Professional Documents
Culture Documents
Underground Gas Pipeline Explosion and Fire: CFD Based Assessment of Foreseeability
Underground Gas Pipeline Explosion and Fire: CFD Based Assessment of Foreseeability
net/publication/276501485
CITATIONS READS
21 1,368
2 authors:
Some of the authors of this publication are also working on these related projects:
Thermal radiation hazard from biofuels pool fires (Sponsor: DST, Duration: 2016-2019) View project
All content following this page was uploaded by K. D. Wehrstedt on 17 May 2021.
a r t i c l e i n f o a b s t r a c t
Article history: Scenarios of underground gas pipeline failure, crater formation, dispersion of gas, explosion and sub-
Received 2 March 2015 sequent fires are investigated with semi-empirical and with CFD (Computational Fluid Dynamics)
Received in revised form modelling. In order to strengthen the accident based learning approaches present investigations are
7 April 2015
performed in the context of recent GAIL (Gas Authority of India Limited) natural gas pipeline incident
Accepted 8 April 2015
occurred in India. The foreseeability of damages to lives of people and assets due to explosion over-
Available online
pressure and thermal radiation are assessed. The released gas is considered as slightly dense-than-air i.e.
1.5 times. Depending on the LFL (Lower Flammability Limit) of gas the dispersion diameter and heights
Keywords:
Natural gas
are predicted which followed the visual evidences appropriately. The model was furthermore tested with
Underground pipeline an even denser medium and was found to be worked well there too. The estimated explosion over-
Dense gas dispersion pressures with the standard methods and also with CFD reproduced the scenario nicely. The effects of
Explosion and fire congestion VBR (Volume Blockage Ratio) in form of vegetation on stable atmospheric boundary layer
CFD flow is analysed and its contribution towards turbulence and hazard enhancement is studied.
Safety distance It is found that the major source of fatalities was higher thermal radiation emitted by pool fires of
methane. The estimated thermal safety distances clearly demonstrate the ignorance/under estimation of
likelihood and consequence of such hazardous events. For such incidents CFD demonstrated a strong
capability to assess the pre or/and post events foreseeabilities within a reasonable amount of time and
with an acceptable level of accuracy meeting the industrial needs for risk analysis.
© 2015 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jngse.2015.04.010
1875-5100/© 2015 Elsevier B.V. All rights reserved.
K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542 527
of source, explosion overpressure and thermal radiation from pool minor leaks were brought into the knowledge of the company of-
fire. The EN 1473 and NFPA 59A codes (Federal Pollution Control ficials well in advance (The Hindu; Firstpost; The New Indian
Act, 2013; Berger, 2009; PHMSA Calendar Year, 2009; NFPA 59A, Express) which were however ignored. So it is likely that the
2013; European Standard EN 1473, 1997e11; American Institute incident was waiting to occur. An existence of minor leak of
of Chemical Engineers/Center for Chemical Process Safety, 1994; 1 mme5 mm cannot be a pure hypothesis as also seen (quality of
Raj, 2008) were used to develop appropriate safety distances rusting of steel pipe) after incident images (Fig. 2) (Woodward and
from explosion and fire risks. Furthermore, the effects of congestion Pitbaldo, 2010).
in form of vegetation on atmospheric boundary layer flow and If it were only a minor leak the formation of a crater (7 m 7 m)
overall increase in hazard magnitude is also studied. It is demon- and complete rupture of dp ¼ 0.5 m pipe probably can not be
strated that CFD helps to establish more realistic scenario de- justified. There had to be an additional source of overpressure
scriptions and provide enough evidences for developing reliable generation inside the pipe itself. That can only be possible with an
safety distances and better risk assessment criterias. overpressuried (above design) supply of gas via the compressors.
Some sources also claim that this overpressure was twice as much
2. The incident as of designed supply (Normal: ~2.8 bar and Over: ~4.8 bar). This
fact clearly indicates that there was an initial blast leading to full
On 22 June 2014 an explosion occurred in GAIL (Gas Authority of bore failure of steel pipe leading to a major spill of gas.
