2017 Bang Deflagration

International Journal of Thermal Sciences xxx (2017) 1e12
Contents lists available at ScienceDirect
International Journal of Thermal Sciences

journal homepage: www.elsevier.com/locate/ijts
Theoretical, numerical, and experimental investigation of pressure

rise due to deflagration in confined spaces
Boo-Hyoung Bang a, b, 1, Chan-Sol Ahn a, c, 1, Jong-Gun Lee a, Young-Tae Kim b,
Myung-Ho Lee b, Brad Horn d, Darren Malik d, Kelly Thomas d, Scott C. James e,
Alexander L. Yarin f, *, Sam S. Yoon a, **
a
School of Mechanical Engineering, Korea University, Seoul, 136-713, South Korea
b
Technology Development Team, Daewoo Inst. of Const. Tech., 60, Songjuk-dong, Jangan-gu, 440-210, Suwon, South Korea
c
Fire Research Center, Korea Institute of Construction Technology, 283, Goyang-daero, Ilsanseo-gu, 10223, Goyang-si, South Korea
d
Baker Engineering and Risk Consultants, INC, 3330 Oakwell Court, Suite 100, San Antonio, TX 78218, USA
e
Depts. of Geosciences and Mech. Eng., Baylor University, Waco, TX 76798, USA
f
Department of Mechanical & Industrial Engineering, University of Illinois at Chicago, 842 W. Taylor St., Chicago 60607, USA
a r t i c l e i n f o a b s t r a c t
Article history: Estimating pressure rise due to deflagration in a fully or partially confined space is of practical impor-
Received 31 January 2017 tance in safety design of a petrochemical plant. Herein, we have developed a new theoretical model to
Received in revised form predict the pressure rise due to deflagration in both fully and partially confined spaces. First, the theo-
19 May 2017
retical model was compared and validated against experimental data from the closed-space experiments
Accepted 19 May 2017
Available online xxx
with hydrogen, methane, propane, and ethane. The theory predicted accurate pressure rises near the
stoichiometric regime for all fuel types; outside the stoichiometric regime, especially, for rich mixtures of
hydrocarbons with air, the theory over-predicted pressure rise since it does not account for soot for-
mation and the associated energy losses by radiation. Experimental investigation of propane and
hydrogen deflagration was conducted in a partially confined space and the theory-based predictions
agreed with the data up to 5%. Parametric numerical study was performed to investigate the effect of the
initial pressure and temperature of gaseous fuels on pressure rise.
© 2017 Elsevier Masson SAS. All rights reserved.
1. Introduction Explosions are categorized as either detonations or deflagra-

tions depending on the flame propagation speed. When the flame
Safety of petrochemical plants is a primary consideration in speed is supersonic, it is associated with detonation wave which
their design. Although explosions at petrochemical plants are rare, incorporate a shock wave as in some bomb explosions. On the other
they can happen due to fuel leakage from outdated or malfunc- hand, deflagration is characterized by a low subsonic flame speed
tioning devices. Typically, leaks are detected early and disaster is and is characteristic of gas explosions studied here in relation to
thus mitigated. However, when detection system fails or the plant hydrogen, methane, propane, and ethane explosions. Fuel gas
is subject to an unexpected situation (e.g., a terrorist attack), fuel leakage and ignition may occur inside a confined space where the
leaks can be considerable and could lead to catastrophic explosions. pressure increase could be sufficient to rupture the confining
To minimize damage from potential explosions, safety factors must structure. A liquid fuel tank located close to a fire experiences
be incorporated into the structural designs for pipes and storage heating and its content may boil, leading to highly pressurized fuel
tanks. Because mostly structural damage is inflicted by a sudden tank that eventually ruptures resulting in an explosion of the
increase in pressure due to an explosion, accurate predictions of boiling-liquid and expanding-vapor. Even in an open space, a
pressure rise are required. pressure rise can happen due to obstacles. For this reason, we
considered deflagration in both confined and partially-confined
* Corresponding author.
systems.
** Corresponding author. Jo and Crowl [1,2] proposed a flame-growth model to predict the
E-mail addresses: ayarin@uic.edu (A.L. Yarin), skyoon@korea.ac.kr (S.S. Yoon). maximum pressure from hydrogen and methane explosions in a
1
Equal contribution.
http://dx.doi.org/10.1016/j.ijthermalsci.2017.05.019
1290-0729/© 2017 Elsevier Masson SAS. All rights reserved.
Please cite this article in press as: B.-H. Bang, et al., Theoretical, numerical, and experimental investigation of pressure rise due to deflagration in
confined spaces, International Journal of Thermal Sciences (2017), http://dx.doi.org/10.1016/j.ijthermalsci.2017.05.019
2 B.-H. Bang et al. / International Journal of Thermal Sciences xxx (2017) 1e12
confined vessel. Zhou et al. [3] compared their model against

