxide 15

- 18 -

Acrylonitrile 20Hydrogen cyanide 20Carbon disulphide 20Sulphur dioxide 20Bromine 40Ammonia (anhydrous or assolution containing more than50% by weight of ammonia) 60

(2) In addition to the above named substances, the following indica-tive criteria are used to identify other toxic substances which,owing to their physical and chemical properties, may cause a majoraccident and are stored or processed in quantities of greater thanI tonne:

LD 50 (oral) (1) LD 50 (cutaneous) (2) LC 50 (3)mg/kg body weight mg/kg body weight mg/l (inhalalation)

25 <LD 50 <200 50 CD 50 <400 0.5 LC 50<2

Note: (1) LD 50 oral in rats.Note: (2) LD 50 cutaneous in rats or rabbits.Note: (3) LC 50 by inhalation (four hours) in rats.

) Highly Reactive Substances

L) The following quantities of '-highly reactive" substances representthe threshold above which compliance with Section 4.1 is required.

Named Substances Quantity.Tonnes

Hydrogen 2Ethylene oxide 5Propylene oxide 5tert-Butyl peroxyacetate 5tert-Butyl peroxyisobutyrate 5tert-Butyl peroxymaleate 5tert-Butyl peroxy isopropyl carbonate 5Dibenzyl peroxydicarbonate 52,2-Bis(tert-butylperoxy) butane 51,1-Bis(ter-butylperoxy) cyclohexane 5Di-sec-butyl peroxydicarbonate 52,2-Dihydroperoxypropane 5Di-n-propyl peroxydicarbonate 5Methyl ethyl ketone peroxide 5Sodium chlorate 25Liquid oxygen 200

19-

General Groups of Substances Quantity Tonnes

Organic peroxides (not listed above)' 5Nitrocellulose compounds. 50Ammonium nitrates 500

(2) In addition to the above --med substances, the followingindicative criteria are u:..d to identify potential explosivehazards, irrespective of materials stored or processed.

- Substances which may explode under the effect of flame orwhich are more sensitive to shocks or friction thandinitrobenzene.

(D) Flammable Substances

The following quantities of "flammable" substances representthe threshold, above which and compliance with Section 4.1 isrequired.

Class of Flammable Substances Quantity Tonnes

1. Flammable Gases:

Gas or any mixture of gases which isflammable in air and is held as a gas. 15

2. Liquefied Gases and Flam"able Liquids inProcess at Pressure and/Temperature AboveAmbient Levels:

A substance or any mixture of substances 25 being the totalwhich is flammable in air and is normally quantity of substan-held in the installation above its boiling ces above the boil-point (measured at 1 bar absolute) as a ing points whetherliquid or as a mixture of liquid and gas at held singly or ina pressure of more than 1.4 bar absolute. mixtures.(e.g. LPG's).

3. Befrigerated Liquefied Gas

A liquefied gas or any mixture of liquefied 50 being the totalgases, which is flammable in air, has a quantity of sub-boiling point of less than 0°C (measured stances havingat 1 bar absolute) and is normally held boiling points belowin the installation under refrigeration OC whether heldor cooling at a pressure of 1.4 bar singly or inabsolute or less (e.g., LNG). mixtures.

- 20 -

4. Highly Flamable Liquids

A liquid or any mixture of liquids not 10,000included in items 1 to 3 above, whichhas.a flash point of less than 21°C.

5. Flammable Liquids at High Temperature and Pressure

Substances which have a flash point lower than 55°C and whichremain liquid under pressure, where particular processingconditions, such as high pressure and temperature, may lead to amajor accident hazard.

- 21 -

APPENDIX III

List of Acutely Toxic and Reactive Hazardous Substances

Group A:

Very Toxic Substances(See definition Appendix II, Page 17)

Aldicarb

4-Amiaodiphenyl

Amiton

Anabasine

Arsenic pentoxide, Arsenic (V) acid andsalts

Arsenic trioxide, Arsenious (III) acidand salts

Arsine (.Arsenic hybride)

Azinphos-ethyl

Azinphos-methyl

Benzidine

Benzidine salts

Beryllium (powders, compounds)

Bis (2-chloroethyl) sulphide

Bis (chloromethyl) ether

Carbofuran

Carbophenothion

Chlorfenvinphos

- 22 -

4-(Chloroformyl) morpholine

Chloromethyl methyl ether

Cobalt (powders, compounds)

Crimidine

Cyanthoate

Cyccloheximide

Demeton

Dialifos

O0-Diethyl S-ethylsulphinylmethylphosphorotioate

00-Diethyl S-ethylthiomethylphosphorothioate

00-Diethyl S-isopropylthiomethylphosphorodithioate

00-Diethyl S-propylthiomethylphosphorodithioate

Dimefox

Dimethylcarbamoyl chloride

Dimethylnitrosamine

Dimethyl phosphoramidocyanidic acid

Diphacinone

Disulfoton

EPN

Ethion

FensulfothionFluenetil

Fluoroacetic acid

Fluoroacetic acid, salts

- 23 -

Fluoroacetic acid, esters

Fluoroacetic acid, amides

4-Fluorobutyric acid

4-Fluorobutyric acid, salts

4-Fluorobutyric acid, esters

4-Fluorobutyric acid, amides

4-Fluorocrotonic acid

4-Fluorocrotonic acid, salts

4-Fluorocrotonic acid, esters

4-Fluorocrotonic acid, amides

4-Fluoro-2-hydroxybutyric acid

4-Fluoro-2-hydroxybutyric acid, salts

4-Fluoro-2-hydroxybutyric acid, esters

4-Fluoro-2-hydroxybutyric acid, amids

Glycolonitrile (Hydroxyacetonitrile)

1, 2, 3, 7, 8, 9-Hexachlorodibenzo-p-dioxin

Hexamethylphosphoramide

Hydrogen selenide

Isobenzan

Isodrin

Juglone.(5-Hydroxynaphthaleue-1,4-dione)

