CFD simulation to predict the thermal radiation of large LNG pool fires

S. Schalike1,2 , K.-D. Wehrstedt1 , A. Schonbucher2,

1

Division 2.2 Reactive Substances and Systems

BAM Federal Institute for Material Research and Testing
Unter den Eichen 87, 12205 Berlin, Germany
2
Institut for Chemical Engineering I
University Duisburg-Essen
Universitatsstr. 5-7, 45141 Essen, Germany
Abstract
Flame temperature (T ), surface emissive power (SEP ) of Liquefied Natural Gas (LNG) pool fires (d = 1 m, 6.1 m,
30 m) are investigated by CFD (Computational Fluid Dynamics) simulation and compared with experimental results.
Time averaged flame temperatures of T = 1320 K, T = 1298 K and T = 1281 K are obtained. Surface emissive power
(SEP ) of 55 kW/m2 , 130 kW/m2 and 230 kW/m2 are predicted.

Introduction

Accidental fires in process industries often occur as

pool fires which are hazardous to people and adjacent
objects due to thermal radiation, largely sooting plumes
and formation of other combustion products (1; 2; 3).
Pool fires are turbulent non-premixed fires burning over
a horizontal pool. In addition to experimental pool
fire tests numerical investigation of those fires using
Computational Fluid Dynamics (CFD) codes becomes
more important. The Surface Emissive Power (SEP ) is
a key parameter to characterize thermal radiation emitted
by a fire. Beside the SEP the temperatures T and
irradiances E of pool fires are of particular interest. To
predict the thermal radiation from LNG (d = 1 m, 6.1 m,
30 m) pool fires CFD methods are used and the CFD
results are compared with experiments.

Flame temperatures and axial velocity decreases due to

continued air entrainment.
_.

Qtop

Ttop
sootparcels

_
HP

hot spots

SEP

SEP

u ~~ 100 - 1000 cm/s

Ths
_.

Characteristics of pool fires

Pool fires are turbulent non-premixed fires burning over

a horizontal pool (4; 5; 6). Pool fires can be divided
into two- or three non-continuous zones, but until now
these length can be calculated with large uncertainty
only. Directly over the pool rim existes a luminous clear
burning zone (Hcl in Fig. 1), which is not covered with
black smoke and has beside hot spots the largest surface
emissive power SEPclma of a fire. In the pulsation zone
(Hpul in Fig. 1) the flame front is still connected to the
flame basis but it is a less efficient combustion zone of
a flame. Due to large eddies of intaken air radial and
axial pulsation occurs and formation of black soot can
be observed. In the top region the plume zone (HP in
Fig. 1) a non-continuous segregated flame is observed.

Corresponding author: stefan.schaelike@bam.de

Proceedings of the European Combustion Meeting 2011,
Cardiff, UK, 28 June - 2 July, 2011

Qba,tot
expansion of
combustion products

_
fuel-parcels (T fb)

air entrainment
(liquid)
fuel

Hpul

uF

ignitable volume VZ

(Tsp)

80% N2

Hcl

pyrolysis gas

pool/tank

fuel vapor
(uf ~
~ 1cm/s)

Fig. 1: Physical processes in pool fires

Thermal radiation models

A widely used thermal radiation model is the Solid

Flame Model (SFM, Fig. 2) (7), from which several
modefications exists (8). In the classical case a flame
is postulated as a cylinder with circular base, with a
homogeneous temperature of T = 1173 K on flame
surface, a emissivity of f = 0.95 for hydrocarbon pool
fires a specific emission of 100 kW/m2 is obtained by
the following equation:

ma

SEP SF M = (T T a )

E = E,F at r SEP

(1)

(3)

Beside the semi-empirical models according to Mudan (10), Fay (5) and Raj (6) especially the model
OSRAMO II (11; 12) and OSRAMO III (3; 11) which
contain mainly quantities with physical meaning are to
be noted.

postulated
cylindric flame

SEPma
SFM

In CFD simulation a domain is represented by a 3-D

hexahedral block structured mesh (Fig. 4) (13; 14).
The fuel is assumed to be already evaporated and the
fuel vapor entering the domain from inlet has a constant
temperature of T = Tb and an experimentally determined
constant mass flux (15). The inlet is surrounded by a
low rim and an adiabatic ground area. The remaining
boundary conditions are set as pressure outlet in a
relative wide distance from the pool to achieve open
boundary conditions. With increasing axial und vertical
distance from the pool cell size increases.

pool

CFD simulation

Fig. 2: SFM:The flame is approximated by a

cylinder
A different semi-empirical thermal radiation model is
the point source model, with which the irradiance can
be calculated directly (PSM respectively API 521, Fig.3)
(9):
kd
f rad AP (hc ) m
00
)
f,max (1e
,
16(y/d)2

EP SM (y/d) =

(2)

fu
r y/d > 4

Fig. 4: Block structured grid for simulation of LNG

pool fires
The time step vary depending on sufficient convergence from t = 106 s to t = 104 s depending on
Courant-Friederich-Levy (CFL) criterion. Large Eddy
Simulation (LES) is used for modeling turbulence (16;
17). In the pool fire simulations presented here, the
following sub models are used:

Dy

AP

Assumed pdf approach with laminar flamelets

containing up to 21 species and elementary reactions (18; 19; 20).

