Thermal Response of Vessels and Pipework Exposed To Fire

Crown copyright material is produced with the permission of the Controller of HMSO and Queen's Printer for Scotland.
0T192 610
Crown copyright material is produced with the permission of the Controller of HMSO and Queen's Printer for Scotland.
THERMAL RESPONSE OF
VESSELS AND PIPEWORK
EXPOSED TO FIRE
This document was written by
Dr J N Davenport
Shell Research Limited

Dr S M Richardson and Dr G Saville
Imperial College
for
The Steel Construction Institute
Silwood Park
Ascot
Berkshire SL5 7QN
This document
London: HMSO
Lc0t5
pages
Health and Safety Executive- Offshore Technology Information
Crown copyright 1992

Applications for reproduction shouldbe madeto HMSO
Firstpublished1992
iSBN 0 11 8820990
This reportis publishedby the Healthand SafetyExecutive as part

of a series ofreports of work whichhas been supported by funds
formerly providedby the Department of Energy and lately by the
Executive. Neither the Executive, the Departmentnor the

contractors concerned assume any liability for the reports nor do
they necessarily reflectthe views or policy ofthe Executiveor the
Department.
Publications in the Offshore Technology Information (OTT) series
are intendedto provide background information and data arising
from offshore research projectsfunded by the Department, or the
Executive, and majorcompanies.
Results, including detailed evaluation and, where relevant,

recommendations stemming from their research projects are
published in the 0TH series ofreports.
HMSO
Standing order service

Placinga standing order with HMSOBOOKS enables a customer
to receive other titles in this series automaticallyas published.
This saves time, trouble and expense ofplacing individual orders
and avoidsthe problem ofknowingwhento do so.
For details please write to HMSO BOOKS (PC 13A/l).

Publications Centre, P0 Box 276, London SW8 5DT quoting
reference 12.01.025
The standing order service also enables customers to receive

automatically as published all material of their choice which
additionally saves extensive catalogue research. The scope and
selectivity of the service has been extendedby new techniques,
and there are more than 3,500 classificationsto choose from. A
special leaflet describing the service in detail may be obtainedon
request.
FOREWORD
This report is one oftwenty six workpackage reports writtenas part ofthe Joint Industry Project on Blast
andFire Engineering forTopsideStructures. TheProjectPhase 1 startedin May 1990 to collate, appraise
anddisseminate informationonblastandfireloads,andontheresistance ofstructures andfacilities tothese
loads. The titles and numbers ofthe reports generated bythis projectare as follows.
List of Reports
01192597 Behaviour
ofoil and gas fires'inthe
(FL2)
presence ofconfinementandobstacles.
General
01192585 Genericfoundationdata to be used in
theassessment ofblast and fire sce(01(a))
01192598 Currentfireresearch, experimental,
narios.
(FL3)
(02 includedTypicalstructuraldetailsforprimary,
inthe above secondary, andsupporting structures/
report)
components.
scenarios.
Theeffectsofsimplification ofthe
(BR1)
explosionpressure-time history.
(BR2)
release rates.
miningstructural response.
01192601 Computerised analysistools for

(BR3)
assessing the response ofstructures
01192588 Legislation, codesofpracticeand

(G3)
OTT 92599
01192 600 Explicitanalytical methods fordeter
01192587 Thepredictionofsingle andtwo phase

(Gl(c))
resources.
Blast Resistance
01192586 Representative range ofblast and fire

(01(b))
theoreticalandpredictivemodelling
subjected to blastloading.
certification requirements.
01192589 Experimental facilitiessuitableforuse

in studiesoffire and explosion hazards
(04)
in offshorestructures.
01192602 The effectsofhigh strain rateson
01192590 The use ofalternativematerialsin the

(05)
designand construction ofblast and
(BRS)
(BR4)
01192603 Analysis ofprojectiles.
Fire Resistance
fire resistantstructures.
Blast Loading
01192604 Experimental data relatingto the

(FR1)
performance ofsteel components at
01192591 Gas/Vapour build-up onoffshore

(BL1)
elevated temperatures.
structures.
01192605 Methodologies and available tools for

thedesign/analysis ofsteel components
(FR2)
01192592 Confmedventedexplosions.
atelevatedtemperatures.
(BL2)
01192593 Explosions in highlycongested

(BL3)
01192606 Passive fireprotectionperformance

(FR3)
requirements andtest methods.
volumes.
01192594 The prediction ofthepressureloading

on structures resulting from anexplo(BL4)
01192607 Availabilityand properties ofpassive

and activefireprotectionsystems.
(FR4)
sion.
01192608 Existingfire designcriteria for
01192595 Possibleways ofmitigating explosions

on offshore structures.
(BL5)
(FR5)
secondary, support and systemsteelwork.
01192609 Fireperformance ofexplosion(FR6)

damagedstructural and containment
Fire Loading
01192596 Oil and gas fires - characteristics and
(FL1)
material properties.
steelwork.
impact.
01192610 Thermalresponse ofvessels and

(FR7)
pipeworkexposedto fire.
Thepurposeofthese reportsistocollate, reviewandcriticallyappraisethecurrentstateofknowledgeineachofthe

listedsubjectareas. They discuss the currentstate oftechnologyas applicableto eachofthese areas. Theyalso
Theobjectivesandscopeofeachreportareoutlinedinthe'Work
indicatetheareasofsignificantresidualuncertainty.
included
in
each
PackageDescriptions'
report
Thereportsarewrittenforusebyengineerswith specialistresponsibilities forspecifictasksassociatedwith safety
againstpotentialfiresandexplosions inthe designofTopsidesonOf hore Structures (egaspartofaFormalsafety
Assessment)or forthe usebytheirexpertadvisors. Theyare intendedtobeboth expertandauthoritative.
Thesereportshavealsobeenusedasthebasisforestablishinginterimguidanceandtoidentifyanynecessaryfurther
researchordevelopmentwork These are alsoproject deliverables andare issuedas separatedocuments.
This Projecthasbeensponsoredand fundedbythe followingtwentynine oianisations:
AEA Technology(SafetyandReliabilityDirectorate)
Agip(UK) Limited
AmeradaHess Limited
Amoco (UK)ExplorationCompany
Aico BritishLimited
BritishGasPlc
BP International Limited
Chevron(UK)Limited
Conoco(UK) Limited
DenNoiske Stats OljeselskapAS
TheUKDepartmentofEneiy
Elf(UK) Plc
EnterpriseOilPie
ESSO Exploration& Production(UK)Limited
HamiltonBrothersOil& (lasLimited
Kerr-McGeeOil(UK) Plc
MarathonOil (UK) Limited
MobilNorth SeaLimited
NorskHydro AS
OccidentalPetroleum(Caledonia) Limited
Petro-CanadaResources
PhillipsPetroleumCo (UK) Limited
Ranger Oil(UK) Limited
SagaPetroleumAS
ShellUKExplorationandProduction
TexacoBritainLimited
Total OilMarinePlc
UltramarExplorationLimited
Unocal (UK)Limited
CONTENTS
Page
EXECUTIVE SUMMARY
1.
DEFINITIONS
2:
INTRODUCTION
The fire response work packages
2.1
2 . 2 Obj ectives of FR7

2.3
2.4
2.5
3.
ON OFFSHORE PLATFORMS
12
Platform layout
Pipes and vessels
The hydrocarbon inventory
Fire scenarios
Fire loading of vessels and pipes
Mitigatory factors
PHYSICAL PROCESSES
4.1
4.2
4.3
4.4
4.5
4.6
4.7
5.
10
10
Outline of the study
FIRES
3.1
3.2
3.3
3.4
3.5
3.6
4.
Background
Scope
5.2
12
12
13
13
14
14
17
External heat fluxes

4.1.1 Heat flux from the fire
4.1.2 Heat transfer to the pipe or vessel
4.1.3 Heat transfer to a water deluge
4.1.4 Heat losses to the surroundings
Heat transfer through the vessel/pipe wall
Heat transfer to the vessel/pipe contents
4.3.1 Heat transfer from wall to vapour
4.3.2 Heat transfer from wall to liquid
4.3.3 Heat, mass and momentum transfer
between liquid and vapour
Venting
4.4.1 Venting through the blowdown system
and a PSV
4.4.2 Venting of a flashing liquid
Failure mechanisms
4.5.1 Superposed pressure and thermal stress
4.5.2 Stress at failure
Summary
Sources of equations
CURRENT MODELLING CAPABILITIES

5.1
9
9
9
The LPG fire response models

5.1.1 Features of the LPG response models
5.1.2 Validation of the LPG response models
Blowdown models
18
18
18
19
20
20
22
23
24
26
28
28
30
31
32
33
34
35
37
37
37
38
40
5.3
6.
5.2.1 Features of the blowdown models

5.2.2 Validation of the blowdown models
Summary of current capabilities
OUTSTANDING REQUIREMENTS
6.1
6.2
6.3
40
41
41
43
The requirements of predictive models
43
Physical processes
Predictive modelling
6.3.1
Non uniform heat fluxes
6.3.2
Models for pipework
6.3.3
Fire protection
6.3.4
Multicomponent fluids
6.3.5
Venting
6.3.6
Thermodynamic treatment
6.3.7
Vessel failure
6.3.8
Validation
6.3.9
Interconnected vessels
6.3.10 Partial damage
6,3.11 Heat transfer to the sea
43
44
45
46
46
46
47
47
48
48
48
49
49
7.
CURRENT POSITION
50
8.
AREAS OF UNCERTAINTY
54
REFERENCES
57
TABLES
FIGURES
APPENDICES
FR7 Work Package
Primary information on predictive models
REVIEW PAPER BY K MOODIE
EXECUTIVE SUMMARY
EXECUTIVE SUMMARY
Fire attack of vessels and pipework containing hydrocarbon on
offshor platforms can lead to escalation if loss of
containment occurs.
The
appropriate emergency
response
on
a
firm
depends
understanding of the vulnerability of
vessels and pipes to different fire types, a ranking of the
equipment most at risk, predictions of the time available for
countermeasures, and the hazard posed by rupture.
Knowledge of these factors is essential for the safe operation
and shutdown of existing facilities, and should influence the
design of new platforms.
This report examines the extent of knowledge of the physical
processes which take place during fire attack, and compares
the predictive capabilities of available models with the
requirements. It gives a "snapshot" of the current position,
concentrating on information which is within the public
domain, and models which have been validated in large fires.
The relevance and applicability of the relationships used to
describe physical processes is discussed, and reference is
made to sources where the appropriate equations can be found.
This report includes summaries of the current position and
areas of uncertainty, but the reader is encouraged to study it
as a whole. Where possible the report is self contained, but
it draws on findings of other Work Packages, to which
reference should be made as appropriate.
The physical processes taking place when pressurised equipment

containing hydrocarbons is exposed to a fire are relatively
well understood, at least when the fluid is simple, and
empirical relationships are available to describe the heat and
mass flows. However, for the fluids of interest, which are
mixtures such as condensate and crude oils,
a physical
is
far
more
and
there
are many
description
complex,
uncertainties.
Modelling the latter subject is very difficult, and has not

yet been attempted as a whole. There are predictive computer
codes for the simpler situations, and in that role they are
satisfactory, but as might be expected, they have only limited
application.
One of the most important questions to be answered is the
The available
likely time to failure of a vessel or pipe.
models do not generally predict this, and the subject needs
further attention.
It is possible that the existing computer codes can be used to

rank vessels and pipework at risk from fire attack, to gain an
EXECUTIVE SUMMARY
idea of the relative importance of factors not yet included,

and to aid the development of more appropriate models.
great deal of research and development work may be needed,

but it must be well focused to meet pre-identifled essential
needs on time.
SECTION 1
DEFINITIONS
1.
DEFINITIONS
Organisations and Official Bodies
API
American Petroleum Institute
ASTM
American Society for Testing and Materials
BAM
Federal
HMSO
Her Majesty's Stationery Office
HSE
Health and Safety Executive (UK)
Institute for Materials Research and Testing
Terms and Abbreviations
BLEVE
A foiling iquid expanding apour xp1osion which
results
from the sudden failure of a vessel containing a
pressurised saturated liquid/vapour at a temperature well
above its normal (atmospheric) boiling point.
BLOWDOWN
The controlled venting of a pressurised vessel or line on

normal or emergency shutdown of a process.
CRITICAL HEAT FLUX (density)
The heat flux (density) above which
occur.
film boiling
CRITICAL PRESSURE
The pressure below which a substance may exist as
in equilibrium with the liquid.
can
a gas
FIREBALL
A fireball is the rapid turbulent combustion of fuel as
an expanding, usually rising ball of flame.
HEAT FLUX (density)

Heat flux density is an expression quantifying the rate
of heat transfer per unit area normal to the direction of
heat flow. A convenient unit in the present context is
kW m2. (1 kW m2 = 317 Etu ft2 h').
DEFINITIONS
SECTION
HYDROCARBON (fire)
A hydrocarbon is a molecule comprised exclusively of
carbon and hydrogen. There are an enormous number of
such species; the petroleum industry commonly deals with
hydrocarbons ranging from one carbon atom (methane,
natural gas) through to waxy and bituminous mixtures with
several tens of carbon atoms.
From the fire point of
there
is
also
a
wide
view,
spectrum of other organic
chemicals, solvents, treatment agents etc., with similar
combustion characteristics producing a rapid temperature
rise and sustained high heat flux density.
See Work
Package FL1 Section 7.
JET FIRE
turbulent diffusion flame resulting from the combustion

of a fuel continuously released with some significant
momentum in a particular range of directions.
LPG (Liquefied Petroleum Gas)

The term applied to those hydrocarbons which are in the
gaseous state at normal temperature and pressure but
which
can
be
liquefied
by
compression
and/or
refrigeration. Propane (boiling point -42C) and butane
(boiling point -0.5C) are the most common.
POOL FIRE
For the purposes of this report, a pool fire is defined
as a turbulent diffusion fire burning above an upward
of
fuel
under
facing horizontal
pool
vaporising
conditions where the fuel vapour or gas has zero or very
low initial momentum. See Work Package FL1 for details.
PSV (pressure safety valve)

A relief valve which prevents the internal pressure of a
vessel or
pressure.
pipe
from
exceeding
an
allowed
limiting
TOPSIDE
In this report, topside structures are taken to include

the jacket, risers and conductors down to the waterline.
The insides
included.
of
legs
below
the
waterline
are
also
INTRODUCTION
SECTION
2.
INTRODUCTION
2.1
The Fire Response york Packages
There are seven Fire Response work package reports:

FRi
Experimental data relating to the performance

of steel components at elevated temperatures
(SCI).
FR2
Methodologies and available tools for the

design/analysis of steel components at elevated
temperatures (SCI).
FR3
Passive
fire
protection
requirements and test methods
performance
(Shell Research
Ltd).
2.2
FR4
Availability and properties of passive

active fire protection systems (SCI).
and
FR5
Existing fire design criteria for

support and system steelwork (Sd).
FR6
Fire
of
performance
explosion-damaged
structural and containment steelwork (SCI)
FR7
Thermal
and
response of vessels
exposed to fire (Shell Research Ltd)
secondary,
pipework
Objectives of FR7
The formal work package objectives are:
To summarise knowledge on the thermal response of

hydrocarbon containing vessels and pipework to fire
effects.
To identify gaps in knowledge.
The FR7 work package is reproduced for reference in Appendix

A.
2.3
Background
The rupture of pipework or process vessels through fire impact

on an offshore platform can lead to escalation. To devise
effective means of protection and control it is essential to
understand the mechanisms involved.
There has to be a firm understanding of the vulnerability of
vessels and pipework to particular fire events, and the
SECTION 2
INTRODUCTION
ability to predict which parts of the plant are most at risk,

so that special protection or additional safety steps can be
adopted, either retrospectively on an existing structure or at
the design stage for a new platform. If a fire does develop,
there needs to be an assessment of the time to possible
failure, so as to implement the safest and most effective
emergency procedures.
2.4
Scope
This study is based on the assumption that a platform fire has
triggered the alarm system, and initiated an emergency
shutdown sequence involving isolation and blowdown of the
process trains, and activation of the water deluge system.
For vessels and pipework containing hydrocarbon, this is
probably a more severe scenario than one in which the process
plant is still in operation, as it precludes removal of heat
by the flowing fluid.
It is recognised that the emergency shutdown may not occur in

practice, for example the alarm system may fail to respond,
and damage to equipment or communication lines might be so
rapid and severe that shutdown does not happen as planned.
However, such events are outside the scope of this study.
2.5
Outline of the study
This study has involved work in two main areas:
The state of knowledge of the physical processes which

occur when vessels and pipework are exposed to fire, and the
ability of models to predict the response.
The relevance and application of this knowledge to the safe
operation of existing and future offshore platforms, and the

identification of gaps or deficiencies where further work may
be required.
The approach has been to appraise
The characteristics of vessels and pipework on the topsides

of offshore platforms, the nature of potential fire loadings
upon them, and the ways in which the effects of fire on
vessels and pipework can be mitigated (Section 3).
The basic physical processes occurring when vessels and

pipework containing hydrocarbons are exposed to fire loadings
(Section 4).
10
INTRODUCTION
SECTION
The capabilities and extent of validation of models

(probably, but not necessarily, involving the use of computer
programs) for prediction of the response of vessels and
pipework containing hydrocarbons to fire loadings (Section 5).
The differences between what is required of a comprehensive

predictive model which has been fully validated and what is
currently available, and ways in which those differences might
be eliminated (Section
6).
The relevanceof fire response modelling work to the safety

of offshore platforms, and in particular the ways in which
future work might be targeted to improve understanding of the
hazards involved. In this way platform design, operational
practices and emergency response may be improved.
The primary information on the current models, which has been

used for this section of the report, is given in summary form
A relatively recent paper, which includes a
(Appendix B).
review of the modelling of the response of vessels engulfed in

fires and of experimental validation of
also given (Appendix C).
Parts 7 and
uncertainty.
present
the
11
such modelling,
current position
and
areas
is
of
FIRES ON OFFSHORE PLATFORMS
3.
SECTION
All
process plants dealing with flammable materials are

vulnerable to fire attack of the pipework or process vessels.
Loss of containment can quickly result in escalation of the
fire and a loss of control.
This is particularly so for
offshore platforms, where restricted space means that vessels
and pipes are very close together.
In addition, there is
often a large inventory of hydrocarbons, and escape to a safe
distance is difficult.
Escalation of the fire through attack of a vessel or pipe can
occur in several ways. Low momentum events such as leakage of
hydrocarbon would increase the fuel available to a local pool
fire or create an additional one, while high pressure gas or
liquid releases might give rise to jet fires and fireballs.
Explosion hazards such as blast waves and missiles can occur
if rupture of the vessel or pipe is sufficiently fast.
3.1
Platform layout
The space available on the topsides of offshore oil or gas
platforms is very restricted, typically some tens of metres
long, wide and high. Almost without exception this means that
the processing vessels and pipework are packed very closely.
Thus the layout is very different from that in a comparable
onshore process plant, and the type and degree of reliance on
fire protective measures is not the same.
At a location such as a refinery, fire escalation can be
minimised through well-defined hazardous and non hazardous
areas and large safety distances, but for the offshore
platform, far more reliance has to be placed on containment of
the fire by barriers and water deluge systems.
3.2
Pipes and vessels

The vessels of interest in the present study range from
ambient pressure storage tanks to high pressure separators,
with varying capacities up to several tonnes. Usually the
vessels are housed within specific modules, but in some cases
they are large enough to span adjacent areas.
There are kilometres of interconnecting pipework, and in
addition, there are the conductors and risers for oil/gas
import and export.
The walls of the vessels and pipelines on an offshore platform

are generally steel, but there may be additional, usually
concentric, layers of different materials such as insulation
on walls which can become very cold (for example downstream of
Joule-Thomson (flash)
and pressure relief valves) or as
12
SECTION
thermal protection on walls which might become very hot in a

fire.
Operating temperatures of the pipes and vessels can range from

30 Celsius to +100 Celsius, with pressures up to several
hundred bar, but generally below 200 bar.
3.3
The hydrocarbon inventory

The hydrocarbons of interest fall into two categories - those
involved in the primary production process, and those which
are used in support of the platform operation.
In the production process the hydrocarbons can range from
natural gas (mainly Cl) through associated gases (mainly Cl to
C4) and condensates to 'live' and 'dead' crude oils.
In addition there will be a variety of hydrocarbons used in
support roles, for example diesel fuel for generators,
methanol for gas processing, and aviation fuel for helicopter
refuelling.
Thus the inventory of the pipes and vessels can be all vapour,
all liquid or vapour plus liquid and can comprise hydrocarbons
There may also be
ranging from Cl to C30 and beyond.
of
water, nitrogen, carbon dioxide and
significant quantities
hydrogen sulphide.
On an oil platform in particular, the inventory may be quite

large. Quite apart from the contents of the process vessels
and pipework, which will amount to many tens of tonnes, there
are large quantities held in the risers (this can usually be
segregated by emergency shutdown devices). Much more detailed
information on platform layout and inventories can he found in
Work Packages Cl and G2.
3.4
Fire scenarios
A fire can be broadly characterised by the
type, physical
state and source of the fuel involved, the ventilation
conditions, the intensity with which the fire burns, and its
duration. The principal fire conditions to be considered on
the basis of severity and in relation to the response of
vessels and pipework relate to hydrocarbon pool fires and jet
fires.
Work Packages FL1 and FL2 provide a comprehensive
review and appraisal of heat flux data relating to these
fires.
13
3.5
SECTION
Fire loading of vessels and pipes

Incident heat flux on a vessel or line will, of course,
increase the wall temperature, and the temperature and
pressure of the contents. Usually, the pressure rise will
eventually trigger lifting of a pressure safety valve (PSV),
and fire detection systems may have initiated external cooling
from a water deluge system.
A PSV
does not prevent weakening and failure of a vessel that

becomes locally overheated and overstressed.
It can only
limit the internal pressure from rising beyond the allowable
accumulation pressure.
vessel or pipe
be protected against such failure by
depressurising
the heat input
and/or
The value of these mitigatory factors are considered in the
next section.
limiting
3.6
Mitigatory factors
Controlled depressurising of a vessel (blowdown) reduces the
internal pressure, and hence the stress in the vessel walls.
It also reduces the potential addition of fuel to the fire
should the vessel rupture.
Industry practice for the blowdown of process facilities has
criteria given in API Recommended
Practice documents 520 and 521 [1,2]. This generally involves
reducing the pressure from initial conditions to 50 % of the
vessel's design pressure or 7 barg, whichever is the lower,
within approximately 15 minutes.
been led by blowdown
blowdown operation is one which vents from the vapour

then depending on how effective it is, the overall
If
change of vessel pressure may be positive or negative.
blowdown is from the liquid then of course the liquid level
drops. The relative amounts of liquid and vapour will change
as a function of temperature and pressure, and indeed as the
If the
space,
liquid temperature increases,

either rise or fall.
The strength of the
If
temperature rises.
with
the
internal
cope
fail, and the process
vessels and pipework.
the
quantity
of
liquid
may
vessel will decrease as the wall

the wall strength is no longer able to
pressure then the vessel or line will
may continue with other neighbouring
important to realise that failure may actually take

at
a lower pressure than was originally present in the
place
It
is
vessel,
and
that
blowdown
14
sequence
which
gives
net
SECTION
reduction of pressure with time is no guarantee of the vessel

integrity.
If blowdown is activated as a result of a fire, it might
increase the likelihood of vessel
failure under
some
circumstances. For example, if the blowdown operation lowers
the fluid level in a vessel with a fire loading,
the
increasing area of unwetted wall makes the risk of vessel
failure greater.
Limiting the heat input from fires by external insulation

reduces both the rise of the vessel wall temperature and the
generation of vapour inside the vessel. The various types of
fire protective materials which can be used to mitigate the
effects of fire loading on vessels and pipework are discussed
in Work Packages FR3 and FR4.
External deluges, generally of water, are often directed onto
the upper parts of a vessel. Convective flux is reduced by
heat transfer to the water, and radiative flux to the vessel
wall is reduced because of high absorption by the water.
Water deluge systems are discussed in Work Package FR4.
the platform can be improved, whereby
the
Layout of
geometrical arrangement of vessels and lines is used to reduce
the risk or consequences of fire spreading from one vessel or
line to another.
All of these mitigatory factors have their uses on platform
topsides, in various situations, but the most appropriate
choice or combination requires knowledge of the most likely
fire hazards and the response of the vessels and pipework both
protected and non-protected.
There are other considerations as well.
For example:
Fireproof layers on the outside of the wall can have low

mechanical strength when exposed to severe fires. The high
temperature properties of the materials have to be studied
carefully, as radiative heat transfer within the insulation
may become important.
Firewalls can be used as barriers for protecting particularly
sensitive areas, but can hinder access or egress in an
emergency. Design of firewalls must take into consideration
the effect of compartmentalising any fire or explosion.
Water deluge systems are common, but have usually not been
tested under jet fire conditions, where relatively high gas
velocities in some fires might seriously affect performance.
Scale and corrosion within the water system can affect
operation. In addition, the deluge system can only work if
15
SECTION
the seawater mains, and the pumps and associated power supply
survive the fire.
Removal of contents tends to work only when the incident heat
flux is relatively small, since the contents have to be
removed sufficiently quickly and safely.
The use of different layouts, for example whether it is better

to cluster high pressure vessels in one area and low pressure
ones in another, rather than mix high and low pressure vessels
together, has not yet been seriously investigated.
One novel form of heat absorption for vessels [3] involves a
concentric inner shell, with apertures at the top and bottom.
If fluid in the vessel boils, some is forced up the annular
space, and back through the aperture into the body of the
vessel.
It is essentially an internal analogue of a water
deluge, providing cooling to the upper part of a vessel, where
it is generally most needed.
16
SECTION 4
THE PHYSICAL PROCESSES
4.