experimental data for a premixed laminar flame case. Razus et al. Mf
Cf ¼ : (5)
[4] experimentally investigated pressure rise resulting from pro- Mf þ Mair
pane explosion in a confined space at different initial pressure and
The equivalent specific heat is evaluated as
temperature. Mitu et al. [5] acquired data on pressure rise due to
ethane explosion in a confined space for the initial pressure (P0) c M þ cv;air Mair
and temperature (T0) ranges of 20e130 kPa and 298e423 K, cv;eq ¼ v;f .f : (6)
respectively. Bang et al. [6,7] predicted pressure rise resulting from meq Mf mf þ Mair =mair
a worst-case scenario of strong explosion of hydrogen in an open
space. Luijten et al. [8] presented a laminar flame speed model to In Eqs. (5) and (6), Mf and Mair are the initial masses of fuel and
predict the maximum pressure rise in a confined area. For defla- air in the mixture, respectively, cv,f and cv,air are the specific heats of
gration, the model proposed by Jo and Crowl [2] was shown to be the fuel and air, respectively, mf and mair are the molecular masses of
more accurate than its competitors listed in that work. However, the fuel and air, respectively, and meq is the equivalent molecular
their model is sensitive to the specified flame speed for a given fuel weight of the fuel-air-products mixture.
concentration, which requires acquisition of the instance-specific Consider the situation when Mair is fixed, while Mf can vary.
data, and is hence impractical for situations with different fuel Then, according to Eq. (4) through (6) the combustion temperatures
types and concentrations. A more versatile model that does not are given by the following expressions:
require flame-speed data is required.
To address this issue, here we propose a general and simple (i) For lean mixtures (Mf < Mf,s)
deflagration model which determines the flame speed. We
Q Mf
demonstrate the accuracy of this model by comparing its pre- Tbv ¼ T0 þ : (7)
dictions to the experimental data for hydrogen, methane, propane, cv;eq Mf þ Mair
and ethane combustion in a confined space. Moreover, we study
experimentally deflagration of propane and hydrogen in a partially
(ii) For a stoichiometric mixture (Mf ¼ Mf,s)
confined space, and use these data to further validate our model.
Lastly, parametric numerical study aims at the effect of the initial
Q Mf ;s
pressure and temperature p0 and T0, respectively, on the overall Tbv ¼ T0 þ : (8)
cv;eq Mf ;s þ Mair
pressure rise.
2. Theoretical and numerical models (iii) For the rich mixtures (Mf > Mf,s)
2.1. Theoretical model for confined areas Q Mf ;s

Tbv ¼ T0 þ ; (9)
cv;eq Mf þ Mair
Combustion can be accompanied by pressure rise [9]. The aim of
the present sub-section is to evaluate and compare it with the with Mf,s being the stoichiometric fuel mass.
experimental evidence. The molar concentration of fuel Cf, molar, a parameter given in the
The balance of the total (physical and chemical) internal energy experiments, is
E reads .
Mf mfuel
EðT0 Þ ¼ EðTbv Þ; (1) Cf ;molar ¼ . : (10)
Mf mfuel þ Mair =mair
where T0 is the initial temperature and Tbv is the temperature of
combustion products. Using Eq. (10), Eqs. (7)e(9) can be transformed into:
Because
(i) For lean mixtures (Mf < Mf,s)
EðT0 Þ ¼ cv;eq T0 þ QCf ; (2)
Q Cf;molar
Tbv ¼ T0 þ . : (11)
EðTbv Þ ¼ cv;eq Tbv ; (3) cv;eq C
f;molar þ mair mfuel 1 Cf;molar
one obtains from Eqs. (1)e(3) that
Q (ii) For a stoichiometric mixture (Mf ¼ Mf,s)

