Journal of Loss Prevention in the Process Industries 68 (2020) 104284

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

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

Safety analysis of instantaneous release of compressed natural gas from

a cylinder
Ibrahim Abdullah Altuwair , Faisal Khan *
Centre for Risk, Integrity and Safety Engineering (C-RISE), Faculty of Engineering & Applied Science, Memorial University, St John’s, NL, A1B 3X5, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: This study presents a numerical model to analyze the sudden failure of compressed natural gas (CNG) cylinder
Compressed natural gas onboard a CNG vehicle. The model is developed using COMSOL. It accounts for the real gas effects, physical
Physical explosion energy, and combustion of the flammable gas. The model is tested using experimental data.
CNG vehicle Safety
The study highlight compression energy as one of the serious concern. An unintentional rupture of a com­
Failure analysis
Explosion modeling
pressed cylinder filled with natural gas would generate a rapid energy release in the form of the pressure energy
(blast). The release of energy and gas would cause rapid mixing and generate overpressure and may also cause
flash fire. A detailed failure frequency analysis is also done to analyze the effectiveness of barriers. This study
identifies critical points for the safe operation of the CNG system onboard a vehicle.

1. Introduction accidents and injuries have been reported due to road accidents. How­
ever, loss prevention in those reported accidents could have achieved if
A unique challenge to the evaluation of the safety of an overpressure better understood of the compressed natural gas properties and associ­
cylinder and the investigation of risks is the dynamic nature of complex ated hazards carefully analyzed. For instance, Leveson noted that com­
alternative fuel systems. The safety issue associated with overpressure plex systems often operated in high-risk states, which can lead to
has become a controversial topic with increasing uses of alternative fuels unacceptable loss events (Leveson, 2009).Therefore, the risk toward
(i.e., hydrogen H2, liquefied petroleum gas LPG, and compressed natural CNG may become a real threat and affect personnel safety, asset, and the
gas CNG). For example, if the high-pressure system designed to be safe, environment. One of the technical features that make the CNG system
without consideration of the system’s evolving goals, risk propagation different from others is the overpressure energy, which raises safety
would occurred. Therefore, in order to prevent such risks, a reflection of concerns and poses a question of whether the CNG onboard vehicle is
humans and impacts on safety requirements need to be reviewed safe or not. This study discussed safety analysis using protection barriers
throughout the high-pressure systems lifecycle. onboard a vehicle. It also analyzes pressure hazards and associated risks
In addiotn, during the era of the oil crisis, scientists have thought of due to instantaneous release from a CNG storage cylinder. This requires
alternative fuels to replace fossil fuels for powering and heating our cars adopting a systematic model to design flaws that will foster risky op­
and homes such as CNG. Thus, CNG was used as a vehicular fuel in Italy erations and guide us for safer alternative fuels is essential.
in the early 1930 (Khan et al., 2015). There are over one billion of a Despite the threat associated with CNG, the economic and environ­
light-duty motor vehicle powered by fossil fuels in operation (Khan mental benefits of CNG attract developers to invest in it. For example,
et al., 2015). A new statistical summary shows that more than 20 million the chemical structure of CNG gives it several advantages compared to
NGVs and more than 30 thousand fueling stations in operation distrib­ fossil fuels such as gasoline and diesel. The CNG consists of Methane
uted in different regions including Asia-Pacific with more than 70% (CH4),Ethane(C2H6),Propane(C3H8),Butane(C4H10),Carbon dioxide
which become the major part of the transport distribution followed by (CO2),Oxygen(O2), Hydrogen sulfide(H2S),Nitrogen(N2),Helium(He),
Latin America with more than 18%, and 7.31% located in Europe. Less Neon(Ne),and Xenon(Xe) reported by CCPS 2000 (Centre for Chemical
than 1% found in Africa and North America which has recoreded the Proce, 2010). One of the CNG advantages is the lower maintenance cost
lowest number of NGV’s with only 0.81% Table 1. (Khan et al., 2016).Also, due to the lower molar mass of CNG which is
Despite the environmental advantage of NGVs, thousands of much lower than gasoline, this gives a homogenous air-fuel mixture,

* Corresponding author.
E-mail address: [email protected] (F. Khan).

https://doi.org/10.1016/j.jlp.2020.104284
Received 15 December 2019; Received in revised form 6 September 2020; Accepted 8 September 2020
Available online 22 September 2020
0950-4230/© 2020 Elsevier Ltd. All rights reserved.
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

