Journal of Loss Prevention in the Process Industries 49 (2017) 162e170

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

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

Research on maximum explosion overpressure in LNG storage tank

areas
Dong Lv a, b, Wei Tan a, Liyan Liu a, Guorui Zhu a, *, Lei Peng b
a
School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China
b
Key Laboratory of Building Fire Protection Engineering and Technology of MPS, Tianjin, 300381, China

a r t i c l e i n f o a b s t r a c t

Article history: Maximum explosion overpressure is an important index in combustible gas explosions for safety
Received 12 December 2016 managers and fire rescue commanders. The main factors affecting the maximum explosion overpressure
Received in revised form in an LNG tank area were deduced by a series of reduced-scale tests. A correlation of maximum explosion
18 June 2017
overpressure in LNG tank area was developed based on the momentum conservation equation and the
Accepted 18 June 2017
Available online 20 June 2017
deduced factors in the explosion tests. The parameters in the correlation were fitted by FLACS simula-
tions of explosions of different LNG storage tank areas. Compared with traditional methods such as the
TNO Multi-Energy method and Baker-Strehlow method, this correlation is targeted to the scenario of
Keywords:
LNG
LNG leaking from the joint of a pipeline and the tank top, which may be the most probable type of leak in
Top leaking LNG tank areas. The correlation was verified by five FLACS simulation cases of the LNG tank areas with
Explosion tank radii of 10 m, 20 m, 30 m, 40 m and 50 m. Comparing the overpressures calculated with FLACS and
Maximum overpressure the correlation, there are fewer relative deviations when the radii of tanks are larger. LNG tanks with a
FLACS radius greater than 40 m have been widely developed in recent years, and the relative deviation of the
correlation and FLACS greater than this size is less than 6.6%. The maximum overpressure of the test scale
explosion obtained from the correlation is 326.09 Pa. It is close to 389 Pa, which is the maximum tested
overpressure in this paper. This correlation can be used as a reference for LNG tank area safety managers,
fire rescue commanders and researchers.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction Planas et al (Planas et al., 2015). described LNG vapor explosion

accidents from 1999 to 2012 in Spain. Yang et al (Yang et al., 2011).
As a clean energy resource, natural gas has achieved rapid presented an LNG tank leakage accident that caused a massive fire
development in recent years. According to the IEA (International and explosion, leading to 131 deaths in Cleveland, Ohio in 1944.
Energy Agency) in 2014, the “Golden Age” of natural gas has been Many other LNG explosion accidents have also been recorded in
firmly established in North America, and it will expand to China other studies (Pitblado and Woodward, 2011; Dadashzadeh et al.,
over the next five years (Beckman, 2014). In addition, liquefied 2013; Dan et al., 2014).
natural gas (LNG) will play an important role in clean energy To elucidate the consequences and further reduce the potential
expansion. In the European Union, the consumption of natural gas damage, researchers have focused on the overpressure rule of
was approximately 44.5*1010 m3 in 2010, and it may increase to flammable gas explosions. The TNO Multi-Energy (Bosch and
approximately 2.5*1011 m3 in 2030 (Planas et al., 2015). Meanwhile, Weterings, 1997) and Baker-Strehlow (Lobato et al., 2006)
numerous gigantic LNG storage areas are under construction in methods are popular empirical methods used to describe it. How-
China. Recently, high-capacity tanks (160,000 m3 or even larger) ever, the calculation results of the empirical formulas are generally
have been built in LNG reserves. LNG vaporizes rapidly and be- less accurate than CFD simulation in specific explosion scenarios.
comes an explosive vapor mixture with air in the case of leakage. Furthermore, using these empirical methods, some essential pa-
Many disasters have been caused by LNG vapor explosions. E. rameters must be selected by experience. For example, using the
TNO Multi-Energy method, a parameter of Confinement Strength
from 1 to 10 must be chosen by experience, which makes the
* Corresponding author. calculation results more subjective. Meanwhile, the TNO Multi-
E-mail address: [email protected] (G. Zhu). Energy method and the Baker-Strehlow method assume that the

