Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

journal homepage: www.elsevier.com/locate/jlp

New correlation for vapor cloud explosion overpressure calculation at

congested congurations
Jingde Li a, Madhat Abdel-jawad b, *, Guowei Ma a
a
b

School of Civil and Resource Engineering, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia
GexCon Australia, 8/64 Fitzgerald Street, Northbridge, WA 6003, Australia

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 4 December 2013
Received in revised form
27 May 2014
Accepted 29 May 2014
Available online 21 June 2014

In this study, we present a newly developed correlation for the estimation of boundary overpressures in
and around congested regions subjected to vapor gas explosions. The GAME correlation, which is based
on the MERGE, EMERGE experimental programs, shows rather moderate correlation with computational
uid dynamics (CFD) results in homogeneously congested congurations, however, a greater level of
inaccuracy is found when it comes to the combination of a number of realistic scenarios. The newly
developed model (connement specic correlation), which consists parameters of volume blockage
ration, the density of the gas, the ame path distance, the connement ratio and the laminar ame speed
of the ammable gas is proposed as a non-dimensional alternative and it shows a closer correlation with
detailed CFD simulation in general particularly for realistic geometries. A linear least square method is
used to achieve the best tting parameters by applying the validated commercial software FLACS. About
400 CFD cases with homogenous congestions are modeled using FLACS for the purpose of testing both
the GAME correlation and the connement specic correlation (CSC). In addition to those 400 CFD homogenous cases, around 700 realistic cases in ten different module scenarios of a Liqueed Natural Gas
(LNG) train along with three simplied models are simulated to validate the CSC; it is found that the CSC
is applicable to both realistic modules with irregular obstacles and homogenous articial modules.
2014 Elsevier Ltd. All rights reserved.

Keywords:
Connement
Congestion
Blockage ratio
Obstacle diameter
Flame path

1. Introduction
Explosions and res in the process industry (Mannan, Aldeeb, &
Rogers, 2002) can result in large nancial and environmental
damages in addition to potential injury and loss of life. Typical
major industrial accidents include vapor cloud explosions (VCE),
Boiling Liquid Expanding Vapor Explosions (BLEVEs) and dust explosions. The VCE is dened as an explosion resulting from an
ignition of a premixed cloud of ammable vapor, gas or spray with
air, in which ames accelerate to sufciently high velocities to
produce signicant overpressure (Mercx & van den Berg, 2005).
Although analytical methods for the calculation of overpressures
arising from accidental inventory releases and subsequent delayed
ignition of resulting gas clouds leading to explosions have long
been in use, these methods hold signicant uncertainty because
they do not adequately account for several important parameters,

* Corresponding author.
E-mail addresses: jaden_li@hotmail.com (J. Li), madhat@gexcon.com (M. Abdeljawad), ma@civil.uwa.edu.au, guowei.ma@uwa.edu.au (G. Ma).
http://dx.doi.org/10.1016/j.jlp.2014.05.013
0950-4230/ 2014 Elsevier Ltd. All rights reserved.

particularly the role that congestion and connement play in ame

acceleration and hence the overpressures arising. In particular,
methods such as the multi-energy method (MEM) can be in error
by more than an order of magnitude because they do not take into
account the geometry detail of most industrial layouts and also rely
on estimates of explosion strength and congestion input by the
engineer. Where such methods are applied conservatively, the
estimated overpressure can be much higher than in a real event,
leading to signicant nancial overspends. Conversely, where these
estimates are under-conservative (which is a possibility with these
methods even when thought to be applied conservatively), the
results can be catastrophic.
Computational uid dynamics (CFD) is by far the most detailed
methodology for quantifying the risk posed by this class of catastrophic events. However, despite signicant advances deployed in
CFD, it remains computationally and labor intensive. There is,
therefore, a need for the development of faster analytical models
that can be applied with far less effort yet still capture the dominant
mechanisms for gas dispersion and ame propagation and ame
acceleration. Here we present a correlation that better accounts for
important details of complex geometries, which enables this