Tbv ¼ T0 þ Cf ; (4)
cv;eq
Q Cf ;molar;s
Tbv ¼ T0 þ . : (12)
cv;eq C
where Q is the heat release per unit mass of reactive mixture and f;molar;s þ mair mfuel 1 Cf;molar;s
cv,eq is the equivalent specific heat of the fuel-air-products mixture
at constant volume. Cf is the dimensionless initial fuel concentra-
tion in the reactive mixture (either lean, stoichiometric, or rich), (iii) For rich mixtures (Mf > Mf,s)
!
Q Cf;molar;s 1
Tbv ¼ T0 þ . . : (13)
cv;eq 1 Cf;molar;s Cf;molar 1 Cf ;molar þ mair mfuel
B.-H. Bang et al. / International Journal of Thermal Sciences xxx (2017) 1e12 3
For combustion, the internal energy balance requires that for a dynamics tool, FLACS [10], has been used to simulate 3D distribu-
spatially averaged temperature, T, and the spatially averaged fuel tions of pressure and temperature for various explosion scenarios.
concentration, C f , the following equality holds The governing equations used in FLACS include the mass (21),
momentum (22), enthalpy (23), fuel (24), and turbulent kinetic
EðT0 Þ ¼ E T ; (14) energy balances (25) plus the equation for the dissipation rate of
the turbulent kinetic energy (26)
which means that
Q vr v m_
ðb rÞ þ b ru ¼ ; (21)
T ¼ T0 þ Cf C f : (15) vt v vxj j j V
cv;eq
Using Eqs. (4) and (15) yields
v v vP v
ðbv rui Þ þ bj rui uj ¼ bv þ bj sij þ Fo;i þ bv Fw;i
T T0 C vt vxj vxi vxj
¼ 1 f ¼ h; (16)
Tbv T0 Cf þ bv ðr r0 Þgi ;
(22)
where h indicates the completeness of combustion.
The equation of state applied to the effective volume occupied
!
by fuel, air, and combustion products, V0, is v v v m vh DP Q_
ðb rhÞ þ b rhuj ¼ bj eff þ bv þ ; (23)
M vt v vxj j vxj sh vxj Dt V
pV0 ¼ RT; (17)
meq
!
v v v m vYfuel
where p is pressure. b rY þ b ru k ¼ bj eff þ Rfuel (24)
Equations (15)e(17) relate pressure to the completeness of vt v fuel vxj j j vxj sfuel vxj
combustion
!
MR QCf v v v m vk
p¼ T0 þ h : (18) ðb rkÞ þ b ru k ¼ bj eff þ bv Pk bv rε; (25)
meq V0 cv;eq vt v vxj j j vxj sk vxj
Equations (4) and (18) indicate that
!
v v v m vε ε2
p ¼ p0 ½1 þ ðε 1Þh; (19) ðb rεÞ þ b ru ε ¼ bj eff þ bv Pε C2ε bv r ;
vt v vxj j j vxj sε vxj k
where p0 is the initial pressure and ε ¼ Tbv =T0 >1. In particular, the (26)
pressure achieved at the end of the combustion process (i.e. h ¼ 1),
pend, is where Fo;i is flow resistance due to sub-grid obstructions and Fw;i is
flow resistance due to walls. bv is volume porosity, bj is area
pend
¼ ε; (20) porosity, r is the density of the gas mixture, xi are the Cartesian
p0
coordinates, ui are the velocity components, m_ is mass release rate,
V is volume where the mass is released, g is gravity, P is pressure,
with ε ¼ Tbv =T0 calculated from Eqs. (11)(13).
and sij is the stress tensor. Q_ is the rate of heat release in the
combustion reaction, Yfuel is the fuel mass fraction and Rfuel is the
fuel reaction rate.
2.2. Numerical model for partially-confined areas
Semi-empirical constants used in the k-ε turbulence model are
taken from Launder et al. [11]. The compressible 3D Navier-Stokes
The thermodynamic theory developed in sub-section 2.1 can be
equations are solved using the finite-volume method. FLACS uses
applied not only to a completely confined space, but also to a
a second-order scheme to resolve diffusive terms and a second-
partially confined space by tuning the completeness of combustion
order hybrid scheme to resolve convective terms. The time
(h); therefore, the model is versatile and useful for various explo-
marching steps are advanced by a first-order backward Euler
sion scenarios. Because the analytical model does not consider any
scheme to ensure numerical stability. The Semi-Implicit Method for
heat losses (e.g. radiation losses due to soot formation), it has a
Pressure-Linked Equations (SIMPLE), the pressure-correction al-
tendency to over-predict pressure rise in rich hydrocarbon fuels, as
gorithm [12] is applied and modified to handle compressible flows
is shown below. Therefore, the thermodynamic theory of sub-
with an extra source terms, which accounts for the work done by
section 2.1 would be expected to be a useful conservative esti-
compression in the enthalpy equation. Iterations are repeated until
mate for such cases. However, an over-estimated pressure rise
a mass residual of less than 104 is obtained.
could impact plant financing if construction costs are derived from
The influence of the wall is modeled using a wall function where
an excessively conservative estimate. For this reason, more detailed
the dimensionless wall distance is:
simulations such as a fully three-dimensional model may be
required.
Structures located away from a potential explosion are sub- rCm0:25 k0:5 y
yþ ¼ ; (27)
jected to much lower pressures compared to the maximum pres- m
sure near the source. In addition, if there are many obstacles
between the explosion and the structure of interest, then the where y is the distance from the wall point to the wall. The wall
structure would be subjected to a much lower pressure than the point is defined as the point closest to the wall where transport
maximum pressure, commensurately reducing the required pro- equations are solved. The wall friction term in the momentum
tection scheme. For this reason, a fully 3D computational fluid equation becomes
number based on fluid flow velocity is 0.5, which means that each
Aw time step is assigned so that fluid propagates only about half the
Fw;i ¼ bv tw;i ; (28)
V length of a finite-volume cell each time step. The simulation
Here Aw is the area of the wall and tw,i is the shear stresses duration was limited to 0.5 s, which is long enough for an exploding
caused by the wall given by gas to achieve maximum pressure. Simulations were conducted on
a 3.30-GHz PC with 8 cores 2 processors and 128 GB of RAM. Fig. 2
8 ui shows the actual experimental space in which an explosive gas was
>
> m if yþ < E þ contained. In FLACS, this container of 7.3 14.6 3.7 m3 di-
>
> y
>
< mensions was included inside the computational domain of
tw;i ¼ rui kCm0:25 k0:5 ; (29) 100 100 50 m3. The Euler equations are discretized for a
>
> ! if yþ Eþ
>
> yþ boundary element. The stagnant air at an ambient temperature of
>
: kEþ þ ln þ
E 303 K was used as an initial condition. The computational geometry
is depicted in Figs. 2e4. Inside the container, about 200 5-cm-
diameter cylindrical stainless steel pipes were regularly aligned to
where Eþ is a constant in wall functions and k is the von Karman
simulate a congested-space scenario as shown in Fig. 3, which also
constant.
indicates the steel walls as thick, black lines with the open side
Laminar burning velocity is a function of pressure
temporarily sealed with a plastic sheet (see Fig. 1).