http://dx.doi.org/10.1016/j.jlp.2017.06.010
0950-4230/© 2017 Elsevier Ltd. All rights reserved.
 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170 163

maximum explosion overpressure is independent of the amount of the assumed leaking tank. The test area includes six tanks that were
vapor/gas. Based on this assumption, a larger explosive gas cloud arranged in two rows and three columns with a steel frame
will only increase the explosive range but not affect the maximum (160 cm
160 cm
58 cm) surrounding the assumed leaking tank.
explosion overpressure. In fact, the maximum explosion over- A remote igniter was positioned in the steel frame. All six tanks
pressure will increase with increasing volume of explosive vapor/ were of the same size with a diameter and height of 82 cm and
gas (Mingshu et al., 2001). For the past few years, LNG tanks have 39 cm, respectively. The distance between adjacent tanks was
become much larger to accommodate economic needs. In this case, 41 cm. The test tank area was a miniature of a 160,000 m3 LNG tank
the increase in peak explosion overpressure due to potentially high area, reduced with the scale of 1:100 in every dimension. Mean-
leaking gas volume cannot be ignored. while, the leaking tank and steel frame were enclosed by a poly-
Researchers have focused on the more accurate methods in ethylene membrane with a thickness of 0.01 mm to prevent the
specific explosion scenarios. As a CFD tool, FLACS is continually flammable gas from diffusing away. The implemented pressure
validated against both large- and small-scale experiments (Middha sensors with output current of 4e20 mA, measurement over-
et al., 2010). The more satisfying results make FLACS widely used in pressure range of 0e1 kPa were placed in the test tank area. The
gas explosion research areas (Davis and Hansen, 2010). Filippo details are listed in Fig. 1. The data acquisition part includes NI
Gavelli et al (Gavelli et al., 2011). studied the explosion conse- (National Instruments) data acquisition and a computer with Lab-
quences of a flammable vapor cloud from LNG spill during the View software.
carrier unloading process. Jingde Li et al (Li et al., 2014) developed
correlations of gas explosion overpressures in and around a con- 2.2. Experimental procedure
gested region by FLACS and verified them with EMERGE (Extended
Modelling and Experimental Research into Gas Explosions) exper- Five scenarios are shown in Fig. 2. The red point and black points
imental programs. numbered nearby are the ignition and pressure sensors, respec-
CFD simulation can obtain accurate results for a certain explo- tively. The ignition was fixed in the middle of two tanks at half the
sion scenario. When the scenario is changed, another simulation tank height in each scenario as shown in Fig. 2. A single number
needs to be reworked over again. In LNG tank areas, the potential indicates only one pressure sensor, which was fixed at half the tank
leaking position, wind speed and direction, ignition position and so height. Two numbers indicate two pressure sensors: an odd-
on affect the overpressure distribution case by case. Therefore, the numbered sensor fixed at half the tank height and an even-
explosion overpressure distribution will change in different sce- numbered sensor fixed at the top of the tank.
narios. Calculating every possible scenario is an enormous work- LNG leakage is rarely caused by the rupture of a tank; it mostly
load. Safety managers, fire rescue commanders and other relevant occurs at the joint of a pipeline on top of the tank. The leaking liquid
researchers might be mostly concerned with the explosion will flow down and evaporate along the outer wall of the leaking
maximum overpressure that might appear in an LNG tank area. tank. Thus, a flammable gas cloud will be generated at the leaking
This work developed a correlation of the explosion maximum side of the tank without the influence of wind. Considering the
overpressure in LNG tank areas based on the momentum conser- worst case, we assumed that the premixed flammable gas cloud
vation equation and the deduced factors in the explosion tests. The would surround the leaking tank.
correlation is targeted at LNG leaking from the joint of a pipeline For each scenario, the weight of the methane bottle was
and the tank top, which might be the most probably type of leak in measured using a G&G electronic balance with maximum range of
an LNG tank area(Kim et al., 2005; Xiaoping et al., 2013). The pa- 30 kg and a division value of 1 g 107 g methane gas that was 1.2
rameters in the correlation were fitted by a FLACS simulation of a times the dose of the stoichiometric concentration leaked into the
methane explosion at different scales of LNG storage tank areas. closed frame through the multi-hole copper tube. The data acqui-
This correlation can predict the LNG explosion maximum over- sition system began to record the pressure data, and then the
pressure conveniently for safety managers and fire rescue remote ignition system was activated. The local wind was south-
commanders. west direction with wind force less than 3 grade (less than 3.4/s).
Photos of the explosion in scenario 5 are shown in Fig. 3 as an
example.
2. Experimental setup and procedure
3. Results and discussion
2.1. Experimental setup
3.1. Overpressure
Fig. 1 shows a schematic of the experiment system with a gas
source the simulating tank area and data acquisition units. The gas
The overpressure in the reduced-scale LNG tank test was
source was a methane bottle providing methane gas into the tank
measured by six pressure sensors in five scenarios. Most detected
area through a plastic pipe. A copper tube with holes to allow the
overpressure values were relatively small, as shown in Figs. 4e8,
gas to mix with air was connected to the pipe outlet, encompassing
except for some scenarios. The rules in high overpressure scenarios
will be studied to develop a maximum overpressure correlation.
The maximum overpressures measured by the six pressure
sensors are listed in Table 1.
In scenario 1, the ignition point was to the right of the flammable
gas cloud. The explosion propagated from right to left. The pressure
values monitored by sensor #1 - sensor #4 in scenario 1 were
relatively lower than those in the other two sensors. Similarly, in
scenario 3, the overpressure in sensor 5 was also much higher than
that in sensor #1- sensor #4. These results illustrate that in a
flammable dispersion range the explosion overpressure increases
in the direction of explosion propagation. In contrast, the pressure
Fig. 1. A sketch of the experimental system. values in the positions of sensor #3esensor # 4 in scenario 2 were
164 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170