Exposure of a vessel or line to a fire involves up to six
essentially independent units (the fire, the wall of the
vessel or line, its vapour contents, its liquid contents, the
vent and the surroundings). The units are interconnected by
appropriate heat, mass and momentum fluxes, as shown below.
Vessel or pipe
vapour contents
Fire
heat
mass
flux
heat flux
heat flux
flux
Vessel or pipe wall
><
mass flux
Vent
including insulation
momentum flux
heat
heat flux
flux
mass
flux
4r
Surroundings
Vessel or pipe
liquid contents
For the purposes of this review, the characteristics of the

fire itself are regarded as known, this information being
provided in Work Packages FL1 and FL2.
Thus the primary concern here is with:
The
temperature response of the wall
of the vessel
or
line, and the insulation if fitted (Section 4.2);
The temperature and pressure response of the contents of

the vessel or line, with thermodynamic and phase consistency
requirements if the contents are two-phase (Section 4.3);
The characteristics of the PSV and the blowdown sequence

(Section 4.4);
The response of the pressure-containing wall of the vessel
or
line,
in particular
the manner
in which
it
might
fail
(Section 4.5).
In what follows, no equations are given. To do so would vastly
lengthen the report without a significant increase in useful

information. The equations themselves are available in the
literature, to which specific references are given (Section
4.6).
17
4.1
SECTION 4
External heat fluxes
The external flux balance comprises the fire loading, the heat
which is transferred to the pipework or vessel, heat losses to
the surroundings, and heat transferred to the water deluge if
present. These are illustrated in Figure 4.1.
4.1.1
Heat flux from the
fire
The principal external heat flux is from the fire to the wall
of the vessel or pipe. The magnitude of the flux is dependent
on the type of fire and relative positions of the flame and
the object.
If
the vessel
or pipe
is
partially
or
fully
engulfed in the fire, heat transfer will be through radiation

and convection; for a non engulfed object the convection term
is insignificant.
Detailed information
Packages FL1 and FL2.
on
this
area
is
presented
in
Work
Fire attack of a vessel or pipe is likely to be accompanied by

venting, through blowdown and/or the PSV to a flare, or
through loss of containment. When venting is to flare, the
additional heat flux can usually be neglected since the flare
tip will almost certainly be far away, and there are likely to
be intervening walls which shield the radiation.
However,
when venting is through a rupture it is possible and perhaps
likely that the efflux will ignite, giving an additional heat
flux.
In principle this could be modelled in a similar way to
venting through a valve or choke (see Section 4,4), but there
are many uncertainties, such as the size of the rupture and
the way it develops with time.
4.1.2
Heat transfer to the pipe or vessel

Transfer of heat to the outside of the pipe or vessel is by
convection and radiation.
Where engulfment does not take
transfer
is
place,
through radiation, the important factors
being
*
Magnitude of the incident radiative flux
Variations of the flux in space and time

Relative geometry of the fire and the vessel
Absorptivity of the intervening air (perhaps
including combustion products)
Absorption by a water deluge
Absorptivity and emissivity of the wall
Wall temperature
Radiative interchange factors have been calculated for several

rather simple geometries (see, for example [4]), but the shape
and position of the flame can be
predict, and may
difficult to
18
SECTIOI'T
be more complex than the cylinder or cone which

assumed, particularly if it is in an enclosure.
is
often
the pipe or vessel which are engulfed in the flame

will additionally experience a convective heat flux, which can
also show large variations spatially and perhaps also with
time.
Prediction of the heat fluxes within a flame is outside
the scope of FR7 - see FL1 and FL2 for more details.
Areas
If we make the assumption that the radiative and convective

fluxes present at the wall of the vessel or pipe are known,
then relatively simple relationships (see, for example [6])
can be used to predict the magnitude of the heat transfer.
However, it is important that due account is taken for the
change in wall properties with time and temperature. There
may be a layer of burnt paint, for example, or soot may be
deposited on the wall surface, which can change the emissivity
In addition, temperature dependant thermal
considerably.
properties are often unavailable for insulating materials particular, measurement of the radiation properties is a
specialist area [7].
in
4.1.3
Heat transfer to a water deluge

Many pipes and vessels
are equipped
system, and additional water may be
emergency firefighting response. Heat
may take place from the external heat
vessel wall.
with
a water deluge
part of the
transfer to the water
flux or from the hot
used as
For efficiency a water deluge is usually applied to the upper

part of the containment, as drainage of water down the wall
under
will
then
cool
the
lower
More
gravity
part.
importantly, fire attack of the containment will give far
higher shell temperatures where vapour is in contact with the
inner surface than where there is liquid, because of the
different efficiencies of heat transfer. Thus the upper part
of the wall is generally more in need of cooling than the
lower part.
Heat transfer to the water deluge is by radiation and forced
which
leads
to
Heat
transfer
convection,
boiling.
correlations are available (see, for example reference [6]),
but the actual performance of the deluge in terms of mass flow
rate,
droplet size and distribution are often unknown,
The
particularly in the presence of a high momentum release.
boiling regime, as discussed later in Section 4.3.2, is also
important.
19
4.1.4
SECTION 4
Heat losses to the surroundings

For most vessels and pipework on topsides the surroundings are
module walls and the atmosphere, but there are obvious
exceptions, notably risers, the substantial parts of which
inevitably pass through the sea. The response of these parts
of the risers may or may not be affected by fire attack on the
topsides
There are two basic modes of heat transfer from the wall of
the vessel or pipe to the surroundings - radiation and natural
convection - which combine additively to give the overall
flux.
The temperature dependance is such that when the wall
is hot, most of the outward heat transfer is by radiation;
when it is cool, most is by natural convection.
Of course, if a vessel
is fully engulfed by fire, any heat

transfer to the surroundings may be negligible compared with
the fire itself. However, if there is partial engulfment, for
example a pipeline, conduction of heat along the line may give
significant levels of radiative flux from non-engulfed areas
to the surroundings.
Modelling the radiative and convective heat losses to the

surroundings is more straightforward than modelling the
incident flux. A reasonable assumption is that the vessel or
pipe is a black or grey body, for which well established
relationships exist between wall temperature and the radiant
flux emitted, eg [6]. However, the emissivity of the surface
may not be known, particularly if is covered by soot or char.
Convection from the wall to surrounding air can be natural or
forced, and correlations for the two cases exist (see, for
example, reference [5]). However the initial air temperature
and velocity can be important uncertainties, particularly if
the fire is very intense or if the vessel or line is partially
enclosed, as is often typical in the modules on topsides.
The completely dominant mode of heat transfer from a wall to
the sea is natural convection.
Radiation is negligible
because of efficient cooling of the vessel or pipe wall.
Forced convection can be important in certain sea states, but
a conservative risk analysis would presume natural convection
alone. The heat flux from such natural convection can be
determined using well established correlations, for example
from [6]. Unlike the case of heat transfer to the atmosphere,
there is no difficulty in specifying the sea temperature, but
the water velocity may be unknown.
4.2
Heat transfer through the vessel/pipe wall

The rigorous determination of heat transfer through the wall
of
the
pipe
or
vessel
to
20
the
fluid
contents
requires
SECTION 4
complete solution of the transient three-dimensional (radial,

axial and azimuthal or circumferential) heat conduction
This is not trivial, and numerical methods
equation.
discretisation
in space and time must generally be
involviftg
finite-difference
methods
are more
applied.
Currently,
but
finite-element
methods
can
also
be
used.
common,
The data needed are the dimensions, specific heat and thermal
conductivity of each part (steel, insulation and so on) of the
The thermal properties can vary significantly with
wall.
temperature, and point values are probably not sufficiently
accurate.
Because the numerical
solution can involve significant
of
quantities
computer time, approximations can sometimes be
made which reduce the dimensionality of the problem.
The
basic geometry is essentially cylindrical, and it is tempting
to suppose that a full three-dimensional analysis of heat
transfer through the wall is not needed.
To examine the validity
consider two scenarios:
of
such
approximations,
1.
vertical vessel in
a fully engulfing pooi
2.
horizontal pipe in
a non engulfing jet
let
us
fire.
fire.
These two cases are illustrated in Figures 4.2 and 4.3.
In the first case, for simplification, we will assume that the

heat flux to the entire vessel is uniform and constant. Thus
the important modes of heat transfer are radial conduction to
the
in
the vessel,
circumferential
liquid and vapour
and
axial
or
conduction.
Of course,
conduction,
longitudinal
these are closely coupled, but are considered separately for
the time being:
As shown below (Section 4.3), the wetted and unwetted surfaces

of the inner wall show the largest temperature difference, but
each area is close to isothermal. Thus, the driving force for
both axial and circumferential conduction along the wall is
small, particularly if the assumption of uniform and constant
heat flux is true.
A useful simplification is to treat the
two cases as isothermal zones, one between the incoming flux
and the vapour within the tank - the other between the flux
and the liquid.
This assumption will give errors if the
vessel is short and fat, rather than tall and thin, as there
is a larger area of heat conduction near the ends than in the
middle of the vessel.
Radial variations can be small within the two zones - one
source is convection of the liquid inside the vessel. Neglect
of radial conduction through the wall is possible, but doing
so means that the heat flux through the wall can not be
21
SECTION 4
evaluated, and in many cases it is variation of response with

this same heat flux which is sought.
The second case, that of a pipe exposed to a non-engulfing jet

The heat flux is almost entirely
fire, is more complex.
radiative, and there can be no assumption of
spatial
uniformity, as one side of the pipe may be receiving up to 100
kW m2, while the other
is receiving nothing. Dividing

pipe into zones may still be possible - clearly there
radial areas such as wetted and non-wetted; high flux and
flux.
However, there will also be large variations along
axial direction. The flux received is strongly dependent
distance from the heat source, and any objects between
source and the pipe effectively cast 'shadows'
the
are
low
the
on
the
Another, perhaps less important, problem in trying to reduce

the dimensionality arises from the less than perfect shape of
Most vessels have
the pipes and vessels.
geometrical
irregularities such as inlets, outlets and supports, and lines
have bends, tees and supports. There can be additional
geometrical irregularities resulting from fire damage (for
example partial loss of insulation), the effects of which need
to be determined and accounted for.
4.3
Heat transfer to the vessel/pipe contents

In general,
comprise
the contents
of
the vessel
or
pipework
a vapour or gas region above a liquid region.
will
(The
case when there are two liquid regions, for example in primary
separators, where there is a liquid hydrocarbon region above a
water layer, will be ignored here). Heat transfer from the
fire through the wall to the liquid causes it to heat up and
then boil; heat transfer to the vapour causes it to heat up
and increase in pressure. As the wall heats up, its strength
decreases, perhaps to the point of failure.
It is in the contents of the vessel or pipework that the most

Heat is
complex heat and mass transfer effects occur.
transferred from the fire through the wall of the vessel or
pipe to the liquid or the vapour or both, depending on the
nature of the fire. Heat may also be transferred from either
or both through the wall to the surroundings or to external
Heat and mass will be transferred between the two
cooling.
as
towards thermodynamic and
regions
phase
they move
and
also be
equilibrium. (Mass
inevitably heat transferred out of the vessel or line through any vent; see
Section 6.4.).
The various heat and mass
fluxes
are
illustrated in Figures 4.4 and 4.5.
will
22
4.3.1
SECTION 4
Heat transfer from wall to vapour

Heat transfer from the wall to the vapour region is largely by
radiation and natural convection, although blowdown may
increase gas velocities sufficiently for there to be forced
convection.
Because there is generally poor heat transfer
between the wall and the vapour (though there is good heat
transfer from the wall to the liquid, as already noted), the
region of the wall exposed to the fire and in contact with the
vapour is generally very hot (and much more so than that in
contact with the liquid).
Hence radiation can often dominate
natural convection.
The radiative flux can be modelled using the same theory as is

used for transfer to the outside of the vessel. However, the
geometry inside the vessel makes this more difficult, as shown
in Figure 4.4. Heat is radiated from the inside wall and some
of it is absorbed by the vapour; the rest is reflected, or
absorbed and then re-emitted by other parts of the wall (some
parts of which can be relatively cool if they are not exposed
to the fire), and the vapour-liquid interface, and so on. It
can generally be assumed that the inside of the wall is a grey
body and that most of the radiation incident on the
vapour-liquid interface is absorbed by the (relatively cool)
liquid. The data needed for the model are:
and
the wall emissivity and absorptivity;

the vapour absorptivity;
the geometry of the vessel or pipework.
The convective flux cannot be determined using an exact

theory. Instead, empirical correlations have to be used (see,
for example reference [5]), generally leading to a heat
The
transfer coefficient and hence to the heat flux.
correlations
of
the
always
require knowledge
physical
properties of the vapour. Because the pressure in the vessel
or pipework can be relatively high, it is important to use
For the
appropriate estimates of the physical properties.
data
fluids
on
a
these
will
multicomponent
typical platform
vary with time.

Given information on both modes of heat transfer, the
temperature of the vapour can be determined. Of course, this
temperature is not in fact spatially uniform: it will be
higher near parts of the wall engulfed in fire and less near
parts of the wall being cooled, for example by a water spray.
These processes lead to stratification, and it may not be
sufficient to assume that natural convection within the vapour
will be
sufficient to
cause good mixing
and spatial
isothermality.
23
SECTION 4
4.3.2
Heat transfer from wall to licuid

As discussed above, if the unwetted wall of the pipe or vessel
is exposed to the fire, it will quickly become hot, producing
radiative heat transfer to the liquid surface. However, the
predominant source of heat to the liquid is through the wetted
wall, the mode of heat transfer depending on the boiling
regime.
As
shown
in Figure 4.6, there are two principal types of

boiling and film boiling (between them lies
boiling
an area known as transitional boiling). With low incident
heat fluxes, conventional nucleate boiling occurs, in which
the liquid is largely in contact with the heated surface, and
bubbles carry the vapour to the liquid surface.
-
nucleate
As the heat flux is increased, the rate of vapour production

rises, together with the difference in temperature between the
heated surface and the bulk liquid boiling point. The maximum
boiling rate occurs at a point known as the critical heat
flux, and at higher fluxes the bubbles nucleating at the hot
surface coalesce to form a continuous layer of vapour. This
is the film boiling regime.
be recognised that the curves shown in Figure 4.6

and 4.7 will not be followed in a real fire. If the critical
heat flux is exceeded, the film of vapour drastically reduces
the heat transfer coefficient between the wall and the liquid,
and unless the incident flux is reduced, the wall temperature
rises sharply, and may approach the value of the unwetted
It should
wall.
In
practice this means that partial fire attack below the

liquid level can pose as severe a hazard as engulfment of the
unwetted wall, if film boiling does indeed occur.
It will generally be the case that the liquid in the vessel or

pipework will be sub-cooled in the initial stages of fire
attack, and under these conditions the critical heat flux is
Thus, the predominant
unlikely to be reached (see [8,9]).
mode of heat transfer will be natural convection from the
wetted wall (or possibly forced convection).
Empirical
correlations can be used (see, for example reference [5]),
generally leading to a heat transfer coefficient and hence to
the heat flux.
As the liquid temperature rises, the values of critical heat

flux decrease, and film boiling becomes more probable.
If
film boiling takes place, an alternative correlation for
convective transfer must be used, as the wall of the vessel or
pipe is no longer wetted, and radiative transfer has to be
included as for the vapour region, but with the simplifying
the
assumption that all the radiation is absorbed by
24
SECTION
(relatively cool) liquid and hence that none passes through

it.
Because of the severe difficulties in examining stable film

boiling, there are few experimental data.
Indeed, none have
been found for materials such as condensate or crude oil.
However, the evidence suggests that film boiling is possible
in a platform fire.
For example, Figure 4.7 gives an
experimental boiling curve,
produced from the data in
references [10] and [11], and showing a critical heat flux of
about 265 kW m2 for toluene at a pressure of 3 bara.
There have been
many attempts at deriving or producing

correlations for the critical heat flux, see for example
references [12] to [14].
However, these have only been
validated for a limited range of fluids.
The critical heat flux is strongly dependent on pressure, and

it has been found [15,16] that the curve depicted in Figure
4.8, which is for butane, is common, with minimum flux at the
critical pressure, a somewhat higher value at very low
pressures, and a maximum when the pressure is about one third
of the critical pressure.
Indeed, for a rather restricted
of
substances
it
was
found
range
[16]) that the flux-pressure
curves could be reduced to a single one by plotting critical
heat flux divided by critical pressure against reduced
pressure.
taking Figure 4.8 as an example, film boiling is more
likely to occur at low or high pressures, and less likely at
intermediate values. This means that for light hydrocarbons,
at least, the risk of rupture of the vessel or pipeline may
actually increase if the net result of the fire and the
blowdown sequence is a reduction in pressure.
Thus,
No critical heat flux data has been found for heavy

hydrocarbons. However, as the molecular weight increases, the
critical pressure falls dramatically [17], and hence it is
reasonable to suspect that the range of absolute pressures
over which film boiling can occur will be greater for heavier
materials. There is little information on critical heat flux
for mixtures, and none found for a multicomponent system.
Given information on heat transfer through convection and
boiling, the temperature of the liquid can be determined. Of
course, the liquid temperature is not spatially uniform in
practice, as it will be higher near the.vessel wall than in
the centre.
There is some limited evidence (see reference
and
also
[18]
Appendix B, ff Bl4) that natural convection
within the liquid is insufficient to cause good mixing of the
liquid everywhere, and that thermal stratification occurs.
This arises because boiling liquid rises up along the wall of
the vessel or tank and forms a lighter, hotter layer above the
25
SECTION 4
bulk liquid: temperature differences of up to around 20 K are

possible. While it is possible for such zonal stratification
to be modelled, presumably with inter-zone heat transfer by
natural convection (though the way in which this would be done
is not entirely obvious), it is not clear that such an
additional complication is warranted. Thus for most purposes,
it is probably sufficient to assume that the liquid is
spatially isothermal.
4.3.3
Heat, mass
and momentum transfer between licuid and vapour
During blowdown there is in general both evaporation of the

relatively lighter liquid components as the pressure drops and
condensation of heavier vapour components as the temperature
When there is a fire load on a vessel or line,
drops.
irrespective of whether there is simultaneous blowdown, there
is always evaporation of lighter liquid components; there can
also be condensation of heavier vapour components on cooler
areas of the wall, for example if there is a water spray and
only partial engulfment.
When the more volatile components of the liquid evaporate,

there is a mass flux from the liquid to the vapour.
In
addition there is a heat flux, caused by the energy change on
evaporation.
Similarly, if the less volatile components of
the vapour condense, there will be a mass flux and an
accompanying heat flux from the vapour to the liquid. There
is also a heat flux between the liquid and the vapour induced
by the (natural) convection in both. This can be modelled
using an empirical correlation to give a heat transfer
coefficient (see reference [5J) and hence the heat flux.
Determination of the mass and heat fluxes requires information
on the thermodynamic and phase behaviour of the fluid in the
vessel or line.
If it is a single component fluid, then
matters are straightforward and relatively simple methods can
be used to determine the boiling point and vapour pressure as
a function of temperature.
However, the fluids are generally multicomponent, which makes
the modelling much more complex.
The phase behaviour of a
fluid
is
described
multicomponent
by a surface in the three
dimensional pressure-temperature-composition phase space.
For
a fluid of a particular overall composition, the relative
amounts of liquid and vapour, and their compositions are
dictated by the temperature and pressure of the system, A
section through such a surface is shown on Figure 4.9
The combined effects of the incoming heat flux and the
blowdown change the pressure and temperature simultaneously,
and the problem is in calculating how the phase behaviour will
be affected. If the conditions are far from critical and do
not involve trajectories in phase space which
26
traverse the
SECTION
then
liquid-vapour boundary
(see
1),
obliquely
Figure
based
for
on
cubic
equations
straightforward methods,
example
of state (such as Soave-Redlich'-Kwong),
can be used - see
references [19] and [20].
In practice, however, trajectories are sometimes rather near

critical (they are more often near critical when the vessel is
being blown down in the absence of a fire loading) and often
traverse the liquid-vapour boundary obliquely, the accuracy of
such methods is often unacceptable.
In that case, more
and
more
complex
computationally
expensive and time
- methods must be used, for
consuming
example those based on
the extended method of corresponding states (the extension
is
being used to model non-spherical molecules, which
for
the
see
for
particularly important
higher hydrocarbons)
example references [21] to [23].
should be noted that physical property data (such as
density, thermal conductivity and viscosity) are required to
model heat and mass fluxes, both inside the vessel or line and
outside it. It is generally important that such property data
are consistent with the thermodynamic data.
This can, of
- but
- be
often
no
means
course,
by
always
guaranteed by
using the same computer program for both thermodynamic and
physical property data.
Examples of such programs include
PPDS (Institute of Chemical Engineers), PROCESS (SimSci: see
Appendix B, ff B27) and PREPROP (Imperial College: see
It
Appendix B, ff B6)
It should also be noted that there can be problems with
multicomponent systems when using empirical correlations,
which have usually been developed for single component
systems. A particular example of this is in heat transfer
during boiling, which generally involves the latent heat of
the fluid. For multicomponent systems, the concept of latent
heat is undefined, so it is not immediately obvious what to
do.
This problem may, however, be more apparent than real.
Provided the energy change on evaporation/condensation does
not vary too much with changes in composition then the latent
heat can be identified with the energy change. When this is
not the case, the penalty incurred by incorrect specification
of a latent heat sometimes manifests itself only in a small
error in temperature difference.
In addition to the thermodynamic and phase information, mass

and energy balances are required on the liquid and vapour to
complete the model. These are of the form:
mass balance on liquid:
rate of change of mass of liquid
= Inet rate of gain of mass by evaporation and condensation
I
Jrate of loss of mass
through vent
27
SECTION 4
mass balance on vapour:
rate of change of mass of vapour

net rate of gain of mass by evaporation and condensationl
irate of loss of mass through vent
I
energy balance on liquid:
rate of change of energy of liquid

=
boiling flux from walli
+
convection flux from wall
net
rate of energy gain by evaporation and condensation

rate of loss of energy through vent
energy balance on vapour:
rate of change of energy of vapour

= radiation flux from wall
+
convection flux from wall
net rate of energy gain evaporation

rate of loss of energy through vent
rate of loss of energy by expansion
and condensation
Furthermore, momentum balances on the liquid and vapour are

required when venting lines. An important distinction between
a line and a vessel is that there is a significant pressure
drop (between the closed end and the vent) in the former but
not in the latter.
To determine the pressure drop,
a
flow
method
can
be
used
on
(based
relatively simple two-phase
the analyses of Friedl, Premoli et al and Lockhart &
Martinelli; see, for example reference [13].
4.4
Venting
4.4.1
Venting through the blowdown system and a PSV

It is assumed in this study that the blowdown system has been
activated by the fire.
However, as heat is supplied to a
vessel or line from a fire, the temperature of the contents
rises and, because the volume of the vessel or line cannot
increase significantly, there may still be a net increase in
occur when the pressure
pressure. Venting through a PSV
in the vessel or line exceeds the lift pressure of the valve.
will
Sufficient
venting may occur that the net pressure falls below

the seat pressure of the valve (the lift pressure being
perhaps 5% greater and the seat pressure perhaps 5% less than
the set pressure of the valve) and the valve closes. The
pressure will then increase again, because heat is still being
28
SECTION 4
transferred to the contents of the vessel or line, and valve

However, seat wear and spring weakening
cycling may occur.
often occur, and operation of the valve commonly deteriorates.
Clearly, if there is sufficient venting that the strength of
the wall (which decreases with increasing temperature) is
never exceeded, then failure of the vessel or line will not
occur.
However, it may be impossible to vent through a
sufficiently large flow area to prevent failure in anything
other than a very small fire.
Despite this, the operation of
the vent is important because it can have a significant effect
on the time to failure.
To model venting, the following data are needed:
the characteristics of the blowdown choke;
the characteristics of the PSV,

in particular
the
variation of the flow area with pressure, which includes
specification of the lift and seat pressures (note that
these may be modified, probably unpredictably, if there
is fire damage to the valve);
the nature of the flow path downstream of the PSV and
choke, where the effects of a significant back pressure
would be important;
the nature of the flow through the valve and blowdown
choke.
The last of these, the nature of the flow, is by no means

trivial.
If venting is only of vapour (not containing
suspended droplets) or only of liquid (which does not flash to
form vapour as the pressure drops through the valve) then
matters are relatively simple.
true whether or not there is choking in the case of

vapour venting, though the common assumption of perfect gas
behaviour is generally inconsistent with the treatment of the
thermodynamics and phase equilibria in the vessel or line and
can give order of magnitude errors in predictions of flow
This
is
rate.
Otherwise, however, matters are not simple and there is still

neither a full theory nor a full set of empirical relations to
cover all circumstances which can - and do - arise in
practice, namely:
venting of vapour containing suspended liquid droplets;

venting of liquid containing vapour bubbles;
venting of liquid which flashes as it is vented.
The first of these can be treated crudely as a modification to

the pure vapour case if the volume fraction of droplets is
The second of these is
small, as is likely in practice.
likely only in a vessel or line in which the volume fraction
of liquid is so large that de-entrainment of vapour bubbles
29
SECTION 4
from the boiling liquid is impossible. It can be treated albeit crudely - by modification of standard two-phase flow
correlations. The third of these, though perhaps unlikely
unless venting is from the bottom of the vessel or line, is
perhaps the least predictable and merits special discussion.
4.4.2
Venting of a flashing liquid

The mechanism of flow of a flashing liquid through an orifice
or short nozzle seems relatively well established; what is
less certain is the rate of flow.
The compressed liquid
of
the
orifice
upstream
expands adiabatically and essentially
until
it
reaches
saturation.
reversibly
Flashing into the
state
does
take place at
however,
equilibrium two-phase
not,
this point, since nucleation to form bubbles of vapour is a
relatively slow operation. The liquid continues, therefore,
to expand as before, passing into the metastable liquid
region. This metastability can be maintained only for a short
of
time
before
bubble
has
period
growth
progressed
for
amounts
of
sufficiently
significant
vapour to form and the
fluid reverts to the equilibrium state.
The time taken for an element of a fluid stream to pass
through an orifice is comparable with the time taken for
nucleation.
Thus there is the possibility of the liquid
passing completely through the orifice before any vapour is
formed. However, the experimental evidence points to an even
stronger conclusion, that no vapour is formed as long as the
stream of liquid is converging. Vapour formation always takes
place close to the point of minimum flow area of the stream of
liquid and the flow rate through the orifice adjusts itself so
that this is achieved.
The limited lifetime of a liquid in its metastable liquid
state produces a situation very similar to the choking of a
gas as it passes through an orifice or nozzle. The limiting
factor in this latter case is the inability of the gas to
exceed the local speed of sound as it passes through the
nozzle. Thus, for small differences between the upstream and
downstream pressures, the liquid will pass the narrowest part
of the orifice without reaching the limit of metastability.
At some larger pressure difference this will not be true and
the attainment of the limit of metastability will take place
Further increases in
actually at this narrowest part.
pressure difference, such as by decreasing the downstream
pressure, will not increase the flow through the nozzle any
further, nor will it change the pressure actually in the
throat of the nozzle. There will be free flashing expansion
of liquid from the nozzle at its exit pressure to the
downstream pressure. This behaviour is, therefore, exactly
analogous to choked flow of a gas through a nozzle, but it is
30
SECTION
important to remember that, in the case of a flashing liquid,

consideration of the speed of sound is irrelevant.
quantitative theory for the amount of metastability that a

can undergo
before
compressed
liquid
phase
separation
takes
has
been
reference
place
(see
necessarily
developed
[24]) and applied to the release of high temperature water
However, although the driving
through nozzles and pipes.
force for nucleation can be expressed unambiguously in terms
of measurable physical properties of the fluid in question,
this is not true of the extent to which nucleation must take
place before phase separation takes over.
This means that this part of the theory can only be
implemented by making a comparison of predictions with actual
experiments.
Nevertheless, the final theory is remarkably
successful at representing the discharge of water through
nozzles.
For more information on venting
references [24] to [30].
4.5
of
flashing
liquid
see
Failure mechanisms
The discussions of the previous sections are devoted to the
determination of the temperature distribution over the surface
of the pressurised equipment as a function of time. It might
also lead to a knowledge of the temperature distribution
through the wall of the vessel or pipe, but this is usually of
secondary importance.
The pressurised equipment will fail, ie rupture, when it is
subjected to a stress in excess of the strength of the
material from which it is fabricated (for a thin-walled vessel
or pipeline under normal conditions, this is usually taken as
the ultimate tensile strength).
This simple statement is,
difficult
to
into
however,
put
practice with a vessel subject
to a non-uniform temperature distribution, because this
creates thermal stresses, in addition to the stresses caused
by the internal pressure, and also results in material of
variable strength over the surface of the vessel.
As a first approximation, one might expect the vessel to fail

at the point at which the total superposed stress - pressure
and thermal - exceeds the material strength, but in practice
the plastic deformation and hence stress relaxation which will
have occurred before
calculation difficult.
failure
is
reached
will
make
this
The effect of plastic deformation will be particularly

noticeable if the vessel or pipe has locally high stresses, or
equivalently, local regions of low strength. Thus, the effect
of a fully engulfing pool fire will be very different from
31
SECTION 4
that of localised fire engulfment.