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

correlation to be more accurate that existing analytical methods

while offering greater implementation speed compared to existing
CFD methods.
Simpler methods such as the TNT-equivalency method (Safety,
1994) and TNO Multi-Energy Method (MEM) (Alonso et al., 2006;
Gugan, 1979; Lobato, Canizares, Rodrigo, Saez, & Linares, 2006;
Vandenberg, 1985), are often adequate for the estimation of far
eld pressures where the explosion eld displays little directionality. The TNT-equivalency method uses the blast generated by an
equivalent amount of TNT to describe the strength of the vapor
cloud explosion and the decay of the blast as a function of distance.
However, it is inherently assumed that the overpressures generated
are equal in all directions with no accounting for directional effects,
and it is hard to achieve the correlation between the quantity of fuel
involved in the explosion and the equivalent-charge weight of TNT
required to model its blast effect (Gugan, 1979). Therefore, this
method has limited applicability in scenarios where the layouts are
directionally irregular. Further, it is difcult to set up a standard to
convert the equivalent charge weight of TNT, in most cases, the
effects of VCE in the near eld can be either overestimated or
underestimated.
The Multi-Energy Method (MEM), which is regarded as a more
reasonable simple and practical method alternative (Mercx, van
den Berg, Hayhurst, Robertson, & Moran, 2000), is also only reliable for the calculation of far eld pressures. MEM uses overpressure results phenomenologically derived from several from
simplied numerical solutions of idealized gas explosions (van
Wingerden, Hansen, & Foisselon, 1999). MEM has shortcomings
similar to those of TNT equivalence method in that they both assume the gas pressure elds are radial. MEM is also ultimately
based on a selected severity from 1 to 10 entirely at the discretion of
the engineer applying the method with no mathematical basis for
the selection.
The shortcomings in MEM led to the development of a Guidance
for the Application of the Multi-Energy method (GAME) (Eggen,
1995). GAME was designed to provide additional guidance and to
extend its applicability to cases where MEM is designed to address.
The phenomenological approach, which is effective for qualitative
research projects (Alfred, 1976; Edmund, 1989; Gurwitsch & GarciaGomez, 2009), is used to derive the GAME correlation based on the
experimental research programs performed during the MERGE and
EMERGE projects (EMEG, 1997; Harris & Wickens, 1989; Mercx,
Johnson, & Puttock, 1995; Schumann, Haas, & Schmittberger,
1993; Wingerdenv, 1988, 1989) at the Dutch research institute TNO.
As seen in the report (Eggen, 1995), satisfactory correlation with
limited experiments were obtained by using GAME correlation, and
the it is a safe approach in the determination of the overpressure in
most situations characterized by articially homogenous congestion and connement.
To set up such experimental tests is a very expensive task and
there is a signicant limit on the quality of possible tests in that it is
very difcult to create realistic elds of congestion and connement at the appropriate scale. Further, the reliability and repeatability of the tests are often very difcult to achieve because some
factors such as initial turbulence, the stability of the wind direction
and speed as well as the exibility of some structural components is
very difcult to characterize or account for. Hence we have chosen
to compare the results from our new correlation as well as results
from the GAME correlation against the highly validated wellestablished CFD software FLACS. This allows us to examine hundreds of cases including those for realistic geometries at realistic
scales which would be impossible to set up without tens of years of
signicant spend.
Our new correlation presented in this study is validated against
widely-accepted CFD commercial software FLACS, which itself has

17

been the validated over the last 40 years against numerous experiments and previous work (Bleyer, Taveau, Djebaili-Chaumeix,
Paillard, & Bentaib, 2012; Hansen, Gavelli, Ichard, & Davis, 2010;
Middha, Hansen, Grune, & Kotchourko, 2010; Middha, Hansen, &
Storvik, 2009). The FLACS CFD solvers account for the parameters of
the congestion (Bakke, van Wingerden, Hoorelbeke, & Brewerton,
2010; Davis & Hanen, 2010; Hansen, Hinze, Engel, & Davis, 2010;
Huser, Foyn, & Skottene, 2009), the ame path distance and the
laminar ame speed of the ammable gas (Chen, Qin, Xu, Ju, & Liu,
2007; Pfahl, Ross, Shepherd, Pasamehmetoglu, & Unal, 2000;
Silvestrini, Genova, & Trujillo, 2008) which were derived by using
the idealized experimental programs' data, the new correlation was
deduced with a set of parameters by means of the linear least
square method to describe the obstructed region and the fuel
properties in the vapor cloud explosion.
By comparing the results from 1100 simulation cases carried out
using FLACS, we are able to compare the estimate the overpressures
from the new correlation and the GAME correlation for vapor cloud
explosions in realistically congested areas, taking into account the
complexity of the geometry; and the congestion and connement
with a well validated benchmark.
2. The GAME correlation and case studies
In this section, the GAME correlation is introduced and investigated by comparing its results with those of FLACS for both
realistic and idealized congurations from CFD simulations.
As originally derived from experiments, two variants of the
GAME correlation were given in the GAME project to determine the
vapor cloud explosion overpressure (Eggen, 1995).
For low ignition energy and no connement in 3D ame
expansion conditions:

2:75


VBR$Lf
0:7
DPo 0:84$
S2:7
l $D
D

(1)

For low ignition energy and connement between parallel

plates (2D expansion):

2:25


VBR$Lf
0:7
DPo 3:38$
S2:7
l $D
D

(2)

where:

DPo the overpressure [barg],

VBR the volume blockage ratio, which is dened as the ratio of

the total volume of the obstacles inside an obstructed region,
Lf the maximum distance of ame propagation obtained by
assuming Lf equal to the radius of a hemisphere with a volume
equal to the volume of the conguration [m],
D the average obstacle diameter, which give a single average
value for the whole obstructed region by assuming a homogenous distribution of obstacle types and obstacle diameters [m],
Sl the laminar ame speed of the ammable gas by assuming a
homogenous stoichiometric ammable cloud in all assessment
[m/s].

2.1. Modules tested in CFD simulations

CFD simulations were carried out to validate the results from
both the GAME correlation and the newly developed correlation.
The overpressures arising from the CFD simulations were extracted for the purpose of comparison with results from both
correlations.