Table 1 community (ISC) (C (2019).ractical, 2019). Also, the consequences of

NVG’s land-based motor vehicle (GlobalNaturalGasVehicle(, 2018). overpressure depend on several factors based on injuries resulted from
World NGV Statistics the blast waves. For instance, the fuel properties strongly affect the
flame speed for a stoichiometric mixture of fuel and air (F/A) (Dorofeev,
Natural Gas Vehicle 27,414,984
Natural Gas Fueling Stations: 32,211 2011). Therefore, variables including fuel properties, concentrations,
Regional Growth NVG wind conditions, congestion scenarios, ignition type, and ignition loca­
Asia-Pacific 19,766,181 tion need to be taken into account when a risk assessment conducted.
Europe 2,005,744 However, it can be more complicated if multiple consequences such as
North America 224,500
Latin America 5,150,210
human loss, structural damage, and environmental effects considered.
Africa 268,349 For instance, the detonation triggered by methane could be more
complicated (Sochet, 2010). Therefore, an appropriate frequency anal­
ysis method required to ensure the reliability of safety risk assessments
completed combustion and benign environment. The CNG also can and barriers effectiveness. Furthermore, an event tree used for frequency
contribute to a higher greenhouse gases reduction. Regardless of the analysis.
benefits of using CNG (i.e., 90% emission reduction), the following key On the other hand, the internal energy of a CNG onboard with a
aspects addressed in the current work to tackle serious safety issues typical volume compressed at a certain pressure (20–36 bar) considered.
associated with CNG: For example, a large internal energy is instantaneously released. This is
due to the physical state of the system therefore, a sudden vessel rupture
• Propose a method to predict accidents using appropriate modeling allows for the compressed gas to expand and release into the environ­
techniques. ment, causing a physical explosion. Thus, the loss of storage tank
• Propose an approach to evaluate the effects of resultant blast wave integrity is not anticipated and should have avoided. Also, the initial
pressure versus the magnitude of the accident (consequence criteria). physical state, which is affected by the heating due to the external effect,
• Propose an alternative model for blast wave analysis. results in violent vapor production and induces substantial pressure
waves if set free.
Developing a strategy that considers the increased overpressure risk Furthermore, empirical methods are used to assess overpressure
onboard a vehicle is given in Fig. 1. events. For example, the equivalent of the trinitrotoluene method (TNT)
is the most straightforward and reliable of all the blast measurements
2. Compressed natural gas (CNG) hazards (Sochet, 2010). Another method of determining the energy release is a
multi-energy method, which is based on experimental data (Sochet,
The safety issues associated with CNG requires comprehensive dis­ 2010). While these methods widely used, the current study has adopted
cussion since current hazard analysis methods, adapted from traditional a more robust and reliable approach to explosion modeling for CNG
accident models, are unable to comprehensively evaluate the potential analysis. A computational fluid dynamic code, COMSOL, that enables
for risk in the case of CNG onboard. Thus, the blast overpressure of an three-dimensional Cartesian Navier-Stokes(NS) flow solver to estimate
explosion is one of the most serious hazards when a compressed gas accident probabilities and overpressures by considering the physics of
failed in related-fire, or non-related fire accident occurrence. It en­ the problem. It can capture complex congestion and confinement in
compasses new knowledge generated by the international gas safety geometry. Fig. 3. The paper aims to apply the study approach for the

Fig. 1. The hazard assessment of compressed gas on-board vehicle.

2
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

safety assessment of the CNG cylinder onboard for controling the po­ walls. Then, the probability of failure derived from the cumulative
tential accident. expression for a normal Gaussian Probability Distribution (GPD) (Fin­
ney, 1971). Also, Khan and Abbasi (2001), who proposed a probit
3. Analysis of past CNG accidents function similar to the Eisenberg expression.
In the framework of the study, particularly hazard assessment, for
There were a number of CNG-related accidents reported earlier by the overpressure storage damage applied. In the following expression
Khan (Khan et al., 2016). Also, the Occupational Safety and Health model, the calculation of loss obtained in terms of
Administration (OSHA) have been documenting accidents regarding Physical analysis using a probit equation for the application of
compressed gases. For example, in 2018, an accident reported by OSHA damage analysis. The use of the probit equation can be suitable for
((1970) and “Vessel ruptu, 1970) highlighted that while conducting an representing the probability distribution function. It is one of the major
inspection and maintenance of compressed equipment, the employee analysis application of the method to physical hazard assessment in the
was struck in the face by high-pressure gas from the pressure relief valve. process industries performed (Finney, 1971).
An inflatable pipe plug into a 24-inch stormwater line installed to stop
Pr ​ = ​ a ​ + ​ b(ln ​ Po ) (1.0)
the flow of water from an adjacent pond. The operator has used an
electric compressor to inflate the plug to 35 psi. The electric compressor The author defined the probability function as (Y) called “Probit
has a regulator and gauge on the manifold and an air supply line. It had Function” and (Po) represents the maximum pressure, a and b are con­
its gauge and shutted off valve, which connected to the pipe plug with a stants (i.e., 23.8 and − 2.92) (Finney, 1971). Thus, a relation between
quick disconnect fitting. The employee was using a shovel to clean the overpressure and damage probabilities can be achieved.
around the pipe in preparation for repair when the inflatable pipe plug
exploded and struck the employee ((1970) and “Vessel ruptu, 1970). The 4. The methodology of the study
operator was killed due to blunt force trauma from the explosion of air
and water pressure as a result of the pipe plug failure. Another incident The methodology explained how the research carried out to easily
was occurred in August 2017, it was an attempt by an operator to use a understand by the readers. For the sake of ensuring the enhanced safety
mixture of compressed gas to cut a catalytic converter off of a junkyard of the CNG system, a risk analysis was developed. As a result of
vehicle. The entire driver’s side of the vehicle was left with a skid steer analyzing the identified hazards, the following perspectives on the
but the operator could access the underside of the vehicle. As the safety of the CNG onboard were deduced:
operator was using a torch to cut, the gasoline vapors started building up
in a small area, therefore an explosion occurred. Also, in December (1) The majority of hazards to the CNG fall in the acceptable risk
2017, two employees were refueling their compressed natural gas (CNG) region.
van when the CNG system failed and exploded. The employees suffered (2) Critical hazards had an unacceptable risk level with a small
leg amputations due to shrapnel wounds as well as sustaining several number.
lacerations ((1970) and “Vessel ruptu, 1970). (3) Hazards that associated with the environmental conditions and
Although high-pressure storage is required to equip with safety de­ the process operation are the main concerns for the CNG onboard.
vices for risk reduction, the released flammable gas could raise addi­
tional safety problems if the ignition source exists (Chuanchuan et al., There is no solid modeling for CNG cylinder onboard a vehicle
2018a). It has been early noticed by many scientists that modeling gas established, therefore the proposed study is critical. However, the
flow considered one of a challenge in the field of simulation, there is no uncertainity of CNG behavior, that is not yet understood, may affect the
general flow model for all fluid flow situations. Several CFD simulations transportation sector due to extreme pressure. The methodology repre­
studies were published and carried out to investigate the gas release and sents the risk analysis steps. Fig. 1 presents the hazard analytical
risk from gas (Houf et al., 2012). Also, a detailed description of blast method, which is the process by which a design analyzed for safety. It is
wave events and the influenced of fluid flow determined (Brinkley and used to identify hazardous states in a scenario. The first step aims to
Lewis, 1959) and the mechanisms of the flow effect and investigation of identify the hazard related to the system for fire-related or non-fire
explosion events thoroughly analyzed (Lewis and von Elbe, 1987) (Kuo, related to the internal pressure elevated inside a CNG cylinder. The
1986). identification of hazards relevant to the accident scenario with over­
Furthermore, Kuo studied the principle of combustion (Kuo, 1986), pressure vessel in a fire. Those were identified based on prior studies.
and different mechanisms of the blast investigated, including the fluid The second step is to determine the consequence of CNG cylinder failure
flow and formation of shock waves (KravtsovJacob et al., 2015) given an by presenting the theoretical modeling schemes. In the consequence
explanation of the explosion due to the velocity of the shock wave. Also, analysis, an alternative model showed for non-related fire (Molkov,
an investigation of the propagation process of the methane explosion Kashkarova). It is designed based on Baker method and provided the
profoundly studied (Cheng et al., 2012). A gas expansion may generated estimation of hazard distances in which the pressure and thermal effects
When chemical composition changed and creating energic material cause death, serious or slight injuries from fire and explosion calculated.
release. This instantaneous release of chemical energy can view and These earlier models were tested for estimation of initiating event
explosion. Thus, the characterization of flow plays an important role in (overpressure) frequency, and calculation of the probability failure
the transition of developing phenomena from combustion to deflagra­ resulting in tank rupture. However, the limitation of physical parame­
tion or detonation. The study follows an experimental study published ters could be a source of error in many of those models. In this study, an
earlier to determine the effect of both the chemical and physical effects appropriate technical method used (i.e., event tree). Finally, risk can be
of CNG. Obtaining the initial blasting stress, displacement, and over­ obtained as a function of the tank rupture probability.
pressure of the fluid blast using theoretical and experimental calcula­ In addition, the study provides a systematic engineering model to
tions are much better than that of ordinary blasting. identify the effect of the blast wave. It based on pressure and potential
Furthermore, the probit analysis is a method of analyzing the dose- energy. More details of the study approach given in the following
effect relation, thermal radiation, and overpressure; however, it can sections.
also be a useful analysis method to determine equipment damages Proposed methods for calculating stored energy of ideal gas theo­
subjected to blast wave (Cable et al., 2019) (Finney, 1971) (Lees, 1996). retically and experimentally presented. A thermal explosion can be
In this study, the Probit function was applied due to the reliability of the determined by the Arrhenius Equation,(AE) (Vagsæther, 2010). The
modeling. As Eisenberg (Eisenberg et al., 1975) presented a model based center of Chemical Process Safety (CCPS) recommended several models
on experimental analysis of storage tank deformation, restricted by rigid to determine the physical strength as the baker model, and Kinetic