Fig. 2. Five scenarios of the explosion tests.

Fig. 3. An experimental explosion process.

relatively low. This reveals a relatively low value of overpressure in that in #5. The left bottom tank was found to be right behind sensor
the opposite direction of gas combustion. #6 in the direction of the explosion propagation. The overpressure
In scenario 3, the overpressure in sensor #6 was higher than in sensor #5 and sensor 6 were also higher than that in sensor #3
 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170 165

Fig. 4. Overpressures versus time in test scenario 1.

Fig. 5. Overpressures versus time in test scenario 2.

and sensor #4 in scenario 5. It was assumed that the rear tank A high blocking rate will contribute to the explosion over-
provides a supporting force to enhance the explosion propagation pressure. In scenario 4, the overpressures were relatively high,
and then enhance the explosion overpressure. which was also observed in sensor #5 and sensor #6 in scenario 2.
Generally, obstacles in the flammable gas may increase the In conclusion, the accelerating burning distance and the rear
explosive overpressure, but a tank located outside of the flammable tank of the explosion propagation direction affect the explosion
gas area will effectively weaken the explosive overpressure. In overpressure effectively. Although the blocking rate is an important
scenario 2, sensor #1 and sensor #2 were in the propagation di- factor influencing overpressure, the factor in a large-scale LNG tank
rection of the explosion, but the overpressures were still very low. area is the same as in the reduced-scale LNG tank area because of
166 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170

Fig. 6. Overpressures versus time in test scenario 3.

Fig. 7. Overpressures versus time in test scenario 4.

 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170 167

Fig. 8. Overpressures versus time in test scenario 5.