P gP
SL ¼ S0L ; (30)
P0 2.3. Experiments for model validation
where gP is a fuel-dependent parameter. In the quasi-laminar To validate numerical predictions from FLACS, the experimental
regime, the turbulent burning velocity is: apparatus in Fig. 2 was filled with an explosive gas. Total flammable
( " #) volume is 391.5 m3. The vertical pipes in Fig. 3 create congestion,
R 0:5 induce turbulence to increase flame speed. As a result, the designed
SQL ¼ SL 1 þ amin ;1 ; (31)
3 congestion due to the pipes produced a relatively high pressure, as
opposed to the case without the pipes. The fuel-air mixture was
where R is the flame radius and a is a fuel-dependent constant. ignited near the center of the rear wall to produce a vapor cloud
The turbulent burning velocity is a simplified form of the explosion (VCE). The partially confined VCE could vent through the
expression [13]: open side producing external blast loads and, in close proximity,
fire exposure as the flame escaped the test apparatus and the
ST u0 expelled unburned gas continued to burn.
¼ 0:875K 0:392 ; (32)
SL SL As shown in Fig. 4, multiple pressure gauges were installed at
different locations on the floor, walls, columns, and roof. To identify
where K is the Karlovitz stretch factor. The Courant-Friedrichs-Levy asymmetries, pressure gauges were installed in pairs with
Fig. 1. Schematic of possible explosion scenarios in petrochemical plant. The bottom left shows an explosion in a congested area and the bottom right shows an explosion in a
confined tank.
Fig. 2. Experimental setup for gas explosion.
Fig. 3. Layout of the 5-cm-diameter cylinders to simulate a congested space.
Fig. 4. Interior overpressure gauge locations.
Table 1
Case study for comparison of Eqs. (19) and (20) with experiments.
Fuel Oxidizer Calculated range with MF Q Reference

[%] (LHV)
[MJ/kg]
Case 1 Hydrogen Air 5e70 120 Sheroeder et al.

Case 2 Methane 4e16 50 Cashdollar et al.
Case 3 Propane 2e10 46
Case 4 Ethane 3e10 47.8 Holtappels et al.
symmetric locations. Pressure, high-speed video (3000 fps), and 3. Results and discussion
high-definition video (30 fps) were recorded for all tests. A total of
17 dynamic pressure gauges captured the pressure history within 3.1. Thermodynamic theory versus experiment data
and outside the experimental apparatus, with 10 and 7 pressure
gauges for the interior and exterior locations, respectively. Because The theoretically predicted pressures of the gas explosions were
of the symmetrical locations of the interior gauges, only 5 sets of estimated using Eq. (19) for the different fuel types (hydrogen,
data are presented for Rear, Wall, Column, Floor, and Roof locations, methane, propane, and ethane) whose initial pressure (p0) and
as described in Fig. 4. Additional 7 pressure gauges were used for temperature (T0) varied depending on the scenario. In addition, the
the exterior locations from MP1 to MP7. These exterior gauges were initial fuel mole fraction, MF, was also varied. Ultimately, the
located at the center line, noted in Fig. 4, with at the distances of maximum pressure rise, pmax, was studied as a function of p0, T0,
z ¼ 11.9, 16.5, 22.6, 30.2, 37.8, 45.4, and 53 m away from the open and MF.
end of the exploding container. Table 1 lists the four explosion scenarios considered. Because
The pressure measurement system consisted of dynamic pres- hydrogen and other hydrocarbon fuels have been widely used in
sure gauges cabled to line-powered signal conditioners. Once the petrochemical plants, pmax as a function of MF (or concentration) is
target fuel concentration was obtained, the fans used to circulate of significant interest to petrochemical engineers. Air is the oxidizer
the fuel-air mixture were shut off for a period of 4 min prior to for all explosions. Heat release in combustion is greatest for
ignition to allow the mixture to become quiescent and eliminate hydrogen (120 MJ/kg) while other fuels belong to the 45e50 MJ/kg
any pre-ignition turbulence. Finally, the cloud was ignited with a range. MF for hydrogen was varied from 5 to 70% (30% stoichio-
low energy (~50 J) exploding fuse wire at the center of the back metric). The MF for other fuels belonged to the 2e16% range (the
wall. stoichiometric values of 9.5, 4.5, and 6% for methane, propane, and
Fig. 5. Comparison of pmax predicted by the theory of sub-section 2.1, in particular, Eq. (19), for hydrogen experiments [14e16] at T0 ¼ 298 K and hydrocarbons experiments [14e16]
at T0 ¼ 293 K. The stoichiometric MF is indicated with an asterisk.
depending on whether combustion takes place in an open, partially