18

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

The commercial software FLACS, which is a reliable tool for

prediction of vapor cloud explosions in petrochemical process area
offshore and onshore, is used here as the benchmark against which
the GAME correlation and the newly developed correlation presented here. FLACS is a specialized CFD solver developed especially
to model dispersion of ammable or toxic gas, gas explosions and
propagation of blast and shock waves, in complex, large-scale,
three-dimensional (3D) geometries.
The CFD simulations were performed for three articial cases
with homogenous congestion along with ve realistic and inhomogeneous congurations as shown in Figs. 1 and 2. For the articial modules 1e3 (Fig. 1), all module sizes are 80  80  80 (m),
and the obstacles in the congurations are arranged orthogonally
by lling the pipes of diameter of 0.5 m. The ve realistic modules
are from a Liqueed Natural Gas (LNG) train. The explosions are
conducted in the fractionation area, the pipe rack area and the
combination areas of the pipe racks and the mercury removal and
dehydration areas, respectively (Fig. 2).
Overall, three articial modules subjected to propane vapor
explosions, ve methane vapor explosions of realistic modules and
another ve propane vapor explosion of the same realistic modules
composes the 13 module cases as seen in Table 1. And for all the
modules, the values of volume blockage ratio, the laminar ame
velocity, the characteristic average obstacle diameter and the gas
composition, as shown in Table 1, are extracted to calculate the
overpressure in the following section.
2.2. Validation results of GAME correlation in case studies
By using the CFD congurations listed above, the applicability
and accuracy of GAME correlation were investigated in this section.
Three models of modules with articial congurations were
initially created with a uniform distribution of cylinders similar to
those in the experiments upon which the GAME correlation was
dened. Due to the homogeneity of the obstacle arrangement and
meshing grid, the CPU time for each calculation of the articial
conguration was relatively short (i.e. within 1 h).
The connement of the modules was controlled with the
insertion of parallel plates, and hence equation (2) of the GAME
correlations was employed. It is seen in Fig. 3 that the values of the
correlation R-squared factor (which indicates how well data points
t a line or curve), for the rst two homogenous cases are 0.78 and
0.51 respectively when applying the linear least square method.
These values show that the GAME correlation is valid for congestion
congurations that are lled with the regular-patterned pipes
within a certain range of the connement. However, as the area of
the top plate decreases, negative R-squared values (0.32) are seen
in the homogenous case 3 with partially conned roof, Fig. 3. The
main reason is that the connement effect is not accounted for in
the GAME correlation. By varying the connement (Fig. 1) with the

other parameters kept constant, the overpressures of those three

homogenous scenarios obtained in the GAME correlation remain
the same, while the pressures are reduced due to the decrease of
connement in the results of FLACS simulations, which means the
overpressures are overestimated by the GAME correlation in the
low connement case.
The results given by the GAME correlation described above
demonstrate that the lack of appropriate denition of connement
within the GAME equation result in an increasing error when this
parameter becomes important. They also show that GAME can only
give satisfactory results when the connement is within a certain
range.
In addition to these three articial congurations, another 5
realistic congurations (Fig. 2) were also investigated, for each of
those realistic cases; the CPU time was increased to the range of
1e3 h each due to the complexity of the geometries and longer
calculation time of the ame turbulence development within the
irregularly congested regions. As those realistic cases (case 4e13)
were included in the comparison, the overall correlation between
the GAME results and FLACS data gives a poor value as seen in Fig. 4,
which can be attributed to the geometric inhomogeneity of the
realistic geometric congurations in addition to the lack of appropriate modeling of connement. The GAME correlations were
derived from MERGE experiments which have a highly regular
pattern of obstacles, all of which are idealized as cylinders and
homogeneously distributed in the obstructed region.
When the repeatability of obstacles, equal obstacle spacing and
the obstacle diameter were carefully chosen the GAME correlation
produced results moderately close to those predicted by CFD.
However as seen in Fig. 4, for realistic modules with inhomogeneous congested volumes, the GAME correlation has poor prediction of overpressures and the GAME correlation often over-predicts
but sometimes under-predicts the overpressures signicantly.
In summary the GAME equations shows a very poor correlation
to numerically simulated results when all realistic modules are
included, it only gives a moderate R-squared value for the idealized
case created with a homogenous distribution of congestion. The
lack of consideration of congestion inhomogeneity and the denition criteria of connement hinders the applicability of GAME
correlation in practical problems. Those issues are improved and
developed in the following section by introducing a new
correlation.
3. Parametric studies and development of a new correlation
Connement was introduced and dened in a connement
specic correlation (CSC); other critical parameters were chosen to
model other factors as was done for the GAME correlation
describing the physical phenomenon of gas explosion. The derivation of the CSC was based on the linear least square method with a

Fig. 1. Articial modules 1e3.

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

19

Fig. 2. Realistic modules 4e8.

Table 1
Parameters in difference modules.
Case no.

Gas composition

D (m)

VBR

Sl
(m/s)

Gas density
(kg/m3)

Cm

1. Module 1
2. Module 2
3. Module 3
4. Module 4
5. Module 5
6. Module 4
7. Module 5
8. Module 6
9. Module 7
10. Module 8
11. Module 6
12. Module 7
13. Module 8

Pure
Pure
Pure
Pure
Pure
Pure
Pure
Pure
Pure
Pure
Pure
Pure
Pure

0.50
0.50
0.50
0.37
0.45
0.37
0.45
0.12
0.34
0.31
0.12
0.34
0.31

0.070
0.070
0.070
0.040
0.058
0.040
0.058
0.080
0.103
0.096
0.080
0.103
0.096

0.46
0.46
0.46
0.4
0.4
0.46
0.46
0.4
0.4
0.4
0.46
0.46
0.46

1.8
1.8
1.8
0.65
0.65
1.8
1.8
0.65
0.65
0.65
1.8
1.8
1.8

1.000
0.925
0.888
0.716
0.707
0.716
0.707
0.917
0.980
0.903
0.917
0.980
0.903

propane
propane
propane
methane
methane
propane
propane
methane
methane
methane
propane
propane
propane

subset of the simulations. In order to appropriately isolate the

important set of parameters, all the CFD cases in this subset
(approximately 400 cases) were simulated as homogenous models;
the distributions of pipes were arranged in regular patterns by
hand.
3.1. Conceptual denition of connement and congestion
Both connement and congestion, which are parameters that
affect turbulence induced ame acceleration, which has a signicant effect on overpressures (Bradley, Lawes, & Liu, 2008; Harrison
& Eyre, 1987; Moen, Donato, Knystautas, & Lee, 1980; van den Berg
& Mos, 2005).
For the GAME correlation, the congestion was dened using the
volume blockage ratio (VBR) divided by the average pipe diameter.
This is a very useful parameter; however the manner in which it is
applied does not take into account the fact that changing the

