Sie sind auf Seite 1von 34

Revised Fire Consequence Models for Offshore

Quantitative Risk Assessment


Ravichandra Pula, Faisal I. Khan, !rian "eitch, and Paul R. Am#otte
$
Faculty of Engineering & Applied Science
Memorial University of Newfoundland, St Johns, N!, "anada# A$% &'(
$
)epartment of "hemical Engineering
)alhousie University, *alifa+, NS, "anada# %&J ,'-
A%stract
.ffshore oil and gas platforms are well/0nown for their compact geometry, high
degree of congestion, limited ventilation and difficult escape routes# A small mishap
under such conditions can 1uic0ly escalate into a catastrophe# Among all the accidental
process/related events occurring offshore, fire is the most fre1uently reported# 2t is
therefore necessary to study the 3ehavior of fires and 1uantity the ha4ards posed 3y them
in order to complete a detailed 1uantitative ris0 assessment# 5hile there are many
conse1uence models availa3le to predict fire ha4ards 6 varying from point source models
to highly comple+ computational fluid dynamic models 6 only a few have 3een validated
for the uni1ue conditions found offshore#
2n this paper we have reviewed e+isting fire models and selected the ones most
suita3le for offshore conditions# 7he selected models were further revised 3y including
the impact of wind, overpressure caused 3y hot com3ustion gases and 3y adopting a
grid/3ased approach to ena3le 3etter modeling and analysis of radiation impact# A
comparison of the performance of the revised models with a commercial software
pac0age 3eing used for offshore ris0 assessment was also carried out and is discussed in
the paper#
Ke#&ords' Fire modeling, 8uantitative ris0 assessment, .ffshore ris0 modeling#
9 "orresponding author: ;hone< =>?/=&=/@?&?, Fa+< =>?/=&=/->-,, f0hanAengr#mun#ca
$
$. Introduction
An offshore oil and gas platform is usually divided into a num3er of modules for
operations such as separation, water inBection, high/pressure compression, and seawater
de/aeration, as well as local and main electrical rooms and an accommodations 3loc0#
Most of these modules are highly congested with the presence of o3stacles in the form of
pipelines and other e1uipment necessary for process operations# 7he level of ris0 in such
conditions, while operating in a remote and harsh marine environment, is very high#
A study 3y the UC *ealth and Safety E+ecutive D*SE, $??E and Mansfield et al#,
$??EaF showed that process and structural failure incidents account for almost @> G of the
ris0 to personnel offshore# ;otential ris0s offshore include< 3lowouts, riser and process
lea0s, fires, e+plosions, vessel collisions, helicopter accidents, dropped o3Bects,
structural failures, and capsi4ing# An e+amination of incidents such as Piper Alpha in the
North Sea and the P-36 production semi/su3mersi3le off %ra4il reveals that most
offshore incidents are in fact process/related#
An offshore development can never 3e completely safe, 3ut the degree of
inherent safety DMansfield et al#, $??E3 and Chan & Amyotte, ,>>,F can 3e increased 3y
selecting the optimum design in terms of the installationHfield configuration, layout, and
operation# 7his is done in an attempt to reduce the ris0 to a level that is As !ow As
Ieasona3ly ;ractica3le DA!AI;F without resorting to costly protective systems# 7his
re1uires the identification and assessment of maBor ris0 contri3utors, which could 3e
accomplished using 8uantitative Iis0 Assessment D8IAF techni1ues early in the proBect
life cycle# 2f a structured approach of identification and assessment is not carried out
early in the proBect, it is possi3le that the engineering Budgment approach will fail to
identify all of the maBor ris0s, and that loss prevention e+penditures will 3e targeted in
areas where there is little 3enefit# 7his may result in e+pensive remedial actions later
during the life of the proBect DJinnem, $??@F#
8uantitative ris0 assessment DChan et al#, ,>>,F involves four main steps< ha4ard
identification, conse1uence assessment, pro3a3ility calculation, and finally ris0
,
1uantification# "onse1uence assessment, which is central to 8IA, involves
1uantification of the li0ely lossHdamage due to any possi3le eventuality# Among the
various possi3le loss/producing events in offshore production facilities, fire is the most
fre1uently reported process/related incident D"ham3erlain, ,>>,F# A fire may result in
anything from no damageHloss, up to catastrophic damageHloss, depending upon the fire
characteristics Dtype of fire, mode of occurrence and potential of escalationF# 7herefore,
fire conse1uence modeling is a 0ey element of conse1uence analysis in 1uantitative ris0
assessment# Even though mathematical models to study the characteristics of process/
related fires in offshore process facilities are reported in the literature, only a few have
3een validated for offshore conditions# "onsidering the importance of fire modeling in
offshore 8IA and the availa3le 0nowledge gap, the current wor0 was underta0en with
the o3Bective of revising availa3le fire models for offshore operations#
7he wor0 is ongoing and has 3een planned to enhance e+isting 0nowledge of fire
conse1uence modeling through the following advancements<
Fire characteristics< Availa3le fire models have 3een reviewed and the most
appropriate ones selected and revised 3y incorporating wind and confinement
effects which are uni1ue to offshore process platforms#
Overpressure impact: 7he importance of overpressure caused due to fire in a
confined or semi/confined space has 3een highlighted 3y 5ighus D$??-F#
*owever, there appears to have 3een no attempt to 1uantify this phenomenon# 2n
the present wor0, a model has therefore 3een developed to study the overpressure
impact# 7his model is em3edded in the 3asic fire conse1uence modeling
methodology descri3ed herein#
Radiation modeling: 2nstead of pointHarea modeling, a grid/3ased approach has
3een employed to ena3le 3etter modeling and analysis of radiation impact at
different locations, the impact of o3stacles, and the effects of flame
impingement#
&
(. Current )tatus of Fire Consequence Modelin*
7here are many predictive models availa3le for the assessment of fire
conse1uence ha4ards 6 varying from point source techni1ues to more comple+
computational fluid dynamic D"F)F calculations# Such predictive models can 3e
categori4ed as follows into< semi/empirical models, field models, integral models and
4one models#
2.1 Semi-empirical models
2.1.1 Point source models
;oint source models D!ees, $??EF do not predict the flame geometry, 3ut rather
assume that the source of thermal radiation is a single point in the flame and that a
selected fraction of the heat of com3ustion is emitted as radiation# 7hese models
generally over/predict the heat flu+ for near/field conditions: however, they can 3e used
relia3ly 3eyond appro+imately five pool diameters from the flame# 7he use of point
source models within offshore structures is limited#
2.1.2 Solid lame models
Solid flame surface emitter models assume a fire to 3e a solid flame with heat
3eing radiated from the surface of the flame DJohnson et al#, $??-, "rac0nell et al#, $??-,
and Iew et al#, $??=, $??@F# 7hey rely mainly on correlations for flame geometry
estimation, average surface emissive power DSE;F of the flame, atmospheric
transmissivity and view factors# 7he various surface emitter models differ in their
methods of assessing atmospheric attenuation of the heat flu+, view factors, and the SE;#
5ell/validated solid flame models provide a 3etter prediction of flame geometry and
e+ternal thermal radiation for offshore fires than is possi3le with point source models#
2n general, semi/empirical models are tas0 specific, are designed to address
specific ha4ard conse1uences, and incorporate correlations fitted to large/scale
e+perimental data# 7hese models are mathematically simple and can 3e easily computer
programmed with short run times#
-
2.2 Field models
Field models are "F) models DJohnson et al#, $??=F 3ased on numerical
solutions of the Navier/Sto0es e1uations of fluid flow Di#e# a description of the