India Limited) owned underground pipeline transporting natural
gas to the nearby refinery in east godavari district of the state of 3.2. Full bore rupture
Andhra Pradesh in India (Fig. 1). The incident began with over-
pressurized gas supply which led to the initial blast of buried Full bore rupture (due to excess supply of gas) seems to be the
pipeline and formed a crater of radius ~7 m as shown in Fig. 2 most likely consequence to justify the crater formation and major
(wikimapia). Due to full bore rupture a massive spill of gas over a spill around the residential areas. Fig. 3 shows an arbitrary source of
radius of ~0.5 km occurred. One of the many potential sources spill and the probable extent of an affected diameter of ~300 m.
might have ignited the gas cloud which exploded and kept burning Significant vegetation in form of coconut and deciduous trees were
the available/supplied gas in form of pool fires for hours long found on the site. Fig. 4 shows one of such location where a total
leading to heavy destruction of life, property and environment. Till congestion in terms of Volume Blockage Ratio (VBR) can be even up
this paper was written a total reported causalities and serious in- to 10%. When considered the total affected area is largely populated
juries were 18 and 40, respectively. The total capital loss has not with coconut trees (d1 ¼ 0.2 m and l1 ¼ 25 m) a VBR of 1e5% would
been assessed yet (The Hindu; Firstpost). be a better approximation.
Fig. 2. Formed crater as a result of overpressurized gas supply leading to first blast and full pipe (bore) rupture.
MW2
equivalent method (Fig. 7) this estimation leads to a scaled distance
Pmax ¼ 2:4 (1)
z* value of 10 m kg1/3 and corresponding source overpressure of ð1 þ MW Þ
200 kPa at 20 m. This high value of overpressure was not in
agreement with the observed damages. Due to this overprediction Pmax
TNT equivalent method is not preferred for UVCE, i.e. near field PS ¼ (2)
Pa
overpressure estimations (Baker et al., 1996; Van den Berg, 1989;
GexCon)]. 1=3
Pa
R¼R : (3)
E1
4.2.2. BakereStrehloweTang method
BakereStrehloweTang model is another popular method for where Pmax, dimensionless average side on pressure P S and the
overpressure estimation of such UVCE (Baker et al., 1996). It is scaled distance R; MW: Mach number; E1: total available energy (J).
based on the flame Mach number MW (Fig. 8) and can be written as: Assuming a less than medium reactivity and flame Mach number
K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542 529
Lf 2:75 2:7 0:7
Pmax ¼ 0:84 VBR SL D : (4)
D
Fig. 7. Explosion overpressure vs. scaled distance according to TNT equivalent model.
Some post explosion flame images also suggest that there was a
pressurized supply of fuel for a specific period of time (may be an
hour) which led to the formation of jet fire of natural gas. The
visible flames (Fig. 10) were twice in length to the average height of
a coconut tree (25 m). For estimating the thermal radiation hazard
from a subsonic jet fire the following equation has been used
(Gmez-Mares et al., 2010).
E ¼ SEP4F;R t: (7)
a jet fire is too low if not impossible. The source of major hazard was proximity of fire.
pool fire thermal radiation. Thus, the above estimated hazardous distance covers entire
residential areas posing a higher thermal radiation and causing
significant fatalities and serious burn injuries.
4.4. Pool fire
Fig. 11 shows the existence of a large pool fire (diameter 5. CFD modelling of dense gas dispersion
d ~ 200 m) after UVCE. Probably the jet and pool fires both coex-
isted for some time. Near to the ruptured pipe large amount of 5.1. Fuel
released gas and immediate ignition led to the jet fire of low mo-
mentum and away from the pipe the spilled gas burned in form of a In the present CFD model Natural Gas (NG) called as fuel
pool fire. mixture to define variable densities was selected as a fuel which
For methane pool fires of such a size no measurements have has the properties listed in Table 1. In general NG consists of mainly
methane which is lighter (0.66 kg/m3) than air (1.185 kg/m3) and
been carried out so far. The measured thermal radiation from
largest size LNG (Liquified Natural Gas) pool fire (d ¼ 35 m) was assumed to be dispersed via a buoyant plume. In reality, NG nor-
mally has 85e90 % methane and rest ethane, propane and butane
much higher 270 kW/m2 than the corresponding hydrocarbon fires
of kerosene, diesel and alike fuels (~25e40 kW/m2) (Woodward and some other heavier components. When considered the pres-
ence of the constituents other than methane in such proportions
and Pitbaldo, 2010; Mudan, 1984; DECHEMA, 2012).