3
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

Energy Model (KEM) (Cable et al., 2019). Table 3

Geometry statistics of CNG cylinder.
5. Compressed natural gas cylinder (CNG) simulation Description Value

Maximum element size 0.0795

All Alternative technologies are being asked to perform in a more Minimum element size 4.50E-04
challenging environment, with greater energy densities and power ef­ Curvature factor 0.3
ficiencies, over a longer lifetime. Those requirements are placing more Maximum element growth rate 1.3
pressure on the academia and industries, but simulation is becoming one Predefined size Fine
Minimum element quality 8.66E-01
of the necessary tools for developing, designing and ensuring quality and Average elements 9.64E-01
safety of alternatives during operation. COMSOL is used to gain insight Triangular elements 2.64E+02
into CNG cylinder onboard a vehicle. It allows us to investigate the Width (m) 1.5E0.0
performance for different operating conditions and design in more Height (m) 4.00E-01
secured circumstances. With the add-on module, we are able to simulate
the characteristics such as fluid flow and pressure. Understanding of
these characteristics, allows us to design the geometries and material ∇.(ρu) ​ = ​ 0 3
choices of CNG cylinder, and feeder with respect to safety and perfor­
mance. In addition, the simulation analyses take into account the cor­ Rew ​ =
ρ|u|lw u ρUnsupportedτlw
= = ​ u+ lw+ 4
relations among parameters at high-pressure with a single-phase system μ uτ μ
and thermodynamically closed. A numerical simulation approach per­
formed as a result of a fully wrapped aluminum composite cylinder. ∇G ​ . ​ ∇G + ​ σw ​ G(∇ ​ . ​ ∇G) ​ = ​ (1 + 2σ w )G4 , lw
1 lref
− 5
Also, COMSOL Multiphysics gives us access for prescribing nonstandard G 2
boundary conditions. For example, it allows you to have domain Wall.
boundaries or a particular mesh that fits the load distribution at all In free-flow domains:
times. Therefore the physical variable (pressure) and dependent variable u|Ɩw=0 = 0.
(flux) are prescribed by COMSOL which illustrate the mixture of the In porous domains:
boundary condition types and create a relation between the variable and
h₁
its gradient using geometric shape order automatically. As described in lw =
2
Tables 2 and 3.
The CNG cylinder filled with CH4 and assembly, as specified by Ɩw = 0.
Global Technical Regulation (GTR). Gravity
The numerical study of CNG onboard is a path to broaden knowledge ) [ ) ) ( )( )]
2
of compressible gas flow in non-uniform regions with emphasis on the ρu.∇ u = ​ ∇. − pl ​ + μ + μT ∇u + (∇u)T − μ + μT ∇.u l
3
development of shock waves. Thus, we plan to use the COMSOL to
develop the CNG flow inside a cylinder. As mentioned for modeling low + F + ρg 6
Mach-number (typically less than 0.3) flows of single-phase fluids, the Inlet
turbulent flow type interfaces are used to simulate flow and pressure u = -U0n.
fields. The equations below solved by fluid flow are the Navier-Stoke ∇ԍ . n = 0.
Equation (NSE). The effect of flow are modeled using the standard Outlet
equation yPlus model type with realizability constraints. Flow close to [ ) ) ( )( )]
walls is also described using wall function. Initial flow, as it has signif­ − pl ​ + μ + μT ∇u + (∇u)T −
2
μ + μT ∇.u l n ​ = ​
icant effect on the development of transient flow field, realistic turbulent 3
( )
value were assign to the domain using approximations for dynamic
− ρ0 ​ + ​ ρhydro n 7
viscosity (μ). The fluid flow described in the following equation of NSE
where the pressure was input as a boundary condition at the outlet, and
рhydro = ρref ɡ . (r - rref)
the inlet condition was the velocity of the fluid flows through the cyl­
∇ԍ . n = 0.
inder. Once all the inputs were made,the mesh of the cylinder was
In these equations, ∇ is the gradient, the fluid properties ρ is the
defined and then refined to accurately estimated how the gas flows
density, u is the velocity, p is a pressure, and F represents the external
would act inside the cylinder. Then, the simulation is ready to excuted.
body forces. Consequently, the gas evolution affected through the
The results were tested using the post-processing techniques included
heating rate. However, if more turbulent gas flow created, then the
with the program specifically a boundary integration of pressure across
pressure grows up and causes stress leading to potential damage equa­
several of the boundaries as well as velocity profile.
tion (7).
ρ(u.∇)u ​ = COMSOL is also using computer aided design (CAD)-based geometry
[ ) ) (
2
)( )]
2 management to construct a mesh, it is possible to capture the geometry
∇. − pl ​ + μ + μT ∇u + (∇u)T μ + μT ∇.u l + F + ρｇ in great detail which contained 886613 domain elements,6688 bound­
3
ary elements,and 264 edges elements Fig. 2. Therefore, we are able to
follow the development of simulation and support advanced modeling of