Table 1
The maximum overpressure results detected at six sensor points. k2
a2 ¼ (3)
Sensor No. Overpressure (Pa)
r þ k3
#1 #2 #3 #4 #5 #6 Substituting equations (2) and (3) into equation (1) gives:
Scenario 1 8.17 12.95 15.89 14.85 62.98 72.07
dv k2 1 dp
Scenario 2 9.89 13.87 15.19 16.37 210.67 121.82 ¼ k1 r þ  (4)
Scenario 3 13.15 14.39 26.95 19.73 103.39 306.65 dt r þ k3 r dr
Scenario 4 31.04 41.60 39.69 18.17 51.27 45.10
Scenario 5 389.76 122.51 89.71 99.09 235.09 293.05 In the LNG tank area, the parameter Rc is the distance between
the ignition source and the boundary of fire extinguishing. When
the explosion flame spreads to Rc, the acceleration a reaches a
maximum value. Integrating parts of equation (4), we can obtain
model similarity in the assumption of this paper. A correlation in the following:
this paper was developed based on the first two factors above.
ZRc
k2 1
k1 r þ dr ¼ k1 R2c þ k2 ðlnðRc þ k3 Þ  lnðk3 ÞÞ (5)
r þ k3 2
0
3.2. Maximum overpressure correlation
In equation (5), if the explosion occurs in an open condition,
The equation for conservation of momentum can be written as only the accelerating burning force exists without the supporting
equation (1): force from the back, and the overpressure caused by accelerating
burning is proportional to Rc2. This is in agreement with Bi Min-
dv 1 vp gshu's research on gas cloud explosion tests (Mingshu et al., 2001).
¼a (1)
dt r vr Assuming an adiabatic compression (Cleaver and Robinson,
1996), we obtain the following:
where a ¼ a1 þ a2 , a is the acceleration comprising a1 anda2 , the
accelerating burning and the supporting force from the rear tank, 1=g g1
1 vp p0 g vp g
respectively. We assume that the acceleration of the burning is ¼ (6)
r vr r0 g  1 vr
proportional to the accelerating distance in the flammable gas re-
gion as shown in equation (2).
ZRc 1=g  g1 
1 vp p g g1

a1 ¼ k1 r (2) dr ¼ 0 pRcg  p0g (7)

r vr r0 g  1
0
The supporting force a2 weakens with distance. It is assumed
that a2 decreases with the inverse proportion of distance from the The pressure of flowing air is lower than the atmospheric
tank center as shown in equation (3). pressure. However, when the flowing gas is blocked, the dynamic
168 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170

pressure will be produced as follows: 3.0 m
3.0 m
3.0 m, 7.0 m
6.3 m
3.8 m for the x, y and z
directions of each scales LNG tank areas. Ambient temperature was
ZRc 20  C. Consider to the worst scenario, wind velocity was assumed as
dv 1 2  DP
dt ¼ u ðRc Þ  u2 ðR0 Þ ^ (8) zero.
dt 2 r0
R0 By calculating the maximum overpressure of a series of tank
areas with FLACS 9.1, the maximum overpressure △P was obtained
By substituting equations (5), (7) and (8) into equation (4) and as shown in Table 2.
integrating, equation (9) can be obtained as follows: Using Origin to fit the parameters k1, k2 and k3 of correlation
1=g   (12), a fitted curve was obtained as shown in Fig. 9.
DP p g g1 g1
Additionally,
¼ k1 R2c þ k2 ðlnðRc þ k3 Þ  lnðk3 ÞÞ  0 pRcg  p0g
r0 r0 g  1 k1 ¼ 18.26
(9) k2 ¼ 2,217,250
k3 ¼ 1366.63
In equation (9), k1 is used instead of 12k1 because the parameter Substituting the parameters k1, k2 and k3 into equation (12), the
k1 only appears once in equation (9) and k1 is just an undetermined maximum explosion overpressure values of different scale tank
constant in this step. areas were calculated. The results are shown in Table 3.
Rc is correlated with the tank radius R0. In Bin Mingshu's A plot of the overpressures calculated by the correlation versus
research (Mingshu et al., 2001), Rc is approximately 1.8e2.2 times the ones calculated with FLACS is given in Fig. 10.
greater than R0. Assuming Rc is proportional to R0, according to Comparing the overpressures calculated with FLACS and the
equation (9), if R0 substitutes for Rc, the coefficients of k1 and k3 will correlation obtained in this work, there are fewer relative de-
be changed. viations between the two series of data when the radii of the tanks
However, are larger. The relative deviation is 6.55% when the tank radius is
40 m, which is the popular dimension for the new developing LNG
 