4,4' - Methylenebis (2-chloroaniline)

Methyl isocyanate

Hevinphos

2-Naphthylamine

- 24 -

Nickel (powders, compounds)

Nickel tetracarbonyl

Oxydisulfoton

Oxygen difluoride

Paraoxon (Diethyl 4-nitrophenyl phosphate)

Parachion

Parathion-methyl

Pentaborane

Phorate

Phosacetim

Phosphamidon

Phosphine (Hydrogen phosphide)

Promurit (1-(3,4-Dichlorophenyl)-3-triazenethiocarboxamide

1,3-Propanesultone

I-Propen-2-chioro-1,3-diol diacetate

Pyrazoxon

Selenium hexafluoride

Sodium selenite

St±bine (Antimony hydride)

Sulfotep

Sulphur d'±chloride

Tellurium hexafluoride

TEPP

2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD)

Tetramethylenedisulphotetramine

-25-

Thionazin

Tirpate (2,4-Dimethyl-2,3-dithiolane-2-carboxaldehyde.O-methylcarbamoyloxime)

Trichloromethanesulphenyl chloride

1-Tri(cyclohexyl)stannyl-lH-L,2,4-triazole

Triethylenemelamine

Warfarin

Group B:

Other Toxic Substances (See definition Appendix 11, Page 17)

Acetone cyanohydrin (2-Cyanopropan-2-ol)

Acrolein (2-Propenal)

Acrylonitrile

Allyl alcohol (2-Propen-l-ol)

Allylamine

Ammonia

Bromine

Carbon disulphide

Chlorine

Ethylene dibromide (1,2-Dibromoethane)

Ethyleneimine

Formaldehyde (concentration > 902)

Hydrogen chloride (liquefied gas)

Hydrogen cyanide

Hydrogen fluoride

Hydrogen sulphide

- 26 -

Methyl bromide (Bromomethane)

Nitrogen oxides

Phosgene (Carbonyl chloride)

Propyleneimine

Sulphur dioxide

Tetraethyl lead

Tetramethyl lead

Croup C.1:

Bighly Reactive Substances and Explosives

Acetylene

Ammonium nitrate (where it is in a state whichgives it properties capable of creating amajor accident hazard)

2,2-Bis(tert-butylperoxy) butaneconcentration > 70%)

1,1-Bis (tert-butylperoxy) cyclohexane(concentration > 80%)

tert-Butyl peroxyacetate(concentration) > 70%)

tert-Butyl peroxyisobutyrate(concentration > 80%)

tert-Butyl peroxy isopropyl carbonate(concentration > 80%)

tert-Butyl peroxymaleate (concentration > 80%)

tert-Butyl peroxyphivalate (concentration > 77%)

Dibenzyl peroxydicarbonate (concentration > 902)

Di-sec-butyl peroxydicarbonateconcentration > 80%)

Diethyl peroxydicarbonate (concentration > 30%)

-.27 -

2,2-Dihydroperoxypropane (concentration > 30%)

Di-isobutyryl peroxide (concentration > 50%)

Di-n-propyl peroxydicarbonate (concentration > 80%)

Ethylene oxide

Ethyl nitrate

3,3,6,6,9,9-Rexamethyl-1,2,4,5-tetroxacyclononane concentration > 75%)

Hydrogen

Methyl ethyl ketone peroxide (concentration > 60%)

Methyl isobutyl ketone peroxide (concentration > 60%)

Peracetic acid (concentration > 60%)

Propylene oxide

Sodium chlorate

Group C.2:

Explosive Substances

Barium azide

Bis (2,4,6-trinitrophenyl) amine

Chlorotrinitrobenzene

Cellulose nitrate (containing > 12.6% nitrogen)

Cyclotetramethylenetetranitramine

Cyclotrimethylenetrinitramine

Diazc adinitrophenol

Diethylene glycol dinitrate

Dinitrophenol, salts

Ethylene glycol dinitrate

1-Guanyl--4-nitrosaminoguanyl-l-tetrazene

- 28 -

2,2',4,4'.,6,6'-Hexanitrestilbene

Hydrazine nitrate

Lead azide

Lead styphnate (Lead 2,4,6-trinitroresorcinoxide)

Mercury fulminate

N-Methyl-N,2,4,6-tetranitroaniline

Nitroglycerine

Pentaerythritol tetranitrate

Picric acid (2,4, 6-Trinitrophenol)

Sodium picramate

Styphnic acid (2,46,6,-Trinitr2resorcinol)

1,3,5-Triamino-2,4,6-Trinitrobenzene

Trinitroaniline

2,4,6-Trinitroanisole

Trinitrobenzene

Trinitrobenzoic acid

Trinitrocresol

2,4,6-Triaitrophenetole

2,4,6-Trinitrotoluene

- 29 -

AUPENDIX IV

Background Information for aMajor Hazard Assessrment (Section 4.1)

Section 4.1.3 specifies the details required to be includedin a major hazard assessment prepared in accordance with theseguidelines.

The details required in Section 4.1.3 are discussed below asseparate items but it is for the developer to determine the mostappropriate or convenient method of presenting the requiredinformation. In particular, general issues (e.g., pressure vesselinspection arrangements) could be referenced and any variations ordepartures from the generally accepted practice will suffice.

The item-by-item guidance given below in relation to Section4.1.3 is by example rather than by lists of topics to be covered. Thelatter approach suffers the twin disadvantages of not being trulycomprehensive while at the same time giving the impression that eachlisted topic is of equal importance. In practice, the depth ofinformation required on each topic will vary according to the-circumstances of the individual installation.

Sections 4.1.3.1 and 4.1.3.2 require factual informationabout the dangerous substances and the installations handling them.Section 4.1.3.3 relates to the-management control of the activity.Section 4.1.3.4 requires information about the sources and nature ofpotential major accidents and the measures taken to prevent andcontrol them.