Since methane burns cleanly and soot absorption is not signif- making an overall density of fuel to be 1.8 kg/m3 making it heavier
than air. Clearly, a dense-than-air cloud will be formed settling the
icant a higher flame temperature (1500 K) led to a higher value of
thermal radiation. In the present incident this probably was the heavier and reactive components close to the ground surface. The
concentration of this dense gas was assumed to vary exponentially
most disastrous reasons for casualties and serious burn injuries
(Fig. 12). For a methane pool fire (d ¼ 200 m, H ¼ 400 m and with time (see Section 5.4) and reaches its maximum value in
t ¼ 1 s. The dispersion of this gas in air is governed by an additional
b ¼ 200 m) a minimum safety distance 2.4 km (EN 1473) from the
pool rim (flame outer surface) was found. Even the conservative scalar transport equation.
estimation by considering an arbitrary pool of methane (d ¼ 50 m,
H ¼ 250 m, b ¼ 50 m) leads to a safety distance value of ~1 km. 5.2. Geometry and grids
If assumed a pool fire of d ¼ 200 m burned with a rate of 0.2 kg/
m2.s a total burn time for 2 106 kg of fuel is more than 5 min. That CFD modelling was done for a scenario of dense gas release from
is clearly disastrous when all the populated areas were assumed to an opening of 0.5 m diameter to mimic the full bore rupture of the
lie within the unsafe limit i.e. either engulfed or stayed in the steel pipe. Fig. 13 shows the same. This opening was located at 5 m
below the ground level in a cylinderical crater having a diameter
and a height both of 7 m, respectively. The dispersed fuel vapors
were collected in a cylindrical area (diameter D ¼ 500 m and height
L ¼ 25 m).
Unstructured grids (Fig. 14) were generated with details of cell
sizes as given in Table 2. The smallest cell size was 0.25 m with a
total number of grids of 56,265. Additional simulations were per-
formed with refined grid and total number of elements of 415,488
to verify the grid dependency of the solution.
Table 1 wall. Ground was defined as adiabatic wall and the surrounding as
Properties of fuel used in the simulations. open pressure, respectively (Table 2).
Fuel Density LFL LFL
3
kg m Volume % kg m3
5.4. Governing equations
a
Natural Gas (FuelMixture ) 1.8 4 0.03
a
(Gant and Ivings, 2005). The following conservation equations were solved for the
dispersion predictions (Ferziger and Peric, 2006; ANSYS, 2013)
Mass Conservation:
vr v rUj
þ ¼0 (10)
vt vxj
Momentum Conservation:
vðrUi Þ v rUj Ui vp vtij
þ ¼ þ þ rgi ði ¼ 1; 2; 3………Þ (11)
vt vxj vxi xj
Scalar Conservation:
Fig. 13. Schematic of CFD model for dispersion.
!
vðrfÞ v rUj f v vf
þ ¼ G þ qf (12)
vt vxj vxj vxj
R
p¼r T (13)
M
Table 2
Mesh and boundary conditions used in the simulations.