0
d
pool

AE y

Moos-Brookes soot model (21).

Discrete Ordinates model for radiation (22; 23;
24).

Fig. 3: PSM: Flame is approximated as point source

on flame axis
The flame approximated by a sphere with
radius r 0 hence the application of the model is
bounded to the far field. The irradiance is connected to
SEP over the view factor E,F , atmospheric transmission at and the absorptivity of the receiver element r
by:

The coupling between thermal radiation and soot reactions is described by a weigthed sum of gray gases
approach (25; 26).
The governing equations were solved with an iterativ solution method with either coupled or segregated
solvers, e.g. the pressure correction methods SIMPLE
(Semi-Implicit Methods for Pressure Linked Equations).

The starting and boundary conditions are listed in Tab.12.

5.1
Tab. 1: Starting conditions of CFD pool fire
simulation
starting conditions
quantity
value
mass fraction N2
0.743
mass fraction O2
0.231
mass fraction Ar
0,012
mass fraction CO2
0.001
mass fraction H2 O
0,013
pa
1013.25 hPa
p pa
0
flow velocity
ux = uy = uz = 0
temperature
298 K
gravitational acceleration
9.81 m/s2
mixing fraction f
f=0

Results and Discussion

Flame temperature

With CFD simulation instantaneous temperature fields of

LNG pool fires can be predicted. In both fields (Fig. 56) pulsation of the flame is visible. Inside the observed
vortices a significant higher instantaneous temperatures
(1800 K < T < 2161 K) then in the adjacent area.
(1400 K < T < 1800 K).

Tab. 2: Boundary conditions of CFD pool fire

simulation
boundary conditions
upper end face
+ lateral area
pressure outlet

p = pa
298 K
open boundary conditions

lower end face

+ pool rim
wall

adiabatic
heat flux to pool rim
q = 0

pool
mass flow inlet

experimentally determined
mass burning rate
p = pa , T = Tb

Fig. 5: CFD predicted instantaneous temperature

field of a LNG pool fire (d = 30 m) t = t0

The main purpose is to determine the temperature T ,

Surface Emissive Power (SEP ). The simulations are
started with an two equation model based on the eddy
viscosity hypothesis like k with a buoyancy correction
term to reach a certain flame heigth which refers to the
developing stage of the fire. Assuming the flame to be
developed, further simulation is continued by using LES.
CFD simulation is carried out with commercial software
ANSYS
FLUENT 12 (27).

Fig. 6: CFD predicted instantaneous temperature

field of a LNG pool fire (d = 30 m) t = t0
The vortices cool down in the higher flame region
where temperatures of T < 1400 K are obtained. The
maximum temperatures accure above the pool rim, but

5.2

not directly at the center line. The simulation shows

broken off vortices with temperatures T < 650 K until
a heights of 90 m.
With CFD simulation also time averaged flame temperatures can be predicted. The maximum time averaged
temperature is located at center line above the pool
rim (Fig. 7). The time averaged maximum flame
temperatures of LNG pool fires decrease with increasing
pool diameter (Tab. 3).

Surface emissive power

The SEP is a derived quantity and its value depends on

flame surface and flame shape. Especially the value of
the flame height, but also atmospheric transmission and
general experimental errors are import. The SEP of a
fire can be obtained by CFD simulation in three ways. In
the first way the SEP is predicted by radiative heat flux
qout leaving each grid cell placed on the flame surface
AF . The value of qout can by obtained by calculating the
component of radiative flux qrad that is normal (~nAF )
to the cell surfaces which define the flame surface AF :
SEP = qout = (1 t ) qin + t T 4 ,

(4)

with qin
Z
L ~s ~n d .

qin =

(5)

~
s~
n<0

To get the SEP at the flame surface it is necessary

to determine a surface AF which presents a realistic
shape of the flame. One possibility is an isosurface of
temperature (T > Ta ).
The procedure can be described as follows:
An instantaneous flame surface AF,CF D is defined
as an isosurface of temperature (interior wall).
The CFD calculated heat flux qout (t) is averaged
over the isosurface AF,CF D for each time interval
t (an usual value is t = 0.1 s) to evaluate area
averaged heat flux < qout (t) >.