Rc tanks in China.
ðlnðRc þ k3 Þ  lnðk3 ÞÞ ¼ ln 1 (10)
k3 Compared with the tests in this paper, the maximum over-
pressure in Table 1 is 389.76 Pa. If R0 is 0.4 m, which is the test scale
The coefficients have not been determined in this step, and the in this research, according to equation (12), △P is 326.09 Pa.
symbols of k1, k2 and k3 can still be used if R0 takes the place of Rc. Overall, the new correlation can be utilized for LNG storage tank
g ¼ 1.4, p0 ¼ 101325 Pa, so, areas as a rapid calculation method.

DP ¼ k1 R20 þ k2 ðlnðR0 þ k3 Þ  ln k3 Þ 3.4. Analysis and discussion

 0:285714286 !
Dp
 354637:5 1þ 1 (11) In order to have a better analysis of our results, some other
101325
datasets were provided. Mercx et al(Mercx et al., 2000). developed
Under normal conditions, the tank area is in an open environ- a vapour cloud explosion overpressure model and compared with
ment, Dp is less than 101325 Pa. Simplifying equation (12) gives the SHELL, CMR, MERGE datasets(HARRISON and EYRE 1987, Hjertager,
following: 1997). In Mercx's correlation, the maximum explosion over-
pressure was postulated as follow.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi  
VBR,Lf b 2:7 0:7
2 4  14:10
106 k1 R20 þ k2 ðlnðR0 þ k3 Þ  ln k3 Þ P0 ¼ a Sl Sc (13)
DP ¼ D
7:05
106
(12) Where:

P0 d maximum explosion overpressure (bar);

VBR d volume blockage ratio;
3.3. Coefficient fitting Lf d flame path length;
D d average obstacle size(m);
In China, the latest built LNG tanks are mostly approximately or Sl d laminar burning velocity of the fuel (m/s);
greater than 160,000 m3. For these high-capacity tanks, it is Sc d a scale factor(m);
impracticable to conduct a real-scale experiment. To validate the a,b d constants which determine best fit correlation, a ¼ 0.84,
new correlation presented in this study, FLACS (Middha et al., 2009; b ¼ 2.75 when calculate the upper bound overpressure ac-
Hansen et al., 2010; Middha et al., 2010; Bleyer et al., 2012) was cording to the Mercx.
used to determine the parameters in the correlation and to verify
the accuracy of the correlation.
A serial of different scales of LNG tank areas were studied by Table 2
FLACS. The radii of the LNG tanks were 10 m, 20 m, 30 m, 40 m and Maximum overpressure of different scale LNG
50 m, respectively. The distance of two tanks was equal to the tank areas.
radius of the tank. The tanks were arranged was two rows and three R0 (m) △P (kPa)
columns, and the leaking gas were around the leaking LNG tank as
0 0
the test scenarios. The simulation volumes were 10 2.25
220 m
210 m
70 m, 340 m
290 m
70 m, 20 11.50
460 m
350 m
100 m, 600 m
500 m
120 m and 30 44.50
700 m
500 m
150 m respectively, with the grid resolutions of 40 54.80
50 67.90
2 m
2 m
1.4 m, 2.0 m
2.1 m
1.8 m, 3.0 m
3.0 m
2.9 m,
 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170 169

Table 4
The comparison of the overpressures calculated from equation (12) and other
datasets.

Tank's radius (m) Overpressures (kPa)