Dispersion model
No. of cells 7 268 6622 7209
Boundary Inlet mass flow Adiabatic Adiabatic Relative
conditions (Velocity: 100 m/s & 250 m/s) no-slip wall no-slip wall pressure
Obstacle model
No. of cells 2054 24,684 7118 2060
Boundary Inlet Velocity Adiabatic Adiabatic Adiabatic
conditions (1 m/s) no-slip wall no-slip wall Relative pressure
Dispersion model
5.7.4. Explosion overpressure estimation
Shear Stress Transport yes e e e
Obstacle model As mentioned in Section 5.7.1 considering the 1/2 LFL a hemi-
keε yes e e e spherical cloud of (d ¼ 500 m) a total source mass of
Pool fire model (density 0.5 volume ¼ 1.8 0.5 (4/3) 3.14 (250)3 ¼ 59
keε yes Eddy Discrete Single 106 kg) and equivalent source energy of (source mass source energy
dissipation transfer step
capacity ¼ 59 106 55 106 ¼ 3.2 1015 Jules). When utilizes
534 K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542
Fig. 15. CFD predicted LFL (0.03 kg/(m3)) isosurfaces of gas for two release rates.
Fig. 16. CFD predicted contours of gas concentrations for two release rates.
multi energy method as discussed in the section with a explosion written as follows.
strength of 6 and 7 (Fig. 9) overpressure values of 50e100 kPa result.
The corresponding safety distance from explosion overpressure for
such gas pipelines should not be less than ~1 km for the populated
MG ¼ exp 〈 ln CC 〉m
p
(16)
Because of the novelty of scenario and unavailability of experi- where Cm and Cp are the measured and predicted concentrations,
mental data on the same the developed model was validated only respectively.
against the far field LFL predictions. As shown in Fig. 17 LFL were Here, only far field measured data from Coyote test (Woodward
over predicted close to the source of spill due to the crater and wall and Pitbaldo, 2010) was taken for the estimation VG and MG. It can
effects. A major part of initial gas momentum was lost on wall be seen in Fig. 22 that present model fulfills the criteria of model
increasing the higher downwind concentration above and near to validation quite well even though the scenarios were not exactly
the carter. However, for the far downwind locations model pre- the same as in measurements. Hence, the model can be useful for
dicted the LFL very nicely. This is due to the negligible effects of near scenario selection and consequence determination for under-
source conditions and steady dispersion of gas at far distances. ground gas pipeline risk assessment.
Furthermore, the model results were also compared with the
widely used models for dispersion modelling. In Fig. 22 the geo- 6. CFD modelling of effects of obstacles
metric variance (VG) is plotted against the geometric mean bias
(MG) for existing and present models (Woodward and Pitbaldo, The interaction of atmospheric boundary layer with the obsta-
2010). Generally, all the dispersion models are said to be vali- cles (Fig. 23) on the flow path has an important contribution to-
dated when VG is below 2 and within the curve. MG and VG can be wards turbulence and combustion enhancement and finally
K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542 535
Fig. 17. Measured and CFD predicted gas concentrations vs. distance at different ver-
tical locations. Fig. 20. CFD predicted normal distances (z) vs. gas concentrations at different hori-
zontal locations.
Fig. 19. CFD predicted gas concentrations on the center plane over the whole domain for two release rates.
536 K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542
Fig. 21. CFD predicted LFL isosurface (left) and contour plot (right) of concentrations of a gas having a density of 4 kg/m3.
Table 4
Safety distances from overpressure and thermal radiation. Limiting values are 2 kPa
and 1.5 kW/m2, respectively.
Fig. 23. Schematic of an obstacle model for studying effects of VBRs on atmospheric
boundary layer.
Fig. 24. Geometry, grids and boundary conditions for obstacle model.
Fig. 25. CFD predicted velocity contours at two axial locations before first and after last row of obstacle for two VBRs (left: 0.08 and right: 0.125).
Fig. 26. CFD predicted velocity contours at two normal locations before first and after last row of obstacle for two VBRs (left: 0.08 and right: 0.125).