Section 4.1.3.1 (Substance Name)

The information required under this sub-heading is concernedwith identifying the dangerous substance which qualifies-the activityor storage for the requirement to make a major hazard assessment thewording should be self-explanatory.

Section 4.2.3.1 (Monitoring Methods)

Where standard analytical methods are used by the firm thisitem need only identify the method and any departures from it.Provision of gas detection equipment could be referred to here or leftif appropriate to the discussion of preventive measures under Section4.1.3.4.

Section 4.1.3.1 (Hazards of the Substances)

Information given under this sub-item should cover (a) theroute of the harm to man (skin contact, inhalation or ingestion fortoxic substances, and flame contact, thermal radiation or blast for

30-

flammable.or explosive substances); (b) the dose-responserelationship, where known, citing standard published references asappropriate (e.g., the Chemical Industries Association's table of thetoxic effects of chlorine at various concentrations); and (c) thenature of the trauma, where this is not obvious (e.g., chloracne fromexposure to dioxin).

In relation to the hazards from the substance to theenvironment, information is not required about the obvious effects offlame or blast from explosive or flammable substances. For toxicsubstances such information as. is readily available should be given orbe referred to, including the route of the harm (e.g., the pollutionof water courses); the effect on flora and fauna which may be exposed;and an indication of the substance's persistence.

Section 4.1.3.1 (Coaposition of Process Streams)

This item requires information about the composition of thesubstance so that the effects of diluents or impurities which have asignificant effect on the hazard can be assessed. For example, wherean organic peroxide is present in an activity, the name and percentageof any stabilizer should be given.

The presence of significant quantities of impurities couldalso affect the behaviour of a substance and these impurities andtheir percentages should also be identified. An example would be thealteration of the toxic properties of chloromethyl methyl ether by thepresence of bis (chloromethyl) ether. It is not necessary to list theminor components of mixtures being processed where these have aninsignificant effect on the potential hazard (e.g., a hydrocarbonmixture might be described as 70% butane, 25% propane, 5% higherhydrocarbons).

Section 4.1.3.2(i) (Location)

This map is required to indicate where the installation islocated, showing its position in relation to local geographicfeatures, such as roads and towns. In general it will be sufficientto use the latest available map or maps on a scale which includes boththe site and the surrounding features. Changes (such as a newmotorway) which have occured since the printing of the map and whichare known to the firm should be shown; it is not intended that thisshould involve any extended research effort. For many sites a staleof 1 to 10,000 will be appropriate.

Section 4.1.3.2(ii) (Plot Plan)

The intention of this item is to identify clearly, both inlocation and quantity, the main parts which contribute to the totalinventory of the dangerous substance. In addition, the plan should beannotated to indicate the lesser quantities which make up the stated

- 31 -

total. For example, an estimate of the quantity of the substancewhich is Dresent in pipework around a particular plant should be made.

It may be convenient to combine the response to this itemwith that to Section 4.1.3.4, and in particular, Section 4.1.3.4 item(ii) .

Section 4.1.3.2(iii) (Process/Flow Description)

The intention of this requirement is that a sufficientdescription of the process be given to enable discussion on lateritems to be understood ard placed in context. The amount ofdescription required wili depend on the complexity of the process.For example, a water-treatment plant using chlorine will require onlya brief account of the water dosing process, whereas a chemical plantproducing qualifying quantities of a very toxic substance as anintermediate will require sufficient information to enable thecritical aspects of the process chemistry to be understood. In thislatter example, the information will be supplemented by theinformation given under Section 4.1.3.4 and it will be for thedeveloper to determine the most appropriate method of presentation.

The conditions under which the substance is normally heldshould be stated, including the physical state and pressure and/ortemperature at the main stages of storage and process, e.g., butane isheld in storage as a refrigerated liquid at OC; vaporized in adirect-fired evaporator and fed as a gas to a process vessel at 10bar, 90°C.

Section 4.1.3.2(iv) (Personnel on site)

The number given should be accompanied by sufficientexplanation to show how it was derived. Account should be taken ofthe number of people who may be present at shift changeover; peoplewho may be employed from the site but who may be present only forshort periods (e.g., sales staff or delivery drivers); regularvisitors to the site (e.g., contractors); and casual visitors. Exactnumbers are not required.

Sectiou 4.1.3.2(v) (Local Land Use and Population Distribution)

The information required is about the use of land or watersurrounding the activity and the location of people who may beaffected in the event of a major accident. This can be provided byannotations on a suitable map, indicating broad categories of land use(e.g., dwellings, other factories, schools, sports facilities,agricultural land, etc.). It is not necessary to give numericalestimates of population in the areas covered by each category, thoughany unusually high densities such as shanty towns, blocks of flatsshould be marked as such.

- 32 -

The choice of the phrase 'in the vicinity of' rather thandefinite distances is intended to allow for flexible interpretation inrelation to the potential hazard. For example, for a flammable liquidtank farm the 'vicinity' might-be less extensive than for bulkchlorine storage. A degree of judgment is thus called for and a briefexplanation should be given of the choice of vicinity illustrated onthe map.

Section 4.1.3.3 (Management)

The aim of this item is to demonstrate that the developerhas a proper management svstem and technical staff to concrol themajor hazard aspects of his activities. The required information isgeneral, insofar as it relates to the management control of the siteas a whole, in contrast to the greater detail that may be needed underSection 4.1.3.4 to demonstrate that there are adequate arrangements toprevent and control particular hazardous events. The extent of theresponse to this item should be seen against its importance inproviding a framework in which the rest of the safety case may be setand which will to some degree color the credibility of the wholesubmission.