conservation of mass, momentum and scalar 1uantities in flowing fluid, 3y means of a
set of partial differential e1uations#F 7o predict fire 3ehavior, these models incorporate
various su3/models to account for the physical and chemical processes occurring in a
fire# All such models re1uire validation against e+perimental data 3efore using them as
predictive tools to estimate the ha4ards of open or compartment fires# !imiting factors in
the applica3ility of these models are related to high ";U demands and user e+pertise#
2.3 Integral models
2ntegral models are a compromise 3etween semi/empirical and field models, and
are formulated mathematically in a manner similar to field models# 7hus, integral
models also solve the conservation of mass and momentum e1uations and contain su3/
models for com3ustion and heat transfer# 7he mathematical treatment is simpler than in
field models, thus reducing computer run times# Some integral models that have 3een
validated against la3oratory/scale e+perimental data are availa3le D5ilco+, $?=( and
Jachon & "hampion, $?@EF#
2.4 Zone models
Kone models D"ham3erlain, $??E and "ham3erlain & ;ersaud, $??=F divide a
module or a compartment into a num3er of 4ones that are assumed physically distinct,
3ut coupled 3y empirical heat and mass transfer e1uations# Even though this is a
traditional approach to model compartment fires, very few 4one models have 3een
validated 1uantitatively for offshore applications# Kone models have wide applica3ility
and validity for the purposes for which they are designed, i#e# 3uildings with reasona3ly
small rooms and predominantly small vertical vents# *owever, they encounter severe
limitations in modeling large offshore compartments# Further research in this area of
modeling would 3e 3eneficial#
(
;rovided they are used within their range of applica3ility, validated semi/
empirical solid flame models are well/suited for the prediction of heat flu+es to o3Bects
outside the flame# 7hese models have 3een successfully used for fire conse1uence
analysis and further, for 8IA DIew et al#, $??=, $??@, Johnson et al#, $??-, and
"rac0nell et al#, $??-F# *owever, they are not directly applica3le to the study of fire
characteristics in offshore facilities without revision 3y incorporation of su3/models as
descri3ed in the current wor0#
+. Characteristics of Offshore Fires
7he characteristics of a fire depend on factors such as the type of fuel, release
conditions, local geometry, ventilation, and air access# ;otential fuels handled offshore
include crude oil, natural gas, and gas containing condensate and water 6 all of which are
continuously produced, processed, separated, dried and stored# 7hese fuels pose a
significant amount of ris0 to personnel, e1uipment and the environment#
Fires on process plants onshore differ from offshore mainly in the level of
confinement# 7he harsh marine environment offshore dictates the need for process areas
to 3e enclosed and shielded against the weather# 2t is well/esta3lished that a fire inside a
confined volume develops differently from an open fire# 7he restriction in air supply is
often the limiting factor with respect to fire si4e, and a severely under/ventilated fire
environment can develop# %urning of hydrocar3ons under such conditions may 3e more
intense than in open fires, as the mi+ture of air and fuel may 3e closer to an ideal
stoichiometric mi+ture# Additionally, heat losses from the fire to the surroundings are
reduced, leading to higher flame temperatures#
MaBor ha4ards associated with compartment fires include those normally
associated with open fires, such as e+ternal thermal radiation and direct flame
impingement on o3Bects# 2n addition, other ha4ards e+ist due to the effect of confinement
D"ham3erlain, $??E and "ham3erlain & ;ersaud, $??=F# Some of these additional
ha4ards to personnel are impaired visi3ility along escape routes due to e+cess smo0e,
E
to+icity from the release of car3on mono+ide due to incomplete com3ustion, and
overpressure impacts from the hot com3ustion gases#
"onsideration of smo0e and car3on mono+ide generation re1uires detailed
chemistry calculations which are outside the scope of the current paper# An important
aspect mentioned a3ove, and often neglected in fire modeling, is overpressure due to
fires in confined spaces# 5ith only small openings in a compartment, the highly
energi4ed com3ustion products released from these fires can generate pressures greater
than am3ient# 7his condition may further lead to an e+plosion creating 3last wave and
missile effects# 7herefore, overpressure effects have to 3e ta0en into account when
analy4ing the ha4ards from offshore fires# 2n the present wor0 we have developed a set
of e1uations for overpressure 1uantification, which are discussed in detail in the
following section#
,. Revised Fire Models
!ea0age or spillage of flamma3le material can lead to a fire that is triggered 3y
any num3er of potential ignition sources Dspar0s, open flames, etc#F# 7hese fires are
3roadly classified into four types, namely pool fires, Bet fires, fire3alls and flash fires,
irrespective of offshore or onshore conditions# 7he availa3le models for each of the four
fire types are now reviewed, and the ones most suita3le for offshore environments are
identified# Model revisions are also discussed#
4.1. Pool fires
A pool fire is a tur!ulent diusion ire !urning a!ove a pool o vapori"ing
h#drocar!on uel $here the uel vapor has negligi!le initial momentum.
7he pro3a3ility of occurrence of pool fires on offshore platforms is high due to
continuous handling of heavy hydrocar3ons on3oard# !i1uid fuel released accidentally
during overfilling of storage tan0s, rupture of pipes and tan0s etc#, forms a pool on the
surface, vapori4es, and upon ignition, results in a pool fire#
=
"onse1uence models for pool fires in open spaces have 3een well/documented
over the past few years DIew et al#, $??=, ;ritchard & %inding, $??,, and Johnson,
$??,F# Although there has also 3een significant wor0 done on compartment fires, most
of these efforts deal with "F) modeling# 7here are also a few physically 3ased 4one
models that have 3een developed for compartment fires D"ham3erlain, $??E and
"ham3erlain & ;ersaud, $??=F#
After reviewing the pool fire models availa3le in the literature, the model
descri3ed 3y Iew et al# D$??=F was Budged to 3e most suita3le for offshore conditions,
with the modifications descri3ed here# First, the effect of wind on flame length was
accounted for 3y means of the wor0 of 7homas D$?E&F: second the effect of confinement
in the form of overpressure generation was also considered# A 3rief description of the
selected radiation model for pool fires DIew et al#, $??=F follows#
Using a solid flame approach, a pool fire is modeled as a sheared elliptical
cylinder which is assumed to radiate in two layers 6 a high emissive power, clean
3urning 4one at the 3ase, with a smo0y o3scured layer a3ove as shown in Figure $# 7he
radiation heat flu+ received 3y a target depends on the atmospheric transmissivity,
geometric view factor and surface emissive power of the fire# 7he correlation used to
1uantify heat flu+, %, is as follows<
& & & ' ' ' & '
S(P )F S(P )F % % % + +
D$F
where is atmospheric transmissivity, )F is the geometric view factor and S(P is the
average surface emissive power# 7he su3scripts ' and & refer to values calculated for the
clear lower layer and smo0y upper layer of the model flame shape, respectively#
Atmospheric transmissivity is calculated using an algorithm developed 3y
5ayne D$??$F# 7his calculation is 3ased on the assumptions that the flame is a 3lac0
3ody source at $(>> C, with ".
,
and *
,
. vapor 3eing the only molecules that a3sor3
radiation in the pathway 3etween the fire and the target#
@
7he view factor D)avis & %agster, $?@?F represents the fraction of the overall
heat output that stri0es the target, and is dependent upon the geometry of 3oth the flame
and the target# For radiation from a finite flame to a differential receiving element, the
view factor is given 3y the integral over the flame surface<