538 K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542
of air with the fuel gas and finally to enhance the combustion. In the 7.3. Eddy dissipation combustion model
present simulations the initial turbulence was low as the inlet ve-
locity of air was kept 1 m/s and the temperature was 293 K. Also In turbulent flows, mixing time is dominated by the eddy
early morning time (5e6 AM) of the incident suggests that the properties and, therefore, the reaction rate is proportional to a
velocity of air replicates the relatively stable weather conditions mixing time defined by the turbulent kinetic energy, k, and dissi-
and corresponds to atmospheric stability class F. Evidently, such pation, ε.
weather conditions prevail the formation of a large cloud which
remain sticking to the ground and subsequent congestion in form of k
rate (23)
trees provides extra necessary turbulence to increase the flame ε
velocity. The important observation was that the irregular distri- In many cases the reaction rates are fast compared to reactant
bution of obstacles enhances the mixing much more than the mixing rates and can be considered as mixing controlled combus-
regularly spaced ones. tion. In the eddy dissipation concept model, the rate of progress of
elementary reaction k, is determined by the smallest of the two
7. CFD modelling of methane pool fires following two expressions:
Reactants limiter
Two simulations for methane pool fires (d ¼ 8 m and 35 m) were
performed to estimate the thermal radiation emitted from them. ε ½I
Rk ¼ A min 0 (24)
The pool sizes were chosen so that they represent 1.) crater as a k nkI
source and 2) the largest methane (LNG) pool fire experiment that where [I] is the molar concentration of component I and I only
has ever been performed. The geometry and grids are shown in includes the reactant components.
Fig. 28. The outer domain sizes were modified according to the pool Products Limiter
size. About 1 million hybrid control volumes were generated with a !
P
minimum cell size of 0.05 m. The solved transport equations were ε ½IW
same as written in Eqs. (10)e(13), respectively. Steady-state cal- Rk ¼ AB P P 00 I (25)
k P nkI WI
culations were performed with the following sub-models.
where P loops overall product components in the elementary re-
action k. The products limiter is disabled when the model coeffi-
7.1. keε turbulence model cient B is set to a negative value. For both single step and multi-step
reaction schemes the value of B set to 1 (ANSYS, 2013).
The equations and descriptions are same as in Section 7.
Fig. 27. CFD predicted velocity contours 1 m above ground for two VBRs (left: 0.08 and right: 0.125).
K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542 539
Fig. 28. Geometry, mesh and boundary conditions for pool fire model.
CFD predicted steady-state temperatures and fuel mass frac- Higher SEP leads to larger safety distances (Dy) for a cylindrical fire
tions for pool fires of methane originated from carter of d ¼ 8 m and of diameter d and height H as described in Eqs. (7) and (8). The
a large pool fire of d ¼ 35 m are shown in Figs. 29 and 30. It can be same have been shown in Figs. 31, 32 and 33. Considering the LNG
seen that increasing the pool size extends the zones of higher pool fire thermal radiation norm according to EN 1473 (for no skin
temperatures (1500 K) near to the base of pool surface. The better burn injury for sufficiently long duration exposure) for a largest
consumption of methane in case of d ¼ 35 m (right of Fig. 30) pool fire (d ¼ 35 m) of methane both measured and computed
proves this. As a result of less radiative losses due to small soot values suggest that for such underground gas transportation
absorption much higher flame temperatures (300 K - 500 K higher) pipelines the safety distances should not be less than
than corresponding liquid hydrocarbon pool fires were predicted 12.5d ¼ 12.5*35 ¼ 437.5 m.
by CFD. That obviously enhances the average SEP as shown in Fig. 31 Extrapolating above outcome of Dy~12.5d to a pool of d ¼ 200 m
to 270 kW/m2 which is 2e3 times higher than the liquid pool fires. provides a safety distance of 2.5 km which seem to be realistic for
540 K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542
Fig. 30. CFD predicted flame temperatures and gas mass fractions at the center plane
(d ¼ 35 m).
Fig. 33. Measured and CFD predicted irradiance E vs. non dimensional distance Dy/d of
LNG (methane) pool fires according to the EN 1473.