Little quantitative
information is available as to the effect of this, and as a
consequence, stress is usually calculated on the basis of
elastic behaviour.
Further work is needed to identify the
true failure mechanism.
4.5.1
Superposed pressure and thermal stress

The principle of superposition in stress analysis maintains
that in a vessel which is subject to two, or more, sources of
elastic stress, the total stress is simply the sum of them.
Thus we can consider pressure and thermal stress separately.
Provided the stresses remain elastic, the determination of
pressure stress is a well established procedure and is the
basis of the pressure vessel design codes, such as BS 5500.
Some vessel designs may be too complex for these codes, in
which case finite element techniques will be required, but
this is an increase in computational complexity rather than a
change in procedure.
The determination of thermal stresses for a vessel which
remains elastic in situations in which there is a substantial
degree of symmetry, e.g. a pressure vessel in which the inside
temperature differs from the outside, but with no temperature
variation over the surface, is also well established and
analytic solutions are often possible.
Such solutions, however, are rarely available when there
is
This
significant temperature variation across the surface.
latter case is expected to be the norm where the temperature
gradient is caused by flame impingement or by a pool fire
under a large vessel. Solution of these cases is possible by
finite element techniques, often using the same computer
program as is used for the pressure stress calculations.
The relative magnitude of thermal stresses and pressure
stresses varies enormously.
For thin-wall vessels, the
thermal stress caused by a temperature difference through the
wall is relatively small, but temperature differences over the
surface of the wall can lead to significant stresses.
an example, a simple finite element model of a horizontal

tank containing a volatile liquid subjected to a pooi fire and
held at constant pressure through the action of a PSV, gave a
maximum thermal stress (which was in the axial direction - see
Figure 4.10) some 3 to 4 times the hoop stress due to pressure
As
[31]
The large thermal stresses arose because the lower part of the
vessel was held at essentially constant temperature (through a
high heat transfer coefficient to the boiling liquid), while
the temperature of the upper part of the vessel increased
32
SECTION 4
This gave rise to a large difference in thermal

rapidly.
expansion between the lower and upper zones, and hence the
high thermal stress. If, on the other hand, the vessel had
contained only gas, the axial thermal stress would have been
quite small compared with the hoop stress arising from the
internal vessel pressure.
The wide variation in the relative magnitude of thermal and

induced stresses makes
it
difficult
to
make
pressure
generalisations because if the direction of maximum stress
changes, so also would one expect the direction of initial
fracture to change, hence leading to a different mode of
failure. The examples given above are the two extremes; there
will be many intermediate cases which will be very difficult
to quantify without a full finite element analysis.
4.5.2
Stress at failure
Pressurised equipment exposed to fire will experience a steady
increase in temperature. This may result in a time-varying
stress (see Section 4.5.1), and we expect that once this
time-dependent maximum stress equals the material strength,
the equipment fails. However, the material strength is itself
decreasing in time as the material weakens with increasing
temperature.
For carbon steel vessels or pipes built according to the usual

codes, with a burst pressure at normal temperatures some 2.5
to 4.0 times the maximum working pressure, and operating at
the maximum working pressure, the reduction in strength is
such that failure would be expected at around 500 to 550 C if
thermal stresses are negligible. If the vessel is operating
below its maximum working pressure, the failure temperature is
correspondingly higher, for example, if the pressure is 50% of
the maximum working pressure, failure might be expected at 550
to 600
C.
The time for failure to occur depends on the severity of the

fire, the extent and type of fire protection, and the pressure
response of the vessel or pipework (including the blowdown
system). This can vary between a few minutes and a few hours,
and consequently the relevant strength criterion is more
appropriately taken as the creep rupture strength, rather than
the short term tensile strength (as the time to rupture goes
to zero the creep rupture strength becomes equal to the short
term tensile strength; the latter is just a particular case of
the former).
is important to realise that the use of creep rupture
stress does not imply the presence of significant creep strain
at failure.
For the time periods of interest, the creep
strain is likely to be small, but the creep rupture stress may
well be significantly lower than the short term tensile
It
33
stress.
SECTION 4
Ignoring this may therefore result in an overestimate
of the time to failure.

For a given vessel or pipe made of a given material at a fixed
internal
more
and
pressure
(or,
stress)
accurately,
temperature, the time to failure due to creep rupture can be
determined by the Larson-Miller method (see, for example
reference [32] and [33]).
This method
is based on the assumption that creep is a rate

process governed by an Arrhenius-type equation and the
experimental observation that time to rupture is inversely
The method is generally
proportional to creep strain rate.
expressed in the form of nomograms.
For a vessel or pipe exposed to fire, the temperature will

not, however, be constant in time. While there is no properly
validated way to account for this, a common and acceptably
accurate way [33] is to use the Robinson Method (34] (or Life
Fraction Rule). This is based on the assumption that the time
to failure resulting from the overall pressure-temperature
history of the vessel can be related to experimental failure
times under particular values of pressure and temperature for
a vessel of similar construction.
In view of the complexity of creep rupture calculations, the

short term tensile failure criterion is often used, in which
failure is presumed once the stress level reaches the short
term ultimate tensile strength.
The overestimate of time to failure is probably not too
important if the rate of temperature rise of the vessel or
In this case, the final stages of the failure
pipe is high.
since strength
process are likely to be extremely fast,
decreases very rapidly with temperature when the material is
hot.
However, for slower temperature rises, for example where
failure takes an hour or more, the creep rupture stress
criterion should be used unless one is sure that this is not
necessary.
4.6
Summary
In the scenario chosen for this study, fire attack occurs
directly or indirectly on a vessel or pipework which may or
may not be undergoing blowdown. The physical processes which
take place are heat transfer to the exterior of the vessel,
through any insulation to the interior wall and from there to
the vessel contents. The temperature of any unwetted wall
rises very rapidly, the wetted wall less so, the fluid
temperatures increase, and fluid boiling occurs, resulting in
an increase in the internal vessel pressure.
34
SECTION 4
The individual steps in this overall process are relatively

well understood, and for a single phase one component fluid it
is not too difficult to derive expressions relating vapour
Not all of these are
pressure to the incoming heat flux.
deterministic. However, there appear to be well established
which
describe
the
behaviour
empirical
relationships
adequately.
If the fluid in the vessel is two-phase, or it develops
two-phase behaviour in its temperature cycle, then it becomes
more difficult to predict heat transfer.
In particular,
interphase heat and mass transfer is complex, the boiling
regime is different, and vessel wall temperatures will not be
the same.
Virtually all fluids to be found on an offshore platform will
be in this latter category, and in addition will be
This causes
multicomponent, rather than simple materials.
more serious problems, not least ofwhich is the fact that the
such as
heat
transfer
empirical
relationships
through
convection are derived for single components, and their
applicability for this case is uncertain.
is
In addition, if removal of vapour through blowdown
the
occurring, it is likely that fractional distillation of
liquid will occur, removing the lighter components and
concentrating the heavier ones, and therefore reducing the
rate of pressure rise.
The lack of temperature-dependent
physical and thermodynamic properties for these materials,
which causes modelling problems, also makes a rigorous
theoretical analysis significantly more difficult.
Nevertheless, if the required properties were determined, the

thermodynamic equations of state of the fluid could be
calculated, and with sufficient computing power it should be
possible to carry out the task with some precision.
Thus, it appears that the physical processes involved in the
fire attack and response of vessels and lines are relatively
well understood, and that satisfactory relationships between
attack and response can be derived. Models to describe vessel
behaviour are, at least, feasible.
4.7
Sources of equations
There are four main references in which most of the equations
needed for modelling the physical processes described here can
be found. The most complete is reference [35]; two others
which are less complete are reference [l8J and [36].
The
fourth, which applies to blowdown as opposed to fire loading,
is reference [37].
35
SECTION 4
Reference
good for:
[35]
and
[36]
are useful for:
radiation heat fluxes;

natural convection heat transfer coefficients.
Reference
is
mass and energy balances in the liquid and vapour

contents of the vessel or line;
transient heat conduction through the wall (though
horizontal variations are neglected).
References
[18]
[37 is useful
for:
the mass flow rate for venting of a vapour with

without small quantities of suspended liquid drops.
36
or
5.
SECTION

Because
of
the
of
complexity
the
physical
processes
(see
Section4), all of which are coupled, most predictive models

of the thermal response to Lire loadings are computer-based.
The main source of information on current capabilities has
been obtained by direct contact with some organisations (see
Appendix B), and by conducting a computer-based literature
survey.
Itis
important to recognise that this list of models is not

all-inclusive. For example, there are some models which touch
on the area of interest in a peripheral way, and have been
There may well be other, more
ignored for that reason.
which
have
important models,
escaped our search.
In this context it is noteworthy that the fire response models

identified are all based on work for the onshore LPG industry.
the application is primarily for propane or butane at
ambient temperature and at its saturated vapour pressure - the
vessels for such storage generally have rather lower pressure
ratings than those found offshore.
Thus,
The LPG models were never intended for use in predicting the
response of a vessel on an offshore platform, and as might be
expected, their predictive ability falls short of the
requirements for that application. However, the models are a
good starting point, and we first need to assess how well they
perform in their intended role.
5.1
The LPG fire response models
Six computer programs for simulating the thermal response of

LPG vessels to fire loading have been identified:
5.1.1
ENGULF II from AEA Technology Safety & Reliability

Directorate;
HEATUP from Shell Research Ltd;
TCTCM from Queen's University Ontario;
PIA from SINTEF Applied Thermodynamics Division;
PLGS-I from the University of New Brunswick;
VT*VESSEL from VERITEC.
Features of the LPC response models

The principal features of the computer programs are summarised
in Table 5.1, with a key given in Tabie 5.3 (see also Table C2
of Appendix C). The table reveals the following:
There is an almost universal ability to simulate the
response of vessels to a fire of reasonably uniform flux not
exceeding 100 kW m2 (i.e. a totally engulfing pool fire), but
37
SECTION
no proven ability to model non uniform fluxes or fires of

greater heat flux density;
There is a widespread ability to simulate radiative and

convective heat transfer from the fire to the wall of the
vessel;
There is some ability to simulate radiative and convective
heat transfer to the atmosphere from the vessel;
There is
a widespread ability to handle all orientations
of vessels;
There is
a widespread ability to simulate two-dimensional,
but not three-dimensional heat conduction in the wall of the

vessel, including insulating layers;
There is a universal ability to simulate two-phase single
component fluids in the vessel;
There is a widespread ability to simulate convective heat

transfer to the vapour in the vessel and convective and
boiling heat transfer to the liquid in it;
is
There
a common tendency to
use
semi-rigorous
and
for
the
vessel contents
thermodynamics
phase equilibria
with perfect gas behaviour for venting vapour, which is
thermodynamically inconsistent (and can lead to large errors);
There is a very patchy ability to simulate all possible

modes of pressure relief valve operation and venting from the
vessel;
There is very little attempt to predict stresses within

the wall of the vessel, or to predict vessel failure times or
modes. There is also no agreement on the failure criterion to
be adopted;
No models
identified.
5.1.2
for
the
fire
response
of
pipes
have
been
Validation of the LPG response models

There has been considerable experimental validation of the
programs, largely with vessels containing propane. The most
complete tests have been carried out by:
Federal Institute for Materials Research & Testing (BAN);

UK Health & Safety Executive, in conjunction with Shell
Research Ltd;
Queen's University Ontario;
University of New Brunswick.
38
SECTION
Detailed information on these research programmes is given in

Appendix
B.
Their principal
features are now discussed
in
turn.
BAM have conducted four series of tests on 4.85 cubic metre

horizontal cylinders containing liquid propane, generally
One series was for an
totally engulfed in pool fires.
unprotected vessel, one for an insulated vessel, one for a
vessel protected by a conventional water spray and one for a
vessel protected by an upgraded water spray.
No complete
program has yet been compared with the experimental data,
however.
The UK Health & Safety Executive and Shell Research Ltd have
used the HSE facilities at Buxton to validate both ENGULF II
(AEA Safety & Reliability Directorate) and HEATUP (Shell
Research Ltd).
Experiments have been conducted on 0.25, 1.0
and 5.0 tonne tanks containing propane in totally engulfing
pooi fires.
Queen's University Ontario have validated their program TCTCM
using experimental data from full scale and one fifth scale
tests on tanks apparently containing propane. The tests were
conducted
by the Transportation Development Centre of
Transport Canada, for whom TCTCM was developed initially.
The University of New Brunswick have conducted a series of
experimental tests on a 37.4 litre horizontal cylindrical
vessel partially filled with refrigerant Rll to validate their
program PLGS-I.
Multiple heater elements were used to
simulate the effect of an external heat load and it is
claimed that totally and partially engulfing pool fires and
jet fires can be simulated by appropriate use of the heater
elements.
It is not clear, however, whether the spatially
non-uniform heat fluxes occurring in partially engulfing pool
fires and jet fires can indeed be simulated in this way,
either in terms of heat flux density or the flux distribution.
It emerges from these sets of experimental tests that:
There are extensive validatory data for cylindrical

vessels in totally engulfing pool fires, but none for vessels
in partially engulfing pool fires or in jet fires;
With a totally engulfing pool fire loading, the computer

codes are capable of good predictions for the pressure and
wall
within
the
and
the
vessel
vessel,
temperature
temperature, until such time as a relief valve opens, venting
the hydrocarbon inventory.
The flow through the valve and hence the response of the
vessel and its contents after the valve has lifted cannot
always be predicted accurately, and there is often no attempt
to predict the time to failure.
39
SECTION
There is a limited amount of validatory data on vessels

with wall insulation and water spray protection;
As expected, there have been no validatory

vessels containing multicomponent fluids;
There are
tests
on
no validatory data for pipework exposed to fire
loading.
5.2
Blowdown models
In addition to the LPC fire response models, there are several

codes which simulate the blowdown of a vessel. These share
many features with the programs for simulating thermal
It is claimed that SAFIRE
(N.B.
response to fire loading.
can simulate thermal response to fire loading, but it does so
in such a crude and unvalidated manner that it was not
appropriate to include it in Section 5.1.)
Four blowdown models have been identified
5.2.1
VENTFLO from British Gas London Research Station;

SAFIRE from Fauske & Associates;
BLOWDOWN from Imperial College;
PROCESS from SimSci.
Features of the blowdown models
The principal features of the four blowdown programs are shown

in Table 5.2 (key given in Table 5.3), and can be summarised
as follows
There is a widespread ability to simulate convective heat
transfer to the atmosphere from the vessel. However, the less
general case of heat transfer to the sea is not covered;
There is a universal ability to handle both horizontal and

vertical vessels;
There is a universal ability to simulate one-dimensional,

but not two-dimensional or three-dimensional heat conduction
through the wall of the vessel;
There is a universal ability to simulate blowdown of

two-phase multicomponent fluids but very little ability to
simulate heat transfer to the contents properly or to handle
multicomponent vapour-liquid equilibria in a rigorous manner;
There is a widespread ability to simulate venting of
vapour but very little ability to simulate multi-phase
venting;
40
5.2.2
SECTION
Validation of the blowdown models

The blowdown models have received far less validation than
those for fire response. Indeed, only one has been tested in
this way.
Experimental tests and validation of the BLOWDOWN model have
been carried out by Imperial College and Shell Research Ltd.
Imperial College have conducted a series of blowdown tests
using a 36.4 litre horizontal or vertical cylinder containing
An
nitrogen or a mixture of nitrogen and carbon dioxide.
extensive additional series of tests has been conducted by
Shell Research Ltd using a 2.85 cubic metre vertical cylinder
containing methane, nitrogen, and mixtures of methane/propane
and methane/propane/carbon dioxide.
a
The
results
of
these
tests
showed
that
rigorous
thermodynamic treatment can give good agreement between
prediction and experiment for the wall temperature, and the
composition, pressure and temperature of a multicomponent
fluid.
There have also been experimental tests conducted on the

blowdown of pipelines but, with the exception of one brief
paper (reference [38]), these are as yet unpublished.
5.3
Summary of current capabilities
The status of current predictive modelling capabilities in

this area can be summarised as follows:
We have not been able to identify a validated model which

can be applied to the fire response of a vessel or pipework
under blowdown conditions. However, models do exist for the
separate processes of fire attack on a vessel, and for the

fluid response to blowdown.
The validated models for fire response which have been
identified are all based on fire attack of a vessel containing
LPC (more strictly, propane or butane) with a simple venting
There has been extensive validation from large scale
system.
tests in fully engulfing pool fires.
The fire response models can be used with confidence to

the pressure and temperature in and the wall
predict
temperature of a vessel containing a one-component but
possibly two-phase hydrocarbon in a fire of reasonably uniform
flux not exceeding 100 kW m2, until such time as a relief
The flow
valve opens, venting the hydrocarbon inventory.
through the valve and hence the response of the vessel and its
contents after the valve has lifted cannot always be predicted
41
SECTION
accurately, however, and the modelling of time to failure has

not been examined thoroughly.
There is no validated code
which can be applied to multicomponent inventories or other
types of fire loading.
several models dealing with blowdown are

Although
described here, only one has been verified. For this case a
rigorous thermodynamic treatment can give good agreement
between prediction and experiment for the wall temperature,
and the composition, pressure and temperature of a (rather
simple) multicomponent fluid. This particular model does not
include fire loading.
The ranking capabilities of the existing models may

some value, despite the severe limitations.
42
be of
6.
SECTION
Sections 4.6 and 5.3 indicate the state of knowledge of
physical processes which occur when a vessel or pipe
exposed to a fire loading, and the current capability
models to predict the response.
It is now necessary
examine once again the situation which is to be modelled,
to assess whether the current capabilities are sufficient.
6.1
the
is
of
to
and
The requirements of predictive models

The scenario which should be simulated is one in which a pipe
or vessel, with or without fire protection, is subjected to
uniform or non uniform heat flux. The vessel or pipe contains
a fluid, which is almost certainly multicomponent, and may be
undergoing simultaneous venting from the headspace or through
removal of liquid.
Thus the requirements of a model for the fire response of a
vessel or pipeline on an offshore platform are as follows:
The type of fire attack must include pool fires and jet
fires, with full or partial impingement and heat transfer by
radiation and convection, or radiation alone.
The model must be applicable to both vessels and pipework.
Various types of fire protection should be modelled, in

particular water spray and passive fire protection.
The fluid in the vessel or pipe must be multicomponent.
The thermodynamics of liquid-vapour equilibria need to be

treated with appropriate rigour.
Various types of pressure relief, including vapour

liquid blowdown must be included.
The models should be able to predict:
Wall temperatures
Temperature and pressure of the contents
Failure time
The models should be adequately validated

separate from those used in model development.
6.2
and
using
tests
Physical processes
While the physical processes taking place are relatively well
understood, there are two main areas which are likely to cause
Both are connected with the fact that the fluids
problems.
involved are multicomponent.
43
SECTION
Firstly, the physical relationships between cause and effect

their
and
require rather detailed physical properties,
variation with temperature. It is not appropriate (or in some
cases meaningful) to describe a multicomponent fluid in these
terms.
In some cases, knowledge of the composition would
allow appropriate properties to be calculated, but there are
other cases where more work is needed to resolve the
descriptive difficulty.
The second main problem is in knowing the applicability of
some of the empirical relationships, for example convective
heat transfer and film boiling. The relationships are largely
derived for single component fluids:
work is needed in the
first place to see whether they can be used, and if not, to
derive alternative expressions.
6.3
comparison of the capabilities of the current fire response

models (Section 5.3) with the requirements (Section 6.1)
reveals several major gaps, as follows:
1.
The programs do not simulate

non-uniform fire loadings.
2.
Only vessels are currently modelled.
3.
Vessels protected by anything other than insulation,

removal of inventory and external sprays are not
simulated, and even in these cases the capability is
response
to
spatially
limited.
4.
The programs do not allow

inventories to be modelled.
5.
Venting of anything other than vapour is not included,

and often there is only a rudimentary model for the
operation of
6.
multicomponent
hydrocarbon
a pressure relief valve.
The programs tend to use semi-rigorous thermodynamics and

phase equilibria to model the vessel contents but
simplified and often inconsistent, for example perfect
gas, thermodynamics to model venting.
7.
Mainly the models do not predict time to failure, partly

because of (5) and (6), and partly because there is
little or no agreement on what failure criterion is
appropriate.
8.
The programs are validated, if at all, only for the

response of a (perhaps insulated) single vessel in a
totally engulfing pool fire.
44
SECTION
9.
The response of interconnected vessels and lines is not

covered.
10.
It is not possible to simulate the response of a vessel

or line which has already suffered damage, such as
Such
deformation, perhaps resulting from the fire.
damage is particularly likely on topsides where other

vessels are close by, if not physically connected by
relatively short lengths of line.
11.
Although existing models can deal with heat losses to the

atmosphere, none simulate convective heat transfer to the
sea.
In order to decide on the future work which is needed, it is

important to examine how relevant and important the gaps are,
and to determine whether they are caused by a lack of
In some cases it is difficult to estimate the
knowledge.
of
effects which are currently not modelled
importance
Nevertheless, having prioritised the gaps, the
correctly.
question of filling them must then be addressed.
6.3.1
Non uniform heat fluxes

As discussed in Section 3.4, the most likely fire scenario for
a vessel on a platform is that of a non uniform flux, for
Obscuration
example a partially engulfing pooi or jet fire.
of the fire by intervening vessels or walls is made more
probable by the congested nature of the platform layout.
Section 4.2 showed that the ability of a model to handle non
uniform fluxes is critically dependent on the dimensionality
of the model and whether the vessel wall is divided into
Some of the existing fire. response models claim to
zones.
deal with a non uniform fire loading, and some certainly have
the appropriate geometric basis. However, there has been no
validation, and any claims must be treated as tentative.
Of course, it may be the case that adequate (although probably

pessimistic) results could be obtained for a non uniform
loading by using existing models, with the maximum flux
density. Another possibility is that some intermediate level
of heat flux would approximate the situation. However, it is
not clear what level of errors would be introduced.
Validation of some existing models with jet fires, for
example, would be a useful step forward which might allow the
importance of this factor to be assessed.
6.3.2
Models for pipework
45
SECTION
The LPG models carry with them the assumption that the
containment is a vessel rather than a pipe.
It might be
that
a
is
a
thin
line
vessel, in which
argued
simply
very long
case a pipe could indeed be modelled, particularly in a
uniform heat flux. However, it is not reasonable to expect
the whole of a pipe to be exposed in this way, and the
problems of modelling partial engulfment (see Section 6.3.1)
would apply even more.
The question of pipes raises the topic of end effects, which
is another unknown area, and applies particularly to vessels
with a small length to diameter ratio.
6.3.3
Fire protection
The ability of a model to deal with a variety of protection
methods is important because this would allow comparison of
the different systems under varying types of fire attack, and
could improve the emergency response and aid selection of the
most appropriate protection methods during the design phase of
a new platform.
Different fire protection methods will have a significant
effect on the behaviour of vessels and pipes in fire loading.
For example, a cementitious coating will greatly increase the
heat capacity of the vessel.
The capabilities and degree of validation of existing models
in this respect vary considerably. At the simplest level, a
fire protective system can be modelled as
a thermal
but
in
is
far
more
the
behaviour
resistance,
practice
complex.
As outlined in Section 3.6, there are many different forms the

One of the simplest is a thermal
protection may take.
barrier, which may be a firewall or a coating on the vessel.
Firewalls can be made from a cementitious material, which of
course contains a great deal of water. This is driven off by
the heat flux, so the properties of the barrier vary
considerably with temperature and time.
Though modelling of a water spray system is not included in

the current models, separate models dealing with this aspect
do exist.
6.3.4
Multicomponent fluids
As discussed in Section 3.3, most of the fluids to be found
offshore are multicomponent, for example condensates and
crudes.
The response of a multicomponent fluid to the
incoming heat flux and the blowdown process will be
significantly different from that of a single component. For
example, if blowdown results in vapour being removed, there
46
SECTION 6
will be fractional distillation of the liquid, reducing the

rate of pressure increase and increasing the time to failure.
It might be possible to use existing LPG models to simulate
the response of single heavier hydrocarbons, provided the
However,
necessary physical property data were available.
there is no validation, and any results would have to be
treated with care.
The current fire response models do not allow multicomponent
fluids to be simulated, because it is not in their scope.
However, this capability is present in blowdown models and a
relatively simple two-component fluid has been validated in
one of them.
6.3.5
Venting
Depending on the vessel contents, geometry and blowdown
method, the material passing through the vent line may be
mostly gas, mostly liquid, or mixtures, as described in
Section 3.4. The precise nature has a profound effect on the
venting behaviour, and can affect the behaviour of the vessel
considerably.
3lowdown of a vessel or pipework is in itself a very complex
area, and it is probably not necessary to include a full
description of it as part of the fire response. Nevertheless,.
some aspects are highly relevant. One example is the likely
distillation of the fluid leading to a lower rate of pressure
rise, as mentioned above. Another is the increased risk of
failure through reducing the level of the liquid in a vessel.
There are other cases where a rigorous description of blowdown
may be far less important. For example, with vapour venting,
the relative amounts of liquid and gas change only slowly, and
changes in the headspace composition are largely irrelevant to
the shell or liquid temperatures. The main effect is on the
overall pressure change with time, and this might be modelled
effectively by treating it as a simple leak.
a full description of blowdown is probably not necessary

for a fire response model. However, the current capabilities
are not adequate.
Thus,
6.3.6
Thermodynamic treatment
Simplifications to the thermodynamics are often used to give
acceptable runtimes to the codes. In some cases the errors
introduced are negligible, for example if the perfect gas
assumption is made about venting, but there is actually a
small quantity of entrained liquid droplets. Increasing the
47
SECTION
amount of entrained liquid would introduce far more
serious
errors.
Rigorous thermodynamic treatment is present in the validated
blowdown model, but not in the current fire response codes.