closed, or completely closed space because heat losses can vary
significantly.
If the burning velocity is not fast enough, some of the unburned
gas mixture is expelled from the partially closed confinement
without being burned (he < 1) resulting in a lower final pressure. In
other words, he ¼ 1 implies a sufficiently high burning velocity
when the entire fuel-oxidizer mixture is rapidly burned, even
though the confinement can be partially open, i.e. this yields a
conservative estimate of the highest possible pressure.
Fig. 5 compares pmax for hydrogen, methane, propane, and
ethane deflagration. The solid curves are the theoretical predictions
of subsection 2.1, in particular, Eq. (19) and the symbols are the
experimental data from Refs. [14e16]. The dashed theoretical
curves account for the radiation losses due to soot formation and
are discussed below. For all cases, 8 < pmax < 10 bar was observed
for stoichiometric mixtures. In Fig. 5a the theoretical predictions for
hydrogen agree not only near the stoichiometric regime, but also
for the lean and rich mixtures (MFs). Close agreement over a wide
Fig. 6. Comparison of pmax from Eq. (19) with the data from the hydrogen experiments
with different values of T0.
range of MFs implies that heat losses were negligible in the
experiment with hydrogen explosions. And, indeed, no soot was
formed in this case and thus thermal losses from the transparent
blue flame were minimal. On the other hand, in Fig. 5b e d, there is
ethane, respectively). All experiments in the previous studies
a close agreement only for the lean and stoichiometric MFs. For
[14e16] were acquired in a closed space suggesting a near complete
fuel-rich mixtures, the theoretical model tends to over-predict pmax
combustion at the end, namely he ¼ 1. They also imply no heat loss
because heat loss due to soot formation and radiation occurs in
because of the closed space. The value of the completeness of
these experiments with hydrocarbons. Most of the soot is
combustion, he, should be considered as a fitting parameter
Fig. 7. Pressure pmax as a function of T0 and p0 for propane (a,b) and ethane (c,d). Experimental data for propane and ethane are from Refs. [4,5,20] and [4,5,20], respectively.
generated in rich mixtures, and also more soot is formed when theoretical predictions for different T0 as correspond to different
more carbon is available. For example, there is more carbon in values of ε ¼ Tbv =T0 , as per Eq. (20).
propane (C3H8) than in ethane (C2H6) or methane (CH4); thus de- When T0 increases, the value of pmax decreases. This is because
viations between the theoretical predictions and experimental data the total number of the initial moles of mixture was decreased by
increase from methane to propane. removal of some of the fuel-air mixture from the confined space, as
Recall that he ¼ 1 was used to calculate the theoretical curves in mentioned above. With less fuel, pmax decreases. From Eqs. (19) and
Fig. 5, which overestimates the amount of burned fuel-air mixture (20), ε ¼ Tbv/T0 ¼ pmax/p0 when he ¼ 1. If T0 increases, ε decreases at
in partially closed confinements, especially in the rich mixtures for a fixed adiabatic flame temperature (~Tbv) in a closed space,
methane, ethane and propane. To account for the radiation heat yielding a decreased pmax for fixed p0 [17e19].
loss due to soot formation for fuel-rich mixtures, he was adjusted so For a stoichiometric mixture, the heat release is maximized. The
that the models matched the data. A linear dependence was used as fact that heat losses happened, manifests itself in a slight over-
he ¼ ea MF þ b, where a ¼ 20 and b ¼ 3.3 for methane; a ¼ 35 and prediction of pmax seen in Fig. 6 the stoichiometric cases. Howev-
b ¼ 4.5 for propane; and a ¼ 24 and b ¼ 3.5 for ethane. The dashed er, the overall good agreement of the theoretical predictions with
curves in Fig. 5b e d shows an improved comparison between the the experimental data for hydrogen confirms that the effects of T0
experiments and the theoretical model with this adjustable he. are accurately accounted for by the model.
Sheroeder et al. [16] presented confined-space experimental Fig. 7 demonstrates the effects of p0 and T0 on pmax during ex-
data for pmax for different stoichiometric mixture temperatures of plosion for 3.2% MF 5.7% (lean fuels). The experimental data for
T0 ¼ 293, 373, 473, and 523 K for hydrogen at a fixed p0. Within a propane and ethane were adopted from Refs. [4,5,20]. The lean MF