For validation of pool fire model apart from the grid indepen-
Fig. 32. Measured and CFD predicted irradiance E vs. non dimensional distance Dy/d of dent study carried out by the author the experimental data listed in
LNG (methane) pool fires. (Mishra; Vela, 2009; Schaelike et al.) are taken. Since large scale
K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542 541
LNG pool fires (diameter ¼ 35 m) for which the measurements are infrastructure and environment in much more realistic manner.
available represent the methane pool fires more closely hence in
Figs. 31e33 only data for LNG are shown. For further details on 10. Additional details on modelling
systematic validation please refer to (Mishra, 2010; Vela, 2009;
Schaelike et al., 2011). The following additional informations concerning modelling
and simulation may be helpful:
9. Applicability of different models
1. Software used: Ansys v14, ICEM CFD (mesh generator), CFX (pre
In the context of the present disaster following phenomenons, and post processing and solver).
i.e. dispersion of a dense gas, vapor cloud explosion, jet and pool 2. Computational time:
fire thermal radiation were found critical to analyze. CFD modelling Dispersion model: On an 8 GB machine with 4 parallel pro-
were done for gas dispersion and pool fires which caused most of cessing it took 30 min to get the steady-state solution.
the fatalities. They are discussed below together with the applica- Obstacle model: In a cluster of 32 CPUs (8 GB each) it took 2 h to
bilities of different models to predict the extent of hazard. get the results for transient case.
Pool fire model: In a cluster of 32 CPUs (8 GB each) it took 8 h to
9.1. Dense gas dispersion get the results for at least 20 s.
9.2. Vapor cloud explosion The foreseeability of an underground natural gas pipeline ex-
plosion and fire incident was investigated with the help of CFD
The overpressure generated due to the explosion can be esti- modelling. As a reference case recent pipeline explosion incident
mated by different semi-empirical and phenomenological models as took place in India was considered. CFD predicted the overall
described in Section 4.2. Out of the four different models discussed diameter of dispersed dense gas to be about 300 m. Such an esti-
both BakereStrehlow and TNO Multi Energy model gives the best mation is consistent with the on-site post incident realities and
estimations of overpressures. The reason behind such better pre- tellings of the victims. An explosion overpressure of 50e75 kPa was
diction is that it counts the essential physical (blockage, congestion) deduced based on the damage characteristics of structures and
and chemical (combustion, laminar flame speed) processes. CFD nearby parked vehicles. Thermal radiation emitted from jet and
modelling for vapor cloud explosion was not done in the present pool fires of natural gas was found to be the major source of fa-
work. Instead curves denoting respective explosion strength (based talities rather than the explosion overpressure. The effects of
on total energy of the hemispherical vapor cloud) were selected and congestion on the enhancement of overpressure was demonstrated
the overpressure were estimated accordingly (see also Section 5.7.4). for a test cube with two VBRs representing the coconut trees pre-
sent on site. Although, this level of congestion can improve the level
9.3. Jet and pool fire thermal radiation of mixing and turbulence to a small extent though can not be totally
ignored. An approximation of LNG (methane) pool fire (d ¼ 200 m)
For both jet and pool fires the equations for predicting the led to the fact that a minimum safety distance of ~2.5 km (EN 1473)
thermal radiation are same except the equation for estimating the from thermal radiation should be taken into consideration for such
flame lengths. Jet fires have more local effects whereas pool fire buried pipelines passing over residential areas.
extends its influence to several hundred meters away from the Gas pipelines transporting flammable or toxic substances close
source. In the considered accident the most disastrous event was to the residential/populated areas require significant reviews con-
the formation of a large pool fire of methane and engulfment of cerning selection of major hazard scenarios, consequences and
nearby residential areas. Therefore, along with semi-empirical resulting risk values from them. If not all at least the risk to the
models CFD modelling were also carried out. Both the approaches public should be estimated by considering the consequences that
i.e. semi-empirical and CFD yield a minimum safety distance to the arose from practical incidents such as one discussed in this article.
public from such gas pipelines should not be less than 2.5 km.
Though, CFD models were much more detailed, realistic and certain References
than the semi-empirical models.
The overall scientific knowledge that this work bring is to add American Institute of Chemical Engineers/Center for Chemical Process Safety, 1994.