The magnitude of errors arising from attempting to model, for
example, multicomponent fluids using simplified thermodynamics
is not known. If a full treatment is needed, the penalty will
be greatly increased runtimes.
To give estimates of the importance of such effects would

additional
either
or
require
information,
experimental
theoretical.
One possibility might be to start from the
existing validated blowdown model and introduce thermodynamic
simplifications of the type which are common in the fire
response models, as this would give an idea of the trade-off
between accuracy and speed.
6.3.7
Vessel failure
The likely time to failure of the vessel is perhaps one of the

more important questions that the models should address.
Current predictions of the time to opening a simple valve are
quite good, but events beyond that are far less certain, and
for a vessel undergoing liquid removal there is no predictive
ability at present.
The mechanism of vessel failure is relatively well advanced,

as discussed in Section 4.5.
However, the failure criteria
which should be adopted are subject to uncertainty, and there
has been very little experimental work which has sought to
provide an answer to this question.
more severe problem in attempting to include failure in a

response model is that the thermodynamic treatment of the
liquid/vapour equilibria and venting is less than rigorous.
This is discussed more in Sections 6.3.5 and 6.3.6 above,
together with possible ways of improving the situation.
6.3.8
Validation
hardly needs to be stated that the validation of any
Validation is particularly
predictive model is essential.
in
the
case
of
fire
response because it is a complex
necessary
with
a
number
of
interconnected
factors.
topic
large
It
6.3.9
Interconnected vessels
None of the existing models deal with interconnected vessels

and pipework (again because it is not in their scope).
General guidelines can not be giyen, because the importance of
48
SECTION 6
this effect will depend critically on the exact details of

geometry, contents and heat flux.
6.3.10
Partial damage
Partial damage of vessels and pipework may be an important
area, but it is probably less urgent. Again, no general guide
can be given.
6.3.11
Heat transfer to the sea

This is of importance for heat transfer from lines such as
risers, and might be relatively easy to implement in existing
models.
49
CURRENT POSITION
7.
SUARY OF THE
SECTION
CURRENT POSITION
Work Package FR7 has considered the state of knowledge of the

response of pressurised vessels and pipework containing
hydrocarbons when exposed to fires, and the ability of
mathematical models to simulate the processes.
In summarising the current position, a clear distinction has

to be made between
the physical understanding and the
modelling capabilities, because accurate predictions from the
models can generally only be expected if their physical basis
is correct.
The physical processes

The current position is that the physical processes describing
heat transfer through the wall of the vessel or pipework to
the fluid are relatively well understood. If the fluid is a
single component, and venting is only of gas, there are
relationships available which make an adequate description of
the temperature and pressure response of the vessel feasible.
However, if the fluid in the vessel is two-phase, or it
becomes two-phase during venting, then it is more difficult to
predict the vessel response. In particular, interphase heat
and mass transfer is more complex.
If the fluid is
multicomponent, a far more rigorous thermodynamic treatment is
necessary, with a concomitant increase in the computing power
required.
Modelling capabilities
The mathematical model needed to describe fully the effect of
fire on a vessel or pipe under blowdown conditions is highly
complex, and has not yet been attempted. However, parts of
the model have been developed and validated for related
purposes. There are several computer codes which deal with
the fire response of LPG vessels with simple PSV venting, and
there are some which cover hiowdown in the absence of fire
attack.
The most useful fire attack models have been extensively
validated for propane or butane using large scale tests in
fully engulfing pooi fires, and can be used with confidence to
and the wall
predict the pressure and temperature in,
temperature of, a vessel containing a one-component but
possibly two-phase hydrocarbon in a fire of reasonably uniform
flux not exceeding 100 kW m2, until such time as a relief
valve opens, venting the hydrocarbon inventory.
There is no validated fire attack code which can be applied to
multicomponent inventories or other types of fire loading.
The flow through the relief valves and hence the response of
the vessel and its contents after a relief valve has opened
cannot always be predicted accurately.
50
CURRENT POSITION
SECTION
One of the most important questions which should be addressed

by a model is the likely time to failure in conjunction with
the failure mode of a vessel or pipe. This is not generally
attemptd by the current models, and has not been examined
thoroughly in the investigations to date.
single verified model of blowdown has shown that rigorous

thermodynamic treatment of the vessel contents can give good
agreement between prediction and experiment for the wall
temperature of the vessel, and for the composition, pressure
and temperature of a (rather simple) multicomponent fluid.
However, this particular model has not been verified for a
fire loading.
If we compare the capabilities of the existing response

models, designed for LPG tanks, with the requirements of
simulating a vessel or pipe undergoing fire attack on an
offshore platform, there are several gaps, which are discussed
within the study
1.
The programs do not simulate response to the spatially

non-uniform fire loadings which can be expected on an
offshore platform.
2.
Vessels are currently modelled, but not pipework.
3.
The response of vessels protected by anything other than

insulation, removal of inventory and external sprays is
not simulated, and even in these cases the capability is
limited.
4.
The programs do not allow multicomponent

inventories to be modelled.
5.
Venting of anything other than vapour is not included,

and often there is only a rudimentary model for the
operation of a pressure relief valve. Blowdown is not
included in the fire response models.
6.
The programs tend to use semi-rigorous thermodynamics and
hydrocarbon
phase equilibria to model the vessel contents but

simplified and often inconsistent, for example perfect
gas, thermodynamics to model venting.
7.
Mainly the models do not predict time to failure, because

of (5) and (6), because there is little or no agreement
on what failure criterion is
thermal stresses are ignored.
8.
appropriate,
and because
The programs are validated, if at all, only for the

response of a (perhaps insulated) single vessel in a
totally engulfing pooi fire.
51
CURRENT POSITION
9.
SECTION
The response of interconnected vessels and lines is not

covered.
10.
It is not possible to simulate the response of a vessel

or line which has already suffered damage, such as
Such
deformation, perhaps resulting from the fire.
damage is particularly likely on topsides where other
vessels are close by, if not physically connected by
relatively short lengths of line.
11.
Although existing models can deal with heat losses to the

atmosphere, none simulate convective heat transfer to the
sea.
In addition to addressing these questions, models for the fire

response of a vessel or pipe in a real process train
undergoing blowdown require a great deal of information to
describe this highly complex situation in sufficient detail.
At perhaps the most basic level, the process equipment has
to
be specified sufficiently well to allow prediction of flows in

a system which may contain several vessels, valves and lines.
The effect of blowdown then has to be examined - the pressure
history of the section of pipe or the vessel undergoing fire
attack will depend on the geometry of the system and whether
venting is taking place from the vapour side or the liquid.
For
Having done this, the surroundings must be considered.
example, the position of fire walls and reflective surfaces
can be important.
The incident heat flux depends, of course, on the flame
position and shape. This will probably vary with time, in a
way which can be difficult to predict, particularly in the
confined spaces of a typical offshore platform.
Even with a steady, fully engulfing fire, the heat transfer to
the vessel or pipe can change considerably with time, for
example if the surface becomes covered by soot or charred
paint. The time- and temperature-dependent properties of such
a layer have to be considered. In addition, there may be a
water deluge or passive fire protection.
The relationships for heat and mass transfer to the fluid
fluid
require a great deal of information about the
For
a
it
is
properties.
single component hydrocarbon
to
obtain
and
but
relatively easy
physical
thermodynamic data,
for the type of gross mixtures expected offshore this is far
more difficult, particularly as the composition may change
considerably with time.
The
current
models
have
considerable
as
shortcomings
tools
for
use
offshore.
In
there
is
a
predictive
particular,
52
CURRENT POSITION
SECTION
general lack of validation, particularly where programs have

been extended to cover situations they were never intended to
In addition, there is insufficient information
encompass.
about the codes from which their applicability and accuracy
could-be judged. Any models designed to simulate the response
of vessels and pipes on an offshore platform should remedy
these deficiencies.
53
8.
SECTION
The areas of uncertainty fall into two categories - those
where there is
insufficient knowledge of the physical
situation - and those where there are gaps between the
the
of
models
and
capabilities
existing
predictive
requirements.
The physical processes

The physical
processes which take place when a vessel or

pipework
exposed to a fire are well understood, but this
does not mean that they are universally applicable.
For
of
the
such
as
heat
example, many
empirical relationships,
transfer by forced convection, or heat transfer in nucleate
boiling, have been derived for simple fluids, and may require
modification when applied to mixtures.
is
Modelling capabilities
The LPG fire response models fulfil their aim well, and their
deficiencies when compared with the requirements of simulating
a vessel or pipe on an offshore platform are largely a result
of the far greater complexity of the latter situation,
There is no doubt that a considerable amount of work would be
needed to develop a model (or models) which could simulate all
the possible scenarios for fire attack offshore.
Equally,
this is probably not necessary.
As a first step, the scenarios which pose the greatest hazard

need to be identified. For example, can partial jet fire
attack be more severe in terms of vessel response than a fully
engulfing pool fire
There is, of course, a need to do case by case analysis rather

than search for generalisations, and it is possible that
careful work with the existing models could be useful. Gross
assumptions in the current codes may reduce the accuracy of
predictions, but not necessarily prevent them from being used
in a ranking exercise.
Another example of using the existing models is to examine
whether the risk of rupture of a thin walled low pressure
vessel is higher than that of a thicker high pressure vessel,
and how important insulation can be. Almost all of the fire
response models stop short of predicting failure time, and
consider only hoop stresses caused by the internal pressure.
However, the output from the computer codes, in terms of wall
in
temperature distribution and pressure history could
principle be used with a nonlinear finite element program to
determine thermal stress and creep, and hence the failure mode
and time.
54
SECTION
A major area of uncertainty arises from simplifications
in
existing computer codes. For an unprotected vessel or pipe

exposed to a substantial heat flux, the failure time is
measured in minutes, and improving the prediction accuracy is
unlikely to change the emergency response. However, if the
vessel is insulated, or the heat flux is lower, the time to
rupture will be correspondingly longer, and the accuracy of
the predictions becomes far more important.
an example, the LPG fire response models use semi-rigorous

thermodynamics for venting and for liquid-vapour equilibria,
As
because this reduces runtimes considerably.

The existing
validated blowdown model, on the other hand, incorporates a
full thermodynamic treatment. It is possible that this latter
code can be used to examine the sensitivity of predictions to
such simplifications.
Uncertainty also arises in areas where effects thought to

relevant have not yet been modelled. Because the situation
so complex, it is difficult to estimate the importance
these factors. A good example is the effect of blowdown
the pressure of a vessel containing a multicomponent fluid.
be
is
of
on
The existing fire response models do not cater for blowdown,

but instead incorporate a simple PSV. It might be possible to
treat blowdown as a simple leak, and incorporate this in
current models. However under blowdown in the presence of a
fire load, a multicomponent fluid will tend to distill, with a
reduction in the amount of light fraction. This may affect
the pressure-time behaviour significantly.
One way of examining the importance of this effect might be to

a
simulator
use
commercial distillation
appropriate heat fluxes and offtake rates.
program,
with
The relationships for heat and mass transfer require a large

amount of physical and thermodynamic data for the fluid, but a
gross mixture such as condensate or live crude can not be
found in the standard reference sources. Some bulk properties
and methods of
such as thermal conductivity or density,
approximating them from composition may be available within
the petroleum processing Industry, but others, such as
critical heat flux as a function of pressure, may have to be
found experimentally.
Although the physical processes, and hence the mathematical
relationships, are understood rather well, there are many
cases where accurate prediction of the response of a vessel or
pipe to an external heat flux requires the correct choice of
process. An example of this is where convective heat transfer
between liquid and vapour can be natural or forced, depending
on the vapour velocity. Another is where the liquid becomes
55
AREAS
OF UNCERTAINTY
SECTION
stratified. These could be regarded as minor details when

considering the scenario as a whole, but could have a
significant effect on the accuracy of any predictions.
The range of applicability, and the limitations of any

predictive model have to be well defined during the design
stage of any new computer code, as they will be limited by the
physical model which is chosen. For example, if a non-uniform
heat flux is thought to be, i.mportant, the model has to be at
least two-dimensional.
The limitations must be well documented, to limit the use of
models in situations which are inappropriate, and to guide
modifications.
There must be a clear distinction drawn between experimental
data used
in program
development, and that used for
validation.
56
REFERENCES
REFERENCES
1.
(Anonymous): Guide for pressure-relieving

systems. API RP521 2nd Edition (1982)
2.
(Anonymous): Sizing, selection and installation of pressure

relieving devices in refineries. API RP520 5th Edition (1990)
3.
4.
Siegel & Howell: Thermal radiation heat transfer. NASA SP-164

vols I, II and III Office of Technology Utilization (1968,
1969 and 1970)
5.
R H
6.
N P Chopey & T G Hicks (editors): Handbook

Engineering Calculations, McGraw-Hill (1984)
7.
J C Richmond: Measurement of thermal radiation properties of

materials. High temperatures - high pressures 11, 355-381
and depressuring
Casteiger: Failure analysis of high-pressure separators

during platform fire conditions, 7th International Conference
on Offshore Mechanics & Arctic Engineering, Houston (1988)
Perry & C H Chilton

Handbook, McGraw-Hill (1973)
(editors):
Chemical
Engineers'
of Chemical
(1979)
8.
H J Ivey & D
J Morris: On the relevance of the vapour-liquid

mechanism
for subcooled boiling heat transfer at high
exchange
pressure. Report AEEW-R137, UKAEA, Winfrith (1962)
9.
J Ivey & D J Morris: The effect of test section parameters

on saturation pooi boiling burnout at atmospheric pressure.
4th Natl Heat Trans Conf, Buffalo, AIChE reprint 160, Chicago
(1962)
10.
0 Happel & K Stephan: Heat transfer from nucleate to the

beginning of film boiling in binary mixtures. Heat transfer,
Proc
Heat Transfer Conf, 5th meeting, 4, 340-344 (1974)
mt
11.
C J Capone & E L Park: Film boiling of Freon 113, normal

pentane, cyclopentane and benzene from cylindrical surfaces at
moderate pressures.
mt j Heat Mass Transfer 22, 121-129
(1979)
12.
C T Sciance, C P Colver & C M Sliepcevich: Nucleate pool

boiling and burnout of liquefied hydrocarbon gases. Chem Eng
Frog Symp 63, 109-114 (1967)
13.
G
Hetsroni
(editor):
Hemisphere (1982)
14.
S Y Ahmad: Fluid to fluid modelling of critical heat flux: a

mt j Heat Mass Transfer 16,
compensated distortion model.
641-662 (1973)
Handbook
57
of
multiphase
systems,
REFERENCES
15.
K Srinivasan & M V Krishna Murthy: Determination of the

bulk-saturated liquid condition for maximum heat fluxes at
boiling crises. mt J Heat Mass Transfer 29, 1963-1967 (1986)
16.
M T
Cichelli & C F Bonilla: Heat transfer to liquids boiling

under pressure. Trans AICHE 41, 755-787 (1945)
17.
18.
NU
19.
C Soave, Chem Eng Science
20.
K E Bett, J S Rowlinson & C Saville:

chemical engineers, Athione Press (1975)
21.
T-H Chung, M Ajian, L L Lee & K E Starling,
C Weast, M J Astle & W H Beyer (editors): Handbook

Chemistry and Physics, 69th Edition, CRC Press (1988)
Aydemir, V K Magapu, A C M Sousa & J E S Venart: Thermal

response analysis of LPC tanks exposed to fire, J Hazardous
Materials, 20, 239-262 (1988)
23,
22.
of
27, 1197 (1972)
Thermodynamics
md
for
Eng Chem Fund
(1984)
Pederson & A Fredenslund: Chem Eng Science 42,
182-186
(1987)
Jullian, A Barreau, E Behar, J Vidal: Application of the

SBR equation of state to high molecular weight hydrocarbons
Chem Eng Science 44, 1001-1004 (1989)
23.
24.
N Abuaf, 0
25.
C B Wallis,
26.
M N Hutcherson, R E Henry & D
C Jones & B J C Wu: Critical flashing flows in

nozzles with subcooled inlet conditions, ASME J Heat Transfer,
105, 379-383 (1983)
mt
J Multiphase Flow
97 (1980)
6,
E
Wollersheim: ASME J Heat
Transfer, 105, 687,694 (1983)

27.
80,
A van den
Akker,
Snoey
& H Spoelstra:
CHem E Symp Ser
E23 (1983)
28.
J C Leung & M A Grolmes: AIChE
29.
K E
30.
H E A van den Akker & W M
First
&J E
Huff:
J,
33,
524 (1987)
Plant/Operations Progress
8,
40 (1989)
Discharges of saturated and

Prediction of
superheated liquids from pressure vessels.
homogeneous choked two-phase flow through pipes. I Chem E Symp
Bond:
Ser 85, 91 (1990)

31.
Johnson, R E Welch, A N Takata, R W Bruce: Fire induced

ASME 74-PVP-6
stresses in tanks containing liquefied gas.
M R
(1974)
58
REFERENCES
32.
F R Larson & J Miller: A time-temperature relationship for

rupture and creep stresses, Trans ASME, 174 (5) (1952)
33.
R K Penny & D L Marriott: Design for Creep, McGraw-Hill
34.
E L Robinson: Effect of temperature variation

strength of steels, Trans ASME 60 (1938)
35.
P K Ramskill: ENGULF
(1971)
on the creep
II: a computer code to model the thermal

of
a
tank
response
partially or totally engulfed in fire,
HSE/SRD Report R480, HMSO (1989)

36..
37.
G V Beynon, L T Cowley, L M Small & I Williams: Fire

engulfment of LPC tanks: HEATUP, a predictive model, J
Hazardous Materials, 20, 227-238 (1988) (see also: Appendix B,
ff B12);
Haque, S M Richardson, G Saville & C Chamberlain: Rapid
depressurisation of pressure vessels, J Loss Prevention
Process Industries, 3, 4-7 (1990) (see also: Appendix B, ff
MA
B6).
38.
L T Cowley & V H Y Tam: Consequences of pressurised LPG

releases: the Isle of Grain full scale experiments, 13th
International LNG/LPG Conference, Kuala Lumpur (1988)
59
TABLES
TABLES
5.1
Programs for simulating thermal response to fire

loading
5.2
Programs for simulating blowdown
5.3
List of symbols for Tables 1.1 and 1.2
I-i.
0
Ii
g-1
(I)
pi
CD
a5
a7
e7
d5
d7
e2
gi
Validation
e7
ei e3
1
d6
d4
d2
c2
c4
b2
a6
d5
d7
dl
bi
b3
a3?
a5
al
B9
TCTCM
Failure
gi
d6
di
d3?
Vessel
contents
Venting
d2
d4
c3
wail
Vessel
b2
a8
a3?
Vessel
geometry
a2
a4
al
External
B3
ENGULF II
heat flux
Page in
Appendix B
Program
gi
e2
dS
d7
a6
e8
d2
d4
d6
c2
c4
b2
dl
al
a3
a5
a7
B12
FIEATUP
c4
c2
b2
g2
e2
d5
d7
d6
d2
d4
a6
gi
el?
d7
dl
ci
d2
d4
d6
alO
a2
b2
al
B17
Bl4
dl
a5
al
TAC7
PLGS-l
d5
dl
gi
ei?
d6
d2
alO
c2
c4
hi
B29
PIA
fi
d3
dl
f2
d2
b2
a6
ci
c4
hi
al
a3?
B31
VT*VESSEL
TABLES
Program
BLOWDOWN
Page in
Appendix B
B6
External
heat flux
Vessel
geometry
a8
a9
b2
bl
ci
Vessel
wall
VENTFLO
SAFIRE
PROCESS
B20
B22
B27
a8
a8
alO
bl b2
hi b2
hi b2 b3
ci
ci
ci
dl d2 d3
dl d2 d3
Vessel
contents
dl d2 d3
d5 d6 d7 d8
di d2 d3
Venting
el
el e3? e4 el e3? e4
e5? e6?
e5? e6?
Experimental
validation
TABLE 5.2
e4
e3
e5
e6
d5
g3
Programs for simulating biowdown
e2
TABLES
(? implying doubt):
external heat flux:

al
totally engulfing pooi fire
a2
partially engulfing pooi fire
a3
jet fire
a4
remote fire
a5
radiative heat flux from fire
a6
convective heat flux from fire
a7
radiative heat flux to atmosphere
a8
convective heat flux to atmosphere
a9
convective heat flux to sea
alO arbitrary heat flux
vessel geometry:
bi
b2
b3
vertical cylinder
horizontal cylinder
arbitrarily inclined cylinder
vessel wall:
cl
c2
c3
c4
one-dimensional heat conduction through wall

two-dimensional heat conduction through wall
zonal wall model
multilayered wall perhaps including insulation
vessel contents:
dl
d2
d3
d4
d5
d6
d7
d8
two-phase vapour-liquid
single component
multicomponent
radiative heat flux to vapour
convective heat flux to vapour
convective heat flux to liquid
boiling heat flux to liquid
rigorous vapour-liquid equilibrium
venting:
el
e2
e3
vapour venting
perfect gas venting
vapour plus suspended liquid droplet venting
e4
liquid venting
liquid plus vapour bubble venting
flashing liquid venting
on-off model of pressure relief valve operation
specific models of pressure relief valve operation
e5
e6
e7
e8
TABLE 5.3
List of symbols for Tables 5.1 and 5.2
TABLES
vessel failure:
assumed failure criterion
fi
f2
wall stress determination
experimental validation:
tank containing propane in totally engulfing pool
gl
fire
tank containing refrigerant in totally engulfing
g2
pool fire
blowdown of vessel containing various different
g3
fluids
TABLE 5.3 (contd.)
FIGURES
FIGURES
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
The heat balance outside a vessel or pipeline

Heat transfer through the wall - case 1
Heat transfer through the wall - case 2
Heat transfer inside a vessel or pipeline
Mass transfer inside a vessel or pipeline
General boiling curve
Experimental example of boiling curve
Variation of critical heat flux with pressure
Transitions in phase space
Wall stresses in a vessel or pipeline
FIGURES
WATER,,"
DELUCE/
VESSEL
WALL
NCtDENT FLUX
b
HEAT
Figure 4.1
LOSSES
The heat balance outside a vessel or pipeline
FIGURES
.4
RADIAL
CONDUCTION
TO VAPOUR
>14
AXIAL
CONDUCTION
RADIAL
CONDUCTION
TO LIQUID
UNIFORM HEAT FLUX FROM FIRE

Figure 4.2
Heat transfer through the wall
case
I,
CD
rt
rt
CD
H)
rt
NON UNIFORM HEAT FLUX FROM FIRE
RADIAL
CONDUCTION
TO LIQUID
TO VAPOUR
CONDUCTION
RADIAL
tt,t
AXIAL
CONDUCTION
'I,
CIRCUMFERENTIAL
CONDUCTION
_______
U)
t1
FIGURES
Convection to vapour
Radiation to vapour
Radiation to liquid
o Radiation absorbed by liquid
Figure 4.4
Radiation reflected by liquid
Reradiation at wall
Convection to liquid
Heat transfer inside
a.
vessel or pipeline
FIGURES
Mass flux
-
Figure
4.5
Associated heat flux
Mass transfer inside a vessel or pipeline
70
FIGURES
2.5
(.4
4-
0
-c
1,5
0
0.5
1.5
2.5
og (temperature difference, deg K)
Figure
4.6
General boiling curve
'71
FIGURES
300
A
0
250 -
AA
c"J
Toluene [10]
Benzene [11]
Pressure = 0.3
200
MN m 2
A A
x150
100
.0
50
0
50
100
150
200
Temperature difference, deg.
Figure 4.7
Experimental example
7.2.
of boiling curve
250
FIGURES
400
c'J
E 300
x
200
C
-c
C
C-)
00
C)
0
0
0.2
0.4
0.6
Reduced pressure
Figure
4.8
Variation
0.8
(F/Pc)
of critical heat flux with pressure
72
[12]
FIGURES
Liquid
Critical point
uJ
onoblique
(I)
Ui
uJ
Two p h a se
Gas
TEMPERATURE
Figure
4.9
Transitions in phase space
FIGURES
U)
a)
U,
U,
a)
U)
aC
-o
C
U)
a)
Cl,
U,
a)
-e-J
(I)
00
0
U)
a)
U)
U)
I)
L
(I,
I
Figure 4.10
Wall stresses in a vessel or pipeline
'16
APPENDIX A
APPENDIX A
THE FR7 WORK
PACKAGE
BLAST AND FIRE ENGINEERING FOR TOPSIDE STRUCTURES

WORK PACKAGES (WP)
WP No:
FR7
WP TITLE: Thermal response of vessels and pipework exposed to
fire.
WP START DATE:
WP END DATE:
Month 1
Month 6
DURATION:
& Month
Months
OBJECTIVES:
To summarise knowledge on the thermal response of hydrocarbon
containing vessels and pipework to fire effects.
To identify gaps in knowledge.