equation (12) MERGE Mercx's correlation

10 9.1 13 8.7
20 20.5 18 14.1
30 34.4 30 18.7
40 51.2 55 22.8
50 71.7 65 26.7

than those from Mercx's correlation. That may result from Mercx's
correlation which are not only based on MERGE dataset, but also
CMR dataset and MERGE datasets. Overpressures from CMR dataset
and SHELL dataset are much lower than MERGE dataset in their
scenario.
Some other researches focused on the gas explosion in LNG
Fig. 9. The fitting curve of explosion overpressures versus tank radii calculated by tanks areas by TNO Multi-Energy method. The maximum over-
FLACS.
pressure is 53.1 kPa in our research, which is close to 50 kPa
calculated by TNO Multi- Energy method.
Table 3
Overpressures calculated with the correlation. 4. Conclusions
R0 (m) 0 10 20 30 40 50
△P (Pa) 0 9143 20,499 34,372 51,213 71,729 This paper focuses on the most probable type of LNG tank leak,
which occurs at the joint of the pipeline and the tank top. The
leaking LNG vaporizes when the LNG liquid flows down along the
outer wall of the tank. The LNG vapor will show a heavy gas char-
acteristic and sink down beside the tank. The explosion maximum
overpressure was studied in this paper based on reduced-scale
tests and the CFD method.
A reduced-scale LNG tank area model at a 1:100 scale in each
dimensional direction to the 160,000 m3 LNG tank area was built to
test the LNG vapor explosion. According to the experiments, the
accelerating burning force, the supporting force from the back tank
and the blocking rate have an important influence on the maximum
overpressure of LNG vapor explosions.
A correlation of maximum overpressure was built based on the
former two factors under the premise of geometric similarity to the
160,000 m3 LNG tank area and the same confined strength. The
parameters in the correlation were fixed using a CFD tool, FLACS.
Furthermore, the results obtained by the correlation are close to the
results of FLACS and the reduced-scale test. The correlation can be
used to conveniently assess the maximum explosion overpressure
of large LNG storage tanks.

Acknowledgement

This research was supported by the Key Technologies R&D

Fig. 10. A comparison of the results of the correlation overpressure vs. FLACS. Program of Tianjin (15ZCZDSF00550) the Open Project of the Key
Laboratory of Building Fire Protection Engineering and Technology
of MPS (KFKT2014MS04) and the “Strengthen Police Force by Sci-
In LNG tank area models assumed in this paper, VBR is 0.528; Lf ence and Technology” special foundation on Basic Research of MPS
is two times of the radius of the leaking tank(Mingshu et al., 2001); (2015GABJC28).
D could be considered as the diameter of the leaking tank; Sl is
0.38 m/s(Goswami et al., 2013). References
The comparison of equation (12), MERGE dataset and Mercx's
correlation on the maximum overpressure of the different scales of Beckman, K., 2014. IEA: Golden Age of Gas Comes to China. from. http://www.
LNG tank areas were listed in Table 4. energypost.eu/iea-golden-age-gas-comes-china/?utm_
content¼buffer121e3&utm_medium¼social&utm_source¼twitter.com&utm_
It can be seen that the overpressures calculated by equation (12)
campaign¼buffer.
is very close to the MERGE's dataset with the same VBR and Bleyer, A., Taveau, J., Djebaïli-Chaumeix, N., Paillard, C.E., Bentaïb, A., 2012. Com-
obstacle size. Andnthe maximum overpressures in Mercx's corre- parison between FLACS explosion simulations and experiments conducted in a
PWR Steam Generator casemate scale down with hydrogen gradients. Nucl.
lation are close to our results when the tanks are small. If the tanks'
Eng. Des. 245 (0), 189e196.
radius is bigger than 20 m, the results from equation (12) are higher Bosch, C.J.H.V.D., Weterings, R.A.P.M., 1997. Methods for the Calculation of Physical
Effects (The Yellow Book). TNO, The Netherlands.
170 D. Lv et al. / Journal of Loss Prevention in the Process Industries 49 (2017) 162e170