Guidelines for Evaluating the Characteristics of Vapor Cloud Explosions, Flash
CFD models in future hazard and risk assessment studies on gas Fires, and BLEVEs. Center for Chemical Process Safety of the American Institute
pipelines in order to demonstrate the safety of public, of Chemical Engineers.
542 K.B. Mishra, K.-D. Wehrstedt / Journal of Natural Gas Science and Engineering 24 (2015) 526e542
ANSYS ICEM CFD and ANSYS CFX-13, 2013. Bundesanstalt fue Materialforschung und -pruefung, Berlin.
Baker, Q.A., Tang, E.A., Scheier, Silva, G.J., 1996. Vapor cloud explosion analysis. Mishra, K.B., Wehrstedt, K.D., 2012. Dispersion characteristics of flammable liquids:
Process Saf. Prog. 15, 106e109. a CFD study. In: Proceedings of 4th International Congress on Computational
Berger, E., April 26e30, 2009. EN 1473 and other EN LNG standards the European Mechanics and Simulation (ICCMS). IIT Hyderabad, 10e12 December, India.
Norms and standards for the LNG Industry. In: AIChE Spring Meeting, April Mishra, K.B., Wehrstedt, K.D., 2013. Diffusive burning characteristics of peroxy-
2009, 9th Topical Conference on Natural Gas Utilization. Tampa, Florida. http:// fuels. Fuel 113, 158e164.
www.allriskengineering.com/library_files/AIChe_conferences/AIChe_2009/ Mishra, K.B., Wehrstedt, K.D., 2014. Spill-over characteristics of peroxy-fuels: two-
data/papers/P145032.pdf. phase CFD investigations. J. Loss Prev. Process Ind. 29, 186e197.
Quelltermberechnung bei strungsbedingten Stoff- und Energiefreisetzungen in der Mishra, K.B., Wehrstedt, K.D., Krebs, H., 2013. Lessons learned from recent fuel
Prozessindustrie Methodenbersicht und industrielle Anwendung, 2012. storage fires. Fuel Process. Technol. 107, 166e172.
DECHEMA, Frankfurt am Main, ISBN 978-3-89746-135-2. Mishra, K.B., Wehrstedt, K.D., Krebs, H., 2014. Amuay refinery disaster: the after-
European Standard EN 1473, 1997e11. Installation and Equipment for Liquefied maths and challenges ahead. Fuel Process. Technol. 119, 198e203.
Natural Gas Design of Onshore Installations (English Version), DIN EN 1473. Mudan, K.S., 1984. Thermal radiation hazards from hydrocarbon pool fires. Prog.
Reccomendations of the German Commission for Plant Safety (KAS-18), Federal Energy Combust. Sci. 10, 59e80.
Pollution Control Act (Bundes-immissionsschutzgesetz), 2013. NFPA 59A, Standard for the Production, Storage, and Handling of Liquefied Natural
Ferziger, J.H., Peri
c, M., 2006. Computational Methods for Fluid Dynamics, third ed. Gas (LNG), 2013 Edition, National Fire Protection Association.
Springer Verlag, Berlin and Heidelberg, Germany. PHMSA Calendar Year 2009 Annual Reports for Gas Transmission and Gathering,
http://www.firstpost.com/india/andhra-massive-explosion-in-gail-pipeline-near- Gas Distribution and Hazardous Liquid; PHMSA Calendar Year 2009 npms
ongc-site-14-dead-1591091.html. submissions for LNG Plants. Pipeline Safety Awareness, US Department of
Gant, SE, Ivings, MJ, CFD Modelling of Low Pressure Jets for Area Classification, HSL/ Transportation.
2005/11. Qi, R., Ng, D., Cormier, B.R., Mannan, S., 2010. Numerical simulations of LNG vapor
http://www.gexconus.com/doc//olav/LPS_Houston_Hansen_2007.pdf. dispersion in Brayton Fire Training Field tests with ANSYS CFX. J. Hazard. Mater.
http://www2.gexcon.com/calculators/new/TNTMethod.php. 183, 51e61.