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

y=x
R = 0.78

5
4

FLACS (Barg)

FLACS (Barg)

3
2
1

y=x
R = 0.51

5
4

FLACS (Barg)

20

3
2
1
0

GAME (Barg)

(a) Case 1

6
5
4
3
2
1
0

y=x
R = -0.32

GAME (Barg)

GAME (Barg)

(b) Case 2

(c) Case 3

Fig. 3. The comparison of GAME correlation overpressure results vs. FLACS results for homogenous cases subject to propane vapor explosions.

congestion inherently changes the connement which is demonstrated in this section below. In addition, the manner in which the
congestion parameter was applied for the GAME correlation appears to weight VBR and the characteristic pipe diameter equally.
Here we investigate the isolation of a unique connement parameter and different weighting of VBR in relation to the characteristic
pipe diameter as part of the overall congestion parameter.
We dened the conceptual connement ratio as the total
blocked edge area of a space divided by the total volume of the
space, i.e. ABlocked/ATotal. Then for a cubic volume (of dimension
1 m  1 m  1 m) with six open sides, the conceptual connement
ratio ABlocked/ATotal 0/6 (m2/m2), while the fully conned cube has
the conceptual connement ratio ABlocked/ATotal 6/6 1 which
means the more the surface area being blocked the greater the
connement of the cube. It follows that for a partially conned
volume with 2 sides fully blocked; the ratio is 1/3.
For the same cube (dimension 1 m  1 m  1 m) with six open
sides and with conceptual connement ratio ABlocked/ATotal 0, by
placing a pipe with dimension of 1 m length and 0.4 m diameter in
the center of the cube, as seen in Fig. 5, the congestion volume in
the cube becomes 0.126 m3 (the volume of the pipe) whereas it was
0 m3 in the empty space. Commensurately, the volume blockage
ratio (VBR Vblockage/Vtotal) increased from 0/1 (m3/m3) to 0.126/1
(m3/m3), meanwhile the conceptual connement ratio of the cube
increases from 0/6 (m2/m2) to 0.25/6 (m2/m2), 0.25 m2 is the total
area of the top and bottom cross section of the pipe which reach the
surfaces of the cube on two size. It is clear that a change in
congestion inuences the connement of the conguration
simultaneously, the connement and congestions should be

100

case 1

y=x

case 2
case 3

10

case 4

FLACS (Barg)

case 5

considered together as two interactional factors to determine the

explosion pressure.
3.2. Denition of parameters
Six parameters are taken into account in determining the
overpressure in a vapor could explosion event. They include the
connement ratio, the volume blockage ratio, the characteristic
obstacle diameter, the ame propagation path, the laminar ame
speed and the gas density. In order to investigate their relative
importance in the determination of the overpressures from explosions, parametric studies were conducted and compared to the
output from CFD simulations using the software FLACS.
3.2.1. Connement effect
In accordance with the 2D expansion of the GAME correlation,
simulations were conducted using a geometric conguration that
had parallel plates, the connement was then dened as

Cm

AB
AT

(3)

The blocked area AB is the sum of obstructed areas on the top

and bottom of the domain simulated; AT is the total area of the top
and bottom surfaces. As seen in Fig. 1, the connement parameter
was regulated by reducing the blocked surface on the top of the
geometries modeled using FLACS, while the averaged diameter and
volume blockage ratio as well as other parameters were xed at
certain values. 24 CFD simulations with 6 different connement
levels were performed to investigate the effect of the connement
parameter on overpressure. It is seen that the pressure varies with
connement according to:

Po eexp8:5$Cm

(4)

where Po is the overpressure calculated at different monitor points

in along the explosion ame path (Fig. 6).

case 6
1

case 7
case 8
case 9

0.1

case 10
case 11
case 12

0.01
0.01

0.1

10

100

Place a pipe inside

case 13

GAME (Barg)

Fig. 4. Overall results of GAME correlation vs. FLACS simulation from all cases.

1m*1m*1m Open air cube

Pipe (1m length, 0.4m diameter)

Fig. 5. Conceptual denition of connement and congestion.

Lf=64.3m

3.5

Lf=67.3m

Lf=70.9m

2.5

Lf=76.8

2
Lf=64.3m
trendline
Lf=67.3m
trendline
Lf=70.9m
trendline
Lf=76.8m
trendline

1.5
1
0.5
0
0

0.1

0.2
Confinement Exp(8.5*Cm)

0.3

Fig. 6. Simulation results of connement effect and trendlines for the CFD cases.

3.2.2. Effects of VBR and the average obstacle diameter D

The GAME correlation uses the equally weighted volume
blockage ratio (VBR) and the characteristic obstacle diameter (D) as
the basic predictors of congestion. However, in order to address the
issue of irregular congestion, the VBR was differently weighted in
the correlation developed here, and the averaged obstacle diameter
was investigated separately.
The volume blockage ratio here was dened as the ratio of
obstruction volume within the domain from the ignition point to
the target point to the total conguration volume, so for each
specic target of interest, there is a unique VBR to calculate the
overpressure.
Applying the results from 8 cases with different VBR resulting in
a total of 32 CFD cases, the parameter of VBR was varied while other
parameters kept constant (e.g. constant Cm 1) to determine the
effect of VBR in correlation with overpressure. Similarly, 25 CFD
simulations with 5 different averaged obstacle diameters (D) were
conducted to investigate the relationship between overpressure
and averaged obstacle diameter while the VBR and the other parameters were xed. The similar slopes are seen in Fig. 7(a) and (b)
respectively, which indicates that for all cases the correlation
among overpressure, VBR and D are:

Po e1:6 lnVBR 6

(5)

21

1:5
D
Po e
H

(6)

where H is the height of the conguration.