Cleaver, R.P., Robinson, C.G., 1996. An analysis of the mechanisms of overpressure overpressure calculation at congested configurations. J. Loss Prev. Process In-
generation in vapour cloud explosions. J. Hazard. Mater. 45 (1), 27e44. dustries 31 (0), 16e25.
Dadashzadeh, M., Khan, F., Hawboldt, K., Amyotte, P., 2013. An integrated approach ~ izares, P., Rodrigo, M.A., S
Lobato, J., Can aez, C., Linares, J.J., 2006. A comparison of
for fire and explosion consequence modelling. Fire Saf. J. 61, 324e337. hydrogen cloud explosion models and the study of the vulnerability of the
Dan, S., Lee, C.J., Park, J., Shin, D., Yoon, E.S., 2014. Quantitative risk analysis of fire damage caused by an explosion. Int. J. Hydrogen Energy 31 (12), 1780e1790.
and explosion on the top-side LNG-liquefaction process of LNG-FPSO. Process Mercx, W.P.M., Berg, A.C.V.D., Hayhurst, C.J., Robertson, N.J., Moran, K.C., 2000.
Saf. Environ. Prot. 92 (5), 430e441. Developments in vapour cloud explosion blast modeling. J. Hazard. Mater. 71
Davis, S.G., Hansen, O.R., 2010. New investigation findings on the 2006 Danvers, MA (1e3), 301.
explosion. J. Loss Prev. Process Industries 23 (2), 194e210. Middha, P., Hansen, O.R., Storvik, I.E., 2009. Validation of CFD-model for hydrogen
Gavelli, F., Davis, S.G., Hansen, O.R., 2011. Evaluating the potential for overpressures dispersion. J. Loss Prev. Process Industries 22 (6), 1034e1038.
from the ignition of an LNG vapor cloud during offloading. J. Loss Prev. Process Middha, P., Hansen, O.R., Grune, J., Kotchourko, A., 2010. CFD calculations of gas leak
Industries 24 (6), 908e915. dispersion and subsequent gas explosions: validation against ignited impinging
Goswami, M., Derks, S.C.R., Coumans, K., Slikker, W.J., Oliveira, M.H.D.A., hydrogen jet experiments. J. Hazard. Mater. 179 (1e3), 84e94.
Bastiaans, R.J.M., Luijten, C.C.M., Goey, L.P.H.D., Konnov, A.A., 2013. The effect of Mingshu, B., Shulan, W., Xinwei, D., Zhenghong, L., 2001. Experimental study on
elevated pressures on the laminar burning velocity of methaneþair mixtures. explosion pressures of unrestricted gas cloud explosions. J. Chem. Ind. Eng. 52
Combust. Flame 160 (9), 1627e1635. (1), 68e71.
Hansen, O.R., Gavelli, F., Ichard, M., Davis, S.G., 2010. Validation of FLACS against Pitblado, R.M., Woodward, J.L., 2011. Highlights of LNG risk technology. J. Loss Prev.
experimental data sets from the model evaluation database for LNG vapor Process Industries 24 (6), 827e836.
dispersion. J. Loss Prev. Process Industries 23 (6), 857e877. Planas, E., Pastor, E., Casal, J., Bonilla, J.M., 2015. Analysis of the boiling liquid
HARRISON, A.J., EYRE, J.A., 1987. The effect of obstacle arrays on the combustion of expanding vapor explosion (BLEVE) of a liquefied natural gas road tanker: the
large premixed gas/air clouds. Combust. Sci. Technol. 52 (1e3), 121e137. Zarzalico accident. J. Loss Prev. Process Industries 34, 127e138.
Hjertager, B.H., 1997. Explosions in Obstructed Vessels.” a CFD Study of Entrained- Xiaoping, S., Yuan, Z., Guoming, C., xi, C., 2013. An analysis of foreign and domestic
Flow Gasifiers Using Two Feed Types. explosion accidents in LNG tank fields and proposals forpreventing measures(in
Kim, H., Koh, J.S., Kim, Y., Theofanous, T.G., 2005. Risk assessment of membrane type Chinese). Nat. Gas. Ind. 33 (5), 126e131.
LNG storage tanks in Korea-based on fault tree analysis. Korean J. Chem. Eng. 22 Yang, X., Dinh, L.T.T., Castellanos, D., Amado, C.H.O., Ng, D., Mannan, M.S., 2011.
(1), 1e8. Common lessons learned from an analysis of multiple case histories. Process
Li, J., Abdel-jawad, M., Ma, G., 2014. New correlation for vapor cloud explosion Saf. Prog. 30 (30), 143e147.