DESCRIPTION OF WORK:
Vessel fire response is needed to aid decisions on placement,
design, emergency response procedures and fire protection.
Vessel and pipework failure is a possible route to escalation.
Prediction of the fire response of fluid-containing vessels
and pipework has
additional complexities over
that of
structures. The heat capacity of the vessel and heat transfer
through the containment walls have to be predicted. The fluid
contents have additional effects as a heat sink and heat
transfer route, in producing large thermal gradients, and as a
source of stress via thermally induced pressure increase.
These factors are complicated, involving the interaction of
heat and mass transfer and fluid flow. Two phase systems (and
those which become two phase in fires) have the added
complexities of interphase heat and mass transfer, boiling
regimes, fluid flow and the uite different circumstances at
wetted and unwetted walls.
Other complications arise from
description of blowdown behaviour during emergency response.
Successful prediction of thermal response requires description
of the time dependent container and fluid behaviour.
The
status of understanding of the above will be examined.
Attention will be given to experimental results and current
predictive models.
identified.
Further knowledge
requirements
will
be
INPUT TO ACTIVITY:
Information from fire loading studies.
Published research and information made available by research
organisations, cooperating organisations and participants.
Al
APPENDIX A
DELIVERABLES:
report on the state of knowledge relevant to offshore fire

scenarios on the thermal response of fire impacted vessels and
pipework.
Requirements for future work.
Rev No:
A2
Date:
November 89
APPENDIX B
APPENDIX B
PRIMARY INFORMATION ON PREDICTIVE MODELS
METHODOLOGY
A list was
compiled of industrial and academic organisations believed to have a direct or indirect

interest in modelling the response of vessels and pipework to
external heat fluxes and to blowdown, and/or experimental work
The organisations, listed in Table Bl, were
in this area.
then contacted by letter, given an outline of the Joint
Industry Project, and the objectives of FR7, and asked for
information in summary form under the following headings:
around 30
in total
Contacts and experience:
Staff involved in areas of direct or indirect

relevance.
Mathematical modelling
The types of models used, their physical basis,

validation, solution techniques, capabilities and
applications, and availability.
Physical modelling
Experimental facilities, scale/scope of experiments,

instrumentation, test programmes, and availability of
facilities and results.
Current position
Weaknesses in current mathematical and physical

models, suggestions for further work, and specific
recommendations, including contributions to larger
programmes.
References
Publications and contracts.
The response to the questionnaire was generally good, although

in some cases the information had to be found or supplemented
using commercial data bases.
TREATMENT
OF THE INFORMATION
The information has been grouped into three categories organisations who have produced experimentally verified
mathematical models, those who have carried out either
mathematical modelling or experimental work, and those who use
current models, but have not carried out development work, or
those who gave a null response.
For the first two categories the information is presented in
report form, and where necessary comments have been added. It
was felt that work in progress should be limited to areas
B!
APPENDIX B
which
will become
public knowledge,
undertaken as part of large projects,
references have been restricted to those
in the open literature.
Information is
organisations in the third category.
for example
that
and similarly the
currently available
not presented for
Organisations who have produced experimentally verified models

are as follows:
AEA Technology (Safety & Reliability Directorate)

Imperial College (Chemical Engineering
Technology Department)
&
Chemical
Queen's University Ontario (Mechanical Engineering

Department)
Page
B3
B6
B9
Shell Research (Thornton Research Centre)
Rh
University of New Brunswick (Fire Science Centre)
Bl3
University of Maryland (Mechanical Engineering

Department)
Bi 6
The organisations who have either developed models or carried

out relevant experimental work are:
British Gas (London Research Station)

Fauske
& Associates
Page
B19
B21
Federal Institute for Materials Research &

B23
Testing (BAM)
Health & Safety Executive (Major Hazards Assessment
Unit)
B25
SimSci International
B26
SINTEF (Applied Thermodynamics Division)
1328
VERITEC
1330
In each case the report lists the following.

[a]
[b]
[c]
[d]
[e]
[f]
[g]
Organisation
Contact(s)
Background
Experimental validation
Additional comments
Reference(s)
B2
APPENDIX B

[a]
Organisation
Safety & Reliability Directorate
Wigshaw Lane
Culceth
Warrington
Cheshire WA3 4NE
(Tel: 0925 232000; Fax: 0925 766681)
[b]
Contact
Mr Philip M Ramskill
Fire & Explosion
[c]
Background
use the programs ENGULF and ENGULF II for
modelling the thermal response of vessels containing LPG
to fire loading.
AEA SRD
[d]
AEA SRD, under contract to HSE, developed a computer
model called ENGULF for simulation of the response of an
LPG tank partially filled with hydrocarbon and totally
ENGULF II is a more
engulfed in a kerosene poo1 fire.
recent version which also claims simulation of tanks
which are partially engulfed in a fire.
ENGULF II is written in standard Fortran 77 and
present available on an Apricot XI microcomputer.
times of 30 minutes are typical. It costs 1500.
is
at
Run
The principal features of ENGULF II are as follows.
The fire load can be:
Partial or total engulfment in a pool fire;
Jet flame impingement (torching) on any part of

the tank;
Incident radiation from
a distant
fire.
For each sort of fire load, the user specifies the

incident heat flux, the fire temperature and/or the flame
emissivity. Heat transfer from the fire to the tank is
assumed to be purely radiative unless there is jet flame
is
impingement, when a simple convective mechanism
assumed.
The tank is assumed to be a horizontal circular

cylindrical vessel with flat ends.
The tank wall is divided into four zones (which are

fixed with respect to time):
Heated liquid;
Unheated liquid;
Heated vapour;
Unheated vapour.
Heat can be lost to the atmosphere from all unheated

zones of the wall.
Within each zone, the heat flux
B3
APPENDIX B
(whether from the fire or to the atmosphere) is spatially

uniform. The fluid in contact with each wall zone is at
a spatially uniform temperature. The temperature of each
wall zone is assumed to be constant through its thickness
(though a correction is applied in order to determine the
and
outer
wall
difference
between
the
inner
temperatures).
Temperatures and pressures are predicted by heat and

based
on
mass
transfer
calculations,
generally
and
established correlations. Thermodynamic properties
correlations
phase
equilibria are determined using
The
(derived from: R C Reid & J M Prausnitz:
properties
of gases and liquids, McGraw-Hill (1977)).
Venting of material from the tank through a relief

valve is modelled by assuming that a perfect gas emerges
through a hole of specified size; the valve is assumed to
be shut until the vapour pressure in the tank exceeds the
set pressure, when the valve immediately opens fully.
The vapour pressure is determined as a function of
temperature for a given one-component fluid by using a
simple analytical correlation. ENGULF has been modified
in an ad hoc manner to deal with a multicomponent fluid:
ENGULF II is likely to be similarly modified in future.
[e]
[f]
Validation of ENGULF II has been against experimental
and
tonne horizontal
results
on
1.0
5.0
0.25,
cylindrical propane tanks fully engulfed in kerosene poo1
fires (<100 kW m2) obtained by HSE and Shell Research
been
runs have
Ltd.
No
experimental verification
flame
for
or
for
conducted
partial engulfment
jet
impingement.
Additional comments
Additional work needs to be done on ENGULF II to reduce
assumed flame temperature when
its
sensitivity to
flame
impingement.
simulating jet
Desirable modifications and additions to ENGULF II would
be:
The capacity to handle multicomponent fluids;
The ability to model external tank cooling and

insulation;
Inclusion of a tank failure model, so that for

example time to failure can be predicted;
Refinement and extension of the venting model, so

that liquid and two-phase releases (whether flashing or
not) and more realistic pressure relief valve behaviour
can be simulated;
Generalisation of tank geometry and orientation;
The ability to model spatially non-uniform heat

fluxes from the fire to each wall zone.
B4
APPENDIX B
The first two of these have been implemented, albeit in

an ad hoc manner, in ENGULF and could be added to ENGULF
II.
(The
handle multicomponent fluids
capacity to
is
achieved by assuming ideal behaviour. Up to 10 components

can be accommodated.) The third has been partially
incorporated using a longitudinal (hoop stress) tensile
model.
The fourth does not yet appear to have been
started; the fifth and sixth do not seem to have been
considered yet.
Whilst the model claims to predict response to jet fires,

it is unclear how non uniform incident heat fluxes can be
This is
incorporated without major program changes.
important since local wall temperatures define the vessel
strength and hence time to, and mode of, failure.
[g]
References
[Bl-B5]
B5
APPENDIX B
Imperial College (Chemical Engineering & Chemical

Technology Department)
[a]
Organisation
Imperial College
Department of Chemical Engineering & Chemical Technology
London SW7 2BY
(Tel: 071-589 5111 x 4429/4455; Fax: 071-584 1170)
[b]
Contacts
Dr Stephen M Richardson
Senior Lecturer
Dr Graham Saville
Senior Lecturer
[c]
Background
Imperial College use a program called BLOWDOWN developed
since 1985 in conjunction with Shell UK Exploration and
Production, based on an extensive set of experiments
conducted partly by Imperial College on their own and
The application is
partly by Shell Research Ltd.
for
simulation
of
rapid depressurisation of
primarily
offshore installations,
and has been used by several
oil/gas companies; it has also been used by some onshore
companies.
[d]
Imperial College have developed a computer program called
BLOWDOWN for simulation of the temperature, pressure,
or
composition response of a high-pressure vessel
blown
whether
or
not.
down,
pipeline being
deliberately
The principal features of BLOWDOWN as applied to vessels

are as follows.
The vessel is divided into two zones. The top zone

contains
all
of
the
vapour
plus
any
suspended
liquid-phase droplets. The bottom zone contains all of
the liquid phase (condensate) that has dropped out of the
top zone under gravity to form a pool of liquid on the
base of the vessel. If no liquid is present, the bottom
zone is eliminated but it reappears if condensate forms
at any time.
Each zone is assumed to be spatially
uniform in temperature, pressure and composition.
The whole continuous depressurisation process is

Each
replaced by a series of discrete time-steps.
subdivided
into
a number
of
time-step is
simple
thermodynamic and heat transfer sub-steps.
Provision is made for two types of choke: an orifice

and a long parallel duct:
The rate of discharge through an orifice is

calculated by requiring the fluid to follow an
isentropic path through the orifice, the fluid
emerging with a speed equal to the local speed of
sound of the gas-phase component;
The rate of discharge through a long parallel

duct is calculated by subdividing the duct into a
B6
APPENDIX B
number of short sections and solving the momentum

and energy conservation equations for each section.
Heat transfer from the wall to the top zone is by

natural and forced convection.
Heat transfer from the
wall to the bottom zone (which exists only if a pooi of
liquid is present) is by nucleate or film boiling.
Heat transfer from surrounding air to the wall is by

natural convection. Other modes of heat transfer can be
relevant in other circumstances.
An energy balance is always performed.

A mass
balance is performed only if liquid is present, since
mass transfer between zones only occurs as a result of
evaporation from the liquid pool in the bottom zone and
of sedimentation of droplets from the top zone to the
bottom.
Thermodynamic and transport properties for BLOWDOWN

are calculated using the physical properties package
called PREPROP which was also developed at Imperial
College. It uses an extended principle of corresponding
states.
The extension is necessary in order to treat
non-spherical molecules through the use of shape factors.
The basic idea behind the thermodynamic calculations is
the relation of the properties of the mixture to those of
a single reference substance. Methane is used for this,
since it is the majority component in natural gas.
The
choice
of
a
states
for
corresponding
package
depressurisation calculations was dictated by the need
for good predictions of phase equilibrium, enthalpy and
density simultaneously.
The principal features of BLOWDOWN as
applied to
are
similar
to
those
as
pipelines
applied to
very
vessels: the differences arise because there is no
pressure drop in a vessel whereas there is a pressure
The principal differences are as
drop in a pipeline.
follows.
The pipeline is subdivided into a number of short

sections, in each of which the momentum, linear momentum
and energy conservation equations are solved.
It is assumed that two-phase flows, when they occur,

are homogeneous, that is that the liquid and vapour
phases have the same axial velocity. The calculation of
pressure drop and holdup in two-phase pipe flow is based
on the analyses of Friedi, Premoli et al and Lockhart &
Martinelli (full details of which are given in: C
Hetsroni
of Multiphase
(editor): Handbook
Systems,
Hemisphere (1982)).
BLOWDOWN is written in standard Fortran 77 and at present

runs on a Compupro (equivalent to.a small VAX) and also
on large mainframe machines. Run times of a few hours
are typical.
[e]
B7
APPENDIX B
facilities
for
has experimental
College
Imperial
Further
small-scale
high-pressure testing.
relatively
large scale testing has been carried out by Shell
Research Ltd.
BLOWDOWN has been tested:
Using vessels ranging in diameter from 5 to 110

centimetres and in length from 50 to 300 cm;
With fluids including nitrogen, nitrogen and

carbon dioxide, methane and methane and propane;
At pressures up to 150 bar;

and:
For a wide range of initial liquid-vapour
ratios.
BLOWDOWN has also been tested, though less extensively,
for pipelines (using the data of: L T Cowley & V H Y Tam:
Consequences of pressurised LPG releases: the Isle of
Grain full scale experiments, 13th International LNG/LPG
Conference, Kuala Lumpur (1988) and also on data made
available to the Piper Alpha Inquiry).
[f]
Additional comments
BLOWDOWN does not simulate the response of a vessel or

pipeline to fire and would require extensive modification
in order to do so (the present algorithm is based on the
notion of pressure decrements). It also has rather long
run times.
It does, however, comprise many of the
desirable or essential features of a model that would
it
to
fire.
In
simulate
response
particular,
incorporates:
An essentially exact treatment of multicomponent
thermodynamics, including behaviour in the critical or
near-critical region;
A treatment of all modes of venting, whether vapour,
liquid (flashing or not) and vapour plus liquid and
whether deliberate (for example through a pressure relief
valve to flare - downstream pipework is incorporated into
the model.) or accidental (through a rupture of any
specified - and possibly varying - size);
The capacity to treat circular cylindrical vessels

oriented vertically
or
with flat or round ends,
horizontally and with venting from the top or bottom;
to
simulate
of
The
depressurisation
ability
pipelines.
[g]
References
[B6,B7]
B8
APPENDIX B
Queen's University Ontario (Mechanical Engineering

Department)
[a]
Organisation
Queen's University
Department of Mechanical Engineering
McLaughlin Hall
Kingston
Ontario K7L 3N6
Canada
[b]
Contact
Dr A Michael Birk
[c]
Background
Queen's University Ontario use a computer program called
TCTCM for simulation of the response of LPC rail tankers
to external fire loading. The program has been validated
using full scale and one fifth scale tests on rail
tankers.
[d]
Queen's have developed a computer program called TCTCM in
conjunction with the Transportation Development Centre of
Transport Canada for simulation of the response of rail
tankers to external fire loading.
The principal features of TCTCM are as follows.
The vessel is assumed to be a circular cylindrical

tank, the axis of which is normally horizontal but can be
varied to simulate the effects of pitch and roll.
Axial variations along the vessel are neglected, as

are end effects.
The heat load from the fire on the vessel is assumed

to arise from a totally engulfing pool fire or from
two-dimensional torching. Heat transfer from a pool fire
is by radiation and convection: the radiation varies
azimuthally around the circumference of the vessel, so as
to simulate the shape of a typical pooi fire (in the
absence of crosswind); the convection is based on
empirical correlations. Heat transfer from torching is
based on empirical correlations for a jet impinging on a
flat plate.
Heat transfer through the vessel wall and coverings

(such as insulation) is determined by finite-difference
solution of the transient heat conduction equation.
Heat transfer from the wall of the vessel to its

contents is by convection and radiation to the vapour and
by convection and boiling to the liquid, with empirical
convection and boiling heat transfer coefficients.
The liquid in the tank is decomposed into two

regions:
B9
APPENDIX B
The liquid boundary, near the vessel wall and

liquid-vapour interface, which is assumed to be in
thermodynamic equilibrium with the vapour;
The liquid core, which is assumed to be subcooled

initially and then, after a period of venting, to be
in
with
the
equilibrium
liquid
thermodynamic
and
boundary
vapour.
Two disposable constants, the liquid boundary thickness
and energy
which
determines the
partition
factor,
relative amounts of heat transferred to the liquid
boundary/vapour and to the liquid core, must be specified
by the user.
Venting is assumed to be of material from the vapour

and liquid boundary. Valve dynamics are not accounted
for, so that valve cycling cannot be modelled exactly;
instead, the valve is assumed to remain partially open so
as to represent the reduced flow capacity during cycling
and fully open when the valve should in practice be fully
The valve models can account for egress of vapour
open.
without liquid or for what is termed frozen liquid flow.
Wall stresses are calculated at the point on the wall

circumference which experiences the highest temperature
and includes pressure-induced hoop stress and stresses
induced by radial temperature gradients in the wall.
Tank failure is predicted according to a maximum

normal stress criterion.
Degradation of wall material
strength with temperature changes is allowed for.
[e]
TCTCM has been validated using experimental data obtained
from full scale and one fifth scale tests on tanks
(apparently) containing propane which were unprotected
and on tanks which were thermally protected (by an
internal layer which enhances heat transfer from the wall
to the contents).
These tests were conducted by the
Transportation Development Centre of Transport Canada and
by the American Association of Railroads.
Some discrepancies between observed and predicted vessel
failures are attributed to flaws in the tank walls, poor
welds or unpredicted hot spots.
Some discrepancies
between observed and predicted valve opening pressures,
particularly with liquid or two-phase venting, are
attributed to the simple valve model used. Otherwise,
there are minor discrepancies between
observed and
predicted temperatures.
[f]
Additional comments
No validatory experiments on the effects of jet fire
impingement have been conducted. The assumed mode of
vessel failure appears not to have been validated.
[g]
Reference
B8]
BlO
APPENDIX B

[a]
Organisation
Shell Research Ltd
Thornton Research Centre
P0 Box
Chester CR1 3SH

(Tel: 051-373 5751; Fax: 051-373 5845)
[b]
Contact
Dr Leslie T Cowley
Discipline Leader, Fire Loading
[c]
Background
Shell TRC use a program called HEATUP which they have
developed themselves, based on an extensive set of
experiments conducted partly by the HSE on their own and
The application is
partly in joint work with Shell.
for
onshore
installations
and
primarily
storage
conditions are those for either propane or butane at
ambient temperature and at its saturated vapour pressure.
Vessels for such storage have somewhat lower pressure
ratings than most encountered offshore.
[d]
Shell TRC have developed a computer model called HEATUP
for simulation of the temperature/pressure response of an
LPG tank partially filled with hydrocarbon and totally
engulfed in a fire.
The principal features of HEATUP are as follows.
The fire totally engulfs the tank with a heat flux

that is constant along the tank length but can vary with
The total
height and also with time, if appropriate.
incident flux is
because
breakdown
into
specified
convective and radiative flux is irrelevant.
The tank is a horizontal circular cylinder with a

sufficiently large length-to-diameter ratio that end
effects can be ignored and only vertical cross-section
heat flow calculations are relevant.
The tank wall can be multi-layered, so that insulated

and scale-fouled vessels can be modelled. Heat transfer
through the wall is by transient radial and azimuthal
heat conduction.
The vapour and liquid contents of the tank are each

assumed to be individually well mixed and hence of
spatially uniform temperature.
The
internal pressure is
the equilibrium vapour pressure of the bulk liquid.
Specific PSVs can be modelled, including the thermal

effects on
response (eg spring weakening).
[e]
There
their
is no attempt to predict vessel failure.
Bil
APPENDIX B
Shell has collaborated with HSE and used their test

HEATUP has been validated as
facilities at Buxton.
follows.
A set of experimental verification runs have been
conducted for 0.25, 1.0 and 5.0 tonne mild steel tanks
in
partially filled with propane totally engulfed
kerosene pooi fires.
The experiments were extensively
instrumented. The fires were allowed to burn until close
to predicted BLEVE conditions.
The only parameters adjusted in HEATUP in order for
it to simulate the experiments successfully were:
The emissivity of the external surface of the

tank;
The speed of development of the fire flux to its

steady value.
[f]
[g]
Additional comments
Desirable modifications and additions to HEATUP would be:
The capacity to handle multicomponent fluids;
The ability to model external tank cooling;
Inclusion of a tank failure model, so that for

example time to failure can be predicted;
Refinement and extension of the venting model,

so that liquid and two-phase releases (whether
flashing or not) can be simulated;
Generalisation of tank geometry and orientation;
The ability to model axially non-uniform heat

fluxes from the fire.
None of these appear yet to have been started.
References
{B9-Bll]
B12
APPENDIX B

[a]
Organisation
University of New Brunswick
Fire Science Centre
Fredericton
New Brunswick E3B 5A3
Canada
[b]
Contact
Professor James
E S Venart
[c]
Background
University of New Brunswick use a program called PLCS-I
which they have developed themselves, based mainly on
conducted
that
experimental measurements
they have
themselves but also on published data, particularly those
obtained by HSE, Shell Research Ltd, and BAM.
[d]
University of New Brunswick have developed a computer
program called PLGS-I for simulation of the response of a
vessel containing LPC to fire.
PLGS-I is written in HP Basic 5.0 and runs on an HP 98220
Fortran PC and mainframe versions are also
computer.
The principal features of PLGS-I are as
available.
follows.
Experimental data have been used to distinguish five

regions within the fluid in the vessel:
A vapour region;
A subcooled bulk liquid region;
An unstable liquid region at the bottom of the

vessel
A free convection boundary layer liquid region

the
walls of the vessel;
up
A stratified layer liquid

(the
region
significance of which is stressed by University of
New Brunswick).
The model is based on these five regions.
It is assumed that the vessel is a horizontal

circular cylinder with a vertical plane of symmetry
A two-dimensional (radial
passing through its centre.
and azimuthal) analysis is used; thus axial variations
are neglected.
The fire is assumed to engulf the vessel totally and

to be spatially uniform but varying
(in a manner
the
in
time.
user)
specified by
Heat transfer from the fire is by radiation and

convection.
The shell of the tank, with or without insulation,

can
be
into
to
24
azimuthal
decomposed
up
sections
and
8
radial
sections.
The
(circumferential)
B13
APPENDIX B
transient heat conduction equation is solved in the

shell.
Heat transfer to the contents is by:

Radiation and convection to the vapour;
Natural convection and nucleate/film boiling to

and:
the liquid (though film boiling is considered
unlikely in practice).
The flow in the free convection boundary layer is

predicted using closed form analytical expressions.
Venting of material from the vessel through a relief

valve is modelled by:
Assuming flow of a perfect gas through an

orifice when the material is vapour;
and:
Using a curve fitting procedure including use
of a discharge coefficient for two-phase releases:
variations
are
needed
in
this
significant
coefficient to model real venting, indicating a
deficiency in the modelling.
[e]
University of New Brunswick have conducted an extended
set of experimental measurements on a small-scale (37.4
litre) windowed horizontal circular cylindrical vessel
partially filled with refrigerant Ru.
The effect of the incident heat flux from a fire is
simulated by the use of ten individually controlled
radiant heater elements located around the vessel.
By
altering the heater controls, it is claimed that pooi and
jet impingement (torching) fire loads can be simulated
(although no basis exists for the jet fire loads
employed).
A pressure relief valve is fitted to the vessel.
Its
position on the vessel can be altered, as can its open
and close pressures (which are usually set at 105% and
95% of the nominal set pressure, respectively).
The experiments have been analysed and revealed the
importance of thermal stratification in the liquid and
critical two-phase flow through the valve.
Further
in
which
the effects
of
experiments are planned
impurities, non-condensihie gases and non-uniform heating
are to he investigated.
[f}
Additional comments
PLGS-I has a definite strength inasmuch as the liquid
within the vessel is not assumed to he spatially
isothermal. It also has limitations, including:
A heavy reliance on the use of adjustable
fire and
constants, particularly in modelling
venting characteristics which limits its use as an a
priori predictor;
and:
An apparent limitation to one-component fluids
B14
APPENDIX B
in the vessel: no mention is made of multicomponent

fluids
and
no
for
appropriate
thermodynamics
physical property and phase equilibria predictions
appears to be available.
[g]
References
[B12,Bl3]
B15
APPENDIX B
University of Maryland (Mechanical Engineering

Department)
[a]
Organisation
Mechanical Engineering Department
University of Maryland
College Park
Maryland 20742
USA
(Tel: 0101 301
[b]
[c]
405 1000, Fax: 0101 301 314 9477)
Contact
Professor Dr Dirse
Sallet
Background
The University of Maryland use programs called TAC6 and

TAC7 which they have developed themselves
simulation of the response of tanks to Lire.
for
the
University of Maryland have developed two computer models
called TACG and TAC7 for simulation of the response of a tank
partially filled with hydrocarbon and engulfed in a fire.
[d]
TAC6 and TAC7 are written in standard Basic, and can be run on
GW-Basic 3.20 and IBM Basic 2.0. Run times of 30 minutes are
typical.
TAC6 is in a sense a subsidiary version of TAC7, which is the
main program. The principal features of TAC7 are as follows:
The fire load can (apparently) be partial or total
The heat load can also be
engulfment in a pool fire.
internal, for example from a runaway exothermic thermal
reaction.
The tank is assumed to be an infinitely long horizontal

circular cylindrical vessel.
The one-component contents of the vessel can be liquid,

gas (vapour) or two-phase liquid plus gas, with an initially
spatially uniform temperature. If the contents are two-phase
it is assumed that the two phases are always in equilibrium
with each other.
With the exception of
the vapour pressure and fluid

viscosities, the physical properties of the contents of the
vesel are assumed to be constant.
Venting of material from the tank through a relief valve

modelled, though details of the model are not given.
Apparently, the only requirement is that the depressurisation
rate is not so high that forced convection effects are
significant. (The subsidiary program TAC6 should be used when
is
B16
APPENDIX 13
there is little or no natural convection.)

how if at all two-phase venting is modelled.
It is not clear
Three modes of heat transfer are modelled:
From the surroundings to the wall of the vessel with a

user-specified heat flux which can vary spatially and
temporally and also with the wall temperature.
Through the wall of the vessel, assuming that there is

no temperature difference radially through the wall
thickness., but allowing for circumferential temperature
variations.
To the vessel contents, both liquid and gas, each of

which comprises a number of discrete spatially isothermal
elements.
The transient mass, momentum and energy conservation
solved in the liquid and/or
equations are
gas using
approximate integral methods in which the mathematical forms
of profiles of variables such as temperature are assumed. It
is not clear that this is always valid.
There is no model of vessel failure.
Two particular features are included:
Thermal
stratification
in the
liquid,
particularly
when the liquid inventory is small.
Radiation from
the unwetted wall of
the vessel
not
only to the gas (which is usual), but also to the liquid

in the vicinity of the liquid-gas interface, where local
heating can significantly alter the pressure.
[e]
TAC7 has been validated using two sets of experimental data.

The first set is of US Army tests on a horizontal vessel
of length 18.3 metre, diameter 3.0 metre, and thickness 1.6
cm, and fitted with a relief valve. The vessel was initially
nearly completely filled with 98% propane and 2% ethane and
placed in a totally (or nearly totally) engulfing JP4 kerosine
pool fire.
The second set is of joint Shell Research Ltd / 1-ISE tests

on a horizontal vessel of length 4.0 metre, diameter 1.7 metre
and wall thickness 1.8 cm and fitted with two relief valves.
The vessel contained commercial propane and was placed in a
totally engulfing pooi fire.
Generally good agreement has been obtained between the

predictions made with TAC7 and the experimental measurements.
There are, however, some odd features in the predicted
results, in particular a strange piece-wise continuity in
time.
[fJ
Additional comments
B17
APPENDIX B
Desirable modifications and additions to TAC7 would be:
The capacity to handle multicomponent fluids.
Inclusion of a tank failure model, so that for example

time to failure could be predicted.
Refinement and extension of the venting model
Generalisation of tank geometry and orientation.