Section 4.1.3.3(i) (Responslible Person and Staffing)

A description of the management structure should be givenwhich covers reporting relationship and the exoerience andqualifications of staff at the different levels. It is important toshow how accountability for decisions which affect the potentiallyhazardous activit7 is assigned to staff who have the appropriate levelof expertise and the relevent professional discipline. Referenceshould also be made to the developer's policy towards the appointmencof competent deputies to cover key positions.

This section should also cover the arrangements whichmanagement have set up for identifying and dealing with safety issuesarising from the potentially hazardous activity, with references asappropriate to the group, division and site safety policies, and therole of safety representatives and safety committees. This sectionshould include an account of how management decisions about thepotentially hazardous activity are made with due regard for theirsafety implications, and how these decisions are monitored.

Section 4.1.3.3 (ii) (Quality Control for Safety)

The importance of the containment of dangerous substances issuch as to require an adequate account of Lhe management of theengineering system. This should include a discussion of thedeveloper's approach to the design of important plant items and safetysystems (e.g., use of standard or company codes); the arrangements forquality assurance; the inspection and testing procedures (who carries

- 33 -

them out? how frequently? who decides on-the necessary action in theevent of deficiencies being.discovered? who monitors these actionsand how?); the provision of operating manuals and instructions and theprocedures for revising them In the event of process changes; theprocedures for ensuring that plant modifications are adequatelydesigned; installed and tested; and che system for generalmaintenance.

The response to this item should not contain lengthyabstracts from written company procedures, but should aim to give acoherent justification for the system of engineering controlsexercised over the potentially hazardous activity.

Section 4.1.3.3(iii) (Training)

Information should be given about the standards of relevanttraining, both on- and off-the-job, for people on site with asignificant role in the control or operation of the major hazardactivity, including the senior management and engineering staffinvolved. A brief account should also be given about how trainingneeds are identified and met.

Section 4.1.3.4 (Potential Major Accidents)

The response to this item will provide the heart of themajor hazard assessment. Though the information required is specifiedin some detail, it will be essential for developers to interpret thisitem with commonsense and flexibility. As noted in the generaldiscussion of the term, 'major hazard assessment", the developer mustinvestigate those potential major accidents which may produce adverseconsequences outside the plant boundary using hazard analysistechniques such as described in the World Bank manual (ref (9)).Sufficient evidence must be provided to support his concluuions.

Section 4.1.3.4(i) (Identification of Potential Major Hazard Events)

This item requires the manufacturer to identify the ways inwhich a major accident might occur. "Major accident' is defined inSection 2 and guidance on its interoretation is given in thissection. Suitable techniques for identification are hazard analysis,hazard and operability studies, reactive chemicals studies,engineering flow chart review, and review of past accidents and nearmisses etc. The concept of the major hazard assessment allows thedeveloper the opportunity to argue in the response to this item thathis activities are not capable of giving rise to a major accident,provided that they are satisfied that such an argument can be fullyand properly justified. In some cases, a submission of this sort maybe relatively easy to support. For example, a toxic hazardoussubstance may be stored in separate cylinders, and the quantity ineach one may not be sufficient to give rise to a major accident. If

- 34 -

the cylinders were disposed around the site so that the possibilityof an incident affecting them simultaneously could be ruled out, thenthe argument that a major accident cannot arise on the site may bereasonably'straightforward. In other cases the evidence necessary tojustify such an argument may prove to be as extensive as a response tothis item which accepts that major accidents may occur, and then goeson to describe how they are prevented; it will be for the manufacturerto decide initially which option better fulfills the objectives of the.major hazard assessment. If it can be shown that the activity is notcapable of giving rise to a major accident, then the remaining partsof Section 4.1.3.4 need not be answered. It is in this section thatthe quantification of the magnitude of hazardous events by means ofhazard analysis techniques are required.

In most cases it will be necessary to describe the potentialsources of a major accident (the World Bank -Manual of IndustrialHazard analysis Techniques" provides guidance in this respect).Storage and process vessels which contain significant quantities ofthe dangerous substance should be examined for the most probable waysin which their inventories may be released, and these should includeconsideration of, spontaneous failure (due to original defects orthose arising in the course of operation); failure due to excursionsfrom normal operating conditions (including such matters as operatorerror, loos of services, and failure of control devices); failure dueto events elsewhere on site (e.g., fire, explosion); and failure dueto external events (e.g., flooding, seismic activity). For a complexchemical or petrochemical works or a refinery, where the number ofvessels and pipes for individual plants may be very large (but builtto common standards), it may be appropriate to be more selective inexamining the potential sources of a major accident by consideringonly the largest vessels in detail and then referring to smallervessels or groups of vessels in general terms.

Sectiou 4.1.3.4(il) (Process Flow Diagrams)

This item calls for judgment by the developer as to whichsections of the plants containing the dangerous substance need to beillustrated ia diagrammatic form. Such diagrams should show theprocess vessels, storage facilities and instrumentation on the plantsconcenred in sufficient detail to enable the discussion under Section4.1.3.4 (i) and (iii) to be readily understood. For vessels in theplants which have been identified in Section 4.1.3.4(i) as significantas regards the potential for a major accident, details should be shownon the diagram of their designed maximum working capacities, theirdesign temperatures and pressures, and their normal cperatingconditions. It may be convenient to combine this with the response toSection 4.1.3.2(ii). For example, a large vertical cylindricalstorage vessel containing the dangerous substance would require adiagram showing the vessel, its connections, instrumentation andexternal safety features such as water sprays and bunds; for complex

- 35 -

plants diagrams of only the largest or most significant vessels arerequired.

Section 4.1.3.4(iii) (Preventative and Control Systems)

The third part of Section 4.1.3.4 requires a response in twoparts. First, relating to the preventive and control measures whichcheck any sequence of events which could foreseeably result in a majoraccident, and secondly, relating to measures which may be taken aftersuch a hazardous outcome to minimize its adverse consequences. It isimportant to realize that it is better to prevent a release occurringrather than trying to control the consequences, which may, in part, beimpracticable.