3.2.3. Maximum distance of ame propagation
The maximum distance of ame propagation (Lf) in the CSC was
dened as the direct distance from the ignition location to the
target point of overpressure in contradistinction to the assumption
in the GAME project that Lf was equal to the radius of a hemisphere
with a volume equal to the volume of the conguration, which
makes the CSC easier and more convenient to use. About 300 CFD
simulation cases were included in the investigation. The trendlines
for all cases show similar slopes In Fig. 8 with the power of 2.2 of
the maximum distance of ame propagation, namely:

2:2
L
Po e f
H

(7)

3.2.4. Mass density and laminar ame speed of gas

For these two parameters, 13 explosion scenarios consist of
approximately 1100 simulation cases were conducted by using two
different gases which were methane and propane with two
different mass densities and two different laminar ame speeds.
The approach taken here was to use the phenomenological method
to analyze all the available data. The last correlation for mass
density and laminar ame speed with power of 0.5 and 2 was
subsequently found by minimizing the total variance of the complete correlation against CFD results.
3.3. New proposed correlation
With the parameters of connement, volume blockage ratio, the
average obstacle, laminar ame velocity and gas density derived in
the manner described above, the new dimensionless correlation
(CSC) is given by:

DPo
0:037$e8:5Cm $1:6 lnVBRt
Pair




2:2
1:5

rgas 0:5 Sl 2
L
D
$
$
$
6$ f
H
H
rair
Ss

Pressure from correlaon Po (Barg)

where:

3.5

Lf=39.9m

Lf=42.3m
Lf=45.6

2.5
2

Lf=46.7m

1.5

Lf=39.9m
trendline

Lf=42.3m
trendline

0.5

Lf=45.6m
trendline

Lf=46.7m
trendline

Volume blockage rao 1.6ln(VBR)+6

(a) Parameter of VBR

Pressure from correlaon Po (Barg)

Pressure from correlaon Po (Barg)

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

14

Lf=45.6m

12

Lf=48.7m
Lf=49.7m

10

Lf=51.7m

Lf=64.3m

Lf=45.6m
trendline
Lf=48.7m
trendline
Lf=49.7m
trendline
Lf=51.7m
trendline
Lf=64.3m
trendline

4
2
0
0

5
10
15
Averaged obstacle diameter (D/H)-1.5 (m)

(b) Parameter of D

Fig. 7. Simulation results and trendlines for the CFD cases.

(8)

22

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

Corner ignion
Cm=1

Pressure from correlaon Po (Barg)

5
4.5

Corner ignion
Cm=0.925

4
3.5

Edge ignion
Cm=0.85

Corner ignion
Cm=1 trendline

2.5
2
1.5
1

20

40

60

80

Corner ignion
Cm=0.925
trendline
Edge ignion
Cm=0.85
trendline

Maximum distance of flame propagaon (Lf/H) 2.2 (m)

Fig. 8. Simulation results of ame propagation maximum distance effect and trendlines for the CFD cases.

DPo the escalation overpressure [barg],

Pair 1 standard atmospheric pressure 101.325 kPa [1 barg],

D the average obstacle diameter [m],
Lf the direct distance from the ignition location to the target
point [m],
Sl the laminar ame speed of the ammable gas [m/s],
Ss the speed of sound [m/s],
Cm the connement ratio,
VBRt the volume blockage ratio of conguration region from
the ignition point to the target,
rgas mass density of gas (kg/m3) (the gas density is assumed
ideally under one standard atmosphere pressure at normal
temperature 26 in this study),
rair mass density of air (kg/m3),
H the height of the conguration (m).

3.4. Validation of the new correlation

In the validation of CSC, about 1100 realistic and idealized vapor
cloud explosion simulations from the modules in Figs. 1 and 2 were
conducted to compare the results of CSC with FLACS data. The
obstacle congurations were lled with equivalent stoichiometric
ammable gas cloud in the simulations; methane and propane
were used as fuels in this study. The parameters are shown in
Table 1.
As seen in Fig. 9, the comparison of the overall results between
the pressures yielded by the CSC and the simulation results from
FLACS is remarkable the R-squared value yielded by this
16
14

y=x
R = 0.8395

Pressure in FLACS (Barg)

12
10
8
6
4
2
0
0

10

12

14

16

Cofinement Specific Correlaon (CSC) calculated pressure (Barg)

Fig. 9. Overall R-squared values of CSC vs. FLACS results for 13 simulation cases subject
to two types of gas vapor explosions.