References
[g]
B14]
B18
APPENDIX B

[a]
Organisation
British Gas
Research & Technology Division
London Research Station
Michael Road
London SW6 2AD
(Tel: 071-736 3344)
[b]
Contact
Dr Trevor K Yerlett
Senior Scientist, Gas Properties
[c]
[d]
&
Measurement Group
Background
British Gas LRS use VENTFLO for simulation of
temperature/pressure response of a venting vessel.
the
British Gas have developed a computer program called
VENTFLO, which is a modified version of SAFIRE developed
by DIERS, for simulation of the depressurisation of a
vessel.
VENTFLO is written in Fortran 77 and is at present
available
in-house or
elsewhere
on
an
IBM/Amdahl
mainframe and also on a VAX. Run times of a few seconds
are typical of steady-state runs; dynamic runs take much
longer.
The principal features of VENTFLO are as follows.

*
The vessel can be:
A vertical cylinder;
or:
A sphere.
bullet;
The vessel can contain gas, liquid or gas and liquid.

The gas and liquid, when they coexist, are assumed each
to be at a spatially uniform temperature.
The fluid can be multicomponent; a database of 200

library components exists; together with user-defined oil
fractions; the fluid can comprise up to 30 components.
The heat flux to (or from) the surface of the vessel

from (or to) the surroundings is assumed to be spatially
uniform. The heat transfer can be:
AdiabatIc (so that the vessel is perfectly
insulated);
To freely convecting air;
At constant and specified temperature;
At constant and specified heat flux;
At varying and specified heat flux;
or:
According to the NFPA method (originating from
the US National Fire Protection Agency).
Both steady state and transient venting can be

simulated.
B19
APPENDIX B
and phase
are
Physical properties
equilibria
determined using in-house and RKS equations of state,
together with the Wilhoit-Harmens ideal gas method for
heat capacity, the Hankinson & Thomson COSTALD method for
liquid density and the Ely & Hanley TRAP? method for
transport properties.
Venting can be one-phase or two-phase, with
in short vents
and
non-equilibrium
two-phase flow
and
flow
in
or
slip
longvents
pipes.
homogeneous
[e]
British Gas LRS give nodetails of the experimental

validation of VENTFLO, other than a general claim that
'the thermal response of a vessel and contents during
rapid blowdown of both single-phase gas and two-phase
fluid has been modelled successfully'.
[f]
Additional comments
VENTFLO clearly needs further experimental validation and
British
Gas
LRS
would
welcome
appropriate
data,
particularly for large-scale fire exposure and blow down
of vessels containing multi-component fluids (such as gas
condensates).
It is not clear that VENTFLO models heat transfer
The assumption of spatially uniform heat
correctly.
transfer over the whole of the vessel wall, whether it
contains one or two phases, is rather drastic.
It is
also not clear that VENTFLO handles venting correctly.
Thus it is not clear that the venting is modelled in a
thermodynamical1y consistent way, nor is it clear that
flashing discharge is dealt with properly.
[g]
References
None available.
B20
APPENDIX B
Fauske
.5c
Associates
[a]
Organisation
Fauske & Associates
16W070 West 83rd Street
Burr Ridge
Illinois 60521 USA
[b]
Contact
Dr Hans KFauske
[c]
Background
Fauske & Associates have an extensive record of work in
the field of vessel venting, much in association with the
DIERS project and the SAFIRE program.
[d]
Fauske have made a significant contribution to the
computer program called SAFIRE, which was developed as
part of the DIERS project. This is a program for the
simulation of the response of a batch reactor or vessel
with an emergency relief system to actuation of the
relief system.
SAFIRE is written in ANSI standard Fortran 77
in
consistent SI units. It is a multi-purpose program:
The vessel can be a vertical or horizontal

circular cylinder or a sphere;
The vent can be located at the top or bottom and

can be a nozzle or a long line;
and:
The fluid in the vessel can be two-phase:
various different two-phase flow models can be used.
SAFIRE does not appear to be able to simulate response to
fire, other than very crudely by specificaion of a
British Gas
spatially and temporally uniform heat flux.
LRS have developed a variant
called VENTFLO
for
simulation
of
blowdown
of
a
vessel
containing
hydrocarbons.
Fauske have also developed a simple criterion for
predicting the onset of liquid entrainment in the
freeboard volume of a storage vessel exposed to fire,
with the entrainment resulting from flow of vapour out of
the vent.
The criterion is developed using potential
from
which
a
two-dimensional
theory,
axisymmetric
distribution
is
obtained
in the freeboard volume
velocity
of a vertical circular cylinder.
[e]
None for fire loading on or depressurisation of vessels

containing hydrocarbons.
[f]
Additional comments
B21
APPENDIX B
It does not appear that SAFIRE as it exists at present is

capable of forming a useful basis for accurate simulation
of the response of vessels to fire loading.
The criterion developed by Fauske may be of some value
for prediction of the onset of liquid entrainment.
[g]
References
[B15-Bl7]
B22
APPENDIX B
Federal Institute for Materials Research & Testing (BAN)

[a]
Organisation
tJnter
den Eichen 87
D-l000 Berlin 45
Germany
[b]
Contact
B. Dro.st
[c]
Background
BAN have conducted
(in conjunction with Technischer

Oberwachungs-Verein Hannover) experimental tests on the
response of tanks containing LPG to fire.
[d]
None.
[e]
BAN have conducted two series of tests on 4.85 cubic
metre horizontal circular cylindrical storage vessels.
The vessels contained propane and were in pool fires of
burning fuel oil with:
Almost (but not quite) total fire engulfment of

an unprotected vessel;
and:
Total fire engulfment of an insulated vessel.
Two similar series of tests were also conducted for pool
fires of burning liquid propane with:
Total fire engulfment of a vessel protected by a

conventional water spray;
and:
Total fire engulfment of a vessel protected by
an upgraded water spray.
Two modes of tank failure were distinguished in the tests

on unprotected vessels:
An increase of overpressure caused by an increase in

the temperature of the tank contents with insufficient
pressure relief by the safety valve;
A decrease of overpressure resistance of the vessel

caused by an increase of the temperature of the tank
shell and a corresponding decrease of yield and tensile
strengths.
The observed bursting pressure for a given tank shell
temperature was found to be greater (by of order 20 bar)
than the predicted overpressure assuming that rupture
occurs when the yield strength of the shell material is
exceeded. Nevertheless, the rapid rise in overpressure
with time means that the conservative nature of the
predictions is of limited practical value.
The insulated vessels were protected conventionally, as
were the valves and fittings.
The tests showed a
B23
APPENDIX B
increase
significant
vessels.
in
the
fire
resistance
of
the
The water sprays used to protect the vessels were of two

types:
and
Three parallel pipes aligned axially on top of

the vessel;
A cage-like pipe structure around the vessel.
Spray rates of 100 to 1000 litres per square metre per
hour were used. The cage system was found to improve the
fire resistance of the vessels considerably.
[f]
Additional
comments
More data on the vessel failure mode would be of value.

[g]
References
[Bl8 ,Bl9]
B24
APPENDIX B
Health & Safety Executive (Major Hazards Assessment Unit)

[a]
Organisation
UK Health & Safety Executive

Major Hazards Assessment Unit
St Anne's House
University Road
Bootle
Merseyside L20 3RA
(Tel: 051-951 4000)
[b]
Contact
Mr David A Jones
Head, Flammables
[c]
Risk Group
Background
use programs, in particular ENGULF and ENGULF

for regular assessment of onshore hazards involving
LPG.
These programs have been developed for HSE by AEA
HSE MHAU
II,
SRD.
[d]
None.
[e]
The Research and Laboratory Services Division of HSE has
experimental facilities at Buxton. The contact there is:
D K Pritchard, Head of Investigation Section, Research &
Laboratory Services Division, Harpur Hill, Buxton SK17
9JN.
The facilities have been used to validate ENGULF II (AEA

SRD) and HEATTJP (Shell Research Ltd) for total engulfment
of tanks containing propane in kerosine pool fires.
These have been for 0.25, 1.0 and 5.0 tonne tanks, the
latter in joint work with Shell Research Ltd.
No
experimental verification runs have been conducted for
partial engulfment or for jet flame impingement.
[f]
Additional comments
A review paper on modelling and validation of the fire
response of vessels by K Moodie of HSE comprises Appendix
2 of this report.
[g]
References
[BlO,Bll,B20,B21]
B25
APPENDIX B
[a]
Organisation
Highbank House
Exchange Street
Stockport
Cheshire SK3 OET
(Tel: 061-429 6744; Fax: 061-480 9063)
[b]
Contact
Mr Michael M Orbach
Technical Manager
[c]
Background
SimSci are a software company responsible for the PROCESS
A new version, PRO/lI is
(chemical) plant simulator.
scheduled for the end of 1990.
[d]
SimSci have developed a very large computer program
called PROCESS for the simulation of process plant. The
program is modular; one of the modules is for simulation
of depressurisation of vessels.
The principal features of the PROCESS depressurisation
module are as follows:
The module can be used to calculate the venting rate

and the temperature and pressure profiles as functions of
time at a sequence of discrete time steps. For each time
the heat input and average venting rate are
step,
calculated and the new temperature and pressure of the
vessel contents are calculated in order to maintain
constant vessel volume.
The heat flux into the vessel from the surroundings,

for example from a fire, must be specified by the user,
as must the valve opening.
Initial conditions can be either:

Specified overall composition, temperature and
pressure, which should be used when the fluid in the
vessel is initially one-phase;
or:
Specified liquid composition, liquid volume and
pressure, which should be used when the fluid in the
vessel is initially two-phase.
Either the relief valve set pressure or valve opening

time must be specified.
It is possible to include the thermal mass of the

vessel in the calculations. It is then assumed that the
walls of the tank are at the same temperature as the
fluid within it, unless there is a blowdown of gas alone,
in which case it is assumed that there is heat transfer
by natural convection between the walls and the gas. The
work done by a gas during depressurisation can also be
included by specifying an isentropic efficiency, though
B26
APPENDIX B
it is not clear how such an efficiency could be specified

in advance.
Venting of material from the vessel through a relief

valve is modelled by assuming that gas alone is vented.
The flow through the valve is then either what is called
supersonic or subsonic: presumably this means that the
flow is either choked (and hence sonic) or unchoked (and
hence subsonic); in either case, the equations used are
essentially those for a perfect gas.
Heat transfer to the contents of the vessel is
modelled in one of four ways:
five adjustable
Purely
empirically, using
constants;
Using API correlations, with the effects of

insulation
an
incorporated
by
adjustable
multiplicative constant;
Isothermally;
or:
Using a natural convection correlation for
blowdown
of
with
an
alone,
gas
adjustable
multiplicative constant incorporated.
[e]
None.
{f]
Additional comments
SimSci present a
'Fire Depressuring
Example' which
illustrates the use of the PROCESS module.
This is
somewhat misleading inasmuch as no heat flux is imposed
from any fire.

The PROCESS module has several limitations, including:
A very heavy reliance on the use of adjustable

constants which makes its use as an a priori predictor
(rather than an a posteriori checker) somewhat limited;
The equations used for venting are almost certainly

inconsistent
with
the
thermodynamically
physical
properties used (at least some of which are derived using
the SRK method);
and: A lack of experimental validation: the experience of
several users seems to be that blowdown of a vessel
containing gas alone under far from critical conditions
can be reasonably accurately simulated but that other
circumstances cannot.
[g]
Reference
[B22]
B27
APPENDIX B

[a]
Organisation
SINTEF
Applied Thermodynamics Division
Norwegian Institute of Technology
N-7034 Trondheim
Norway
(Tel: 010-47-7-59 3920; Fax: 010-47-7-59 2480)
[b]
Contact
Professor Bjorn
F Magnussen
[c]
Background
SINTEF ATD have experience of modelling fire loads on
depressurising vessels using a program called PIA.
[d]
SINTEF ATD have developed (in conjunction with Statoil) a
computer program called PIA (based on a finite-difference
code called KAMELEON, with pre-and post-processing) for
the analysis of the behaviour of process equipment being
blown down in hazardous conditions such as fires.
The principal features of PIA are as follows.

It appears that the vessel is assumed to be a
vertical circular cylinder. The fluid contents of the
vessel are treated one-dimensionally (axially) whereas
the walls of the vessel are treated two-dimensionally
(axially and radially).
The mass, linear momentum and energy conservation

equations (of whatever dimensionality) are discretised
The momentum equation is
using finite-differences.
applied only to the gas in a vessel: any liquid is
presumed to be stagnant.
Physical properties and phase composition of the

are
fluids
determined
a
multicomponent
using
thermodynamics package developed at SINTEF ATD, the basis
of which is not specified.
Two-phase flows are assumed to be homogeneous, so

that the liquid and vapour have the same axial velocity.
Heat transfer between the wall of the vessel and its

contents is assumed to be either forced or mixed
convection: for boiling, the heat transfer coefficient is
simply set to a large value. Heat transfer between the
wall
is
modelled
surroundings and the vessel
by
specifying a heat flux which is claimed to be variable in
space and time: the heat flux from a fire can be
predicted using simplified fire models incorporated in
the code.
The fire load is assumed to be symmetrical
about the axis of the vessel.
The properties of the wall can vary radially, so that

for example the effects of insulation can be simulated.
B28
APPENDIX B
Wall stresses developed during blowdown and as a

result of fire loading are determined using so-called
The stresses are assumed to be
'super elements'.
symmetric about the vessel axis and are predicted using
thin-plate/thin-shell linear elasticity theories.
No attempt is made to predict vessel failure.
PIA is also capable of handling pipework, though

whether or not in fires is unclear.
{e]
There has been no validation
of
PIA.
[f]
Additional comments
PIA has several limitations, including:
A crude, often ad hoc, treatment of heat

transfer;
The assumption of axisymmetry of fire loading

for vessels;
An unspecified treatment of venting;

and:
A lack of experimental validation.
{g]
Reference
[B23]
B29
APPENDIX B
VERITEC
[a]
[b]
Organisation
VERITEC
Veritas Offshore Technology
Det Norske Veritas
P0 Box 300
N-1322 Hovik (Oslo)
Norway
&
Services A/S
Contact
Mr Erik Skramstad
[c]
Background
VERITEC model fire loads on depressurising vessels using
a program called VT*VESSEL and on structural members
using a program called NV613.
[dl
VERITEC have developed
VT*VESSEL for the simulation of blowdown of a

vessel subjected to a fire load;
and:
NVG13 for the simulation of the thermal response
of structural members to heating, for example by a
fire, or cooling.
The principal features of VT*VESSEL are as follows.

The program is said to perform a simultaneous
calculation of wall temperature, wall strength and
internal pressure as a function of time.
The vessel is divided into two zones, an upper one

containing vapour and a lower one containing liquid.
Each zone is at a spatially uniform temperature.
The fluid in the vessel can be multicomponent: a

process simulation program called HYSIM is used for
and
of
physical
properties
phase
equilibria
multicomponent fluids.
For a vessel totally engulfed in a pool fire, vessel

failure is assumed to occur when the vessel wall is
For a jet flame impingement on the
totally plastic.
failure
is
assumed to occur when the hoop stress
vessel,
concentration exceeds the yield strength (modified to
account for work hardening). No information is presented
to verify the realism of the input conditions.
The flame temperature and a (convective) heat

transfer coefficient from the fire to the vessel wall
must be specified.
Venting of material from the vessel is modelled by

assuming that a perfect gas emerges through a hole of
specified size.
Both choked flow and unchoked flow can be handled.
[e]
B30
APPENDIX B
There appears to have been no validation of VTVESSEL or

NV613.
{f]
Additional comments
VT*VESSEL has several limitations, including:
It is not clear that the vessel failure models

have been validated; also, VERITEC claim that creep
rupture is not a significant mechanism, but their
models appear to involve creep rupture;
and:
Venting is not handled properly.
NV613 has several limitations, including:
Doubtful thermodynamics, particularly for

multicomponent fluids;
and:
Rather crude heat transfer modelling.
{g]
References
[B24,B25]
B31
APPENDIX B
REFERENCES
31.
I
B2.
B3.
Hunt & P K Ramskill: The behaviour of tanks engulfed in

the developmentof a computer program,
Chem E Symposium Series 93, 71-86 (1985)
fire
M
-
P K Ramskill: The development

multicomponent liquid mixture in
Report R414, HMSO (1987)
of
ENGULF
II
to
model
a fire engulfed tank, HSE/SRD
P K Ramskill: A description of the "ENGULF" computer codes
codes to model the thermal response of an LPG tank either

fully or partially engulfed by fire, 3 Hazardous Materials,
20, 177-196 (1988)
B4.
P K Ramskill: ENGULF
II: a computer code to model the thermal

response of a tank partially or totally engulfed in fire,
HSE/SRD Report R480, HMSO (1989)
B5.
P K Ramskill & C Marriott: A user's guide to ENGULF II: a

computer code to model the thermal response of a tank
partially or totally engulfed in fire, HSE/SRD Report R481,
HMSO (1989)
Saville & R Szczepanski: Methane-based equations

for a corresponding states reference substance,
Engineering Science, 37, 719-725 (1982)
36.
37.
A Haque, S M Richardson, G Saville &
depressurisation of
Process Industries,
of state
Chemical
Chamberlain: Rapid
pressure vessels, 3 Loss Prevention
3, 4-7 (1990) (see also: Appendix B, ff
36).
B8.
A M
Birk: Modelling the

external fire impingement,
(1988)
response
of
B9.
G V Beynon, L T Cowley, L M Small

engulfment of LPG tanks: HEAT1JP, a
Hazardous Materials, 20, 227-238 (1988)
810.
K Moodie, K
Bli,.
&
exposed to
20, 197-225
Williams: Fire
predictive model, 3
I
Billinge & D P Cutler: The fire engulfment of LPG

storage tanks, I Chem E Symposium Series 93, 87-106 (1985)
Moodie, L T Cowley,
engulfment tests on a
20,
Bl2.
tankers
3 Hazardous Materials,
3 E
55-71 (1988)
R B Denny, L M Small & Williams: Fire

5 tonne LPG tank, 3 Hazardous Materials,
Sumathipala, R A Doyle, W Qi, N U Aydemir,

Steward & A C Sousa: Experiments on the thermo-hydraulic
response of pressure liquefied gases in externally heated
tanks with pressure relief, Paper 51B,
A I Ch E Summer National Meeting, Minneapolis (1987)
FR
Venart,
B32
APPENDIX B
313.
NU
Aydemir, V K Magapu, A C M Sousa & J E S Venart: Thermal

response analysis of LPG tanks exposed to fire, J Hazardous
Materials, 20, 239-262 (1988)
B14.
D Dancer and D W
315.
M A Crolmes & J
Sallet: Pressure and temperature response of

in
containers and pressure vessels which are
liquefied gases
to
accidental
heat input, J Hazardous Materials, 25,
subjected
3-18 (1990)
relief
C Leung: Code method for evaluating integrated

phenomena, Chemical Engineering Progress, 81, 47-52
(1985)
316.
Epstein, H K Fauske & G M Hauser: The onset of two-phase

venting via entrainment in liquid-filled storage vessels
exposed to fire, J Loss Prevention Process Industries, 2,
45-49 (1989)
Bl7.
H H Klein: A computer model for sizing two-phase emergency

relief systems, validated against DIERS large scale test data,
I Chem E Symposium Series 102, 143-155 (1987)
B18.
B Droste & W Schoen: Full scale fire tests with unprotected

and thermal insulated LPG storage tanks, J Hazardous
Materials, 20, 41-53 (1988)
Bl9.
320.
K Moodie &
Schoen & B Droste: Investigations of water spraying systems

for LPG storage tanks by full scale fire tests, J Hazardous
Materials, 20, 73-82 (1988)
S
dispersion of
215-245 (1987)
B2l.
B22.
F Jagger: Flow through relief valves and the

the discharge, I Chem E Symposium Series 102,
K Moodie:
Experiments and modelling:- an overview with

reference
to
fire
J
Hazardous
particular
engulfment,
Materials, 20, 149-175 (1988)
(Anonymous):
Depressuring
calculations,
PROCESS
Technical
Bulletin 38 (1987)
Fagertun, T E Hals & B F Magnussen: PIA - a
computer program system for simulation of blowdown of process
equipment under hazardous conditions, Eurotherm Seminar 14,
Louvain-la-Neuve (1990)
J A
323.
T Evanger,
B24.
D M Solberg & 0 Borgnes: Thermal response of process equipment

to hydrocarbon fire, A I Ch E Loss Prevention Symposium (1982)
B25.
J Nylund: Fire survival of process vessels containing gas,

Chem E Symposium Series 85 (1984)
B33
APPENDIX B
Table Bi
Organisations contacted
AEA Technology (Harwell)

Babcock Woodall-Duckham
BP Exploration
BP Research Ltd
British Gas (Engineering Research Station)
Brown & Root Vickers
CHAM
Norske Stats OljeselskapA/S (Statoil)
Fauske & Associates
H&G Engineering
Health & Safety Executive (Major Hazards Assessment Unit)
Imperial College (Chemical Engineering & Chemical Technology
Department)
John Brown Engineers and Constructors
MW Kellogg
Matthew Hall Engineering
Queen's University Ontario (Mechanical Engineering Department)
Shell UK Exploration and Production
SINTEF (NBL)
Technica
Det
UKOOA
University of Maryland (Mechanical Engineering Department)
VERITEC
WS Atkins Engineering Sciences
B34
APPENDIX C
APPENDIX C - REVIEW PAPER BY K NOODlE

Appendix C comprises the paper: K Moodie: Experiments and
modelling:- an overview with particular reference to fire
This
which
is
engulfment.
reproduced by kind
paper,
of
at
the International
was
permission
TJKHSE,
presented
on
in
the
Hazards
Symposium
Major
Transportation and Storage
of Pressure Liquefied Gases, New Brunswick, Canada (August 10
13, 1987), and published in the Journal of Hazardous
Materials, 20, 149-175 (1988).
The paper is a general review of modelling and validation,
with the specific topic of modelling storage tanks engulfed in
fires in Section 4, and the experimental validation of such
modelling as Section 5.
Cl
APPENDIX C
EERfl4ETS
K.
AND
DELLING:
- AN
OVERVIEW WITH PARTICULAR
REFERENCE
TO FIRE
DIE
Health and Safety Executive,

Research & Laboratory Services Division, Herpur Hill, &lxton, SK17 9JN,
UK
The paper considers general modelling and experinental requiroments

before discussing the probloms associated with losses from contairment of
pressurised liquefied gases. Computer models for assessing the consequences
of fire on rp type storage vessels are reviewed together with the
experimental data currently available for mode], validation purposes.
1
rRODUCTION
of gases stored and transported as liquids under their own

pressures ie. as pressurised liquefied gases, have increased
The quantities
vapour
considerably over the last two decades. In most industrialised countries this
trend is likely to continue for m.any years yet, because liquefied gases are
used by the chnjcal and petrochamical industries in vast quantities, as the
raw materials for many products eg. feedstocks and fertilisers from chlorine,
aTmDor,ia etc or as hydrocarbon fuels (LNG,
.itane). The processes and
prcducts utilising liquefied gases in a typical industrialised country are
chtiical production (organic and inorganic), transportation (fuels), domestic
L,
industrial heating, and electricity production.

There are three main categories of liquefied gases, 1) pressure storage
ie. liquefied gas under pressure and at ambient
tnperature; 2) semi
under
refrigerated storage ie. liquefied gas
pressure at low temperature and
3) fully refrigerated storage ie. liquefied gas at aospheric pressure and
very low temperature. It is the first and second of these categories that are
of primary concern because they are a greater hazard. Generally speaking
pressurised liquefied gases in commercial use are either toxic or flaimrtable,
and
although some gases
have both properties e.g. acrylonitrile.
Further
sub-
divisions are the relatively small inventories of acutely toxic and persistent
chemicals (carcinogens). The more camoonly used of the liquefied gases are
listed in Table 1, which is reproduced from Vilain(l).

It is because of their potentially harmful properties, together with the
large inventories often stored, that many liquefied gases are deemed to be
potential major industrial hazards. The traditional, mainly retrospective
C2
APPENDIX C
TABLE 1
Boiling point and typical storage (pressure) of
Paraffins
B.P 0C
StOrage
Olefins
Methane
164
Etharle
88.6
F.R.
Ethylene
B.P 0C
Storage
80b
103.7
propylene
47.4
Propane
810
1 Butene
Butane
6.3
36.].
34 b
2 Butene(cis) 1 Pentene
3.7
29.9
Pentane
0.5
42
chnicals
canalon
(piped)
Industrial Hydrogen
oxygen
B.P0C
Storage
Others
B.P 0C
Storage
F.R.
P
A
259.1
F.R.
(Vinyl
chloride
209.8
Nitrogen
F.R.
F.R.
Ethylene
chloride
13.4
83.5
A
218.4
Ethylene
oxide
13.5
12 b
fully refrigerated
liquefied under pressure

anbient storage
C3
Chlorine
34.6
89 b
12 Butadiene
1isnonia
33.3
1012
Oenzene
Acetylene.
84
bcy1inders
A
Acrylo
nitrile
10.85
80.1
77
APPENDIX C
safety approach of learning fran
experience, developing codes
of
practice,
stipulating operating instructions have becane
devising precautions and

unacceptable because of the scale of possible accidents. Thus, in parallel
with the increasing inventories and accidents, methodologies have developed
for assessing the hazards and risks posed. The use of fault tree analysis,
hazard analysis, hazard and operability stodies etc, have now becane
widespread in an attempt to anticipate possible hazards and where possible
a consequence, without being
reduce the likelihood of their occurrence.
ccmplacent, it may be argued that accident controls have more or less kept
pace with the growing use and the increased hazard potential of the substances
involved. Nevertheless a number of major accidents involving releases of both
flammable and toxic gases have occurred during the past two decades.
No objectives of hazard assessments are, firstly to quantify the hazard
and the risk of
occurring, and secondly to reduce the risk. The two are
connected but are not identical. 4.ach of the value of a hazard assessment
lies in the discipline produced and in identifying the saker links in the
overall system. The hazard assessment of a particular process, storage or
transport system examines the consequential events resulting fran what are
considered to be credible failures.
integral part of this assessment is
the need to utjlise models and modelling techniques to both assess and predict
the hazards and their consequences. Thus a considerable amount of research
effort, both nationally and internationally, is being devoted to developing
suitable models and validating them against acceptable experinental data.
The paper discusses modelling and experinental requirements, the various
modelling approaches currently used, and model validation in the light of
available data. The consequences arid physical processes involved in losses of
Particular
containment of pressurised liquefied gases are discussed.
consideration is given to the fire attack on storage and transport vessels
it
containing flammable pressurised liquefied gases as the majority of the papers

presented at this conference relate to this aspect.
2
EERIME1'T2S AND JDELLING
2.1 Basic Philosophy

'Models'
in the context of this
paper
quantitative techniques used for predictive
are taken to
mean
systematic
purposes, whether these be
graphical techniques, analyticallyderived equations, correlations of

experimental data, or canputer programmes. The proof of any model is in its
ability to make reasonably accurate predictions within its acknowledged
limitations. Hover, practical constraints may be imposed upon a model or
modelling technique. For instance, if the model is one of a number which
could be called upon during a hazard assessment, then the factors that assume
C4
APPENDIX C
importance are, ease of use (the user may not be totally familiar with the
basis of the model), efficiency (in terms of computing requirements), the
user's confidence in its physical basis and in its predictions. Other
desirable features may to versatility, numerical stability, and the knowledge
that it has been checked against acceptable experimental data.
Models for assessing what may happen for a clearly defined set of circum
stances may differ considerably in their predictions. One reason being that
each model makes various assumptions about the nature of the complex physical
processes involved. Often the physics of the processes are not well understood and their description in the model involves a good deal of empirical
input. This input often comes from experiments carried out on a much smaller
scale than that of concern in the industrial context. Furthermore the
appropriate scaling laws may not be well understood or cannot be completely
satisfied on a different scale. In these circumstances one approach is to
carry out 'largescale' experiments in order to provide data to, 1) enable
predictive models to be validated, 2) further the understanding of the
physical processes and test model hypothesis and 3) provide information on

scaling criteria so that better use can be made of laboratory or small scale
tests.
user may also need to know that the model is 'valid', ie. that the
predictions of the values needed to assess a particular hazard agree with the
results of representative experiments to an extent which is consistent with
the uncertainty in the available data, and the accuracy required for the
hazard assessment. The situation is by no means clear cut, and it is to be
expected that competing models will differ in the degree of validity they
have, according to the parameter that is being considered. For instance, a
model may predict one property with acceptable accuracy but may not do so for
others, or it may in achieving acceptable results for one parameter require
unrealistic values to be used for others. Perfect agreement uld not of
course be a reasonable objective and acceptable tolerances or error bands need
to be imposed from the outset.
Moffat(2) asks the question, "theri a good theory and a good experiment
differ, how much difference can be overlOoked before one must conclude that a
disagreement exists?" He concludes that the question cannot be answered until
The
the uncertainty on both counts is properly documented.