s
d
)F
,
, $
cos cos


D,F
where
$
and
,
are the angles 3etween the normals to the fire surface and the
receiving element, respectively, and d is the distance from the receiver point to the flame
center#
7he surface emissive powers for the clear lower layer and the smo0y upper layer
are correlated separately as follows<
( )
S R ' R &
S(P & S(P & S(P + $
D&F
( )
*
m
+
'
e S(P S(P

$
D-F
where &
R
is the uno3scured ratio, S(P
S
is the surface emissive power of smo0e,
S(P

is the ma+imum surface emissive power of a fuel, m
+
is the e+tinction coefficient, and
* is the pool diameter#
Overpressure calculation
7he model descri3ed so far predicts only the radiation heat flu+ received 3y a
target o3Bect# A model for the estimation of overpressures generated 3y highly energi4ed
com3ustion gases in a compartment is given 3y a com3ination of the ideal gas law and
radiative heat transfer e1uations with the following assumptions<
aF negligi3le convective heat transfer,
3F ideal gas 3ehavior of the com3ustion gases,
cF small compartment openings, and
dF linear distri3ution of temperature variation within the defined space#
7he algorithm for overpressure calculation is as follows<
?
$# "alculate the flame temperature, ,
lame
, using the surface emissive power
estimated from the radiation model descri3ed earlier<
( )
- -
o
lame
rad
, , -
D(F
,# Similarly calculate the temperature at one corner of the compartment, ,
cc
, and
using assumption dF a3ove, estimate the average temperature of the gases, ,
gases
<
,
cc lame
gases
, ,
,
+
DEF
&# Finally, use the ideal gas law to estimate an appro+imate value of the
overpressure, P
o
, generated 3y the gases in the compartment<
room
gases
o
)
nR,
P
D=F
where -
rad
is the heat emitted 3y the flame per unit area, is the Stefan/%olt4mann
constant D(#EE?9$>
/@
5Hm
,
C
-
F, ,
o
is the initial compartment temperature, n is the moles
of com3ustion gases in the compartment, R is the gas constant, and )
room
is the volume of
the compartment under study#
7his overpressure impact model is em3edded in the pool fire radiation model as
well as in the other fire models descri3ed in su3se1uent sections#
4.2 Jet fires
A jet fire is a tur!ulent diusion lame resulting rom the com!ustion o a uel
continuousl# released $ith some signiicant momentum in a particular direction.
Jet fires represent a significant element of ris0 associated with maBor incidents on
offshore installations, with the fuels ranging from light flamma3le gases to two/phase
crude oil releases# %etween hori4ontal and vertical Bet fires, the former is the most
dangerous 3ecause of the high pro3a3ility of impingement on o3Bects downwind
D"arsley, $??-F# 7his can lead to structural, storage vessel, and pipe/wor0 failures, and
can cause further escalation of the event Di#e# domino effectF# 7he heat flu+es released
from these fires are very high, ranging from ,>> 6 ->> 05Hm
,
depending on the type of
fuel# Almost all the fuels handled offshore can form Bet fires provided that the release
$>
occurs under conditions such that the fluid has some initial momentum Dsuch as a lea0
from a pressuri4ed gas lineF#
Jertical Bet fire models De#g# "ham3erlain, $?@=F are commonly used to assess
the ha4ards from flares# 7he model of "ham3erlain D$?@=F has 3een e+tended to
hori4ontal Bet fires 3y Johnson et al# D$??-F# 7his model was developed with offshore
conditions in mind, and was therefore chosen as the 3ase Bet fire model in the current
wor0# A 3rief description of the selected radiation model DJohnson et al#, $??-F follows#
A hori4ontal Bet fire is modeled as the truncated frustum of a cone, which emits
thermal radiation from its surface as shown in Figure , DJohnson et al#, $??-F# For
hori4ontal releases, the 3uoyancy of the flame dominates over wind momentum, causing
the flame to rise a3ove the hori4ontal plane# %ecause o3Bects in the direction of the
release receive radiation from emitting paths roughly e1ual to the flame length Dwhich is
much larger than the flame widthF, a different surface emissive power is assigned to the
ends of the solid flame from the SE; used for the sides of the flame# 7hus, the thermal
radiative flu+ at a target o3Bect is given as<
( ) +
end end side side
S(P )F S(P )F %
D@F
where the su3scripts side and end refer to values calculated for the side and end of the
model flame shape#
Jiew factors and atmospheric transmissivity are calculated as per the
formulation for pool fires# 7he surface emissive powers for the side and end of the
model flame are given as follows<
( )
,
$
+.
side
e S(P S(P

D?F
( )
'
+R
end
e S(P S(P

$ D$>F
where
A
- F
S(P
s

, is the ma+imum surface emissive power, and


$- # > ,$ # >
>>&,& # >
+

/
u
s
e F , is the fraction of the overall heat emitted as radiation#
$$
7he factor + is the gray gas a3sorption coefficient, .
2
is the ma+imum width of the
flame, R
'
is the length of the frustum, - is the net heat released 3y com3ustion, A is the
surface area of the flame, and u
/
is the velocity of the Bet#
7his radiation conse1uence model, with the inclusion of the overpressure impact
model descri3ed in the previous section on pool fires, ena3les prediction of the
characteristics of hori4ontal Bet fires on offshore platforms#
4.3 Fireballs
A fireball is a rapid tur!ulent com!ustion o uel0 usuall# in the orm o a rising
and e1panding radiant !all o lame.
5hen a fire such as a pool or Bet fire impinges on a vessel containing pressure/
li1uefied gas, the pressure in the vessel rises and the vessel wall wea0ens# 7his can
eventually lead to catastrophic failure of the vessel with the release of the entire
inventory# 7his phenomenon is 0nown as a 3oiling li1uid e+panding vapor e+plosion
D%!EJEF# 2n such releases, the li1uefied gas released to the atmosphere flashes due to
the sudden pressure drop# 2f the released material is flamma3le, it will ignite: in addition
to missile and 3last ha4ards, there is thus a thermal radiation ha4ard from the fire3all
produced# 2t is this thermal radiation which dominates in the near field# Although the
duration of the heat pulse from a fire3all is typically of the order of $> 6 ,> s, the
damage potential is high due to the fire3alls massive surface emissive power# !arge/
scale e+periments carried out 3y Io3erts et al# D,>>>F with propane as the fuel, measured
a ma+imum average surface emissive power ranging from ,=> 6 &&& 05Hm
,
upHdownwind and ,=@ 6 -$& 05Hm
,
crosswind# )ue to the high tur3ulence involved, a
fire3all can also 3e e+pected to cause significant overpressures#
Modeling of fire3alls has 3een carried out 3y several researchers De#g# ;rugh,
$??-, "rac0nell & "arsley $??=, and Io3erts et al#, ,>>>F# Alternative scenarios for
fire3alls to the one descri3ed in the previous paragraph, e#g# fire3alls from delayed
ignition of continuous Bet releases, have also 3een dealt with 3y "rac0nell & "arsley
$,
D$??=F# For the present study, the model of Io3erts et al# D,>>>F was selected as the 3ase
fire3all model# 2n this approach DIo3erts et al#, ,>>>F a fire3all is modeled as a sphere as
shown in Figure &# 7he radiation heat flu+ incident on a target is evaluated using a solid
flame model as the product of atmospheric transmissivity, geometic viewfactor and the
surface emissive power<
S(P )F %
D$$F
Surface emissive powers measured e+perimentally 3y Io3erts et al# D,>>>F are
used, while view factors and transmissivity are evaluated using the correlations
descri3ed in the previous section on pool fires# %ecause a fire3all is a transient event, the
heat flu+ varies with time 6 initially increasing, and then falling off after reaching a
ma+imum value De#g# Figure -F# 7his is 3ecause the fire3all si4e grows in the initial
stages, reaches a ma+imum, and then reduces as the fire3all rises#
4.4 Flash fires
A flash fire is a transient ire resulting rom the ignition o a gas or vapor cloud0
$here a dela# !et$een the release o lamma!le material and su!se%uent ignition has
allo$ed a cloud o lamma!le material to !uild up and spread out rom its release point.
A flash fire is usually characteri4ed 3y a Lwall of flameM DIaB & Emmons, $?=(
and "";S, $??-F progressing out from the point of ignition at a moderate velocity until
the whole flamma3le cloud has 3urned# Similar to fire3alls, flash fires can occur either
3y ignition of a flamma3le vapor cloud formed from an instantaneous release, or 3y
delayed ignition of a cloud from a continuous release, provided the tur3ulence in the
cloud is low enough that a fire3all does not occur# 7he instantaneous or continuous
releases considered in ris0 studies would physically correspond to a spreading transient
puff or a long steady/state plume#
5hen the cloud ignites, the initial damage will 3e caused primarily 3y thermal
radiation# *owever, flash fires may generate more damaging L0noc0/onM events,
especially if they 3urn 3ac0 to the source# 7he 0noc0/on events can 3e a pool fire, Bet
$&
fire, %!EJE etc# Further, the presence of o3stacles along the pathway and the high
degree of congestion on offshore platforms can lead to significant flame acceleration#
Such increases in flame speed can in turn lead to significant overpressures and ultimately
a partially confined or confined vapor cloud e+plosion# 7he effects of these escalation
events are li0ely to 3e more severe than the flash fire itself#
"onse1uence modeling of a cloud fire in an uncongestedHunconfined
environment where overpressure is not a principal ha4ard has 3een well/documented
DIew et al#, $??(, $??@ and "rac0nell & "arsley, $??=F# 7hese flash fire models are
3ased on gas dispersion modeling coupled with the pro3a3ility of ignition, where the
3oundary of the fire is defined 3y the unignited clouds downwind and crosswind
dimensions at flamma3le limit concentrations Dusually the lower flamma3le limit, !F!,
of the cloudF# After reviewing the availa3le models, the approach developed 3y Iew et
al# D$??@F was selected for the present study, and the overpressure impact model
previously descri3ed was incorporated as a revision# 7his stage of the modeling wor0 is
ongoing within our research group#
-. .ama*e /ffect Calculations
7he conse1uence models discussed so far provide estimates of the radiation heat
flu+ at a target o3Bect and the overpressure in a confined area# 7hese are, of course,
important parameters to 0now, 3ut they must 3e translated into anticipated harm to
personnel on3oard# )ose response evaluation is therefore needed to 1uantify the damage
DfatalityF from thermal radiation and overpressure# 7o facilitate this analysis, personnel
harm is e+pressed in terms of pro3it functions DChan & A33asi, $??@F, which relate the
percentage of people affected in a 3ounded region of interest 3y a normal distri3ution
function#
7he pro3it function, Pr, for heat radiation lethality is given as<
2 % t ln3 . . Pr
4 & -
(E , &@ &E + D$,F
where % is the radiation heat flu+ and t is the time of e+posure#
$-
7he pro3it function for li0elihood of death due to overpressure Dlung ruptureF is
given as<
2 3P ln . . Pr
o
+ ?$ E $ ==
D$&F
where P
o
is the overpressure#
Finally the pro3a3ility DpercentageF, P, of damage is 1uantified using the formula
D"";S, ,>>>F<
1
1
]
1