comparison is 0.8395. For individual cases as shown in Fig. 10, the

correlation factor R-squared is within the moderate range of
0.44e0.90, which means the CSC applies to all realistic and idealized modules very well. In terms of the homogenous case 1e3 with
different connement, the factors of R-squared in CSC are 0.886,
0.608, 0.464 respectively, which are closer to the CFD results than
for the GAME correlation. Overall, the CSC equation corresponding
to the connement criteria can be effectively utilized to the practical problems with different congestion and connement
conditions.
4. Discussion
Although the GAME correlation includes the volume blockage
ratio and the characteristic pipe diameter together in an attempt to
account for congestion, the connement of the combusting gas is
not thoroughly examined in the experimental programs. It is
shown in Section 3 that a change in congestion necessarily affects
the connement, and the change can be signicant. Connement
plays a major role in the evolution of combustion and the overpressure magnitudes generated. In real explosion situations, the
two parameters of connement and congestion should be both
explicitly accounted for in the calculation of explosion
overpressure.
Secondly, the acceptable results can only be obtained from the
GAME correlation when it is used in the situations where obstacles
are homogeneously distributed; indeed the hydraulic average
obstacle diameter is taken into account to investigate the congestion effect on the overpressure calculation. However, in the practical scenarios where there is inhomogeneous distribution of
obstructions, the explosion overpressure will vary signicantly and
this will result in overpressure being overestimated as a result of
using GAME correlation based on the averaged parameters of
obstacle diameter.
Another critical parameter to determine the vapor explosion
pressure is the ame path length, which has, for the GAME correlation, been assumed to be equal to the radius of a hemisphere with
a volume equal to the volume of the obstructed region (Mercx, van
den Berg, & van Leeuwen, 1998). Hence, for cases that ignition locations at the edge/corner of the congurations and the aspect ratio
of congurations larger than 1, it creates the question that how to
convert or determine the ame path length appropriately. This
uncertainty results in an uncertainly in the overpressure through
the use of the GAME correlation, which also makes the GAME
correlation difcult to be applied in realistic scenarios.
Finally, and interestingly, one notes that the GAME correlation is
dimensionally unbalanced, i.e. the dimensions of the right hand
side of the equation do not match those for the left hand side. This
indicates that there are either too few or too many parameters and
indicates that there will also exist cases where this equation is not
appropriately applicable.
On the other hand, the applicability of the newly developed
correlation is greatly improved, and the issues about connement
and congestion are addressed.
The greatest difference between the two correlations is that the
connement Cm which was not accounted for separately in the
GAME correlation was introduced in CSC. In order to invoke two
different GAME equations to calculate the overpressure, the gas
explosions were classied into three categories, namely 1D, 2D and
3D expansions depends on the degree of connement. However,
the way in which the GAME distinguishes between the 2D and 3D
ame expansion lacks detail; for instance, congurations with explosion charge conned by parallel planes were deemed to be 2D
expansion to enable Eq. (2) applying to all such cases, which results
in very large errors for those cases with partially conned top.

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

y=x
R = 0.608

FLACS (Barg)

5
4
3
2

4
3
2
1

1
0

y=x
R = 0.765

0.02

0.005

0.01

0
0.02

y=x
R = 0.485

4
3

CSC (Barg)

16

14

4
3
2

12
10
8
6
4

y=x
R = 0.644

0
0

10

20

CSC (Barg)

CSC (Barg)

(10) Case 10

(11) Case 11

14

y=x
R = 0.787

12

18

(9) Case 9

14

20

CSC (Barg)

(8) Case 8

(7) Case 7

0.05

CSC (Barg)

1
0

y=x
R = 0.561

0.02

0.05

(6) Case 6

0.03

CSC (Barg)

FLACS (Barg)

FLACS (Barg)

0.04

0
0

0.05

0.01

(5) Case 5

y=x
0.07 R = 0.767
0.06

0.03

CSC (Barg)

y=x
R = 0.90

12

10

FLACS (Barg)

FLACS (Barg)

0.08
y=x
R 0.585

FLACS (Barg)

0.01

(4) Case 4

FLACS (Barg)

FLACS (Barg)

FLACS (Barg)

FLACS (Barg)

0.02

0.02

(3) Case 3

0.04

0.015

CSC (Barg)

y=x
R = 0.486

0.025

0
0

5
CSC (Barg)

0.05

0.03
y=x
R = 0.441

0.005

(2) Case 2

0.025

0.01

CSC (Barg)

(1) Case 1

0.015

CSC (Barg)

0.02

y=x
R = 0.464

FLACS (Barg)

FLACS (Barg)

y=x
R = 0.886

FLACS (Barg)

23

8
6
4
2

10
8
6
4
2

0
0

10

10

CSC (Barg)

CSC (Barg)

(12) Case 12

(13) Case 13

Fig. 10. The comparison of CSC overpressure results vs. FLACS results for 13 cases subject to methane and propane vapor explosions.

Therefore, a criterion was established in the CSC, the connement

was well dened by considering the ratio of blocked area to total
surface area of the conguration, the improved correlation of the
CSC equation's results and the FLACS data are seen in Fig. 10.
Secondly, in the CSC, the volume blockage ratio and the average
diameter were investigated separately with unequal weightings to
quantify the congestion. Unlike the parameters D and VBR in GAME
correlation, the volume blockage ratio of conguration was dened
as it in the calculation volume from the ignition point to the target,
the overpressure calculation in the conguration with inhomogeneous congestion therefore can be accurate since for each specic
target of interest, it has individual VBR to determine the
overpressure.

Moreover, the CSC is relatively easy to use by redening the

maximum distance of ame propagation (Lf) as the direct distance
from the ignition location to the target point of overpressure,
whereas in the GAME project that Lf was assumed to be equal to the
radius of a hemisphere with a volume equal to the volume of the
conguration, where for edge/corner ignition cases and the congurations with aspect ratio of larger than 1, the GAME correlation
may not be applicable. And thanks to the balanced inputs' dimension to represent different fuels, the gas mass density was introduce
as well as the laminar ame speed of gas (Sl) to improve the accuracy of the new correlation.
However, all the simulations conducted so far have used
methane and propane as fuels for explosions. Further tests using