The first part of the response should concentrate on thosepreventive or control measures which are critical in counter-actingsignificant hazards, though many of these measures will also' beappropriate to preventing or controlling lesser events.

The interpretation of certain key words in this part ofSection 4.1.3.4 may usefully be discussed in relation to examples;say, a large pressure storage sphere of LPG, representing the risk offire and explosion, and for an example of toxic risk, a large pressurestorage sphere of ammonia.

'Measures taken to prevent" relate to ensuring the safeoperation of the plant under nor-ijl operating conditions or withinspecified process limits. They are those measures which are intendedto prevent the initiation of a sequence of events which could lead toa significant hazardous outcome and would include consideration ofdesign, engineering standards, constructional and quality assurance,inspection and maintenance, and control systems insofar as these wereconcerned wi'th controlling the process during normal operation. Thesepreventive measures derive their validity from the way in which theoverall management of the plant and company exercises control overthen and ensures their effectiveness. Thus, a sufficient account ofthe relevant parts of the management system, expanding on the generalinformation given in Section 4.1.3.3 will be needed to support anyarguments about the probability of initiating a sequence of eventswith a hazardous outcome. In relation to the ammonia sphere example,"measures taken to prevent", would include a discussion of thearrangements for checking ammonia purity and special vesselinspections in connection with the problem of stress corrosioncracking.

'Measures taken to control' relate to the interventionspermitted by the design of the plant (e.g., valves) or safety hardware(e.g., dump tanks) which may counter an event so that the dangeroussubstance is retained within the plant. The operation of a 'measuretaken to control" assumes that a sequence of events has been initiated

- 36 -

and the control is intended to prevent the sequence proceeding to amajor accident. In relation to the LPG sphere example, a spillage ofLPG near the tank which ignited, could be prevented from escalating toa BLEVE (boiling liquid expanding vapour explosion) of the wholesphere by the effective operation of a water spray system on thesphere. The discussion of this control measure should includereference to the the capacity of the water storage system, the waterapplication rate; how the company ensures that the system will in factoperate effectively when required to do so; and the systemreliability.

The second part of the response to this item covers..measures to minimize the consequence", which relate to those measureswhich can be taken after the major accident has actually occurred.Examples of minimization include: bunding, water curtains, foamblankets, and emergency procedures. In relation to the ammoniasphere, the effects of a release of ammonia from a pipe at asufficient rate to lead to a lethal concentr&tion of gas at (say) thenearest domestic dwellings might be minimized by the application of awater curtain around the sphere, and, through the emergency services,by evacuation of those people who are downwind of the escape.

Section 4.1.3.4 (iii) refers to the prevention or control ofmajor accidents identified in Section 4.1.3.4(i). It is therefore notnecessary to include information about measures directed solely atpreventing small releases or those which have only minor consequences,unless these have the potential to escalate to a major accident.

Section 4.1.3.4(iii) by concentrating on preventivemeasures, is intended to draw out discussion on the positive aspectsof the major hazard assessment. However, because the sequence ofcause, effect and consequence are closely intertwined, it is notpossible in practice to confine such a discussion to preventionwithout mentioning the potential consequences of the potential majoraccidents discussed in the major hazard assessment. Estimates ofconsequences will in any case be required in order to formulateadequate advice to the authority responsible for drawing up off-siteemergency plans (see Section 5).

Sectiou 4.1.3.4 (iv) (Emergency Procedures)

Section 5 requires the developer to produce a plan fordealing with emergencies on-site. The operator/developer should beable to present the whole document for examination if necessary, butthat is not the intention of this part of the major hazardassessment. On the other hand, it will not be sufficient merely tostate that the emergency plan exists. The response to this itemshould describe the procedures in outline; indicate the nature andextent of emergencies with which the plans are intended to cope,drawing as necessary on the information in other parts of the major

- 37 -

hazard assessment; mention those arrangements which may be critical tothe success of the plans such-as access for emergency services, theprovision of adequate supplies of fire-fighting water, the remotesiting of emergency control points and the evacuation of non-essentialsite personnel; and confirm that the plans have been discussed withthe relevant outside bodies and practiced with them.

Section 4.1.3.4 (v) (Meteorological Conditions)

Data should be obtained from the nearest MeteorologicalOffice weacher station as to the prevailing weacher conditions in thevicinity of the site, and if necessarly confirmed by actualmeasurements on the site.

Section 4.1.3.4 (vi) (Numbers at Risk)

This estimate should include those people normally workingon the plant concerned and any major concentrations of people in theimmediate vicinity of the plant, e.g., office buildings. As forSection 4.1.3.2 (iv), exact numbers are not necessary.

-38 -

MEN= V

S5 of ae sRv

Process unit Process Step Nature of Hazard Maans of Control

I Thin film evaporator. Rzunaay polymerization of - Use of thin flIm evaporation.Cherical A at >1700 C in Ue. Stem at 150°C, &ster dcusingpresene of traces of system, relief to blowdoncatalyst. drun Temp. Alarms and inter-

locks.

_ qll quantities only inprocess.

2 ?eaction step- Highly cwthermic reaction - Continxuos reaction systenitration of with catalyst. with swll quantities ofChemical B react4ons in large qusnity

of liquid catalyst recycle.3 Reaction step- Highly exothermic reaction Reactants metered by dosing

nitration of with H2 S04 as catalyst. pump systen with special safe-Chemical C guards. Reaction takes place

in pmp after static mlxr.Temp. alan and interlocksto prevent rmaway reactions.

- mll quantities of reactants

4 Reboilers of Ru&my polymerization _ Small qunatity of mteriaLdistillation colum=s reaction. with tepenrature control.in pzification stages. Sfe discharge of reboLler

contests to an enclosed areaamay -frm other process eauip-m.nt and personel via arupcure disc relief systemaxn vnt argled at 45@ tohorizontal <10kg products.