In controlled
experiments may be estimated and reduced by appropriate design. However in
modelling the sources of error lie in the assumptions and simplifications used
to arrive at a tractable model, in the empirical constants used, and with more
complex models in errors associated with the numerical end computational
procedures.
it
C5
APPENDIX C
2.2 Model
Validation
majority of the pysical processes of interest, in the context of

this paper, are statistical in nature rather than
eg. the
flows.
character
isation
structure and nature of turbulent
consequently any
must resort to statistical principles. Ideally the mean of any property
should be defined from ensemble averages, but in practice this is technically
impractical on the large scale, as
requires many replicate runs of an
in
assess
the
experiment order to
variability. Howaver it may be possible on
the rmall (laboratory) scale under carefully controlled conditions. This
raises the question of what to accept as an alternative averaging procedure.
The position is by no means clear but usually the parameters are assumed to be
statistically stationary, and timeaveraged quantities at a point are substituted.
It may be argued that the aim of model validations should be to determine
whether predictions of a particular parameter fall within the observed error
bands of that parameter with the frequency indicated, and if not whether the
failure is attributable to sampling fluctuations or is due to the failure of
the model hypothesis.
Models may be assessed in a number of ways such as by direct comparisons
of model predictions with the equivalent parameters derived experimentally.
There are no hard and fast rules and tselectedt trials or data sets are chosen
for comparison. There may be good reasons for doing this ie. some trials
within a series may not have produced a complete data set possibly because of
instrurentation failure. Nevertheless, for an objective assessment of model
performance one needs to be satisfied that there has been no jndicious choice
of data.
Other assessment methods are comparisons of nondimensional quantities
(arising from the framework of a model) with observations, or parametric
sensitivity stndies of models. Both of these may prove particularly useful
rm,
The
it
for understanding the scaling properties of models and the examination of the
validity of the modelling of the physical processes and process parameters.
Finally the optimisation of empirical constants within a model by comparison
with experimental data, although not
speaking model validation, can
strictly
shed
light on the way the models perform.
2.3
rimental Reouirements
?ny experimental data set

nature must ideally be regarded
relating to events that are stochastic in

as one of a number of replicate trials under
the nominally identical conditions necessary to obtain ensemble averages.

Usually a large number of replications (possibly 100) are required to obtain
stable estimates assuming that identical conditions can be reached in the
C6
APPENDIX C
place. In many situations, particularly where turbulent fluid flows are

of importance, true replication cannot be achieved, but in dition there is a
practical constraint of cost, especially with large scale trials. Often the
test conditions for small scale laboratory tests, although not necessarily an
exact simulation of the real event, can provide replicated conditions at a
reasonable cost. Experimental trials are often planned within a framework of
extending art already existing data base. Thus taken as a whole, the extant
first
and laboratory experiments provide a range of observations which may be

appraised collectively, and as such should provide more information on aspects
such as scaling criteria than is possible from an experiment conducted in
field
isolation.
Any test data not only needs to be as comprehensive as possible within
the constraint of a budget, but it must also provide core' data which can be
used for testing as wide a range of models as possible. The test conditions
must be clearly established and defined as precisly as is practical. The
results provided should include estimates of random and systematic errors for
individual sensors, combinations of sensors, the data collection system, and
the processed information.
LOSS OF CO'rAINMENT
it
is the ability to store, transport and process certain

Commercially,
gases in liquid form which make them attractive. Howaver from a safety
is this which makes them so hazardous, since for a given vohnrie
viewpoint
of material stored, liquefied gases generate the largest vapour clouds on
release. The manner of the release contributes significantly to the degree of
hazard. The worst case is a pressurised liquefied gas release at ambient
temperature as the sudden release to athospheric pressure causes a proportion
of the superheated liquid to flash evaporate rapidly to vapour. If the
blast wave into the
expansion is rapid enough the vapour may drive
of
hazard
air.
Thus
the
aim
assessment
must be to
surrounding
any
primary
maintain the integrity of the storage system. Howaver, the consequences of
any loss of containment must automatically be considered, and steps taken to
minimise them.
it
3.1
Partial and
Total
Losses
loss of containment, particularly with

pressurised liquefied gases, are often classified in terms of whether a total
or a partial failure occurs. The former implies sudden rupture of the vessel
leading to the isrnediate release of the whole of the contents. A partial loss
implies a jettype of release, the nature of which is dependent upon the size
and location of the failure. A relatively small leak from a vapour space may
The
consecuences
of
any
C7
APPENDIX C
result in an all vapour release in the form of a highvelocity jet nission.

whilst a larger breach in the sane region
result, at least initially, in a
high-quality twophase flashing jet, due to liquid carry over brought about by
boiling and swelling of the contents as the pressure falls. A loss fran the
liquid region may result in either an all liquid jet which may subsequently
flash evaporate, or the establishment of twophase flashing flow through the
breach itself.
release of an initially saturated liquid fran a pressure vessel,

either as a result of a total or a .partial failure, is subjected to a decaying
pressure field. Consequently the liquid accelerates as well as flashing. The
latter is a relatively rapid process because the heat required for the phase
transition is available within the liquid itself. Hence, an adiabatically
flashing liquid always results in a twophase mixture at a lower tnperature
than the original (vessel) temperature. Consequently the process is governed
by an interaction of hydrodynainic and thermodynamic phenomena, in which non
equilibrium effects predominate in the early stages of the process.
The
3.1.1
A
Modelling
total losses
total loss is usually termed

BLEVE
The
overpressures
(boilingliquidexpandingvapourexplosion).
generated are
liquidvapour release following
usually minimal, however tank or vessel fragments may be projected over large
distances because a relatively high proportion of the energy released is
to the fragments.
Upon catastrophic failure the vessel's contents are assumed to be
released in two stages. An initial rapid (adiabatic) depressurisation occurs
during which the pressureenergy causes a fraction of the liquid to flash to
vapour (flash fraction). A large cloud is thus formed ccmprised of both
imparted
liquid droplets, some of which fall as rainout.

two phases may
be in thermal equilibrium nor moving at the seine radial
velocities. During the secoud stage air entrairment, turbulent mixing, and
heat transfer effects predominate, resulting in further liquid evaporation and
an increased size of vapour cloud.
Theoretical studies of the expansion processes are described in the
literature(34). Generally speaking these assume that isentropic flash
evaporation of the superheated fluid occurs initially with both thermal and
velocity equilibrium existing between the vapour and liquid phases. Enhanced
turbulent diffusion coefficients are often used to model the increased
turbulent mixing which occurs during the entum-daninatedexpansion phase of
the cloud formation process.
Experimental studies are reported(5G) in which scaling, cloud growth,
vapour and
not necessarily
energy
partitioning and turbulent mixing effects have
C8
been examined. In the
APPENDIX
majority of these studies releases of fluids such as propylene, propane,

butane and freons were made from spherical or cylindrical vessels under
conditions.
Sane experiments have included measurenents of
overpressures and combustion behaviour. The practical difficulties of scaling
up from relatively small experimental releases to the industrial scale are
recognised. However the costs of large scale trials are likely to renain
prohibitive unless undertaken on a collaborative basis.
In some industrial accidents, unexpectedly high overpressures were
reported, which has been suggested(7) may be as a result of a high initial
liquid superheat combined with rapid depressurisatiOri. This may lead to the
superheat limit locus being exceeded resulting in homogeneous nucleation and
an increase in the intensity of the explosion. Although this is possible in
theory, the experimental evidence available at present is not wholly
controlled
it
sUppxDrtive.
3.1.2
Modelling partial losses

The jet release resulting from
partial vessel failure may be

inlet quality and the geometry
either
of the
single or twophasedepending upon the

Mass flow rates, pressure losses, etc for single phase releases are
obtained using existing well documented and validated modelling procedures for
breach.
liquids and gases. The storage pressures utilised in practice will

result in sonic or choked flow limiting the mass efflux rate with all gaseous
releases.
If in these circumstances the exit pressure
greater than
both
is
ambient,
then
further
expansion occurs outside of
resulting underexpanded jet may becomesupersonic
decay of
the breach, and
the
locally. The structure and
underexpanded jets has been reviewed recently by Ramskill (8).

The dispersion of jet releases has been studied extensively. No stages
to the process are usually identified, firstly a high velocity jet or
momentumdominatedphase, and secondly a plume dispersion phase when the jet
velocity is comparable to the windspeed. During this latter phase, buoyancy
and abtiospheric turbulence control the mixing and dispersion processes. A
number of recognised modelling procedures are currently used for assessing the
dispersion, these include similarity or integral procedures. Complex 3D
turbulence field models have also enjoyed a limited amount of success. ?re
recent developments have attmopted to incorporate combustion effects into the
turbulence modelling schene.
All vapour jet releases, because of the rapid mixing are, in the case of
flarmnables, soon diluted to concentrations well below their tEL. Thus when
is usually only the dispersion of
plumetype behaviour becomes predominant
toxic materials that is of importance. This may not be the case with two
it
C9
APPENDIX C
phase
jet releases, as fig 1 illustrates. This shows a release of liquid

at 5 bar pressure in a caliti stable atoosphere. Despite the initial
propane
jet momentum and air entrainment these have not proved sufficient to prevent
buoyancy effects fran dominating, and a large flnmable heavy gas cloud has
developed at ground level.
In contrast with single phase
jet releases, the currently available

knowledge of the dispersive properties of twophase flashing jets is united
as a recent review by Appleton(9) has highlighted.

Considerably mare
parameters influence the structure of the jet. For instance the geometry of
the opening influences its
composition because
governs the degree of
flashing which occurs prior to release.
Nonequilibrium conditions are
considered to exert a significant influence on the spatial and tanporal
droplet size and velocity distributions in the liquid phase. The observed
sudden expansion of the jet is assumed to produce a change in the turbulence
it
levels and momentus exchange rates.
Current modelling methods have either applied the basic conservation

equations globally under equilibrium cooditions, or have applied the same
principles to the two phases separately.
The latter approach has been
developed extensively in the nuclear industry for examining multiphase blow
down phenomena in pipes, as reviewsd
wolf(lO). He concludes that numerical
codes offer considerable potential, bat whilst canputing costs
high,
attention is likely to focus upon devising more efficient and economical
solution algorithms, together with better defined interphase transfers
coefficients especially with regard to their realisatjon and range of
applicability. Experimental work may also focus upon this area aided by
advances in optical techniques.
3.2
rain
Flammable and Toxic Peleases
The direct risk to people and nearby plant structure from fire is burning
flame or by thermal radiation. The extent of any burning cloud
by direct
following a total sudden release of a pressurised liquefied flammable gas is

substantial, and as a consequence so is the fire or thermal radiation source.
This is also true of a burning twophase flashing jet release, where the
observed flame lengths are considerably greater than those associated with an
all vapour jet from the same size of orifice and with the same driving
pressure. The surface eiissive powsrs of burning jets or fireballs, because
of their highly turbulent structure, are usually considerably greater than
those fran liquid hydrocarbon pool fires. Values in excess of 200kW/m2 have
been observed from burning jets of LPG.

Toxic releases can result in the production of large lethal clouds, which
do
APPENDIX C
4'
0
0
0
0.
ci
0.
0.
a-
4'
0
0
9.
LL
APPENDIX C
harmful for distances well beyond the local confines of the

Storage site. The nature of the hazard presented is dependent upon the
conditions of exposure. Toxic releases, unlike flamnable ones, are harmful
only if direct exposure of the target occurs, however muoh greater dilution
(to a few ppo) is required before the substance becanes harmless.
Toxic releases are almost invariably denserthanair and their effective
dilution and dispersion is very dependent upon the abnospheric conditions
pertaining at the tine of release. This Irt.ay riot of course be true of toxic
releases which occur during a fire when the buoyancy within the fire plune can
be expected to daninate the early stages at least of the dispersion process.
may be potentially
Toxic releases, initiated as a result of a total vessel failure, lead to

the immediate formation of large vapour clouds, with relatively little
dilution, On the other hand, jet releases fran partial vessel failures have
high initial momerituni, which act to increase the rate of atmospheric
entrainment and hence the degree of dilution achieved.
A considerable amount of effort has been expended over the past few years
on improving our understanding of the dispersion processes of negatively
buoyant and passive releases. Both experimental and theoretical studies are
(11),
reported in the literature, as reviewed by
eatley
)DELLING FIRE
Fire
ENGULFED SWRAGE
T7KS
is the most likely event to initiate a vessel failure which leads to
release of the contents.

Although other initiators, such as Impact,
from
an
overpressure
explosion, or earthquake damage must also be examined in
storage and transport assessments. Fire attack on a pressure vessel creates
an overpressure and eventual failure due to structural weakening, as all
steels at sufficiently high temperatures, undergo a reduction of their yield
and ultimate strengths. However, at sustained elevated temperatures steels
also undergo creep, and creep strengths may become the limiting design factor.
Fire attack may be one of three modes or a caTibination of modes: 1) total

or partial pool fire engulfment, usually the result of a 'largeliquid' leak,
2) jet flame impingement, resulting fran the burning of a single or twophase
leakage under pressure or fran the controlled flaring of discharges through
pressure relief systems, 3) thermal radiation fran nearby or offsite fire
events. There are consequential effects with jet flare impingement such as
increased ventilation of any additional pool fire attack, high thermal
stressing, and distance effects, ie. noise and thermal radiation.
An assessment of the effects of the received heat upon the vessel and its
contents requires knowledge of, the modes of heat transfer, any thermal
stratification which occurs within the contents, and the bulk or subcooled
C12
APPENDIX C
-'
pit
44
VJC)
a)
CW
-4
Cl)
-4
'S
U
U
a_)
0
r-4
U
U)
00U)0
a) --4 a)
.J
.4
't0
(C)
Q ,..4
4)
I(C)LIU)
(C)>i--4
<C
U)
.)Qt1.4
s
Qua)
W'-4a)
Q4-
'
.i
-'a)
-U
c:-2
0
'n
.&J
-vU
LI
.-4
U).
LI V 1-
'0 0
U),-4>,
U)
COtC)
C
C
-.4 U)
LI
CCO.
LI
C4)
j
-i
CC
CC
4-J
'
'S
'S
CO
.-4 U)
Ca)
.-s
..4 ..4
-.1
jj
UC
,-, ...i
'
0j.ICfl
CO(I)
C-4.-4
V C.
CC
CWCO'C)
.4 ..
.u
'C)
C)
C.-4
--.4
CCC
.'
''
>
U)
C))
C))
U)
U)
-----.n
iFI .11
P--L"
4
:!:: :::::
:I!::
(Y)
U)
"
Ca)
LI
Ca)
fl
LI
.8
4)
'S
C.
4)
(0
4J
LI
4-
(C)
Cl)
-4 -.4 4)
-wa
'
'0
C1
a) --4
0
4)
-U
a)
CU)X
-4
a)
4)4)
Oc3
4)
aj
LI
C)
Ca)
.1 ..-4
.c
-.
'C
4'-lO
-0
'C
cn.-i
...
%-'
4)
I1
t:t
:c!!::: ::::::: :&::::!:::E:I:::::: :::::iz
APPENDIX
the pressure relief system has operated. Interactions between

the vent operation and the thermodynamic and hydrodynamic states of the fluid
contents cannot be ignored, as the establishment of twophase flow within a
vent system has a significant influence on its overall performance and relief
boiling
once
capacity.
currently available models known to the author, and which assess the
thermal response of both the vessel and its contents when subjected to fire
attack, are listed in Table 2. All of the models are presented as crmpoter
packages, some of which are 'menudriven'. iarw will operate on micro's as
The
mainframe cemputers, although the run times may be

an overall assessment
prohibitive especially when repeat runs forming part
well
as mini
and
of
are
contemplated.
4.1 Engulfing Fire
specifically for total pool

fire engulfment, some however can
relatively easily to model
a
partial engulfment or radiant heating. In pool fire models assume that the
tank is engulfed to varying degrees by flame and hot cembustion products.
Thus the tank receives both radiant heat frcm the flames and convective heat
Most
2 cater
be modified
of the models listed in Table
absolute values and percentages of these two

ccponents depends in practice upon a number of parameters such as, 1) the
nature and type of fuel, 2) its cembustion characteristics, 3) the local
meteroingical conditions, 4) the size and extent of the fire, 5) optical
thickness of the flame, 6) the size of the tank, and 7) the thermal properties
of the tank. Thermal data pertaining to various types and sizes of pool fires
frco
the
hot gases.
The
are available in the literature as reviewed by Mudan(l2). Modelling methods

for specifying the fire flux require the average heat flux received by the
tank to be specified over the area of engulfment. This may take the form of a
direct beat flux specification or as a radiative cemponent requiring a
representative flame temperature and a surface absorptivity to be specified.
Clearly such approaches are not an accurate representation of the cemplex
interactions which govern a true fire. However, they are considered to
provide a sufficient degree of realism. They enable the user to exercise an
engineering jugement of what the fire conditions are in any particular
situation. Spatial and temporal heat flux variations (which characterise a
typical fire) are ignored in view of the thermal capacity of the tank and its
used by Sousa at al(13) allows circumferential

radiative and convective heat flux canponent variations to be specified.
Hydrocarbon pool fires are usually quoted as having average heat outputs
contents.
An enhancement
in the region of
100120 kW/m2, the lower value is used in most codes of
C14
APPENDIX C
as the basis for sizing pressure relief valves to cope with a fire
attack (rio account being taken of nonuniform heating effects, nor of the
practice
tiperatures likely to be attained).

The burning jet which results fran the ignited release of a vapour or
vapour/liquid at pressure, may have a higher heat flux output than a pool fire
due to increased turbulence and mixing. Should such a jet engulf a tank then
the consequences nay be more severe than those of pool fire engulfment,
because of the localjsed nature of the fire attack. A similar characterisa
tion is required fran a modelling view point. The total heat flux is again
maximum wall
considered to be canprised of radiative and convective canrxnents. Howaver

the energy partitioning may differ, and may not be constant. The heat flux
distribution under the 'walljet' is nonuniform arid the hot gases can reach
zones of the tank woll away fran the direct impinganent zone. The physical
modelling of this situation is highly complex even if steadystate conditions
are assumed to exist. The assumptions of one and twodimensional heat
conduction through the tank walls are not valid and a threedimensional
solution must be sought, usually based on finite difference or finite elnent
solution procedures. Alternatively if a uniform heat flux is assumed over the
heated area then a network analysis approach may provide representative
results, as suggested by Ramskill(14).
The received radiative heat flux fran a distance source is usually
treated as an area or point source of constant flux, which is calculated by
developed methods. Atmospheric tranamissivity can produce beat flux
variations but these can also be estimated and accounted for. The surface
absorptivity of the receiver may also be considerably less than unity, the
value usually assumed for flame engulfment. Flames which are selective
anitters may require a more rigorous analysis in order to obtain a representative heat flux. In some instances the target may be close to a boundary
consequently the respective view factors of the source to target and vice
versa become significant when assessing the radiative heat exchange.
ll
4.2 Wall Heat Conduction
of the current models assess the heat conduction through and within
the vessel walls by solving the 2D heat conduction equations in the radial
and circumferential directions, using finite difference solution procedures.
Although an alternative lD solution procedure with single nodal points in the
liquid and vapour walls is used by Ramskill(14) for analysing total fire
Most
engulfments. In practice considerable tanperature variations, especially in

the vapour space tank walls, are to be expected and have been observed. These
are not of course only dependent upon the nature of the fire, as compositional
c1s
APPENDIX C
163
changes
in the internal fluid
Operation
of the
tank
influence the
boundary
etc., can
layer,
all create
modes of heat transfer, and

feedback effects which will
wall conduction.
4.3 Heat Transfer to Contents

As the tank liquid and vapour walls becase fire engulfed, heat is
transferred by convection and radiation to the liquid and vapour contents of
the tank. The modes of heat transfer to the liquid are, depending on the bulk
liquid and inner wall tenperatures, either convective, SUbcoOled/saturated
boiling or film boiling.

Initially heat is conducted into the liquid, but after a short time, the
nucleate
predaninate and convection becanes the dcwinant mode of heat

Both convective heating and nucleate boiling create a buoyancy
driven flow in the region of the walls, which can form thermally stratified
layers in both the liquid and vapour regions. Current modelling procedures
buoyancy
forces
transfer.
utilise npirically
based free convective heat transfer correlations to

represent these regimes. Once saturated or subcooled nucleate boiling is
predicted then alternative enpirical heat transfer coefficients are asployed
such as those proposed by Rosenhow(l5). Sane correlations require knowledge
of various fluid properties, which are thasselves dependent on the thermo
dynamic state of the fluid. Thus either analytical property routines or a
data base must be referred to at each time step of the solution procedure in
order to calculate the appropriate heat transfer coefficients.
In practice the heat transfer coefficients are also dependent upon other
parameters, such as the type of surface and its finish, its inclination and
curvature. The geanetrical scale, the internal pressure, and the presence of
dissolved gases are also known to influence local heat transfer values,
Butterworth(l6). These factors are not accounted for specifically in the heat
transfer relationships normali:y used but there may be an implicit dependence.
Models which allow heat transfer coefficients to vary around the vessel walls
in a specified manner (often in relation to the wall tnperature distribution)
have
implied dependence.
Sane models allow for the
an
possibility that stable film boiling may exist,

heat transfer correlation and increased wall
tenperature. Film boiling of liquid propane is considered unlikely during the
normal heating and boiling cycles associated with typical hydrocarbon pool
fires heat fluxes of around 100 kW/m2. this is particularly true if
subcooling is present, as this increases the critical heat flux necessary to
trigger film boiling thus taking it further away fran normal pool boiling
fluxes. Howaver it has been suggested by Venart(-7) that stable film boiling
by using an appropriate
C16
APPENDIX C
flashing and frothing accompanying the opening of

the PRy, because the increased vapour likely to be present in the near wall
region may allow wall burnout to occur.
Radiative heating of both liquid and vapour by heat transmitted through
the hot vapour walls is included in the majority of the modelling procedures.
Radiative heating of the liquid results from that proportion of the radiation
flux which is transmitted through the vapour. It is usually considered to
form the major proportion of the radiative heat transfer, because of the
transparent nature of the vapour.
could
result from the
sudden
Operation and Bulk Boiling

Evaporation from the liquid/vapour interfacial layer governs the internal
pressure within a vessel, so that
operation may occur before bulk boiling
has been achieved. However the pressure drop following PRy opening may be
sufficient to induce liquid flashing due to superheating. This can result in
a twophase swelling of the liquid region as vapour bubbles rise through the
liquid. In vessels which are nearly full initially the liquid swelling could
have a considerable influence on the quality of the fluid entering the vent
A cQmnonly used assumption is that a homogeneous vapour/1iquid
System.
mixture enters the vent. Experimental evidence from SalletO-8) suggests that,
for normal venting operations, liquid carryover does occur and gives a high
quality flow into the vent. 'Ittis becomes all vapour as venting progresses and
the liquid level falls. Oirrent models assume (as do the current codes for
relief valve sizing) that only all vapour choked flow takes place through the
vent. In some models le. Beijnon,(-9) this can be treated as a superheated
vapour rather than one at the equilibrium temperature (as is current vent
4.4
PRV
sizing practice).
4.5 Mass and Energy Conservation
a quantity of mass is lost from the vapour space through the vent
system a similar anount is evaporated off from the liquid region. In all
models this is calculated from the conservation equations for mass and energy,
Different solution
which are coupled together and solved numerically.
of
the
sets
of equations to
procedures are used depending upon the complexity
be solved, saturation conditions are assumed to. exist throughout the liquid
region during boiloff, and an appropriate saturation curve defined to
facilitate the solution procedure. This avoids numerical instabilities and
gross errors particularly when coarse time steps are used. None of the
solution procedures can be solved graically or by hand calculations.
The computer programmes predict the thermal response of the vessel and
As
Cu
APPENDIX C
a function of tine, particularly vail temperatures, time to

internal
opening,
pressure and liquid level. In some cases failure
criteria are activated, which predict the t]iae to vessel failure, these are
based upon the reduced mechanical properties of the vessel walls at elevated
its
contents as
PRV
temperatures.
4.6 Tank Fret
tion
Systems
are several methods in current use for protecting LTh storage

vessels froni the sxst effects of fires.
These include water spray
protection, covering the tanks with thermal insulation and burying or sounding
of the tanks. All three methods reduce the heat flux into the contents of the
tank and maintain the tank walls at relatively safe temperatures particularly
in the vapour space region.
efficient insulation may reduce the heat flux
by a factor of 10 or more. Vater spray protection, when applied uniformly at
the currently reccnended rates, can be expected to maintain wall temperatures
There
at around 100C.
The presence of thermal insulation can be incorporated into modelling