24

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

different ammable gases and mixed gases are required to test and
validate the CSC model for those cases.
It is noteworthy that for CFD, the advent of use of mixed gases
and multiple species is a recent development and only 10 years ago,
the use of pure propane or pure methane was the industry
standard.
The effect of Carbon Dioxide in these reactions is predominantly
as a thermal sink which can slow down the combustion rate and
thereby reduce overpressures. However a notable effect requires
large amounts of the gas to be present. The same is true for humidity (i.e. water in vapor form).
Hence the impact of air humidity was not thoroughly considered
in the simulations. The cases we modeled here, and the current
common practice with CFD in industry has been to date, to ignore
humidity in the overwhelming majority of cases as its effect on
overpressure is considered to be relatively small. The exception is
for mitigation measures involving deluge, or events involving rain
where the evaporation of the water can have a signicant effect. The
inclusion of this effect would make an interesting expansion to the
current work. But this requires detailed studies dedicated to this
effect.
Similarly the effect of Carbon Dioxide mixed in with the reactants has not been considered here. CO2 is also often ignored in
CFD modeling of explosions unless it exists in signicant quantities
mixed with the reactants which is possible but not frequent.
5. Conclusions
In this paper, a new correlation to quantify the overpressure is
developed based on the linear least square method by using 400
CFD simulations of homogenous geometries. The method is applicable only to propane and methane and represents a rst step in
developing robust rapid correlations.
The newly proposed correlation termed CSC has satisfactory
results when it is applied to all scenarios consisting of realistic and
idealized homogenous modules in two different explosion blast
sources.
CSC consists of a relation between parameters describing the
obstructed region (the average obstacle diameter, volume blockage
ratio and connement) and describing the fuel properties. Data
from numerical simulations using the CFD software FLACS served as
a reference for comparison with the results of the CSC. The numerical results for approximately an additional 700 simulations on
more realistic geometries were compared with both GAME and CSC.
The difference between the two approaches relates to one signicant parameter, connement. Connement is introduced in CSC
to set up a reference in different conned scenarios. The concept to
quantify congestion in an obstructed conguration as well as the
volume blockage ratio is redened. Additionally, the gas mass
density is taken into account in the calculation and the CSC is
derived as dimensionless.
Because the new correlation have been tested against over 1100
simulation monitor points carried out using CFD it appears to have
less restriction on its applicability than does the GAME correlation.
In any case, this correlation as well as GAME has applicability as a
benchmarking tool only. However based on the discussion above
also it is not recommended using the GAME correlation outside the
connes of the experiments from which it was derived.
Indeed it would be advantageous to test and develop this newly
proposed correlation by comparison to further experiments.
References
Alfred, S. (1976). The phenomenology of the social world. London: Heineman
Educational Books.

Alonso, F. D., Ferradas, E. G., Perez, J. F. S., Aznar, A. M., Gimeno, J. R., & Alonso, J. M.
(2006). Characteristic overpressureeimpulseedistance curves for vapour cloud
explosions using the TNO multi-energy model. Journal of Hazardous Materials,
137(2), 734e741. http://dx.doi.org/10.1016/j.jhazmat.2006.04.005.
Bakke, J. R., van Wingerden, K., Hoorelbeke, P., & Brewerton, B. (2010). A study on
the effect of trees on gas explosions. Journal of Loss Prevention in the Process
Industries, 23(6), 878e884. http://dx.doi.org/10.1016/j.jlp.2010.08.007.
van den Berg, A. C., & Mos, A. L. (2005). Research to improve guidance on separation
distance for the multi-energy method. TNO Prins Maurits Laboratory research
report 369.
Bleyer, A., Taveau, J., Djebaili-Chaumeix, N., Paillard, C. E., & Bentaib, A. (2012).
Comparison between FLACS explosion simulations and experiments conducted
in a PWR steam generator casemate scale down with hydrogen gradients. Nuclear Engineering and Design, 245, 189e196. http://dx.doi.org/10.1016/
j.nucengdes.2012.01.010.
Bradley, D., Lawes, M., & Liu, K. X. (2008). Turbulent ame speeds in ducts and the
deagration/detonation transition. Combustion and Flame, 154(1e2), 96e108.
http://dx.doi.org/10.1016/j.combustame.2008.03.011.
Chen, Z., Qin, M., Xu, B., Ju, Y. G., & Liu, F. S. (2007). Studies of radiation absorption
on ame speed and ammability limit of CO2 diluted methane ames at
elevated pressures. Proceedings of the Combustion Institute, 31, 2693e2700.
http://dx.doi.org/10.1016/j.proci.2006.07.202.
Davis, S. G., & Hanen, O. R. (2010). New investigation ndings on the 2006 Danvers,
MA explosion. Journal of Loss Prevention in the Process Industries, 23(2), 194e210.
http://dx.doi.org/10.1016/j.jlp.2009.10.007.
Edmund, H. (1989). Ideas pertaining to a pure phenomenology and to a phenomenological philosophy. Second book: Studies in the phenomenology of constitution.
Dordrecht: Kluwer.
Eggen, J. B. M. M. (1995). GAME: Development of guidance for the application of the
multi-energy method. TNO report PML 1995-C44.
EMEG. (1997). Explosion Model Evaluation Group, specications of test cases for gas
explosions e Test case C1. EME project, DGXII, Brussels, Belgium.
Gugan, K. (1979). Unconned vapor cloud explosions. Chemical Engineer-London,
340, 54e56.
Gurwitsch, A., & Garcia-Gomez, J. (2009). The foundation of phenomenology:
Edmund Husserl and the quest for a rigorous science of philosophy. Collected
Works of Aron Gurwitsch (1901e1973), 1(192), 463e470. http://dx.doi.org/
10.1007/978-90-481-2831-0_17.
Hansen, O. R., Gavelli, F., Ichard, M., & Davis, S. G. (2010). Validation of FLACS against
experimental data sets from the model evaluation database for LNG vapor
dispersion. Journal of Loss Prevention in the Process Industries, 23(6), 857e877.
http://dx.doi.org/10.1016/j.jlp.2010.08.005.
Hansen, O. R., Hinze, P., Engel, D., & Davis, S. (2010). Using computational uid
dynamics (CFD) for blast wave predictions. Journal of Loss Prevention in the
Process Industries, 23(6), 885e906. http://dx.doi.org/10.1016/j.jlp.2010.07.005.
Harris, R. J., & Wickens, M. J. (1989). Understanding vapour cloud explosions e an
experimental study. In 55th Autumn meeting of the Institution of Gas Engineers.
Harrison, A. J., & Eyre, J. A. (1987). The effect of obstacle arrays on the combustion of
large premixed gas air clouds. Combustion Science and Technology, 52(1e3),
121e137. http://dx.doi.org/10.1080/00102208708952572.
Huser, A., Foyn, T., & Skottene, M. (2009). A CFD based approach to the correlation of
maximum explosion overpressure to process plant parameters. Journal of Loss
Prevention in the Process Industries, 22(3), 324e331. http://dx.doi.org/10.1016/
j.jlp.2008.12.001.
Lobato, J., Canizares, P., Rodrigo, M. A., Saez, C., & Linares, J. J. (2006). A comparison
of hydrogen cloud explosion models and the study of the vulnerability of the
damage caused by an explosion of H2. International Journal of Hydrogen Energy,
31(12), 1780e1790. http://dx.doi.org/10.1016/j.ijhydene.2006.01.006.
Mannan, M. S., Aldeeb, A. A., & Rogers, W. J. (2002). Understanding the role of
process chemistry in res and explosions. Process Safety Progress, 21(4),
323e328. http://dx.doi.org/10.1002/prs.680210409.
Mercx, W. P. M., & van den Berg, A. C. (2005). Vapour cloud explosion (Chapter 5).
TNO yellow book: Methods for the calculation of physical effects due to releases of
hazardous materials (2nd ed.). TNO e The Netherlands Organisation of Applied
Scientic Research.
Mercx, W. P. M., van den Berg, A. C., Hayhurst, C. J., Robertson, N. J., & Moran, K. C.
(2000). Developments in vapour cloud explosion blast modeling. Journal of
Hazardous Materials, 71(1e3), 301e319. http://dx.doi.org/10.1016/S03043894(99)00085-0.
Mercx, W. P. M., van den Berg, A. C., & van Leeuwen, D. (1998). Application of correlations to quantify the source strength of vapour cloud explosions in realistic
situations. Final report for the project: GAMES. TNO report PML 1998-C53.
Mercx, W. P. M., Johnson, D. M., & Puttock, J. (1995). Validation of scaling techniques
for experimental vapor cloud explosion investigations. Process Safety Progress,
14(2), 120e130. http://dx.doi.org/10.1002/prs.680140206.
Middha, P., Hansen, O. R., Grune, J., & Kotchourko, A. (2010). CFD calculations of gas
leak dispersion and subsequent gas explosions: validation against ignited
impinging hydrogen jet experiments. Journal of Hazardous Materials, 179(1e3),
84e94. http://dx.doi.org/10.1016/j.jhazmat.2010.02.061.
Middha, P., Hansen, O. R., & Storvik, I. E. (2009). Validation of CFD-model for
hydrogen dispersion. Journal of Loss Prevention in the Process Industries, 22(6),
1034e1038. http://dx.doi.org/10.1016/j.jlp.2009.07.020.
Moen, I. O., Donato, M., Knystautas, R., & Lee, J. H. (1980). Flame acceleration due to
turbulence produced by obstacles. Combustion and Flame, 39(1), 21e32. http://
dx.doi.org/10.1016/0010-2180(80)90003-6.