,
_

+
,
(
(
(
$ (>
Pr
er
Pr
Pr
P
D$-F
where er is the error function#
A grid/3ased approach has 3een adopted to facilitate modeling and analysis of
radiation impact at different locations in a predefined area# 2n this approach, the area
under study is divided into smaller grids, and the heat flu+ and radiation damage are
estimated at each grid point with the results plotted as contours#
0. Results and .iscussion
2n this section, the revised pool fire, Bet fire and fire3all models are simulated
under specific scenarios for radiation heat flu+es and overpressures# 7he simulation
results for each of the fires are presented graphically in three different ways< radiation
flu+ contours, radiation damage contours and the variation of radiation damage along the
centerline#
)cenario $ 1 Pool Fire< A crude oil storage tan0 of -/m diameter and $>>/0g capacity
on an offshore platform catches fire, leading to a pool fire# 7he mass 3urning rate of the
crude oil is >#>(>= 0gHs# 5ind speed is ( mHs in an easterly direction#
)imulation Results<
*eight of pool fire N (#( m
Angle of tilt from vertical N (E degrees
SE; of lower clear flame N $$, 05Hm
,
$(
SE; of upper smo0y flame N ,( 05Hm
,
Assuming the pool has formed in the center of a (> m 9 (> m room, the radiation
heat flu+ in that space can 3e represented 3y the contours shown in Figure (# Similarly,
lethality Dradiation damageF contours in a $>> m 9 $>> m