Fig. 7: CFD predicted time averaged temperature

field of a LNG pool fire (d = 30 m)

The heat flux < qout (t) > is averaged over the
steady burning time (t 10 s) which results in a
time averaged heat flux < q out (t) > SEP CF D .
It is assumed that a steady state burning time of
10 s shows real burning behaviour.

Tab. 3: Maximum time averaged flame temperatures

of LNG pool fires
pool diamter [m] T max [K]
1
1320
6.1
1298
30
1281

Tab. 4: Maximum time averaged flame temperatures

of JP-4 pool fires
pool diamter [m] T max [K]
2
1300
8
1280
16
1250
25
1230
Fig. 8: CFD predicted instantaneous SEP for an
isosurface T = 500 K of a LNG pool fire (d = 30
m)

The time averaged flame temperaturs of LNG pool

fires are up to 50 K higher, then those of other hydrocarbon pool fires e.g. JP-4 (Tab. 4).

Fig.8 shows the instantaneous SEP , depending in

general on pool diameter and fuel, calculated by using

an isosurface of temperature T = 500 K for LNG pool

fire.
Time averaged SEP is determined: SEP = 55 kW/m2
(d = 1 m), SEP = 130 kW/m2 (d = 6.1 m) and
SEP = 230 kW/m2 (d = 30 m). The maximum
specific emissivity SEPmax accounts for 301 kW/m2
thus being much higher then time averaged SEP . The
SEPOSRAMO II (JP-4)

JP-4

300

SEPCFD (JP-4)
SEPSFM (JP-4)

Hgglund

Adknowledgements
The authors thank Max-Buchner-Forschungsstiftung
for financial support.

SEPLNG (expected)

Kerosin
Hoftijzer

SEP [kW/m]

250

Shokri

LaTable

Jap. Soc. Safety Eng.

200
LNG
May

150

References

Montoir
AGA

[1] C. Balluff, W. Brotz, A. Schonbucher, D. Gock,

N. Schie, Study of hazardous fires of liquid
hydrocarbons as a contribution to the safety of
chemical plants, Chem.-Ing.-Tech. 57 (1985) 823.

Minzer
Schnbucher

100

50

0
0.1

10

[2] H. Persson, A. Lonnermark, Tank fire review of fire

incident 1951-2003, Brandforsk Project 513-021,
Swedish National Testing and Research Institut,
Boras, Sweden, 2004.

100

d [m]

[3] M. Hailwood, M. Gawlowski, B. Schalau,

A. Schonbucher, Conclusions drawn from the
buncefield and neaples incidents regarding the
utilization of consequence models, Chem. Eng.
Technol. 32 (2009) 207231.

Fig. 9: Measured and CFD predicted SEP of

kerosine and LNG pool fires (28; 29; 30; 31) (32;
33; 34; 35)

[4] W. Brotz, A. Schonbucher, Heat and mass transfer

in pool flames, Chem.-Ing.-Tech. 50(8) (1978) 573
585.

SEP of LNG pool fires increases until a pool diameter

of d = 35 m. With further increasing pool diameter a
maximum of SEP is expected due to soot formation and
soot blockage effect (Fig.9).
The mass burning rates of LNG pool fires are by
factor 2.3 higher in comparison with other hydrocarbon
pool fires. The measured mass burning rates are in
comparison with calculated (over thermal back flow)
mass burning rates by factor 2.5 smaller. This effects
are caused by higher absorption coefficent of LNG vapor.
For kryogenic liquids the air entrainment leads to an
additional increase of the mass burning rate due to the
large temperature difference. CFD simulation shows an
temperature increase of the orignally cold LNG vapor
short above the pool surface.

parison with other hydrocarbon pool fires. This facts

can be explained by the mass burning rates which are
in comparison with other hydrocarbon pool fires much
higher caused by the temperature difference between
boiling point and ambient temperature. To determine the
maximum of SEP as a function of the pool diameter
further investigation with pool diamters d > 35 m should
be done.

[5] J. Fay, Model of large pool fires, J. Hazard. Mater.

136 (2006) 219232.
[6] P. Raj, Large hydrocarbon fuel pool fires: Physical
characteristics and thermal emission variations with
height, J. Hazard. Mater. 140 (2007) 280292.
[7] P. Seeger, Untersuchung der Warmeubertragung
von einem brennenden Objekt auf die Umgebung, Arbeitsgemeinschaft der Innenministerien
der Lander, 1972.
[8] Methods for the calculation of physical effects,
(Yellow Book), CPR 14E, (Part 2, chap. 6: Heat
flux from fires, 6.1 - 6.130), 3. Aufl., 1997.