5 Ration step - Hgh1y emtkDermic in - Over designed ecternaal coaydrolysis of Qaieal 10 m reactors. ScAla cooJir systee for beat ofD at 160°C an 4 atms. up fram 100 Utre pilot reactio of 41 kcal/iol cf.

plant and opertirg actual value of 14 kcal/mul.All reactants are perience on 3 m3

idd at start of reactor with si-ila - .nual blow-dam systen onbatch and baated to reactants. reactor.operatirg tenperatureof 150LC. - Semi-autatic blcwdw systen

from contoml roam.

- Co,uter control to inimiehumn error.

- Relief valve system to bLda.

- 39 -

Process Unit Process Step Nture of Hazard Means of Control

6 TWD step reaction If inufficient active - Extentet batch reacrtionproc-ss. First step catalyst (i.e., due to cycle fron 1/2 hour to 3involving catalytic poisoxlg or inadequate hoursreduction of Chemical addition), and coo low aE to Chemical F is the reaction tesperature, a - keep reaction teaperature >hazardous step. budld up of an intermediate 80°C

product occurs. Thisintrmediate can cause - control flow of H2a rutaway reaction if thetemperature is raised too - shut off flow of H2 if temp.quikly. <80C

- Note this systen has been - use an ecess of catalystscaled up fron 100 litrepilot plant. Sane size - remve 10 of catalyst atreactors in use on similar each batch and replace withraw mterials. fresh catalyst

- 3-4 m3 reactors.

Catalytic reduction Sim-lar reaction hazard as - Fully casputer controlled.of ankical G in a Tbit 6, if the intermediatecontimnous reactor at foms due to insufficient - Hydrogen addition regulated60 atmosphere's active catalyst and too accordirg to heat balancepressure. low temperature. This is calalations.

the first tIm this reactorsyten has been used to - Auto shut-off of H2 if heatproduce ahmical H. Ras halance shows generation ofbei used for simil reac- intermediate.tiors and an extensivecoiputer controllsafety, - Initially there will be nosystem has been developed. recycle of catalyst but may

recycle as 3ie =eperience.

- Peactor blow-down systeminstalled.

- Safety interlocks.

- Significant volume of

diemicals present.

APPENDIX 3

Summary of Potential Sources of Ignition

LIST OF IGNITION SOURCES

Industrial plants contain a great number of possible ignitionsources. It is common practice to reduce as far as practical pos-sible ignition sources in areas where there is a risk of flammablereleases. This is achieved by enfor^ing regulations that;

- restrict activity in the area (using a "permit to work system")

- only allow suitable equipment in the area

- restrict access to the area (by fencing off the area)

Lees has listed likely ignition sources. The list is repeated herewith suggestions on how they are controlled within the restrictedarea.

1. Burner flames.Layout of plant, trip systems.

2. Burning operations.Perm't to work.

3. Hot soot.

4. Cigarettes.No-smoking in restricted area.

5. Smouldering material.

6. Hot process equipment.Design.

7. Distress machinery.Design and maintenance.

8. Small process fires.

9. Weldings and cutting.Permit to work.

10. Mechanical sparks.Special non-sparking tools are available.

11. Vehicles.Ordinary vehicles excluded from areas of flammable hazard.

12. Arson.Security procedures implemented in hazardous areas.

13. Self heating.This occurs through slow oxidation of a solid. A typical sourceis oily rags. "Good Housekeeping" should prevent this.

14. Static electricityThis complex problem cannot bG reasonably treated here. Thereader is referred to specialist tests such as "Electrostaticsin the Petroleum Industry" by Klinkenberg and van der Minne.

15. Electrical equipment.Electrical equipment to be posittoned in these areas have to bebuilt to special standards. rn the UK and in many othercountries the areas at risk are classified according to theseverity of the risk with a different electrical equipmentstandard for each classification.

APPENDIX 4

Selected Data on the Properties ofSome Hazardous Materials

-ISOLATED STQRAGE

The list of storage quantities given below has also been prepared bythe EEC (1982), and applies only where the storage is effectively'isolated from all other operations. It is reproduced here toprovide an indication of the levels of concern.

The quantities set out below relate to each instaliation or group ofinstallations belonging to the same manufacturer where the distancebetween installations is not sufficient to avoid, in foreseeablecircumst-ances, any aggravation of major accident hazards. Thesequantities apply in any case to each group of installations belong-ing to the same manufacturer where the distance t1tween the instal-lations is less than 500 metres.

I Substances of groups - Quantities (tonnes)

S Substances or groups I For application of 1 For appli^ation ofof substances Regulation 4 1 Regulations 7 to 12 0

(Column 1) 1 (Column 2) 0 (Column 3)------------------------------------------------------------------ ,

Acrylonitrile 350 5,000* Ammonia 60 0 600Ammonium nitrate 500* 5,000'Chlorine 10 200Flamamble gases

I I '

I i - 50 1 300Highly flammable

I liquids ll l 10,000 ° 100 000Liquid oxygen 200 2,0001Sodium chlorate 25 250'Sulphur dioxide 20 1 500

---- ---- --- ---- --- ---- ---- --- ---- --- ---- ---- --- ---- ---I ,

'Where this substance is in a state which gives it propertiescapable of creating a major accident hazard.

MANUFACTURING PROCESSES

The quantities which are liable to cause hazards if accidentallyreleased from a processing plant, are usually significantly smallerthan in the table above, and many more substances are involved.