area are represented in Figure
E, while the thermal radiation damage along the centerline is shown in Figure =# As the
pool fire is tilted (E degrees from the vertical toward the downwind direction, the
radiation and hence the damage must 3e higher downwind than upwind# 7his is evident
from the contours in Figures ( and E# Figure = shows that $>>G lethality is e+pected up
to a distance of O&> m and minor damages up to a distance of O(> m downwind from the
center of the pool# )istances greater than (> m downwind and -( m upwind are
identified as safer 4ones#
.verpressure from the pool fire in a (> m 9 (> m 9 $> m compartment was
calculated as >#,( 0;a, which has negligi3le impact# 2t must 3e remem3ered, though, that
this result is for a fuel loading of $>> 0g# !arger fuel inventories, such as are routinely
found on offshore installations, would generate higher overpressures Das demonstrated in
the ne+t scenarioF#
)cenario ( 1 2et Fire' A lea0 of >#$(, m in a pressuri4ed natural gas storage tan0
causes the inventory to e+it as a hori4ontal Bet at a rate of $$#, 0gHs# 7he fuel ignites
immediately after release and results in a horiontal jet fire lasting for appro+imately &>
minutes#
)imulation Results'
!ength of flame N &@ m
Frustum length N ,E m
Angle of tilt from vertical N @> degrees
SE; from sides of flame N ,,& 05Hm
,
SE; from end of flame N ,-, 05Hm
,
$E
Assuming the fuel release occurs from one end of a E> m 9 E> m room, the
radiation heat flu+ in that space is represented 3y the contours in Figure @# Since the
flame length is &@ m, the amount of damage is e+pected to 3e very high# 7hus, the
lethality Dradiation damageF contours are represented in a ,-> m 9 ,-> m area as shown
in Figure ?, while the thermal radiation damage along the centerline is shown in Figure
$>#
7he contours shown in Figures @ and ? are elliptical, as is e+pected from the
initial and 3oundary conditions that define the physical scenario# 7his confirmation
facilitates the relia3le prediction of conse1uences, which in turn provides an opportunity
to identify safer 4ones for escape when an unwanted event occurs# For e+ample, Figure
$> illustrates that $>>G lethality is predicted for distances up to ,(> m downwind,
whereas minor damage e+ists up to a distance of a3out (>> m# Safer 4ones in all
directions are apparent from the lethality contours in Figure ?#
For this scenario, the overpressure generated 3y the com3ustion gases in a E> m
9 E> m 9 $> m compartment was calculated to 3e ,&E 0;a 6 i#e# significantly higher than
atmospheric pressure due to the fact that the flame Bet occupies a relatively large portion
of the enclosure volume# 7his clearly highlights the necessity of overpressure
considerations for the congested plant found on offshore oil and gas platforms#
)cenario + 1 Fire%all' Fire impingement on a propane gas storage tan0 causes a
pressure rise inside the tan0 and eventually leads to a %!EJE# A fuel mass of $,=, 0g is
released and ignited, resulting in a fireball#
)imulation Results'
)iameter of fire3all N EE m
7ime of e+istence of fire3all N =#? s
*eight of fire3all center a3ove ground N -?#( m
SE; from the fire3all N &,> 05Hm
,
$=
7he fire3all is assumed to occur at the center of a (> m 9 (> m