confined space at p0 being constant, the only way to vary T0 was to range was where the theory of sub-section and the experiment
extract a certain amount of the fuel-air mixture; otherwise, both T0 compared favorably, as mentioned above.
and p0 would increase because of confinement. Fig. 7a and c shows that as p0 increases from 0.4 to 1.2 bar
The pressure pmax at T0 ¼ 293 K is shown in Fig. 5 and not (propane) and 1.3 bar (ethane), pmax also increases with a slope that
included in Fig. 6. The comparison for hydrogen, as mentioned increases with MF. More powerful explosions (greater pmax) are
above, revealed the best agreement between the theory of sub- achieved when the fuel-air mixture approaches the stoichiometric
section 2.1 and the experimental data because hydrogen combus- ratio. For a given value of ε ¼ Tbv/T0, the relation between pmax and
tion does not produce soot with the corresponding radiative heat p0 is linear according to Eqs. (19) and (20). This linear relation was
losses (the case of he ¼ 1). It should be emphasized that in Fig. 6 the validated by the experimental data. Fig. 7b and d shows that when
Fig. 8. Pressure pmax for varying initial pressure p0 at a fixed T0 ¼ 298 K for hydrogen, (b) methane, (c) propane, (d) and ethane.
T0 increases from 293 to 423 K for propane and ethane at a fixed p0, with theoretical and numerical predictions. Both experiments with
pmax decreases for all values of MF because of the required reduc- propane (Case 1e1 and 1e2) and hydrogen (Case 2) were con-
tion in fuel mass. ducted for fuel-air mixtures near the stoichiometric ratios of
4.34e4.36% and 22.5%, respectively, resulting in the maximum
3.2. Parametric studies in the framework of the analytical pressures of pmax ¼ 43 and 690 kPa for propane and hydrogen,
thermodynamic model respectively. As shown in Table 2, the same test was repeated twice
(Case 1e1 and 1e2), which yielded the same pmax and nearly the
Fig. 8 and 9 compare the effect of p0 and T0, respectively, on the same impulses and the effective duration. Therefore, the mea-
maximum pressure pmax for varying MF of hydrogen, methane, surement was quite reliable with an uncertainty level of less than
propane, and ethane. The range of MF variation spans lean to 8%.
stoichiometric ratios. Rich MF ratios were excluded because the Recall that these experiments were conducted in a congested
theoretical model tends to over-predict pmax there. Figs. 8 and 9 are space with an open wall at one-side (i.e., partially confined, and in
useful to field engineers to estimate pmax given the initial pressure the CFD calculations too), as described in Figs. 3 and 4, while a
and temperature p0 and T0 because it can be directly evaluated from completely confined space was assumed in the theoretical model of
these graphs without any numerical simulations. sub-section 2.1 (Fig. 5).
For all cases, the highest value of pmax was observed at the Comparing the partially- and completely-confined scenarios
stoichiometric MF ratios. In Fig. 8, pmax increases when the initial offers an insight into two different combustion situations. For
surrounding pressure, p0, increases. In Fig. 9, p0 ¼ 1 bar, while T0 example, pmax ¼ 700 kPa (MF ¼ 22.5%) predicted by the analytical
changes from 300 to 500 K, which is a plausible temperature range model for the completely confined case in Fig. 5a, while the
at a petrochemical plant. Again, propane yields the highest pmax, experimental measurement for the partially-confined space
followed by ethane, methane, and hydrogen, which corresponds to revealed pmax ¼ 690 kPa. This minimal difference in these values of
MFs of 4.5% for propane, 6% for ethane, 9.5% for methane, and 30% pmax implies that the combustion process between is nearly iden-
for hydrogen. tical for hydrogen in these two cases. However, for propane, there is
a significant difference in pmax between the completely and
3.3. Numerical predictions by FLACS in comparison with the incompletely confined cases. From Fig. 5c, pmax ¼ 900 kPa
theoretical predictions of sub-section 3.1 and experimental data (MF ¼ 4.34%) for the completely confined space, while
pmax ¼ 43 kPa for the partially confined space. This difference is due
Table 2 summarizes our two experiments used for comparison to the lower burning velocity of propane that decreases the fuel-air
Fig. 9. Pressure pmax for varying initial temperature T0 at a fixed P0 ¼ 100 kPa for (a) hydrogen, (b) methane, (c) propane, and (d) ethane.
Table 2 Table 3 compares the values of pmax from the experiments,