.ILimits of flammabi-lityCompound o------------- -- --

!Lower (%v/v) 'Upper (%v/v)

Acetone i 2.6 13.0Acetylene * 2.5 100.0Ammonia 15.0 i 28.0Amylene 1.8 8.7Benzene * 1.4 8.0

'~~~~~~~~~~~~~~~~~~~~~~~ I

n-Butane 1.8 d.4i-Butane 1.8 8.4Butene-1 i 2.0 10.0Butene-2 1.7 9.7Carbon disulphide 1.3 50.0

Cyclohexane * 1.3 * 7.8Decane 0 0.8 i !j.4

' .thane * 3.0 * 12.41 Ethylene 2.7 : ̂6.0I Ethylene dichloride i 6.2 i 15.9I-------------------------------------------------------

I Ethylene oxide 3.0 100.0Heptane i 1.2 6.7Hexane 1.4 7.4' Hydrogen 14.0 75.0Methane i 5.0 15.0

n-Pentane 1.8 7.8Propane * 2.1 9.5Propylene * 2.4 1 11.0Styrenene 1.1 6.1Toluene 1.3 i 7.0

--------------------------- …--------------------------

I Vinyl chloride * 4.0 22.02,2-Dimethylpropane * 1.3 * 7.52,3-Dimethylpentane * 1.1 6.8

…-------------------------.-----------------------------

TABLE OF CONSTANTS FOR VARIOUS TOXIC GASES

Constants in equations Pr 'At' Bt ln (Cnte), where c is in ppmand t is in minutes.

I At Bt n ! Referenoe,________________________,-____________________________________-__,

Chlorine 1 -5.3 0.5 2.75 DCMR 1984

Ammonia 1 -9.82 0.71 2 DCMR 1984

Acrolein -9.93 2.05 1 1 USCG 1977

Carbon Tetrachloride 0.54 1.01 0.5 USCG 1977

Hydrogen Chloride -21.76 2.65 1 USCG 1977 1J1

Methyl Bromide ' -19.92 5.16 1 USCO 1977

Phosgene 1 -19.27 3.69 1 USCG 1977

Hydrogen Xluoride -26.4 3.35 1 USCG 1978 1(monomer)

FLAMMABLE PROPERTIES

The models presented in Section 4 require upper flammability limits(UFLs) and lower flammability limits (LFLs). Below are some datafor common hazardous materials presented as volume per cent in air.

If data are not available for a particular material it is possibleto estimate a flammability limit by taking data for a similarmaterial and applying

(LFL)A = (LFL)BMA

MA and MB are the relevant molecular weights.

TOXIC PROPERTIES

During the last 10 years, probit equations have been derived forestimating, from dose re'ationships, the probability of affecting acertai.i proportion of the exposed population. These have been basedalmost exclusively on animal test data, which are' often very impre-cise anyway, and can sequentially they are not very reliable. Inparticular they do not possess the accura-y which could be ascribedto the formula as used in calcu'lations of consequential effect, andthey only apply to acute effects of accidental exposures. :n recentyears, some of the original equaticns have been subJect to muchscrutiny and criti2ism, and this has resulted in alternative ecua-tions being proposed.

In the folowi4ng table are listed the latest, or only, equationsavailable for a variety of toxi4c gases, but it is important that theorigin of each equation is correctly referenced so that the effectof using alternative equations can be quickly established. Thedifference in results can be exceedingly large in some instances.

Probit Equations take the form

Pr A+ t ln (Cnte)

where Pr is the orobability function (expressed in units of standarddeviation, but with an offset of +5 to avoid the use of negativevalues)

At, Bt, and n are constants, and

C is concentration of pollutant to which exposure is made, and£s ppm V/v

te is the duration of exposure to the pollutant, measured inminutes.

Note: if the unit of C and/or te are changed, e.g. to mgm/litrev/V, then the values of the constants will change. All units shouldtherefore be quoted.

b) Other toxic substances

The substances showing the following values of acute toxicity havephysical and chemical properties capable of producing majoraccident hazards.

-------------------------------------------------------------

LD50 (oral)(1) 1 LD50 (cutaneous)(2) LC50(3)mg/kg body weight ' mg/kg body weight mg/l inhalation!

----------------- j-------- - - - - ------

25 < LD50 < 200 50 < LD50.< L00 0< LC20 < 2 1

(1) LD50 oral in rats.(2) LD50 cutaneous in rats or rabbits(3) LC50 by inhalation (four hours) in rats.

c) Flammable substances:

i) flammable gases:substances which in the gaseous state at normal pressure ismixed with air become flammable and the boiling point ofwhich at normal pressure is 200C or below;

ii) highly flammable liquids:substances which have a flash point lower than 2 0°C, andthe boiling point of which normal pressure is above 20GC;

iii) flammable liquids:substances which have a flash point lower than 550C andremain liquid under pressure, where particular processingconditions, such as high pressure and high temperature, maycreate major accident hazards.

d) Explosive substances

Substances which may explode under the effect of flame or which ismore sensitive to shocks or f^iction than dinitrobenzene.

INDICATIVE CRITERIA

The European Economic Community (EEC 1982) have drawn up a list ofcriteria which indicate the potential for harm from a wide varietyof hazardous material. These cover toxic and flammable substances,and according to defined criteria indicate the levels for very toxicsubstances, for toxic substances, and for flammable and explosivesubstances.

These criteria are a useful way of examining the potential of aparticular substance to cause harm if released into the environmentin an accident. LD50 represents the concentration required of atoxic material if infested to cause a 50% chance of fatality, andsimilarly LC50 represents the gaseous concentration inhaled (forfour hours) of a substance which may cause a 5301 chance of fatality.Of necessity the data used for these determinations is anilial data,but they are assumed to be similar to the levels likely .,an.

a) Very toxic substances

- substances which correspond to the first line of the tablebelow.

- substances which correspond to the second line of the tablebelow and which, owing to their physical and chemicalproperties, are capable of producing major accident hazardssimilar to those caused by the substance mentioned in the firstline:

-----------------------------------------------------------------, LD50 (oral)( 1 ) , LD50 (cutaneous)( 2 ) , LC50( 3 ): mg/kg body weight : mg/kg body weight mgIl inhalation!