room and the
radiation contours are given in Figure $$# Iadiation damage contours are presented for a
$>> m 9 $>> m area as shown in Figure $,# Figure $& shows the radiation damage
variation along the centerline# Although the fire3all lasts for only a3out @ s, the large
amount of fuel released results in O?> G lethality close to the fire3all with minor damage
persisting until a distance of (> m from the epicenter# )istances greater than this (>/m
radius are therefore safer for escape in the event of such a fire3all#
7he overpressure created 3y the fire3all in a (> m 9 (> m 9 $> m compartment
was calculated as ,#- 0;a, a value which is approaching the point at which personnel
mo3ility is affected# 7he fact that the fire3all overpressure is higher than that predicted
for the pool fire Dscenario $F is e+plaina3le in part 3y the much higher mass of fuel
released in the fire3all scenario 6 which in turn is significantly less than the total fuel
mass 3urned in the Bet fire Dscenario ,F, which has the highest overpressure of the three
scenarios studied#
3. Com4arison of Models
2n this section, the results of the revised pool fire and Bet fire models are
compared with those from the commercial software ;!A7.
P
, a well/0nown pac0age for
simulating hydrocar3on lea0age and ignition on offshore installations# A description of
the modeling approaches for pool fires, Bet fires and fire3alls that are em3edded in this
pac0age are given 3y Jones & 2rvine D$??=F# 7he fire model e1uations in ;!A7.
P
indicate the use of a point source approach# 7hese pool fire and Bet fire point source
models were simulated for the same data and specifications as in scenarios $ and , in the
previous section# 7he results are represented as radiation contours and radiation damage
contours in Figures $- 6 $=#
!.1 Pool fire res"lts comparison #scenario 1$
"omparing the results of the revised pool fire model as shown in Figures ( and E
with those in Figures $- and $( indicates that<
$@
7he radiation contours from the ;!A7.
P
model are circular, whereas the revised
model gives elongated contours due to the actual physical mechanism of flame
tilt#
7he ;!A7.
P
model over/predicts the thermal radiation damage relative to the
revised pool fire model# 7his is to 3e e+pected given the point source approach in
the former model and the use of a solid flame in the latter#
!.2 Jet fire res"lts comparison #scenario 2$
"omparing the results of the revised Bet fire model as shown in Figures @ and ?
with those in Figures $E and $= indicates that<
7he radiation contours predicted 3y the revised model are elliptical in shape,
whereas those from the point source model are co/centric circles#
7he damage contours display the same features as the radiation contours
descri3ed in the previous point#
7he over/predictions as a result of using a point source model are clear from the
damage contour plots#
Although the comparisons a3ove are admittedly limited, it is felt that the a3ility
of the revised pool fire and Bet fire models to simulate aspects of physical 3ehavior has
3een demonstrated# 7hese aspects include pool fire flame tilt due to the prevailing wind
direction and the elliptical shape of a hori4ontal Bet fire due to the momentum impulse
created in such a scenario# 2n addition to radiation impact conse1uences, the revised
models ena3le consideration of the overpressure impact caused 3y hot e+panding
com3ustion gases and un3urned fuel gases#
5. Conclusion
Ievised conse1uence models for pool fires, Bet fires and fire3alls have 3een
simulated and the results presented as contours 3y using a grid/3ased approach# 7he
radiation and damage DlethalityF contours for pool fires show that these conse1uences
are concentrated more toward the downwind direction due to flame tilt caused 3y the
wind# 7he results for hori4ontal Bet fires demonstrate the e+pected 3ehavior of lethality
$?
contours that are elliptical in shape, rather than circular as o3tained from point source
models# 7hus, the solid flame models employed in the current wor0 more closely match
the physical characteristics of fire scenarios that arise on offshore installations#
7he damage contours o3tained 3y adopting a grid/3ased approach permit the
development of a clear picture of potential impact 4ones# 7his can facilitate proper
selection and specification of safe separation distances to prevent inBury to people and
damage to near3y e1uipment# .verpressure generation from fires has also 3een shown
to 3e a critical consideration in developing the impact 4one map for an offshore facility#
Ackno&led*ement
7he authors gratefully ac0nowledge the financial support of the Natural Sciences and
Engineering Iesearch "ouncil of "anada#
,>
References
"arsley, A#J# D$??-F# A model for predicting the pro3a3ility of impingement of Bet fires#
56hem( S#mposium Series 7o. 138, $==/$?&#
"";S D$??-F# 9uidelines or evaluating the characteristics o vapor cloud e1plosions0
lash ires and :'()(s# New Qor0< American 2nstitute of "hemical Engineers#
"";S D,>>>F# 9uidelines or chemical process %uantitative ris+ anal#sis# New Qor0<
American 2nstitute of "hemical Engineers#
"ham3erlain, R#A# D$?@=F# )evelopments in design methods for predicting thermal
radiation from flares# 6hemical (ngineering Research and *esign, 0-, ,??/&$>#
"ham3erlain, R#A# D$??EF# *a4ards posed 3y large/scale pool fires in offshore
platforms, Process Saet# and (nvironmental Protection, 3,, @$/@=#
"ham3erlain, R#A# D,>>,F# "ontrolling hydrocar3on fires in offshore structures#