A
procedures as an ditionai thermal resistance surrounding the tank.
procedure to model water spray protection is included in the compiter codes
described by Ramskill(l4). The water spray protection is represented as an
evaporating layer of water upon the tank surface. A uniform coverage over the
whole of the tank is implied. This may be difficult to achieve in practice,
because of wind effects, the need to ensure that all the spray nozzles are
rking properly, and structural features unique to a particular installation.
Range of Applications
The utility of a computer model is enhanced
it can be used relatively
for
a
of
fluids.
This
is
in
most
cases as the physics
easily
range
possible
of the models are general, and apply to most fluids, hydrocarbon or otherwise.
4.7
if
A range of fluid
and
in
property routines are therefore incorporated into some codes,

some cases the heat transfer correlations are also modified. A further
in some
ability to predict the response

vessels containing either multicomponent fluids (ie. petrol) or permanent
gases. A model to predict the response of vessels filled only with permanent
enhancement included
codes(14)
is the
of
gases
is given by Nylund(20).
Most codes have been designed primarily to predict the behaviour of

cylindrical vessels mounted horizontally or nearly so, yet many installations
contain spherical vessels or cylindrical ones mounted vertically. Models may
not be readily applicable to these alternative designs, as the geometrically

dependent behaviour is not necessarily the sane. Howaver, the thermodynamic
C18
TABLE
3 Recent fire engulfment experiments
Responsjble
Body/Author
Date
Bray
1964
Tank Size/
Sca1e
5T
ContentsEnvirorent
Water
Outdoors
Measureiients
cwr (V/L)
BFT
AARRPI
1973
l/5 scale
AAR-RPI/ABR
1973
'Full
Anderson
Proparze
scale
Outdoors
Propane Outdoors
64T
Transport
1980
1/5th scale
Canada
Propane
Outdoors
HSE/Shell
Cutler
1980/81
Water
Outdoors
Propane,
l/4T
Williams
cwr
(V/L)
IWT (V/L]
Unprotected railcar.
Tested to destruction.
BET, P, WHF JP4
FHF, Fr, FD
cwr
(V/L)
GWT
l/4T &
1981/82
Droste
lT
2T
1983
Water
Propane
Outdoors
(V/L)
G?T (V/L)
Fr, p
cwr (V/L)
Propane1
FT
'DIERS
Fauske
1983
Venart
:198386
Lab
40
Unprotected and
protected
BET, P
BET,
fuel.
1985
Billinge
1985
Freon's Indoor
Laboratory
tests
1/4T
Propane
Outdoors
BET, P
Full kerosene fire

engulfment 5tests
Unprotected.
3tests to destruction.
Pool
fires.
Test rig electrically

heated. Exnined
I.C, WHF
{therrnal response of
cwr (V/L)
BFT,P
Water
Indoors
cWT (V/L)
3.Om Dia.
Gases
Outdoors
P,' FT
ST
Propane
Outdoors
cwr (V/L)
1/5th scale
0.5m3
1985
Nylund
HSE/Shell/
1985/86
'Moodie
'Cowley
Water spray
protected tanks.
Nm'erous tests.
Full kerosene fire engulf

5tests different
Unprotected.
ment
fills.
PD, Th, FT
tp
l1
alk fluid
BET
P.
IC
FT
s.11
Pressure
WHF
temp
temp
Tank weighed by load

Fire temperature
FHF
PD
cells
Wall
heat flux
Fire heat flux
Flare data
Thermal radiation
VVapour LLiquid
ciq
vessel
contents.
Pool fire test to assess

BET, FD, TR flare effects. Radiation
WHF, FT
feedback.
EWI' (V/L)
BET, P, ir
WHF, FHF,
cwr (V/L) Outer

IwT (V/L) Inner
to test
Tests
FT
Davis Eng
(Explosafe)
fire.
Protected tanks
Insulations.
Nurerous tests.
:and
HSE
pool
JP4
FT
HSE/Shell
Cutler
Moodie
Water spray protection

test. Kerosene/oil fire.
tjnprotected tanks
BET, P
FT
Appleyard
Cmnents
APPENDIX
physical processes are similar, as are the overall thermodynamic

responses. The first of the codes attributed to Ramskil].(14) utilises a
'volume equivalent cube' to represent different shapes of vessels. The DIERS
developed medel SAFIRE, the venting aspects of which are described by
Forest(2l), also applies to different shapes of vessels.
and
LPG STORAGE
FIRE
JLF
EypERIMETS
the best of the author's knowledge the published experimental work

carried out over the last few years is as listed in Table 3. Only experiments
directly related to simulating intotal the fire engulfment of LPG type
containers are included, irrespective of scale. Many other fire engulfment
experiments have been undertaken before the period considered.
Sane
experiments go back to 1925 and are the basis of the pressure relief valve
sizing methodologies utilised in the main codes of practice.
The largest fire engulfment test so far reported appears to be that
described by Anderson et al(22), in which a 64 tonne rail car was fire
To
engulfed.
recorded,
Vessel wall temperatures for both inner and outer surfaces ware
together with the bulk liquid and vapour temperatures, and the
average heat flux fran both the
the .V flare. Measurements ware

of the liquid level in the tank and the thermal radiation fran the
PRV flare.
The vessel failed catastrophically some 24.5 minutes into the
test, with an estimated 40% of the LPG still remaining in the tank. Fragments
fran the vessel ware reported at considerable distances fran the test site.
The maximum wall temperature near the point where the vessel started to fail
was of the order of 650C and the internal pressure was 24.1 bar. A series of
-/5th scale trials of this rail car test are also reported in the
literature(23)
fire and
also made
Fire tests on amaller capacity vessels (2 tonnes) have
ben reported by
Federal Institute for Material testing, Berlin(24). Three fire tests ware
for Lp(, stored in accordance with the appropriateDIN standard, and
again vessel skin temperature variations with time ware recorded. In all
these tests tank failures occurred within 7 to 12 minutes fran the beginning
of the fires, depending upon the initial temperature of the TIG.
Recent
studies by the same orgaruisation include fire tests on water spray protected
the
canpleted
and insulated tanks.

Extensive
alC1.7).
laboratory simulations have been
pressure vessel was

and
undertaken
An electrically heated 380 an diameter, 40 1
contained
used,
it was
fitted with
Freon 11
observation
by
Venart et
capacity cylindrical
windows
at both ends
or Freon 12 to simulate the LPG loading. It was

instrumented
so
various aspects of the thermal response could
extensively
that
C20
APPENDIX C
be studied and compared with theoretical predictions.
Although not pressurised

theoretical and experimental
liquefied gas storage, Nyiund(20) has made

studies of the behaviour of pressurised gas
vessels
to
both
total and partial fire engulfment. His
process
subject
predictions of vessel failures within a relatively short time after the
beginning of the fire were confirmed experimentally.
The
undertaken by HSE(2528) covers a range of tank sizes up to 5
tonnes capacity. Kerosene pool fires were used in all of the tests to provide
rk
realistic total fire engulfment. The tests were conducted at an open moorland
site. Uninsulated tanks of l/4t, it and 5t capacity were tested. l/4t tanks
were used to assess the effectiveness of various types of insulation materials
such as mineral weol blankets and epoxybased intumescent coatings. A l/4t

tank was also used
assess the efficiency of water spray protection systems.
Vessel wall temperatures were measured for both the vapour and liquid wall
regions, together with the bulk liquid and vapour temperatures, and the
internal pressure. The 5 tonne tests were iire extensively instrumented than
the mualler capacity tanks, this enabled other parameters such as the heat
flux arid the internal vessel stratification to be measured. The discharge
rate through the pressure
relief system was
also obtained
from the weight loss
5t tests
5t tests.
recorded from load cells (it and

only). Vent stack exit velocity
measurements were made during the
The radiation emitted by the
flare was also characterised, from radiometer data and photographic records.
fire test performed on a scale

model of a rail tankcar, to study the influence of a burning flare of LP( on
the tank and engulfing fire. The test used a l/5th scale waterfilled tank
Davis
Engineering(29)
report an
indoor
with JP4 as the fire fuel.

The instrumentation measured tank wall
wall
heat
bulk
temperatures,
fluxes,
temperature, fire temperatures, heat flux
to remote targets, and entrained air dynamic pressures. The results indicated
that in the absence of cross winds effects, the burning flare had little or no
effect on the tank engulfed in a sooty fire. However the overall fire size
was increased.
A series of outdoor l/5th scale tests are also reported by Appleyard(30).
Fire engulfment trials were conducted with unprotected tanks (Enigma Series)
and with Explosafe protected tanks (Nova Series).
The main parameters
measured were the internal pressure, wall temperatures, and the liquid level.
Bray(3l) undertook a series of fire engulfment trials, on a water filled
vessel some twenty years ago. The p.irpose was to ascertain the water spray
rate necessary to prevent excessive skin temperatures. The test tank was
extensively instrumented with thermocouples. A kerosene/oil spray burner was
used to provide the source of heat. The results of Bray' s werk have provided
C21
APPENDIX C
a design basis for water spray protection systems in the UX.

The DIERS project(32) was concerned mainly with the sizing of vent
systems for vessels containing range of substances both reacting and non
reacting. Included are design methods for sizing relief valves to cope with
fire engulfment(2l).
is understood that some laboratory scale experiments
were carried out for validation purposes, but these have not yet been
reported.
5.1 Comments on Test Data
Collectively the experimental results currently available provide a
comprehensive data set, ranging from laboratory simulations, through to
realistic pool fire engulfment trials on propane filled tanks from l/4t to 64t
capacity. ditional data is available from fire tests with insulated tanks,
flares, water spray protected tanks, and gas filled pressure vessels.
5.1.1
Flame and
wall tesperatures
influence of prevailing meteorological conditions was apparent in

of
the
many
pool fire engulfment trials, as considerable spatial and temporal
flame temperature variations were observed at different locations within the
fire zones. Although this to some extent reflects the highly turbulent nature
of the flame and combustion processes, it meant that there were considerable
variations in the observed tank vapour wall temperatures, both across the
tanks and from end to end. Behaviour of this nature presents difficulties
The
interpretating model predictions, especially failure predic

tions, because of the lD or 2D assumptions contained in the models. A
similar degree of variability was not usually observed between liquid wall
when making and
temperature measurements, presumably because of the thermal mass of the
liquid.
Wall temperature measurements for unprotected pool fire engulfed LEK
tanks, at the time of failure or when was thought imminent, showed that the
vapour region walls reached peak temperatures of around 600C. The actual
values are dependent upon the internal pressure and the strength of the
vessel. This is illustrated in Table 4, which
the maximum wall temperatures and the corresponding pressures. Also given in Table 4 are the times to
PRV opening, percentage fills, initial propane loadings, and the fire heat
fluxes. The interdependence between these parameters is complex, but
generally speaking for a given heat flux, the mnaller the vessel and the lower
the percentage
then the shorter the time to failure or potential
failure. The times to initial
opening appear to be both specific heat
flow rate and test dependent, although they are similar for all the reported
it
lists
fill,
C22
APPENDIX C
TABLE 4
Surtrnary
of experimental results unprotected tanks
II
il
I
I
Originator(s)
Initial Maximum 'Pressure at( Time to Calorimeter

max. Taap PRV first Heat Flux
Tank Size Quantity)
(bar g)
(kWm2)
& Percentage)of r& )Tnp (0C)1
opening
Fill
(sacs)
(litre)
HSE
l/4t,
It,
Nauinal
40%
20%
40%
it, 80%
5t, 20%
5t. 40%
5t, 60%
5t, 80%
185
308
789
1,635
2,250
3,676
5,900
7,644
600
570
620
680
635
657
610
572
35
7
8.8
6.3
II
10
ii
11
190(1)
297
262
226
373
415
401
312
64t, 95%
121,500
650
25
160
2,820
2,820
245
573
340
2,820
460
420
660
90
it,
(Anderson
Droste
2t,
50%
50%
2t, 50%
2t,
Davis
Engineering
500
I
.98m3
800
99
101
96
104
570
13
(Enigma)
(2)
(Bray
21.9m3, 90%
19,700
(1Inc1udestime for fire to establish itself

(2)Water
'
(2)
0.51m3
1
Appleyard
i3t
AAR
100
150
only
C23
120
APPENDIX C
peal fire tests on unprotected tanks.

5.1.2 Pressure relief
All
usually
of
the tests
in
were undertaken with
accordance with
pressures rose
pressure
a specified code of
relatively quickly
relief systems fitted,
practice.
once the fires had
started.
The
internal
The initial
often followed by a sharp drop in pressure before rising again to

maintain a pressure near to the set pressure. The level to which the pressure
Subsequently rose depended on the fill level and whether twophase flow was
occurring in the vent. In some tests(24) the relief valve capacity seemed
inadequate and the internal pressure rose wall above the set pressure before
Opening was
failure occurred.
5.1.3 Bulk fluid behaviour

The experimental results confirm that
the internal pressure was governed

Principally by coritions at the liquid interfacial layer. In particular the
evidence from iial1 scale laboratory trials(17) and the large tank car fire
trial(22) shows that thermal stratification exists in the liquid regions at
least up to the onset of bulk boiling. Liquid stratification can be impertant
when calculating the tines to initial
opening, the liquid behaviour
P1W
or
the
maximu
levels
of
fill which can be tolerated.
following
opening,
Considerable thermal stratification was reperted to exist within the

vapour space for all tests where measurements ware made. Similar patterns
ware observed in cost cases with the temperatures and the degrees of
stratification rising initially until the ?.V's first opened. They then fell
!mTentarily towards saturation, presumably as frothing or bulk boiling was
initiated, before rising again as venting prngressed. The resulting superheat
expected to decrease maximum vent discharge rates if all vapour
venting is taking place, which if not accounted for may lead to overprediction
of mass discharge rates. In practice there is some compensation if droplet
carry over or frothing occurs, as these result in a twophase mixture entering
the vent system. This has been observed in msall scale experiments reported
by Venart(i7).
In some tests, bulk liquid, boundary layer temperatures and wall beat
fluxes were measured. The majority of this data provides support for the heat
transfer modelling procedures discussed previously. The dominant node of
convective heat transfer is that of nucleate boiling.
Although the
differences between wall temperatures
the saturation temperature suggest
that in many instances considerable vapour generation was occurring in the
near wall region. High vapour containing boundary layer flows have been
levels can
be
C24
APPENDIX C
observed
in the
laboratory trials rerorted by venart(17).
proceedures usually assume that saturated bulk boiling occurs

this was confirmed by observation from many tests
during venting.
there have been some exceptions.
In particular during the 5 tonne
At
HSE/SHELL/BG tests liquid superheats in excess of 200 ware observed.
to
a
mixture
of
saturated
and
present this enomena is attributed
liquid
superheated vapour. The latter being entrained from the vapour region or as a
result of sloshing during venting. Although such behaviour may be peculiar to
one set of tests,
cannot be easily dismissed in modelling terms, as
may
have implications for future model developments and their applicability.
Modelling
ilst
it
it
5.2 Scaling and Averaging
currently available test data frcn pool fire engulfment trials

provides
range in scale (based on tank internal diameter) of about 1:8
The data allows those
(0.38m:3.05m) for horizontal cylindrical vessels.
to
be
examined
which
are
considered
aspects
particularly sensitive to scaling,
such as fluid properties, boundary layer development, wall to liquid heat
transfer coefficients, and possibly thermal stratification. The scaling
criteria can be established where possible and compared with theoretical
predictions. The laboratory and field data may be used to test the validity
of the scaling criteria through parametric sensitivity stiies, thus allowing
predictions to larger scales to be made with reasonable confidence. As a
consequence the range of model applicability may be extended upwards to
inclnde the largest sizes of storage tanks.
There are three components to the overall beat transfer process; heat
from the fire, conduction through the walls, and natural convection (both
liquid and vapour). In all three, appropriate scaling laws are derived by
nondinensionalising the governing equations.
There is a considerable anount of appropriate data in the literature
which can be utilised to test these scaling laws. Free convective flows, in
particular, have been comprehensively stodied and their scaling criteria are
wall established. Flow pattern predictions for free convective flows in
partially filled horizontal cylinders have been obtained using hydrodynamic
computer codes as reported for instance by Hadjisophocleous(33).
The questions of variability between data sets and averaging principles
have not been fully examined. Full scale statistically meaningful replicate
data sets do not exist nor are they likely to in view of the costs involved.
chat evidence there is suggests that ateospheric turbulence variations are
having a significant influence on the heat flux profiles to the tank and its
thermal response. However, the influence may go deeper and effect both the
The
C25
APPENDIX C
extent and stability of the convective flows in the liquid region. precise
laboratory simulations of field tests may help to clarify some of these
influences, particularly the consequences of asynTnetric heating, and the most
appropriate averaging procedures to
opted. They may also help to verify
scaling criteria.
The nature of the flow patterns which exist within the vessel during
field tests have not been examined to anything like the extent that they have
in the laboratory. There are considerable practical difficulties viewing
inside a tank during a fire engulfment test, and of measuring velocities and
phase compositions. Nevertheless the problu needs to be dresscd in view of
the potential benefits to be gained from such information.
5.3 Future Developoents

Experimental results suggest that thermal stratification in the vessel,
and twophase flow effects in the vessel and vent syst following PRV
actuation, have a significant effect on its thermal response which cannot
be ignored for modelling purposes.
The 'exacthess' of both average wsll taperature predictions and other
fire dependent parameters may need to be reapraised in view of the degree of
variability noted during and between 'similar' tests.
Future model developoents can be expected to dress further the problns
of different modes of fire attack, such as partial pool fire engulfment and
jet flare impingnent. Modelling the latter aspect may require comprehensive
threedimensional models in view of the nature of jet flares and the number of
different modes of impingnent. Suitable experimental data with which to
validate such models is not available.
The majority of current models are intended specifically for horizontal
alys
cylindrical vessels, yet many installations are either spherical or vertical

cylindrical ones. Models are costly to develop and any enhanced versatility
in this area is desirable ccxmercially.
Further model developoent is
therefore likely, as also is experimental validation in view of the current
lack of data for other than cylindrical vessels. The extension of models to
handle multicomponent fluids and gases rather than single component fluids is
also likely to receive further attention. A number of models have already
been developed, but experimental validation is sought.
6
CCNCLtJS IONS
Considerable progress has been made in recent years on analysis

techniques, model developoentand experiinental verification in connection with
the assessment of major hazards.
C26
APPENDIX C
'The developaent of suitable models to predict the consequences of one

aspect, namely fire attack on transport and storage vessels containing
pressurised liquefied gases has not lagged behind, nor have the necessary
experimental validation exercises.
REFERENCES
Vilain "ophase flows in major technological hazards". WI Lecture

Series. March '86.
R J Moffat "Contributions to the theory of uncertainty analysis for single
sample experiments" AFOSRH'rlM Conf. on complex turbulent flow, Stanford
University, California, 1980.
H C Hardie and D 0 Lee "Expansion of clouds from pressurised liquids'.
Sandia Labs, New Mexico, SAND7452l0. 1974.
B Maurer at al "Modelling of vapour cloud dispersion and deflagration
after bursting of tanks filled with liquefied gas". Proc. 2nd Int. Sy!np.
on Loss Prevention, Heidelberg, Sept 1977.
H Gisbrecht et a]. "Explosion hazard analysis of flammable gas released
spontaneously into the atmosphere". Chemie. Ingenieur Technik. Vol.52,
.1
2
3
4
5
6
7
8
9
No.2, 1980.
R Bettis et aJ. "Expansion and evolution of heavy gas and particulate
clouds", J.Haz.Mats. Vol.14, pp 213232, 1987.
R C Reid "Some theories on boiling liquid expanding vapour explosions".
Fire, March 1980.

P K Ramskill "A study of axisyitinetric underexpanded gas jets". UKAEI-'. SRD
R302, March 1985.
P R Appleton "A study of axisyirinetric twophase flashing jets.
SRDR303,
10
UEIEA.
1984.
Wolf "Twophase flows in major technological hazards".

WI Lecture
March 1986.
C
"Aspect of the dispersion of denserthanair vapours relevant
to gas cloud explosions". EEC Rept. No. SR/007/80/JK,'H. 1984.
K S Mudan "Thermal radiation hazards from hydrocarbon pool fires". Prog.
Energy. Combust. Sci. Vol 10, pp 5980. 1984.
A C M Sousa et al "Thermal Modelling of LPG tanks engulfed in flames".
Heat and Technology Vol. 3, No. 3/4, 1985.
P K Rarnskill "A description
ENGULF (and ENGULF
computer code
Series.
11
12
13
14
15
eat1ey
of
II) in a
to model the thermal response of a tank (partiallyor) totally engulfed in
fire. UKEA Repts Nos SRD/HSE/R 354, SRD/HSE/070/WPI, 1987.
W M Rosenhow "A method of correlating beat transfer data for surface
boiling of liquids". Trans. ASME Section C. Vol 74, pp 969975,1952.

Butterworth and G F Hewitt "Twophase flow and heat transfer". AERE
Harwoll series. Oxford Univ. Press, 1979.
J E S Venart "Experiments on the physical modelling of L3 tank cars under
accident conditions." ]flt. Conf. Storage and transport of LEG and ENG,
Brugge, 1984.
0 w Saflet "Pressure relief valve sizing for vessels containing compressed
liquefied gases. I.Mech.E.Symp. paper No C274/79, 1979.
G V Beijnon et al "Fire Engulfment of LEG Tanks: HEATUP
a predictive
model".
mt. Symp. Transport and Storage of press. Liq. gases. tiNE
Fredericton. Canada. Aug 1987.
T Nylund "Fire survival of process vessels containing gas". I.Chem.E.Symp.
Series No 85. Chester, UK, 1984.
H S Forrest "Energency relief vent sizing for fire exposure when two-phase
flows must be considered", 19th Loss Prevention Symp. A.I.Ch.E., Houston,
1985.
C Anderson et al "The effects of a fire enviroanent on a rail tank car
filled with LEG, tE, ABR Lab, Maryland, 241, 1974.
16 D
17
18
19
20
21
22
C27
APPENDIX C
175
23
24
25
test project Repart on analysis

scale Lire tests". RPIMR, RP11526, US6 P21 1773 Eng. (PF)
B Droste et a]. "Failure mechanisms of propane tanks under thermal stresses
including fire engulfment" mt. Conf. storage and transport of LEG and
U1G, Brugge, 1984.
A F Roberts et a]. "Fire engulfment trials with insulated LEG storage
"Railroad tank car safety research and
of
-175th
tanks" 4th
26
27
28
29
30
31
32
33
K Moodje
mt. Loss Prey. Symp. I.Chem.E. Series 82, Harrogate, UK, 1983.
"The fire engulfment of
storage tanks" I.Clem.E. Syrap.
et al
Series 93 UMIST, UK, 1985.

K Billinge et a]. "The use of water sprays to protect fire engulfed LEG
storage tanks" 5th mt. Loss Prey. Symp. Cannes, Sept 1986.
K Moodie
"Total pool fire engulfment trials on a 5tonne LEG tank,
HSE Internal Sept No. IB/L/FP/87/27, 1987.
et al
AM
3irk
"Fire test to study the significance of a burning relief valve
flare froo a
tankcar engulfed in fire". Transport Canada Rpt No TP

6694E, 1985.
5 D Appleyard "Testing and evaluation of the LL)SAFE System as a method
of controlling the boiling liquid expanding vapour explosion", Sept No TP
2740. Transport Canada, 1980.
G A Bray "Fire protection of liquefied petroleum gas storage tanks", IGE
Journal, UK, 1964.
H G Fisher "DIERS An overviewof the prograrme".
19th mt. Loss
Prevention Symposium. Al.Ch.E. Houston. 1985.
G V Hadjisophocleous et a]. "Simulation of transit convective flows in
filled and partially filled cavities". Int.Symp. transport and storage of
press. liq. gases. UNB Fredericton, Canada. Aug 1987.
C28

Thermal Response of Vessels and Pipework Exposed To Fire

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Thermal Response of Vessels and Pipework Exposed To Fire

Transféré par

Droits d'auteur :

Formats disponibles

Crown copyright material is produced with the permission of the Controller of HMSO and Queen's Printer for Scotland.

This document was written by

Shell Research Limited

Health and Safety Executive- Offshore Technology Information

Crown copyright 1992

This reportis publishedby the Healthand SafetyExecutive as part

Executive. Neither the Executive, the Departmentnor the

Results, including detailed evaluation and, where relevant,

Standing order service

For details please write to HMSO BOOKS (PC 13A/l).

The standing order service also enables customers to receive

01192598 Currentfireresearch, experimental,

01192601 Computerised analysistools for

01192588 Legislation, codesofpracticeand

01192 600 Explicitanalytical methods fordeter

01192587 Thepredictionofsingle andtwo phase

01192586 Representative range ofblast and fire

01192589 Experimental facilitiessuitableforuse

01192602 The effectsofhigh strain rateson

01192590 The use ofalternativematerialsin the

01192603 Analysis ofprojectiles.

01192604 Experimental data relatingto the

01192591 Gas/Vapour build-up onoffshore

01192605 Methodologies and available tools for

01192593 Explosions in highlycongested

01192606 Passive fireprotectionperformance

01192594 The prediction ofthepressureloading

01192607 Availabilityand properties ofpassive

01192608 Existingfire designcriteria for

01192595 Possibleways ofmitigating explosions

secondary, support and systemsteelwork.

01192609 Fireperformance ofexplosion(FR6)

01192610 Thermalresponse ofvessels and

Thepurposeofthese reportsistocollate, reviewandcriticallyappraisethecurrentstateofknowledgeineachofthe

The fire response work packages

2 . 2 Obj ectives of FR7

Outline of the study

External heat fluxes

CURRENT MODELLING CAPABILITIES

The LPG fire response models

5.2.1 Features of the blowdown models

The requirements of predictive models

FR7 Work Package

Primary information on predictive models

REVIEW PAPER BY K MOODIE

The physical processes taking place when pressurised equipment

Modelling the latter subject is very difficult, and has not

It is possible that the existing computer codes can be used to

idea of the relative importance of factors not yet included,

great deal of research and development work may be needed,

Organisations and Official Bodies

American Petroleum Institute

American Society for Testing and Materials

Her Majesty's Stationery Office

Health and Safety Executive (UK)

Institute for Materials Research and Testing

Terms and Abbreviations

A foiling iquid expanding apour xp1osion which

The controlled venting of a pressurised vessel or line on

HEAT FLUX (density)

turbulent diffusion flame resulting from the combustion

LPG (Liquefied Petroleum Gas)

PSV (pressure safety valve)