Peak values of pmax after ignition. theoretical predictions, and the CFD model at various physical lo-
Case Fuel MF [%] pmax Impulse Effective duration [ms] cations (i.e., rear wall, side walls, columns, floor, and roof as illus-
[kPa] [kPa-ms] trated in Fig. 4). For methane Case 1, the theoretical prediction is
1e1 Propane (C3H8) 4.34 43 1448 73 pmax ¼ 89 kPa assuming he ¼ 0.1. In a completely confined space,
1e2 4.36 43 1565 77 he ¼ 1, but this value is reduced here, since combustion takes place
2 Hydrogen (H2) 22.5 690 2850 10 in a partially confined space. The corresponding CFD prediction for
methane is pmax ¼ 78 kPa, which also over-predicts the experi-
mentally measured value of pmax ¼ 43 kPa.
For hydrogen Case 2, the theoretical pressure is pmax ¼ 727 kPa
Table 3
and from the CFD predictions it is pmax ¼ 715 kPa, while for the
Experimental values for pmax compared to the theoretical analytical predictions from
Eqs. (19) and (20) and the FLACS model. experimental value is pmax ¼ 690 kPa. These favorable comparisons
indicate that the theoretical and numerical models are quite ac-
Location Experimental Theory CFD
curate in cases where a burning velocity is high and soot formation
[kPa] [kPa] [kPa]
and the associated thermal losses are minimal.
Case 1 Case 2 Case 1 Case 2 Case 1 Case 2
Table 4 shows the grid-dependence studies carried out for
Interior 1 Rear 43 690 89 727 74 654 computational discretizations ranging from 1 to 7 million nodes.
2 Wall 43 680 78 715 Maximum pressures of pmax ¼ 78 and 715 kPa for propane and
3 Column 40 690 51 707
hydrogen, respectively, showed convergence at about 5 million
4 Floor 38 430 63 482
5 Roof 34 470 64 487 nodes. Thus, for all computations 5 million nodes were used.
Free-field 6 MP1 23 170 - 70 220 Figs. 10 and 11 show snapshots from the explosions in the ex-
7 MP2 19 140 58 175 periments and in the CFD simulations, respectively, for Case 1 and
8 MP3 15 71 43 107
Case 2 as combustion progresses in time. The images for propane
9 MP4 11 47 30 77
10 MP5 9.5 36 20 51
are up to 50 ms and those for hydrogen up to 20 ms. These results
11 MP6 7.7 28 14 38 reveal that the almost stoichiometric hydrogen explosion is more
12 MP7 6.3 23 12 25 powerful than that of propane. Fig. 12 compares pmax for these two
cases (i.e., propane and hydrogen) as a function of the distance from
the explosion apparatus. The location numbers from No. 1 to No. 5
Table 4 correspond to the Rear Wall, Side Wall, Column, Floor, and Roof, as
Summary of grid-dependence studies. described in Fig. 4 while the physical locations corresponding to No.
Case Fuel pmax [kPa] 6e12 are at z ¼ 11.9, 16.5, 22.6, 30.2, 37.8, 45.4, and 53 m away from
the open-end of the exploding container with an interval location
Experimental Number of nodes [106]
of approximately 4e8 m between pressure gauges. At No. 12 (which
1 (low) 3.2 (med) 5 (hi) 7 (very hi) is 53 m away from the container or ignition point), the pressure is
1 Propane (C3H8) 43 65 72 78 79 assumed to be atmospheric.
2 Hydrogen (H2) 690 310 658 715 718
4. Conclusion
mixing and slows down the chemical reaction in the partially A theoretical thermodynamic model for prediction of the
confined space. In a completely confined space, pmax was as high as maximum explosion pressure pmax was developed for lean, stoi-
900 kPa at MF ¼ 4.34%, both experimentally and theoretically (i.e. chiometric and rich fuel-air mixtures. The model was validated
via the analytical model of sub-section 2.1) (see Fig. 5c), implying against the experimental data for hydrogen, methane, propane, and
that the burning velocity of methane is practically as fast as that of ethane explosions in a confined space. The effect of initial gas
hydrogen in a completely confined space. temperature and pressure on pmax was also studied. The
Fig. 10. Propane (Case 1) and hydrogen (Case 2) explosions observed by the high-speed camera.
Fig. 11. Propane (Case 1) and hydrogen (Case 2) explosions from the numerical simulations.
Acknowledgement
This research was primarily supported by Daewoo Institute of