Proceedings o Annual Oshore ,echnolog# 6onerence, $,$$/$,$@#
"ham3erlain, R#A#, & ;ersaud, M#A## D$??=F# A model for predicting the ha4ards from
large/scale compartment Bet fires# 56hem( S#mposium Series 7o. 1;1, $E&/$=-#
"rac0nell, I#F#, & "arsley, A#J# D$??=F# "loud fires 6 a methodology for ha4ard
conse1uence modeling# 56hem( S#mposium Series 7o. 1;1, $&?/$(>#
"rac0nell, I#F#, )avenport, J#N#, & "arsley, A#J# D$??-F# A model for heat flu+ on a
cylindrical target due to the impingement of a large/scale natural gas Bet fire# 56hem(
S#mposium Series 7o. 138, $E$/$=(#
,$
)avis, %#"#, & %agster, )#F# D$?@?F# 7he computation of view factors of fire models#
<ournal o 'oss Prevention in the Process 5ndustries, (, ,,-/,&-#
*SE D$??EF# .ffshore accidentHincident statistics reports# O,O86.8=;, *ealth and Safety
E+ecutive, !ondon, UC#
Johnson, A#)# D$??,F# A model for predicting thermal radiation ha4ards from large scale
!NR pool fires# 56hem( S#mposium Series 7o. 13>, (>=/(,-#
Johnson, A#)#, %rightwell, *#M#, & "arsley, A#J# D$??-F# A model for predicting the
thermal radiation ha4ards from large/scale hori4ontally released natural gas Bet fires#
Process Saet# and (nvironmental Protection, 3(, $(=/$EE#
Johnson, A#)#, E33inghaus, A#, 2manari, 7#, !ennon, S#;#, & Marie, N# D$??=F !arge/
scale free and impinging tur3ulent Bet flames< numerical modeling and e+periments#
Process Saet# and (nvironmental Protection, 3-, $-(/$($#
Jones, J# "#, & 2rvine, ;# D$??=F# ;!A7.P software for offshore ris0 assessment< a
criti1ue of the com3ustion features incorporated# <ournal o 'oss Prevention in the
Process 5ndustries, $6, ,(?/,E-#
Chan, F# 2#, & A3assi, S#A# D$??@F# Iapid 1uantitative ris0 assessment of a
petrochemical industry using a new software pac0age MA'"IE)# <ournal o 6leaner
Production, 0, ?/,,#
Chan, F#2#, & Amyotte, ;#I# D,>>,F# 2nherent safety in offshore oil and gas activities< a
review of the present status and future directions, <ournal o 'oss Prevention in the
Process 5ndustries, $-, ,=?/,@?#
,,
Chan, F#2#, Sadi1, I#, & *usain, 7# D,>>,F# Iis0 3ased process safety assessment and
control measures design for offshore process facilities, <ournal o ?a"ardous @aterials,
7,, $/&E#
!ees, F#;# D$??EF# 'oss prevention in the process industries# !ondon< %utterworths#
Mansfield, )#;#, Clet4, 7#A#, & Al/*assn, 7# D$??EaF# .ptimi4ing safety 3y inherent
offshore platform design# Proceedings o 1886 O@A( 6onerence - )olume 55 3Saet#
and Relia!ilit#2, June $E/,>, $??E, Florence, 2taly#
Mansfield, )#, ;oulter, !#, & Clet4, 7#A# D$??E3F# 2mproving inherent safety< a pilot
study into the use of inherently safer designs in the UC offshore oil and gas industry#
?S( Oshore Saet# *ivision Research Pro/ect Report, *SE .ffice, !ondon, UC#
;ritchard, M#J#, & %inding, 7#M# D$??,F# F2IE,< A new approach for predicting thermal
radiation levels from hydrocar3on pool fires# 56hem( S#mposium Series 7o. 13>, -?$/
(>(#
;rugh, I#5# D$??-F# 8uantitative evaluation of fire3all ha4ards# Process Saet#
Progress, $+, @&/?$
IaB, ;#;#C#, & Emmons, *#5# D$?=(F# .n the 3urning of a large flamma3le cloud# <oint
,echnical @eeting o the .estern and 6entral States Sections o the 6om!ustion
5nstitute, San Antonio 7'#
Iew, ;#J#, *ul3ert, 5#R#, & )eaves, )#M# D$??=F# Modeling of thermal radiation from
e+ternal hydrocar3on pool fires, Process Saet# and (nvironmental Protection, 3-, @$/
@?#
Iew, ;#J#, Spencer, *#, & Madison, 7# D$??@F# Sensitivity of ris0 assessment of flash fire
events to modeling assumptions# 56hem( S#mposium Series 7o. 1;;, ,E(/,=@#
,&
Iew, ;#J#, Spencer, *#, *oc0ey, S#M#, & !ines, 2#R# D$??(F# Ieview of flash fire
modeling# ?S( 6ontractor Report .SA4RS&A>>>4>1=, *SE %oo0s, !ondon, UC#
Io3erts, 7#, Rosse, A#, & *aw0sworth, S# D,>>>F# 7hermal radiation from fire3alls on
failure of li1uefied petroleum gas storage vessels# Process Saet# and (nvironmental
Protection, 35, $@-/$?,#
7homas, ;#*# D$?E&F# 7he si4e of flames from natural fires# 6om!ustion, 7, @--#
Jachon, M#, & "hampion, M# D$?@EF# 2ntegral model of a flame with large 3uoyancy
effects, 6om!ustion and Flame, 0+, ,E?/,=@#
Jinnem, J#E# D$??@F# Evaluation of methodologies for 8IA in offshore operations#
Relia!ilit# (ngineering and S#stem Saet#, 0$, &?#
5ayne, F#)# D$??$F# An economical formula for calculating atmospheric infrared
transmissivites# <ournal o 'oss Prevention in the Process 5ndustries, ,, @E/?,#
5ighus, I# D$??-F# Fires on offshore process installations# <ournal o 'oss Prevention in
the Process 5ndustries, 3, &>(/&>?#
5ilco+, )#"# D$?=(F# Model for fires with low initial momentum and nongray thermal
radiation, A5AA <ournal, $+, &@$/&@E#
,-
8omenclature
$
Angle 3etween the normal to the fire surface and the line
Boining the fire and the receiving element DdegreesF
,
Angle 3etween the normal to the target surface and the line
Boining the fire and the receiving element DdegreesF
Angle of tilt DdegreesF
Atmospheric transmissivity
Stefan/%olt4mann constant D(#EE?9$>
/@
5Hm
,
C
-
F
A Surface area of the flame Dm
,
F
3 Flame liftoff DmF
d )istance from the receiver point to the flame center DmF
) ;ool diameter DmF
erf Error function
0 Rray gas a3sorption coefficient Dm
/$
F
m
0 E+tinction coefficient Dm
/$
F
! !ength of the pool fire DmF
!
c
"lear layer length DmF
n Moles of com3ustion gases DmolF
; ;ercentage damage DGF
;
o
.verpressure DNHm
,
F
;r ;ro3it num3er
1 Iadiation heat flu+ D05Hm
,
F
8 Net heat released 3y com3ustion D05F
I Ras constant D@#&$- JHmolSCF
I
!
!ength of the frustum DmF
SE; Average surface emissive power D05Hm
,
F