Table 2
The model parameters.
Mesh size parameters Value Dependent Variables Value Model Parameters Value

Maximum size 2.01 cm Velocity U a 0.402

Minimum size 0.009 cm Pressure P ßa 6.59
Maximum element growth rate 1.3 Wall Distance Initialization
Curvature factor 0.3 Reciprocal wall distance G Reference position (rref) X=0
Resolution 1 Wall distance y+ Y=0

4
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

Fig. 2. 3D modeling of CNG cylinder simulation.

method. In the estimation of overpressure, the experiment showed some

variation but other distances showed consistant measurements.
In addition, there is also another vital analysis method, which
depended on the historical data for failure probability estimation such as
fault tree (FT) and event tree (ET). The event tree is the most used logical
method to graphically illustrate the combination of the failure of event
and incident sequences, expressed in annual probability. It widely
applied to explosion risk analysis of oil and gas applications, including
fatality rate and potential loss. Meanwhile, the development of the
thermal effect model also determined by the Arrhenius equation. The
equation illustrates the structural changes due to temperature distribu­
tion, causing structure change and thermal expansion. It contributes to
the unstable behavior of fluid affecting the thermodynamic states of the
Fig. 3. The wall resolution of the CNG cylinder simulation. flow. Also, a model (Molkov, Kashkarova) designed based on the cor­
relation of the mechanical energy of the physical explosion and thermal
the associated physical and chemical process. This helps to model with energy (EPhysical) = Ethermal/0.052). Furthermore, Zel’dovich and Knut
minimum geometrical approximations and minimum uncertainity. (Vagsæther, 2010; Eisenberg et al., 1975) discussed the fire and explo­
Besides the detailed representation of geometry, successful applica­ sion modeling and risk assessment for determining the thermal distri­
tion of COMSOL to CNG cylinder simulation lies in the accuracy with bution from point to point between simultaneous heat release and
which fluid flow is modeled. The final geometry of the CNG cylinder has energy transport. Knut has adopted a mathematical model for the
1 domain,6 boundary, 12 edges, and 8 vertices, Fig. 3. This showed that thermal evolution of the gas explosion. The author believed that the
as the flow moved from center to outer edges of the cylinder, the velocity thermal explosion initiated by a sudden increase in temperature resulted
droped. This due to the roughness at the cylinder wall.Therefore, the from mechanical effects that trigger the temperature, a discussion of the
density and pressure are increased. Thus, as the pressure reduced, the process occurring over significant different time scales (induction time)
compressible gas would have a higher velocity through cyinder. also given (Shepherd, 2009). Many studies discussed the concept of in­
In the CNG simulation, a description of the characterization of the duction time, as Schultz and Shepherd compiled theoretical and exper­
impact of the mechanical energy on the CNG demonstrated and the imental data on induction times for a large variety of gases (Schultz and
probability of failure as a function of internal pressure provided by the Shepherd, 2019). Also, a proposed engineering model based on the
probit model. Observation of some changes in the behaviors of the CNG assumption of the energy of the combustion of gas released into the
was noticed. In the inlet of the cylinder, a growing boundary layer atmosphere added to the mechanical energy of CNG (Chuanchuan et al.,
observed, which gives an explanation of the stream acceleration as the 2018b). The model showed the relationship between the overpressure in
displacement thickness of the boundary layer grows, and it loses its blast wave and distance. Additionally, isothermal expansion model,
identity as the boundary layer thickness reaches a value approximately model, and modeling of kinetic energy can be used for the evaluation of
equals to the cylinder radius. Also, the changes in the flow structure the physical effect caused by fragment. Tables 4, 5, and 6.
indicate that the pressure on the CNG cylinder reaches its fully devel­ The cylinder physical parameters including; P1: Pressure, Vg: Vol­
oped value while the velocity remains constant. It demonstrates the ume, ƴ: Adiabatic index = 1.3, MC: Mass of gas, and Mv: Mass of the
fluctuation of the pressure at the wall from a high value near the inlet to cylinder.
its low value for fully developed flow downstream. It is an indication of As a result, the changes in energy determined and an agreement of
the momentum exchange between the high and low gas particles and both experimental and empirical models. It is vital to notice, however,
causes a pressure increase. It is worthy of mentioning that the temper­ existing methods for determining a blast wave affected by the coefficient
ature effect on stored energy is a source of pressurizing on the CNG value of the mechanical energy (α) (Molkov, Kashkarova)—the uncer­
cylinder due to thermal expansion, which can be considered as addi­ tainty of the source of energy stimulating the blast wave as it propagates
tional to the maximum pressure. On the other hand, the generation of a outwards. Also, empirical methods in previous sections were not suffi­
hot spotty observed due to compression. Thus, failure is possible. cient to obtain precise outcomes. In the study, we attempt to adopt a
reliable methodology leading to predict the physical explosion of the
6. Thermal and physical energy hazards CNG cylinder onboard precisely. The model proposed by Molkov
(Molkov, Kashkarova) valid for all compressed gases and compared to
The hazard has been theoretically and experimentally determined
includes physical energy. The experimental data showed the effect of Table 4
mechanical energy coeffiecient on the estimation of overpressure for the Cylinder physical parameters (Eisenberg et al., 1975).
gas model. Table 7 presented some of the experimental data on over­ P1 (Pa) Pa (Pa) Vg (m3) ƴ MC (kg) Mv (kg)
pressure as a function of distance(s) from the storage tank and the
2.00E+07 1.01E+05 0.047 1.3 0.65 83.4
estimation of overpressure was made by several methods, such as Baker