Gmez-Mares, M., Munoz, M., Casal, J., 2010. Radiant heat from propane jet fires. Exp. Raj, P.K., 2008. A review of the criteria for people exposure to radiant heat flux from
Therm. Fluid Sci. 34, 323e329. fires. J. Hazard. Mater. 2010 (159), 61e71.
Hanlin, A.L., 2006. A review of large-scale LNG spills: experiments and modeling. Schaelike, S., Wehrstedt, K.D., Schoenbucher, A., 2011. CFD simulation to predict the
J. Hazard. Mater. A132, 119e140. thermal radiation of large LNG pool fires. In: 5th European Combustion Meeting
Hanna, S.R., Chang, J.C., 2001. Use of the Kit Fox field data to analyze dense gas ECM-2011. Cardiff University, Cardiff, UK, pp. 1e6.
dispersion modeling issues. Atmos. Environ. 35, 223e142. Buncefield Explosion Mechanism Joint Industry Project: Phase II Dispersion and
Hanna, S.R., Hansen, O.R., Ichard, M., Strimaitis, D., 2009. CFD model simulation of Explosion Characteristics of Large Vapour Clouds, vol. 1 and 2, 2014. SCI,
dispersion from chlorine railcar releases in industrial and urban areas. Atmos. Berkshire, UK. SCI Ref: ED024.
Environ. 43, 262e270. http://www.thehindu.com/todays-paper/gail-pipeline-explosion-in-ap-kills-15/
Johnson, D.M., 2014. Characteristics of the Vapour Cloud Explosion Incident at the article6156994.ece.
IOC Terminal in Jaipur. GL Noble Denton Report Number: 11510. http://www.newindianexpress.com/photos/nation/GAIL-Explosion-Takes-16-Lives/
Khabeev, N.S., 1999. Simulation of vapour explosions. Appl. Energy 64, 317e321. 2014/06/28/article2304789.ece.
Koopman, R.P., Ermak, D.L., 2007. Lessons learned from LNG safety research. Van den Berg, A.C., 1989. The multi-energy method: a framework for vapor cloud
J. Hazard. Mater. 140, 412e428. explosion blast prediction. J. Hazard. Mater. 12, 1e10.
Liu, X., Godbole, A., Lua, C., Venton, P., 2014. Source strength and dispersion of CO2 van den Bosch, C.J.H., Weterings, R.A.P.M., 2005. Methods for the Calculation of
releases from high-pressure pipelines: CFD model using real gas equation of Physical Effect: Due to Releases of Hazardous Materials (Liquids and Gases)
state. Appl. Energy 126, 56e68. Yellow Book, CPR 14E, Director-general of Labour. TNO, The Netherlands.
Luketa-Hanlin, A., Koopman, R.P., Ermak, D.K., 2007. On the applicatio1 of compu- Vela, I., 2009. CFD Prediction of Thermal Radiation of Large, Sooty, Hydrocarbon
tational fluid dynamics codes for liquefied natural gas dispersion. J. Hazard. Pool Fires (PhD thesis). University of Duisburg-Essen, Germany.
Mater. 140, 504e517. www.wikimapia.org.
Mannan, S., 2005. Lee's Loss Prevention in the Process Industries, Hazard Identifi- Woodward, J.L., Pitbaldo, R., 2010. LNG Risk Based Safety: Modeling and Conse-
cation, Assessment and Control, Third ed. Butterworth-Heinemann, Oxford, quence Analysis, AIChE. John Willey and Sons, ISBN 978-0-470-31764-8.
ISBN 0-7506-7858-5. Xie, Q., Tua, R., Jianga, X., Lia, K., Zhou, X., 2014. The leakage behavior of supercritical
Mishra, K.B., 2010. Experimental Investigation and CFD Simulation of Organic CO2 flow in an experimental pipeline system. Appl. Energy 130, 574e580.
Peroxide Pool Fires (TBPB and TBPEH). In: BAM Dissertation Series 63.