SE;
Ma+imum surface emissive power D05Hm
,
F
SE;
!
Surface emissive power from clear lower layer D05Hm
,
F
SE;
S
Surface emissive power of smo0e D05Hm
,
F
SE;
U
Surface emissive power from smo0y upper layer D05Hm
,
F
,(
t 7ime of e+posure DsF
7
cc
7emperature at one end of the compartment DCF
7
o
2nitial compartment temperature D$(
o
"F
7
flame
Flame temperature DCF
7
gases
Average gas temperature DCF
u
B
Jelocity of the Bet DmHsF
U
I
Uno3scured ratio
U
+
5ind velocity in the downwind direction DmHsF
U
4
5ind velocity in the crosswind direction DmHsF
U
+4
5ind velocity vector DmHsF
JF Reometric view factor
J
room
Jolume of the compartment under study Dm
&
F
5
$
5idth of frustum 3ase DmF
5
,
5idth of frustum tip DmF
,E
Fi*ure Ca4tions
Fi*ure $ Flame geometry for a tilted pool fire#
Fi*ure ( Flame geometry for a hori4ontal Bet fire Dfrom Johnson et al#,
$??-F#
Fi*ure + Flame geometry for an e+panding, rising fire3all#
Fi*ure , 7ime dependent heat flu+ for a fire3all#
Fi*ure - Iadiation contours from revised pool fire model#
Fi*ure 0 ;ercentage lethality Dthermal radiationF contours from revised
pool fire model#
Fi*ure 3 ;ercentage lethality Dthermal radiationF vs# distance plot from
revised pool fire model#
Fi*ure 5 Iadiation contours from revised hori4ontal Bet fire model#
Fi*ure 7 ;ercentage lethality Dthermal radiationF contours from revised
hori4ontal Bet fire model#
Fi*ure $6 ;ercentage lethality Dthermal radiationF vs# distance plot from
revised hori4ontal Bet fire model#
Fi*ure $$ Iadiation contours from revised fire3all model#
Fi*ure $( ;ercentage lethality Dthermal radiationF contours from revised
fire3all model#
Fi*ure $+ ;ercentage lethality Dthermal radiationF vs# distance plot from
revised fire3all model#
Fi*ure $, Iadiation contours from ;!A7.
P
pool fire model#
Fi*ure $- ;ercentage lethality Dthermal radiationF contours from ;!A7.
P

pool fire model#
Fi*ure $0 Iadiation contours from ;!A7.
P
Bet fire model#
Fi*ure $3 ;ercentage lethality Dthermal radiationF contours from ;!A7.
P

Bet fire model#
,=
Fi*ure $
Fi*ure (
,@
, 5ind vector
Fi*ure +
Fi*ure ,
,?

Fi*ure - Fi*ure 0
Fi*ure 3
&>

Fi*ure 5 Fi*ure 7
Fi*ure $6
&$

Fi*ure $$ Fi*ure $(
Fi*ure $+
&,

Fi*ure $, Fi*ure $-
&&

Fi*ure $0 Fi*ure $3
&-
Jirtual flame shape
Jirtual flame shape

Das könnte Ihnen auch gefallen