-- - - - - - - -- - - - - - - -- - - - - - ------------------__ _ _ _ _ _ _ _ _ _ Ir I l I

,1 LD50 < 5 LDS0 < 10 LC50 < 0.1 i

2 i 3 < LD50 < 25 1 10 < LD50 < 50 0. < LC50 < 0.5 '

(') LD50 oral in rats.(2) LD50 cutaneous in rats or rabbits(3) LC50 by inhalation (four hours) in rats.

APPENDIX 5

Bibliography

Abraham, G. (1963)."Jet diffusion in stagnant amblent fluid". Delft HydraulLcsLiboratory, No. 29 Hydraulics Laboratory, Delft, Holland.

Cox. R.A. (1980). "Methods Cor Predicting the AtmosphericDispersion of Massive Releases or Flammable Vapour".Proc. Energy. Comb. Sci. Vol. 6, pp 141-149. Pergamon Press.

Cox, R.A. and Roe, D.E. (1977)"A Model of the Dispersion of Dense Vapour Clouds".2nd Intl. Loss Prevention SypoPiuUM, Heidelberg, 1977.

Cox, R.A. and Carpenter R.J. (1979)."Further Development of a Dense Vapour Cloud Dispersion Modelfor Hazard Analysis". Paper presented at the SyMposium 'SchwereGase' at Battelle Institute, Frankfurt am Main, September 1979.

Cude, A.L. (1975)."The Generation Spread and Decay of Flammable Vapour Clouds".IChemE Course "Process Safety - Theory and Practice", .TeesidePolytechnic, Middle.sborough, 7-10 July 1975.

Ellison, T.H. and Turner, J.S. (1960)."Mixing of Dense Fluid in a Turbulent Pipe Flow".JFM.8., 514-544.

Fauske, H.K. (1965)."The Discharge of Saturated Water through Pipes". CEP Symp.Series 61, p.210.

Keffer, J.F. and Baines, W.D. (1963). "The round turbulent Jetin a cross-wind". J. Fluid Mech. 15, 481-496.

Lapple, C.E. (1943)."Isothermal and Adiabatic Flow of Compressible Fluids". Trans.Am. Inst. Chem. Engrs. 3.9, 385.

Moody, L.F. (1944).Trans. Am. Inst. Soc. Mech. Engrs 66 671.

Morrow, T.B. et al. (1980). "The Dispersion of Chemical Vapouraemitted from Marine Chemical Carrier3". 3rd Int. Symp. on LossPrevention and Safety Promotion in the Process Industries.Basle.

Ooms, G. (1972)."A New Method for the Calculation of the Plume Path of GasesEmitted by a Stack". Atmospheric Environment, Pergamon Press, 1972.

Ooms, C., Mahieu, A.P. and Zells, F. (t974)."The Plume Path of Cases Heavier than Air". 1st Int. Lo0sPreventlon Symposium, The Hague/Delft, 1974.

Pasquill, F (1974). "Atmospheric DifCu3Lon". 2nd Edition,Ellis Horwood, Chichester.

Perry, J.H. (1963).Perry, J.H. (Ed), "The Chemlcal Engineers Handbook". 4thEdition, McGraw-Hlll.

Perry, R.H. and Chilton, C.H. (1973). Chemical EngineersHandbook. 5th Edition. McCraw-Hill. Kogakusha.

Peterson, R.L. and Cersak, J.E. (1979). "Plume Rise for VarylngAnbient Turbulence, Thermal Stratlfication and Stack ExitConditions in Numerical and Laboratory Evaluatlon". 5th Int.Conf. on Wind Engineering. Fort Collins. Colorado.

Turner D.B. (1969). Workbook of Atmoapheric DispersionEstimates. U.S. Dept. Health, Education and Welfare.?HS Pub. Ho. 995-AP-26.

F. Briscoe and P. Shaw Evaporation trom Spills of Hazardous Liquids onLand and Water. SRD R100 1978.

R.A. Cox and D.R. Roe, "A Model of the Di3persion of Dense VapourClouds". Loss Prevention and Sa:ety Promotion p359, (1977).

R.A. Cox and R.J. Carpenter, "Further Developoent of a Dense Vapour CloudDispersion Model for Hazard Analyysls". Heavy Gas and Risk Assessment,ed. S. Hartwig, (1980).

D.J. Hall, C.F. Barrett and M.C. Ralph, "`xperi=ents on a Model of anEscape of Heavy Gas". Report LR 217, Warren Spring, Stevenage, Herts.,(1975).

R.G. Picknett, "Field Experiments on the Ben.aviour of Dense Clouds",Report Ptn. 1154/ 78/1, Chemical Defence Estatbishment, Porton Dowr., Wilts.

A.P. van Ulden, "On the Spreading cf a Heavy Gas Relea3ed near theGround". 1st Tnternational Loss Prevention Sy=pcs3uU, The Hague/Delft(1974).

"Flare Radiation" API RP521.

J. Hajek and E. Ludwig. Petrochem. Eng. Part 1. (1960).

W. Jost "Explosion and Combustion Processes in Gases". McGraw-Hill, New

York (1946).

M. Zabetakis and E. Burgess, Bureau of Mines Report. Invest 5707. (1961).

J. Moorhouse and M.J. Pritchard. "Thermal Radiation Hazards from LargePool Fires and Fireballs". A literature Review. The Assessment of MajorHazards Symposium, Manchester (1982).

A.F. Roberta. "The Effect of Conditions Prior to Loss of Containment 'inFireball Behaviour". The Assessment of Major Hazards Symposium,Manchester, (1982).

D.E. Jarrett, "Derivation of the 8ritish Explosives Safety Distances".

Ann. N.Y. Acad. Sci. No 152, (1968).

Toegepast - Naturweteschappeli'k Crderzoe.. (TNO): Methods for thecalculation of the physical effects of the escape of dangerous material("TNO Yellow Book) (1979).

Perry and Green: Perry's Chemical Engineers' Handbook, (Sixth Edition,1984).