5
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

Table 5 7.52 mol N2 = 10.52 mol required I mole CH4 + 2 mol O2 + 7.52 mol N2.
The outcomes of studied models (Zel’dovich et al., 1969) (Eisenberg et al., Therefore, Vu = 62.25 m3. Thus, the radius of the hemisphere with a
1975). stoichiometric gas-air mixture would be ru= (3Vu/2π)1/3 = 4.48m. The
Parameters Empirical Models Simulation Model radius of the hemisphere occupied by the combustion products (Vb =
Velocity (m/s) Energy model 172.28 171.1
8.75); then, we got rb = 7.82m. The total mechanical energy of Em =
Moore model 179.65 5.30 MJ, and chemical energy of Ech = 92.5 MJ. Table 5 gives the ap­
Energy (j) 2.28E+06 2.07E+06 plications of the selected type of compressed gas storage as a function of
distance(s). For calculation, we used the following parameters as y
which is the ratio of specific heats for example (air, γ = 1.4); atmo­
Table 6 spheric pressure, which is assumed equal to Ps = 1.013 × 105 Pa; ρa is
Overpressure P‾as a function of distance r‾. the density of surrounding air. The density is calculated as ρa = PsMa/
Application H2 H2 H2 CNG
RT, where R/Ma = 8.314 J/mol/K/0.0289 kg/mol = 288 J/kg/K. The
air temperature was 298.15 K. The ratio of specific heats is γ = 1.3. The
170 L, 35 MPa 33 L, 70 MPa 12 L, 70 MPa 84 L, 30 MPa
density of air calculated as ρa = 1.179 kg/m3. Using the relation between
P‾st 54 80 80 46 the squared ratio of the speeds of sound and the ratio of pressures gives,
r‾v 0.069 0.058 0.058 0.0041
therefore, the overpressure, P‾st = 50.1 (Baker et al., 1983).
rv (m) 0.34 0.2 0.14 0.15
rb (m) 7.91 5.46 3.89 3.08
The results of parameters compared with other experimental results
obtained previously by former literature Table 7. Then α = 0.14 used to
find the blast wave overpressures at different ranges, and the maximum
pressure test is as follows: 140 kPa at a distance of 1.22 m and 80 kPa at a
Table 7
distance 2.44 m. The value of overpressure used with the radius
Comparison of CNG model calculations and experimental data (Molkov,
Kashkarova).
equivalent volume spherical vessel determined. Also, due to reflection
from the ground, the energy stored in the cylinder (Baker et al., 1983),
Reference Overpressure, P (kPa), at different distance r (m)
assuming if the rupture occurred at the ground surface (rigid plane).
Scaling r= r= r= r= r= Thus, it reflects the energy at the ground surface. The author noticed
Em, α = 0.12 1.22 2.44 4.88 9.75 15.24 that the fraction might determine using α = 1.8 instead of 2 as an ex­
Ech, β = 0.09 143.9 61.2 49.65 21.8 11.96
Experiment (Weyandt, 140 56–80 30–69 14 12
plosion usually leads to cratering, and a part of mechanical energy is
2005) consequently lost. We believe that in our case, a correction of this co­
Em, α = 0.14 (Molkov, 140.84 38.5 12.77 4.86 2.68 efficient may be required (i.e., for CNG), using trial and error method.
Kashkarovb) The following analysis showed the observation of the change in energy
CNG, Em,α = 1.8 518.25 98.42 40.73 10.18 6.02
due to the fractions of α and β. At α = 0.45. It enabled us to predict the
CNG, Em,α = 0.14 127.6 34.88 10.87 4.4 2.43
CNG Em, α = 0.19 140.34 51.3 47.1 12.63 9.89 fraction of energy associated with physical and chemical energies. It
CNG Em, α = 0.21 141.5 59.62 45.12 25.42 13.5 represented theoretical analysis on a blast wave overpressure as a
Ech, β = 0.18 function of distance when the storage tank ruptured. Starting at α = 1.8.
Therefore, overpressures for each distance gave a reasonable prediction
compared to the experimental test. With different values of coefficient
data generated experimentally. Therefore, a model application, in this
(α), an increase in overpressure at further distances from the tank
study, to determine the overpressure storage CNG in blast wave (r‾P)
observed.
used, as shown in equation (8).
Fig. 5 observed the use of the methodology of non-related fire for
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
)̅
(
Ps tank rupture compared with previous experimental data. These data
8 were provided for overpressure gases. For example, experimental data
3
r ​ p= r
αE1 + βE2
and CNG pressure with (α at 0.14, 0.45, 0.45, and β at 0.18). The pre­
E1and E2 are mechanical energy and chemical energy, β, and α co­ diction of energy coefficient (α) for the designated distance to match the
efficients (0.09 & 0.12, respectively (Molkov, Kashkarova)). Molkov has experimental data was defined. Molkov (Molkov, Kashkarova) proposed
mentioned that the blast wave subject to energy release rate more than a methodology to determine the total energy. The mechanical energy
on the amount of energy, therefore he added the energy of combustion with the fraction of α = 0.45 illustrated and the amount of mechanical
of gas released into the atmosphere to the mechanical energy of com­ energy contributing to the blast wave estimated as αEm = 0.84 MJ. The
pressed gas (Molkov, Kashkarova). In the case of CNG if burns in the fraction stored in mechanical energy and chemical energy contribute to
presence of oxygen, carbon dioxide, and water produced. This process of the blast wave can be estimated as αEm = 0.4 MJ. The total amount of
combustion releases energy. If the energy released during a chemical chemical energy strengthen the blast wave is determined for both (α =
reaction, an exothermic reaction produced. The combustion of CNG gas 0.21 and β = 0.18), βEch = 19.35 MJ (see Fig. 4).
releases (CH4 = 50.1 kJ/g). It is an equivalent of 802.3 kJ/mol CNG.
Thus, To find the amount of air needed for combustion of this gas, since
the major chemical in CNG chemical structure is CH4 the following re­
action considered:

CH4+2O2+7.52N2→CO2+2H2O + 7.52N2 9

CH4 is Methane, O2 is Oxygen, CO2 is Carbon dioxide, H2O is water,

and N2 is Nitrogen. For calculations of the volume occupied by an un­
burned stoichiometric gas-air mixture, vu = unburn volume (2/3πru),
and vb = burned volume (2/3πrb) given below. The mass of stored gas is
m = 1.85 kg. Thus, the number of moles contained in this amount of
hydrogen is n = m/Mg = 1.654 kg/2.306 × 10− 3 kg/mol = 802.2 mol. In
the assumption of normal conditions, 1 mol of an ideal gas occupies
22.4 L. Thus, the number of moles, N: N = 1 mol CO2 + 2 mol H2O +
Fig. 4. CNG and experimental comparison with α and β

6
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

Fig. 5. Event tree risk model for CNG release.

Due to the limitation of previous methods and do not provide 7. Failure frequency analysis
detailed information about the contribution of all elements to the blast
wave, a blastwave approach proposed. In this study, the integration of An event tree used to assess the probabilities of outcomes designed.
total energy of compressed gas (CNG) described in the following; A proposed timeline matrix (TLM) based on occupational hazards
and expert judgment for the definition of on-set of hazardous events
∫r
presented. Evaluating the performance of layers, for example, the
f(a) = Et ​ dr 10
detection systems include the probability of responses within the time-
0

Et is the total energy includes internal energy and kinetics energy

equation 2.0, therefore, Table 8
( ) Timeline evaluation.
p ρu
Et = − 11 Time-Line 0 1–3 3–10 10–30 30–60 Consequence
y− 1 2
0 IE YES YES YES YES Minor
y is the ratio of specific heats, P is pressure,ρ is density, and u is 1-3 (min) NO DAS YES YES YES Injuries to people
velocity. In addition,the correlation between energy and stress used to 3-10(min) NO NO AFS YES YES Damage to equipment
define the total energy. The proposed model depends on the type of 10-30(min) NO NO NO FFR YES Structural Damages
application. 30-60(min) NO NO NO NO FBI Catastrophic Damage

7
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

frame before the onset of hazardous event induced conditions at the Table 10
targets described in Table 8. Does the protection layer obey the time The property damage categories (72 (2018). Standard, 2018) (Teeter et al.,
frame or not? If not, then possible technical malfunction. In case if the 2000).
protection layer follows the time frame, then a potential source of error Property Damage Remarks
may result from mislead or administrative mistakes. Damage Factor
The likelihood of categories and the property damage categories as 1. Slight 0–1 Minor damage, there’s no need for repairs.
low, moderate, and high per event/year obtained Table 9. At a low level 2. Light 1–10 Significant damage with some equipment repairs
(0.01–0.001), the possible death of one in a hundred to one in a thou­ 3. Moderate 10–25 Significant damage, requiring replacements for
many types of equipment.
sand while the property of 3.3% represents moderate when one possible
4. Heavy 25–60 Structural damages.
fire over one time the useful life of the process within 33 years. One 5. Major 60–100 Major widespread damages.
possible fire in 5–15 years at a probability of 6–20% considered high
level.
Table 10 illustrates the damage factor from 0 to 100 categorized as
Table 11
slight, light, moderate, heavy, and major. Comparing those values with
The Data set for probabilities of each Protection layer.
the probabilities in Table 8, we may say that a low damage with damage
Event Probabilities Reference(s)
factor of (0–1) equivalent to (1/1000–1/100). Thus, minor damage,
which may not be required repairing. IE 0.066 (72 (2018). Standard, 2018), (Teeter et al., 2000), (
In addition, the probability of the event may change with response Schneider, 1997)
DAS 0.089 (72 (2018). Standard, 2018), (Teeter et al., 2000), (
time Table 8. As the response to the event delay or failed, a high risk
Schneider, 1997)
most likely to happen. Reduction of the failure probability (event/year) AFS – –
may contribute to risk mitigation or reduction. Thus, response time is FFR 0.75 (72 (2018). Standard, 2018), (Teeter et al., 2000), (
essential. The table shows a different level of damage based on the Schneider, 1997)
FFR- 0.25 (72 (2018). Standard, 2018), (Teeter et al., 2000), (
probability of event per year from low to high. Thus, protection layers
(Fail) Schneider, 1997)
versus time play an essential role in the type and level of the expected FBI 0.4 (72 (2018). Standard, 2018), (Teeter et al., 2000), (
consequences. For example, structural damage may result if one possible Schneider, 1997)]
fire event occurred over one time the useful life of the process within 33 FBI- 0.6 (72 (2018). Standard, 2018), (Teeter et al., 2000), (
years, Table 9. If this compared to the result provided in Table 8, the (Fail) Schneider, 1997)

structural damage required (10–30 min). This means if we can reduce

the response time (i.e., training, technical improvement), the likelihood
of failure per year will be declined. Table 12
Table 11 shows the data set for the layers probabilities collected Threshold of harmful pressure effects on people adopted from (Global­
based on prior references. The outcomes of the current study applied NaturalGasVehicle(, 2018).
those values to determine the likelihood of the CNG event. Harmful effect Overpressure,kPa
The significance of the data provided in Table 12 is the human body, The threshold for skin laceration from flying glass 6.9–13.8
which is relatively resistant to static overpressure compared to a rigid The threshold for serious wound from flying glass 13.8–20.7
structure such as buildings. The main parts of the body directly sus­ 1% eardrum rupture probability 16.5
ceptible to the damaging effects of overpressure are the eardrums and 50% serious wound from flying glass 27.6–34.5
50% probability of eardrum rupture 43.5
lungs. Lung damage can be fatal and an example of the consequences in
Near 100% serious wound from flying glass 48.3–55.16
terms of probability of injury or fatality. Table 12 shows the threshold 90% probability of eardrum rupture 84
harmful pressure effect, which can be classified starting from flying glass 1% probability of fatality due to lung hemorrhage 100
at 6.9 to the probability of 99% of fatality at 200 KPa. 50% probability of fatality due to lung hemorrhage 140
The table has given only for learning. 99% probability of fatality due to lung hemorrhage 200