Construction and Technology. This work was also supported by the
National Research Council of Science & Technology (NST) grant by
the Korea government (MSIP) (No. CRC-16-02-KICT), NRF-
2016M1A2A2936760, NRF-2017R1A2B4005639, and NRF-
2013M3A6B1078879.
References
[1] Jo YD, Crowl DA. Flame growth model for confined gas explosion. Process Saf
Prog 2009;28:141e6.
[2] Jo YD, Crowl DA. Explosion characteristics of hydrogen-air mixtures in a
spherical vessel. Process Saf Prog 2010;29:216e23.
[3] Zhou B, Sobiesiak A, Quan P. Flame behavior and flame-induced flow in a
closed rectangular duct with a 90 bend. Int J Therm Sci 2006;45:457e74.
[4] Razus D, Brinzea V, Mitu M, Oancea D. Temperature and pressure influence on
explosion pressures of closed vessel propaneeair deflagrations. J Hazard
Mater 2010;174:548e55.
Fig. 12. Experimental and numerically predicted pmax as a function of distance. Note [5] Mitu M, Giurcan V, Razus D, Oancea D. Temperature and pressure influence on
ethaneeair deflagration parameters in a spherical closed vessel. Energy &
that No. 1e5 locations correspond to the interior pressure gauge locations, while No.
Fuels 2012;26:4840e8.
6e12 correspond to the exterior gauge-pressure locations, at z ¼ 11.9, 16.5, 22.6, 30.2,
[6] Bang B, Park H, Kim J, Al-Deyab SS, Yarin AL, Yoon SS. Analytical and numerical
37.8, 45.4, and 53 m, respectively.
assessments of local overpressure from hydrogen gas explosions in petro-
chemical plants. Fire Mater 2016.
[7] Bang B, Park H-S, Kim J-H, Al-Deyab SS, Yarin AL, Yoon SS. Simplified method
experimental data for propane and hydrogen deflagration in a for estimating the effect of a hydrogen explosion on a nearby pipeline. J Loss
Prev Process Ind. 2016;40:112e6.
partially confined space were acquired and also used to further [8] Luijten C, Doosje E, de Goey L. Accurate analytical models for fractional
validate the theoretical model, as well as the results of direct nu- pressure rise in constant volume combustion. Int J Therm Sci 2009;48:
merical simulations. For hydrogen at MF ¼ 22.5%, the measured 1213e22.
[9] Zeldovich YB, Barenblatt GI, Librovich VB, Makhviladze GM. The Mathematical
maximum pressure was pmax ¼ 690 kPa, while that predicted by Theory of Combustion and Explosions. New York: Consultants Bureau, Plenum
theory was 727 kPa. This close agreement indicates that the effects Press; 1985.
of a fully confined versus partially confined case are immaterial for [10] Gexcon A. Flacs v10. 4r2 user's manual. Confidential report, Gexcon AS, Ber-
gen, Norway. 2015.
rapidly exploding hydrogen. It also shows that soot formation and [11] Launder BE, Spalding DB. The numerical computation of turbulent flows.
the associated radiative heat losses are negligible for hydrogen. Comput Methods Appl Mech Eng 1974;3:269e89.
Overall, the developed theoretical thermodynamic analytical [12] Patankar S. Numerical heat transfer and fluid flow. Boca Raton: CRC Press;
1980.
model provides a conservative and accurate prediction for the [13] Bray KNC. Studies of the turbulent burning velocity. Proc R Soc Lond A Math
maximum explosion pressure pmax for the methane-, ethane-, Phys Eng Sci R Soc 1990:315e35.
propane- and hydrogen-air mixtures at different values of the [14] Cashdollar KL, Zlochower IA, Green GM, Thomas RA, Hertzberg M. Flamma-
bility of methane, propane, and hydrogen gases. J Loss Prev Process Ind.
initial fuel mole fraction, pressure and temperature, MF, p0 and T0,
2000;13:327e40.
respectively. The CFD model, FLACS, simulated congested experi- [15] Holtappels K., Liebner C., Schro € der V., Schildberg H.-P., Report on experi-
mental scenarios for propane and hydrogen explosions. The CFD mentally determined explosion limits, explosion pressures and rates of ex-
model had a tendency to over-predict the pressure rise compared plosion pressure rise-Part 2 Ethane, Ethylene, Propane, n-Butane, Ammonia
and Carbon Monoxide, (2006).
to experiments, but the differences remained within 10% and 5% for [16] Schro€ der V., Holtappels K., Explosion characteristics of hydrogen-air and
propane and hydrogen, respectively. hydrogen-oxygen mixtures at elevated pressures, in: International
Conference on Hydrogen Safety, 2005. explosion index of LPGeair and propaneeair mixtures. Fuel 2008;87:39e57.
[17] Gieras M, Klemens R, Rarata G, Wolan ski P. Determination of explosion pa- [19] Razus D, Krause U. Comparison of empirical and semi-empirical calculation
rameters of methane-air mixtures in the chamber of 40dm 3 at normal and methods for venting of gas explosions. Fire Saf J 2001;36:1e23.
elevated temperature. J Loss Prev Process Ind. 2006;19:263e70. [20] Razus D, Movileanu C, Brinzea V, Oancea D. Explosion pressures of hydro-
[18] Huzayyin A, Moneib H, Shehatta M, Attia A. Laminar burning velocity and carboneair mixtures in closed vessels. J Hazard Mater 2006;135:58e65.

2017 Bang Deflagration

Uploaded by

Copyright:

Available Formats

2017 Bang Deflagration

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2017 Bang Deflagration

Uploaded by

Copyright:

Available Formats

International Journal of Thermal Sciences xxx (2017) 1e12

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

Theoretical, numerical, and experimental investigation of pressure

1. Introduction Explosions are categorized as either detonations or deﬂagra-

conﬁned vessel. Zhou et al. [3] compared their model against

2.1. Theoretical model for conﬁned areas Q Mf ;s

one obtains from Eqs. (1)e(3) that

Q (ii) For a stoichiometric mixture (Mf ¼ Mf,s)

Fig. 2. Experimental setup for gas explosion.

Fig. 3. Layout of the 5-cm-diameter cylinders to simulate a congested space.

Fig. 4. Interior overpressure gauge locations.

Fuel Oxidizer Calculated range with MF Q Reference

Case 1 Hydrogen Air 5e70 120 Sheroeder et al.

depending on whether combustion takes place in an open, partially

Table 2 Table 3 compares the values of pmax from the experiments,

This research was primarily supported by Daewoo Institute of

You might also like