Conclusion

[9] American Petroleum Institute, Report API 521,

Guide for pressure relief and depressuring systems,
1969.

With CFD simulation instantaneous and time averaged

flame temperatures T , T , specific emissions SEP , SEP
can be predicted. The about 50 K higher instantaneous
and time averaged flame temperatures are caused by the
higher mass burning rate of LNG pool fires in com-

[10] K. S. Mudan, Thermal radiation hazards from

hydrocarbon pool fires, Prog. Energy Combust. Sci.
10 (1984) 5980.

[11] R. Fiala, D. Gock, X. Zang, A. Schonbucher,

Calculation of thermal radiation of large fires with
the OSRAMO spherical radiation model, Chem.Ing.-Tech. 63 (1991) 760761.

[25] N. Lallemant, A. Sayre, R. Weber, Evaluation of

emissivity correlations for h2o-co2-n2/air mixtures
and coupling with solution methods of the radiative
transfer equation, Prog. Energy Combust. Sci. 22(6)
(1996) 543574.

[12] R. Fiala, D. Gock, X. Zhang, A. Schonbucher,

Experimentally validated spherical radiation model
33(4)
OSRAMO Part 1: Theoretical principles, TU
(1992) 137140.

[26] W. H. Dalzell, A. F. Sarofim, Optical constants of

soot and their application to heat-flux calucalation,
J. Heat Transfer 91 (1969) 100104.

[13] H. Chun, Experimentelle Untersuchungen und

CFD-Simulationen von DTBP-Poolfeuern, Phd.
Thesis, BAM Dissertationsreihe Band 23, 2007.

[27] ANSYS FLUENT 12, User Guide, AEA Technology, 2002.

[28] B. Hagglund, L. E. Persson, FOA Rapport, C30126D6 (A3), 1976.

[14] H. Chun, K.-D. Wehrstedt, I. Vela, A. Schonbucher,

Thermal radiation of di-tert-butyl peroxide pool
fires-experimental investigation and CFD simulation, J. Haz. Mater. 167 (2009) 105.

[29] R. W. G. Hoftijzer, TNO-Ber., 9092 6 (1980) 1.

[15] I. Vela, CFD prediction of thermal radiation of

large, sooty, hydrocarbon pool fires, Phd. Thesis,
Universitat Duisburg-Essen, 2009.

[30] A. Shokri, Society of Fire Protection Engineers,

Engineering Guide for Assessing Flame Radiation
to External Targets from Pool Fires, Bethesda,
Maryland, 1999.

[16] S. B. Pope, Turbulent Flow, Cambridge Univ.

Press,, 2000.

[31] M. Meazawa, Report on Burning of Petroleum

Fires, Japan Institute for Safety Engineering, 1982.

[17] J. H. Ferziger, M. Peric, Computional Methodes

for Fluid Dynamics, Springer-Verlag, Berlin Heidelberg 2. Aufl., 1997.

[32] W. G. May, W. McQueen, Radiation from large

liquefied natural gas fires, Comb. Sci. Technol. 7(2)
(1973) 5156.

[18] N. Peters, Laminar flamelet concepts in turbulent

combustion, 21th Symp. (Int.) on Combustion
(1988) 12311260.

[33] H. Malvos, P. K. Raj, LNG pool fire spectral data

and calculation of emissive power, Proc. Saf. Prog.
26(3) (2007) 237247.

[19] S. K. Liew, K. N. C. Bray, J. B. Moss, A stretched

laminar flamelet model of turbulent nonpremixed
combustion, Combust. Flame 56(2) (1984) 199
213.

[34] LNG Safety Program, Interim Report on Phase II

work, American Gas Association, 1974.
[35] G. A. Minzer, J. A. Eyer, Large-scale LNG and LPG
pool fires, Inst. Chem. Eng. Symp. Ser. 71 (1982)
147163.

[20] J. Warnatz, U. Maas, R. W. Dibbel, Combustion,

Springer-Verlag, Berlin Heidelberg 3. Ed., 2006.
[21] S. J. Brookes, J. B. Moos, Measurements of soot
production and thermal radiation from confined
turbulent jet diffusion flames of methane, Combust.
Flame 116(1/2) (1999) 4961.
[22] G. D. Raithby, E. H. Chui, A finite-volume method
for predicting radiant heat transfer in enclosures
with participating media, J. Heat Transfer 106
(1990) 4958.
[23] S. R. Mathur, Y. Murthy, Coupled ordinates method
for multigrid acceleration of radiation calculations,
J. Thermophys. Heat Trans. 13(4) (1999) 467473.
[24] M. Modest, Radiative Heat Transfer, McGraw-Hill
Series in Mechanical Engineering, New York, 1993.