Lee (Brinkley and Lewis, 1959) states, “Explosions cause fewer

serious accidents than fire but more than toxic release. When toxic re­ damage levels. To decide whether the risk is controlled or not. It
leases occurred, however, they often inflict greater loss and damage than compared to the risk tolerance profile(s).
fire. The explosion usually regarded as having disaster potential greater Fig. 7 shows the tolerable risk based on given values tabulated in
than that of fire but less than that of toxic release.” Fire and explosion Table 11 (see Fig. 6). Thus, the probability of each I.D line can be
risk modeling for CNG cylinder designed and possible consequences determined, and the level of risk defined remarked as a red dot occurred.
were defined systematically based on selected protection layers. Fig. 8 Consequently, the decision can be made if the existing risk met the risk
represented the event tree (ET), illustrated protection layers, and the I.D. tolerance criteria or not. Fig. 8, where the damage level can be classified
lines include: (S1–S11) along with the damage level (1–5) for each one. from low level to high level from (1–5). The maximum value of 5.0 in­
Possible serious damage to types of equipment if the failure of all of dicates critical damage at the frequency. S9 represents the possibility of
the protection layers occurred. However, the likelihood tolerable limit VCE within 5–15 years. The possibility of having a moderate conse­
for CNG scenarios provided. It illustrates the likelihood of fire and ex­ quence is 0.044. In other words, there will be one possible fire over one
plosion scenarios for given protection layers versus the life and property time the useful life of the process within 44 years.
Fig. 9 illustrates the outcomes of given scenarios. The frequency and
Table 9 damage curve gives a possible incident scenario for the CNG cylinder
The likelihood of categories (72 (2018). Standard, 2018) (Teeter et al., 2000). based on the given event tree scenarios. The possible accident events
Level Probabilities Event/ Remarks obtained and the damage levels were measured. The potential damages
Year initiated at the lower level and propagated to the maximum level depend
LOW 0.01 1/100 Possible fire control of the events. on the properties of the fuel. The maximum value of 5.0 indicates critical
0.001 1/1000 One possible fire within 1000 years damage at the frequency of (i.e., 6.48E-03). It represents the possibility
MODERATE 0.033 1/33 One possible fire over one time the of an explosion accident within a thousand years, while the possibility of
useful life of the process within 33 years having a damage level of 3.0 is (i.e., 4.47E-02). It indicates the likeli­
HIGH 0.066 1/15 One possible fire within 15 years.
0.2 1/5 One possible fire within 5 years
hood of the fire or/and explosion within a hundred years.

8
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

Fig. 6. Event tree (ET) for fire and explosion risk modeling for CNG onboard. Note: (Initial events(E),Pressure release(A1),Rapture of stabilizer(B1),Gas sensor(G1),
Ignition (I1),Fire sensor (X1), Delayed Ignition (D1),Extinguisher system (F1), Fire fighting system(H1)).

8. Conclusions overpressure sharply decreases from the point of explosion. To ensure

the safety of the vehicle and people onboard, additional safety measures
The instantaneous release CNG causing blastwave is studied. The are required. The safety measures may include blast barriers around the
study used COMSOL to design the CNG cylinder. The CNG cylinder was CNG tank. This will significantly limit the overpressure damage risk.
designed based on common industrial standards. The risk methodology Providing training and improving knowledge could support the future of
to assess overpressure impact due to the blast wave from high-pressure CNG technology, specifically onboard vehicles.
natural gas storage (CNG) is also tested. An event tree used to determine
the sequence of events and assess the magnitude of damage, which is CRediT authorship contribution statement
presented in the case of CNG.
The results of the CNG release are compared with the experimental Ibrahim Abdullah Altuwair: Conceptualization, Methodology,
data. The mechanical energy coefficient, α, and the chemical energy Software, Formal analysis, Investigation, Writing - original draft,
coefficient β are observed to be in good agreement. It is noted that when Writing - review & editing. Faisal Khan: Conceptualization, Method­
α decreases, the probability of the damage increases, and the likelihood ology, Formal analysis, Writing - review & editing, Supervision, Project
of explosion becomes high. However, when β increases, the blast wave administration, Funding acquisition.
strength is relatively less affected.
The analysis of the potential consequences of the rupture of the CNG
cylinder using selected damage criteria on people and property is also Declaration of competing interest
studied. It is observed that the blast wave has a local impact. The overall
damage of the vehicle and surrounding area is significant; however, The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence

9
I.A. Altuwair and F. Khan Journal of Loss Prevention in the Process Industries 68 (2020) 104284

References

Baker, W.E., Cox, P.A., Westine, P.S., Strehlow, R.A., 1983. Explosion Hazards and
Evaluation. Elsevier Scientific Publishing Company.
Brinkley Jr., S.R., Lewis, B., 1959. Seventh Symposium International on Combustion. the
Combustion Institute, Pittsburgh, pp. 807–811.
Cable, A., Chetehouna, K., Gascoin, N., Settar, A., Van Herpe, K., Kadoche, M., 2019.
Experimental and numerical study of sprinkler fire protection systems: effect of
initial pressure on system performance. Journal of loss prevention 60, pp76–86.
Centre for Chemical Process Safety CCPS, 2010. Guidelines for Vapor Cloud Explosion,
Pressure Vessel Burst, BLEVE, and Flash Fire Hazards, 2nd. Wiley, New York.
Cheng, V., Wenhu, H., Jianguo, N., Yuanyuan, Y., 2012. High-resolution numerical
simulation of a methane explosion in bend ducts. Saf. Sci. 50, 709–717.
Chuanchuan, S., Li, M., Gai, H., Yingzhe, W., Jinyang, Z., Yan, L., Jun, H., 2018a.
Consequence assessment of high-pressure hydrogen storage tank rupture during fire
test. J. Loss Prev. Process. Ind. 55, 223–231.
Chuanchuan, S., Li, M., Gai, H., Yingzhe, W., Jinyang, Z., Yan, L., Jun, H., 2018b.
Consequence assessment of high-pressure hydrogen storage tank rupture during fire
test. J. Loss Prev. Process. Ind. 55, 223–231.
Dorofeev, S.B., 2011. Flame acceleration and explosion safety applications. J. Hazard
Mater. 52 (1), 1–150.
Eisenberg, N.A., Lynch, C.J., Breeding, R.J., 1975. Vulnerability Model: A Simulation
System for Assessing Damage Resulting from Marine Spills. Enviro Control Inc.,
Rockville, MD. Rep. CG-D-136-175.
Finney, D.J., 1971. Probit Analysis. Cambridge University Press, Cambridge. https://doi.
org/10.1002/bimj.19720140111 (accessed 20.05.19).
GlobalNaturalGasVehicle(, N.G.V., 2018. NGVStatisticsReports (accessed 10.03.19).
http://www.iangv.org/stats/NGV_Global_Stats1.htm.
Houf, W.G., Evans, G.H., Merilo, E., Goethe, M., James, S.C., 2012. Releases from
hydrogen fuel-cell vehicles in tunnels. Int. J. Hydrogen Energy 37 (1), 715–719.
Khan, F., Abbasi, A., 2001. Risk analysis of a typical chemical industry using ORA
procedure. J. Loss Prev. Process. Ind. 14, 43–59.
Fig. 7. The exposure categories of the CNG outcomes. Khan, M.I., Yasmin, T., Shakoor, A., 2015. Technical overview of compressed natural gas
(CNG) as a transportation fuel. Renew. Sustain. Energy Rev. 51, 785–797.
Khan, M.I., Yasmin, T., Khan, N.B., 2016. Safety issues associated with the use and
operation of the natural gas vehicle: learning from accidents in Pakistan. J. Braz. Soc.
Mech. Sci. Eng. 38 (8), 2481–2497.
Kravtsov, Alexander N., Jacob, Zdebski, Svoboda, Pavel, Pospichal, Vaclav, 2015.
Numerical analysis of explosion to deflagration process due to methane gas
explosion in underground structures. IEEE (accessed 05.05.19). https://doi: 10.11
09/MILTECHS.2015.7153696.
Kuo, K.K., 1986. Principles of Combustion. Wiley, New York. Retrieved from. https:
//openlibrary.org/books/OL2541323M/Principles_of_combustion. (Accessed 25
March 2019).
Lees, F.P., 1996. Loss Prevention in the Process Industries, second ed. Butterworth-
Heinemann, Oxford, UK.
Leveson, N., 2009. Moving beyond normal accidents and high-reliability organization.
A system approach to safety in complex systems 30 (2–3), 227.
Fig. 8. The damage level of each I.D line(s).
Lewis, B., von Elbe, G., 1987. Combustion, Flame, and Explosion of Gases, third ed.
Academic, pp. 566–573.
Molkov, V., Kashkarov, S. (ND). Blast Wave from A High-Pressure Gas Tank Rupture in A
Fire: Standalone and Under-vehicle Hydrogen Tanks. Hydrogen Safety Engineering
University of Ulster, Newtownabbey, UK.
Molkov, V., Kashkarov, S. “Blast Wave from A High-Pressure Gas Tank Rupture in A Fire:
Standalone and Under-vehicle Hydrogen Tanks” Hydrogen Safety Engineering
University of Ulster, Newtownabbey, UK.
OSHA, 1970. Vessel Ruptures” Incident Statistics Report for the USA. OSHA, Washinton
(accessed 04.05.19). https://www.osha.gov/pls/imis/accidentsearch.accident_deta
il?id=46526.015.
Schneider, P., 1997. Limit states of process equipment components loaded by a blast
wave. J. Loss Prev. Process. Ind. 10 (3), 185–190.
Schultz, E., Shepherd, J., 2019. Validation of Detailed Reaction Mechanisms for
Fig. 9. The illustration of the probability outcomes for CNG cylinder. Detonation Simulation. In: Explosion Dynamics Laboratory Report, FM9vols. 9–5.
California Institute of Technology, Pasadena, CA.
Shepherd, J., 2009. Detonation in gases. Proc. Combust. Inst. 32 (1), 83–98.
the work reported in this paper. Sochet, I., 2010. Blast effects of external explosions. Eighth International Symposium on
Hazards, Prevention, and Mitigation of Industrial Explosions. Japan, Yokohama.
https://www.researchgate.net/publication/268050882. (Accessed 20 January
Acknowledgment 2019).
Teeter, S., Catchpole, D., Dalzell, G., Daniel, R., 2000. Fire system integrity assurance.
Authors acknowledge financial supports from the Natural Sciences Report No. 6.85/304, London, UK The International Association of Oil and Gas
Producers (OGP) (accessed 05/7/2019). https://www.ogp.org.uk.
and Engineering Research Council (NSERC) of Canada through Discov­
Vagsæther, K., 2010. Modelling of Gas Explosions. Faculty of Technology Telemark
ery Grant and Canada Research Chair Tier I program in Offshore Safety University College, Norway.
and Risk Engineering. Weyandt, N., 2005. Analysis of Induced Catastrophic Failure of A 5000 Psig Type IV
Hydrogen Cylinder. Southwest Research Institute report (Motor Vehicle Fire
Research Institute).
Appendix A. Supplementary data Zel’dovich, Y.B., Librovich, V.B., Makhviladze, G.M., Sivashinsky, G.I., 1969. Explosion
and Reacting Systems Gas Dynamics, vol. 10. 2nd International Colloquium,
Supplementary data to this article can be found online at https://doi. pp. 24–29.

org/10.1016/j.jlp.2020.104284.

10