J. Li et al. / Journal of Loss Prevention in the Process Industries 31 (2014) 16e25

Pfahl, U. J., Ross, M. C., Shepherd, J. E., Pasamehmetoglu, K. O., & Unal, C. (2000).
Flammability
limits,
ignition
energy,
and
ame
speeds
in
H2eCH4eNH3eN2OeO2eN2 mixtures. Combustion and Flame, 123(1e2),
140e158. http://dx.doi.org/10.1016/S0010-2180(00)00152-8.
Safety, C.f. C. P. (1994). Guidelines for evaluating the characteristics of vapor cloud
explosions. New York: Flash Fires, and BLEVEs AIChE.
Schumann, S., Haas, W., & Schmittberger, H. (1993). Dust explosion venting e
investigation of the secondary explosion for vessel volumes from 0.3 M(3) to
250 M(3). Staub Reinhaltung Der Luft, 53(12), 445e451.
Silvestrini, M., Genova, B., & Trujillo, F. J. L. (2008). Correlations for ame speed and
explosion overpressure of dust clouds inside industrial enclosures. Journal of
Loss Prevention in the Process Industries, 21(4), 374e392. http://dx.doi.org/
10.1016/j.jlp.2008.01.004.

25

Vandenberg, A. C. (1985). The multi-energy method e a framework for vapor cloud

explosion blast prediction. Journal of Hazardous Materials, 12(1), 1e10. http://
dx.doi.org/10.1016/0304-3894(85)80022-4.
Wingerden, C. J. M. v (1988). Investigation into the blast produced by vapour cloud
explosions in partially conned areas. TNO Prins Maurits Laboratory report no.
PML 1988-C195.
Wingerden, C. J. M. v (1989). Experimental investigation into the strength of blast
waves generated by vapour cloud explosions in congested areas. In 6th International symposium loss prevention and safety promotion in the process industries, Oslo, Norway, 1989, proceedings (pp. 26-1e26-16).
van Wingerden, K., Hansen, O. R., & Foisselon, P. (1999). Predicting blast overpressures caused by vapor cloud explosions in the vicinity of control rooms.
Process Safety Progress, 18(1), 17e24. http://dx.doi.org/10.1002/prs.680180105.